University of Southern California Dissertations and Theses

Out -of -pocket health expenditures by older adults in relation to age, race, and insurance

(USC Thesis Other)

Out -of -pocket health expenditures by older adults in relation to age, race, and insurance

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UM I a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note w ill indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6” x 9” black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UM I directly to order. ProQuest Information and Learning 300 North Zeeb Road, Ann Arbor, M l 48106-1346 USA 800-521-0600 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. OUT-OF-POCKET HEALTH EXPENDITURES BY OLDER ADULTS IN RELATION TO AGE, RACE, AND INSURANCE ©2000 by Susan Tracy Stewart A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA in Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY, (GERONTOLOGY / PUBLIC POLICY) May 2000 Susan Tracy Stewart Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. U M I Number: 3018035 __ ___ ® UMI UMI Microform 3018035 Copyright 2001 by Bell & Howell Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. Bell & Howell Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY O F SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES. CALIFORNIA 90007 This dissertation, written by S . u 5 c m ____ under the direction of /ljsjt. Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School in partial fulfillment of re quirements for the degree of DOCTOR OF PHILOSOPHY Dean of Graduate Studies Q a te March 9, 2000 DISSERTATION COMMITTEE S/ P ofj , csnCst i cc*- ^ Chai rperson Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Acknowledgments I would like to express my deep gratitude to Eileen Crimmins, who patiently and skillfully helped to mold me into an independent researcher. I thank her for her encouragement and guidance, her generosity and understanding, and her invaluable feedback and advice. I also extend my appreciation to my other committee members for their support, interest, and feedback: I thank Jeff McCombs for his valuable comments, and for nourishing the seed of my interest in health economics and encouraging me to continue research in this area. I am grateful to Kate Wilber for her advice, guidance and encouragement, and for her provision of real-life experience in public policy that helped to stimulate my interest in policy research. I also thank Teresa Seeman and Fred DeJong for their helpful feedback on my research proposals during the time they spent on my committee. I am thankful to Liz Zelinski for giving me the research experience that introduced me to and taught me to work with my dissertation data, and for providing a supportive home for me in her lab. I thank her and all of the faculty members who taught me so much during my years at USC and contributed to my personal growth and my development as a scientist. I would also like to thank the staff members at Andrus for their warmth and help. Special thanks to Gerry Jones and Sasha Bucur for their invaluable help with statistical programming. I am especially grateful to my husband, confidant, and closest friend, Iain. Through both good and difficult times, he unconditionally offered his love, support, understanding, and encouragement. I thank him for the many times he helped me work through my thinking, checked calculations, and gave me feedback on rough drafts. Iain is an incredible Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. role model for hard work and dedication, and it is by following his example that I was able to complete this herculean task. I extend deep appreciation to my parents, Norm and Louise Moore, for their constant love, their patience, and their unfailing support of my endeavors. Particular thanks to my mother for her in-depth reading of the final manuscript. I also thank June Stewart, Harry Sutcliffe, Bill Stewart and Linda Carr for encouraging my interest in academia, and for their support and love. I greatly appreciate the support and friendship of my colleagues in the doctoral program, which has helped to sustain me and make my time at Andrus pleasant. Special thanks to my good friend Christianne Lane for her steadfast support and encouragement. Christi’s sense of humor and good cheer provided much needed relief from the stresses of work, and she was always there when I needed someone to talk to. I also appreciate her feedback on parts of the manuscript, as well as the proofreading done by Patty Housen and Robin Engberg. Finally, I am grateful for the financial support of the USC Graduate School through a Haynes Foundation dissertation fellowship, and for financial support from the Leonard Davis School of Gerontology and the Associates of the Andrus Gerontology Center. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. iv Table of Contents List of Tables.............................................................................................................................vi List of Figures.......................................................................................... xi A bstract................................................................................................................................ xii Chapter 1. Introduction............................................................................................................ 1 1. The Importance of Studying Older Adults’ Out-of-Pocket Costs 2. Rationale for Focussing on Age, Race, and Insurance 3. Outline of Remaining Chapters Chapter 2. Literature Review and Conceptual Framework.............. 10 1. Review of Literature on Out-of-Pocket Health Spending 2. Conceptual Framework 3. Research Questions Chapter 3. Samples and M easures................................................................................... 31 1. Description of Samples 2. Description of Dependent Variables and Imputation Procedures 3. Measurement of Independent Variables Chapter 4. Methodological Approach.................................................................................53 1. Descriptive Analyses 2. Multivariate Analyses Chapter 5. Out-of-Pocket Spending: Descriptive Analyses...............................................64 1. Comparison of Imputed and Non-Imputed Dependent Variables 2. Descriptive Analyses of Out-of-Pocket Spending for All Respondents 3. Comparison to Other Reports of Out-of-Pocket Spending Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 6. Age and Out-of-Pocket Health C osts............................................................. 85 Chapter 7. Race and out-of-pocket health costs............................................................. 122 Chapter 8. Insurance and out-of-pocket health co sts..................................................... 172 Chapter 9. General Discussion............................................. 227 References C ited................................................................................................................... 253 Appendix Survey questions regarding health conditions in PSED and AHEAD.. . . 271 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. vi List of Tables 3.1 Percent with Missing and Imputed Data on Survey Questions Measuring Out-of-Pocket Costs in PSID, Summary Variables Created for Analyses............ 36 3.2 Number Imputed on Questions on Out-of-Pocket Spending in AHEAD 39 3.3 Means and Distributions on Predictor Variables in PSID and AHEAD (weighted).....................................................................................................................51 4.1 Percentage of Respondents with No Use and Percentage of Users with Zero Costs for Each Spending Variable in PSID and AHEAD...............................63 5.1 Comparison of Out-of-Pocket Costs Before and After Imputation for All Users of Each Service with Nonzero Costs in P S ID ...............................................66 5.2 Comparison of Out-of-Pocket Costs Before and After Imputation for All Users of Each Service in PSID...................................................................................67 5.3 Comparison of Out-of-Pocket Costs Before and After Imputation for All Respondents in P S ID ..................................................................................................68 5.4 Comparison of Total Out-of-Pocket Costs as the Sum of Non-Imputed and Imputed Cost Variables, for All Respondents in P S ID ............................................69 5.5 Comparison of Out-of-Pocket Costs Before and After Imputation for All Users of Each Service with Nonzero Expenses in AHEAD (weighted)................72 5.6 Comparison of Out-of-Pocket Costs Before and After Imputation for All Users of Each Service in AHEAD............................................................................. 73 5.7 Comparison of Out-of-Pocket Costs Before and After Imputation for All Respondents in AHEAD............................................................................................ 74 5.8 Comparison of Total Insurance Costs and Total Out-of-Pocket Costs as the Sum of Non-Imputed and Imputed Cost Variables, for All Respondents in P S ID ......................................................................................................................... 75 5.9 Comparison of Means Spending in PSID and AHEAD to Other Reported Measures of Older Adults’ Out-of-Pocket C osts.....................................................78 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. vii 5.10 Use and Uncovered Costs Among Users for Different Health Services in PSID and AHEAD...................................................................................................... 79 5.11 Proportion of Household Income Spent by Individuals on Total Out-of-Pocket Costs Excluding and Including Insurance Premiums in PSID and AHEAD .... 81 5.12 Percentage Spending Over 10%, 30% and 50% of Household Income on Total Health Care Costs (Excluding and Including Insurance Premiums) in PSID and AHEAD...................................................................................................82 5.13 Distribution of Out-of-Pocket Spending on Different Services in PSID .................83 5.14 Distribution of Out-of-Pocket Spending on Different Services in AHEAD .... 83 6.1 Distribution of Respondents by Age in PSID and AHEAD (weighted)..................97 6.2 Distribution of Out-of-Pocket Spending on Different Services by Age Group in PSID ........................................................................................................... 110 6.3 Distribution of Out-of-Pocket Spending on Different Services by Age Group in AH EA D .................................................................... I l l 6.4 Hierarchical OLS Regression Results for Out-of-Pocket Spending by Age in PSID and AHEAD: Ambulatory Care, Prescription Medications, and Total Out-of-Pocket Costs (in Logged Dollars)................................................. 114 6.5 Hierarchical Two-Stage Regression Results for Outpatient Surgery, Dental Care, and Equipment Use and Out-of -Pocket Expenditures (in Logged Dollars) by Age in PS ID ....................................................................... 115 6.6 Hierarchical Two-Stage Regression Results for Hospital and Nursing Home Use and Out-of -Pocket Expenditures (in Logged Dollars) by Age in PSID and AHEAD.............................................................................................. 116 6.7 Hierarchical Two-Stage Regression Results for Home Care Use and Out- of-Pocket Expenditures (in Logged Dollars) by Age in PSID and AHEAD . . 117 7.1 Means and Percentages on Independent Variables by Race in PSID and AHEAD...................................................................................................................... 139 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. viii 7.2 Mean and Median One-Year Spending for all Respondents on Each Health Service by Race in PSID ..........................................................................................141 7.3 Mean and Median Two-Year Spending for all Respondents on Each Health Service by Race AHEAD...................................................................................... 143 7.4 Percentage of Household Income Spent on Total Out-of-Pocket Costs Excluding and Including Insurance Premiums in PSID ..................................... 144 7.5 Percentage of Household Income Spent on Total Out-of-Pocket Costs Excluding and Including Insurance Premiums in AHEAD...................................145 7.6 Percentage of Income Spent on Total Health Care (excluding insurance) by Race in PSID and AHEAD..................................................................................... 146 7.7 Percentage of Income Spent on Total Health Care Including Insurance in PSID and AHEAD.................................................................................................. 146 7.8 Percentage Reporting Use of Each Health Service by Race in PSID ....................148 7.9 Percentage Reporting Use of Each Health Service by Race in AHEAD.............. 148 7.10 Distribution of Out-of-Pocket Spending on Different Services by Race in PSID .......................................................................................................................149 7.11 Distribution of Out-of-Pocket Spending on Different Services by Race in AHEAD.............................................................................................................. 150 7.12 Percentage of Users Reporting Uncovered Costs for Each Health Service by Race in P S ID ..................................................................................................... 151 7.13 Percentage of Users Reporting Uncovered Costs for Each Health Service by Race in AHEAD...................................................................................................151 7.14 Mean and Median Spending for Users of Each Health Service by Race in PSID .......................................................................................................................153 7.15 Mean and Median Spending for Users of Each Health Service by Race in AHEAD.............................................................................................................. 155 7.16 Hierarchical OLS Regression Results for Out-of -Pocket Spending on Prescription Medications and Total Out-of-Pocket Spending (in Logged Dollars) by Race in PSID and AHEAD..............................................159 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. IX 7.17 Hierarchical Two-Stage Regression Results for Use and Out-of -Pocket Spending (in Logged Dollars) by Race in PSID and AHEAD: Hospital, Nursing Home, Outpatient Surgery, Dental, Equipment, and Home Care. . . . 160 7.18 Hierarchical OLS Regression Results for Out-of-Pocket Expenditures on Physician/Therapist/ER Visits in PSID ( in Logged Dollars) (n = 9 6 3 )........... 162 7.19 Hierarchical OLS Regression Results for Out-of-Pocket Expenditures on Physician/Therapist/ER Visits in AHEAD (in Logged Dollars) (n = 5936).... 163 8.1 Means and Percentages on Independent Variables by Insurance Type in PSID and AHEAD................................................ 194 8.2 Percentage Reporting Use of Each Health Service by Insurance Type in PSED.......................................................................................................................197 8.3 Percentage Reporting Use of Each Health Service by Insurance Type in AHEAD.....................................................................................................................197 8.4 Percentage of Users Reporting Uncovered Costs for Each Health Service by Insurance Type in PSID............... 198 8.5 Percentage of Users Reporting Uncovered Costs for Each Health Service by Insurance Type in AHEAD......................................................................................198 8.6 Mean and Median One Year Out-of-Pocket Spending on Each Service and Overall by Insurance Type, for All Respondents in P S ID ............................201 8.7 Mean and Median Two Year Out-of-Pocket Spending on Each Service and Overall by Insurance Type, for All Respondents in AHEAD........................203 8.8 Mean and Median One Year Out-of-Pocket Spending on Each Service and Overall by Insurance Type, for Users of Each Service in P S ID .................. 205 8.9 Mean and Median Two Year Out-of-Pocket Spending on Each Service and Overall by Insurance Type, for Users of Each Service in A H EA D.............. 207 8.10 Percentage of Household Income Spent on Total Out-of-Pocket Costs by Insurance Type in PSID ..................................................................................... 208 8.11 Percentage of Household Income Spent on Total Out-of-Pocket Costs by Insurance Type in AHEAD................................. ............................................209 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. X 8.12 Percent Spending Over 10%, 30%, and 50% of Household Income on Total Out-of-Pocket Costs (Including Premiums for Insurance and Medicare Part B) by Insurance Type in PSID ........................................................ 210 8.13 Percent Spending Over 10%, 30%, and 50% of Household Income on Total Out-of-Pocket Costs (Including Premiums for Insurance and Medicare Part B) by Insurance type in AHEAD....................................................210 8.14 Hierarchical OLS Regression Results for Out-of-Pocket Spending by Insurance Type in PSID and AHEAD: Ambulatory Services, Prescription Medications, and Total Out-of-Pocket Spending (in Logged Dollars)...............215 8.15 Hierarchical Two-Stage Regression Results for Hospital and Nursing Home Use and Out-of -Pocket Expenditures (in Logged Dollars) by Insurance Type in PSID and AHEAD.................................................................... 216 8.16 Hierarchical Two-Stage Regression Results for Outpatient Surgery, Dental Care, and Equipment Use and Out-of -Pocket Expenditures (in Logged Dollars) by Insurance Type in P SID ....................................................217 8.17 Hierarchical Two-Stage Regression Results for Home Care Use and Out-of-Pocket Expenditures (in Logged Dollars) by Insurance Type in PSID and AHEAD......................................... 218 9.1 Percentage of Variance in Out-of-Pocket Costs Accounted for by Each Group of Variables in Conceptual Framework, for Each Service Type in PSID and AHEAD......................................................................................239 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. xi List of Figures 2.1 A Conceptualization of Andersen’s Behavioral Model of Health Services Use .. 26 2.2 Conceptual Framework for the Prediction of Out-of-Pocket Health C osts........... 28 3.1 Questions on Health Care Utilization in PS ID .......................................................... 35 3.2 Questions Regarding Health Care Utilization in AHEAD....................................... 39 4.1 Illustration of the Difference Between Individual and Aggregate Approaches .. 58 6.1 Hospital Use and Out-of-Pocket Costs by Age in PSID and AHEAD: Moving Average ...........................................................................................100 6.2 Nursing Home Use by Age in PSID and AHEAD, and Out-of-Pocket Nursing Home Costs by Age in PSID: Moving Average.......................................109 6.3 Home Care Use and Out-of-Pocket Costs by Age in PSID and AHEAD: Moving Average........................................................................................................110 6.4 Ambulatory Service Use and Out-of-Pocket Costs by Age in PSID and AHEAD: Moving Average.................................................................................... I l l 6.5 Outpatient Surgery: Use by Age in PSID and AHEAD, and Out-of-Pocket Costs by Age in PSID: Moving Average................................................................ 112 6.6 Use and Out-of-Pocket Costs for Prescription Medication by Age in PSID and AHEAD: Moving Average............................................................................... 113 6.7 Use and Out-of-Pocket Costs for Equipment by Age in PSID: Moving Average.................................................................................................................... 112 6.8 Use and Out-of-Pocket Costs for Dental Care by Age in PSID: Moving Average.................................................................................................................... 113 6.9 Total Out-of-Pocket Costs by Age in PSID and AHEAD: Moving Average .. 114 6.10 Burden of Total Out-of-Pocket Costs (excluding insurance) as a Portion of Household Income by Age in PSID and AHEAD: Moving Average 115 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. xii Susan Tracy Stewart Eileen M. Crimmins Out-of-Pocket Health Expenditures by Older Adults in Relation to Age, Race, and Insurance Out-of-pocket health expenditures by older Americans are examined in relation to age, race, and type of insurance coverage supplemental to Medicare. Expenditures for eight different health goods and services are examined, as well as total out-of-pocket costs both excluding and including insurance premiums. Data are from two large national survey samples of older adults: the 1990 Elderly Health Supplement to the Panel Study of Income Dynamics (N = 1031 age 66 and older), and the 1995-96 wave of the AHEAD survey (N = 6237 age 72 and older). Out-of-pocket costs did not increase with age for most services, and most of the positive age effects that were significant were explained by health. The exception was for long-term care services, for which use increased dramatically in the oldest age groups even when health was controlled. Blacks and Hispanics had lower mean out-of-pocket costs than Whites, however race effects either attenuated or became non-significant in hierarchical analyses. Greater eligibility for Medicaid among minorities accounted for some race effects, and others were explained by education and economic status. Race remained significant as a predictor of lower costs in some hierarchical models, which may reflect race differences in access, cultural dispositions, and the intensity of treatment among users. Those with insurance supplemental to Medicare had the highest out-of-pocket costs for several services. There was evidence of moral hazard but not adverse selection Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. among those with insurance. Those enrolled in Medicare HMOs had lower overall out-of- pocket costs than those with supplemental insurance, and this was not explained by favorable selection or decreased access to services. Those dually-eligible for Medicaid had significantly lower out-of-pocket costs than those with supplemental insurance, overall and for the majority of services. The burden of total out-of-pocket costs (excluding insurance premiums) as a portion of income increased with age, was lower for minority elders, and was highest for those with only fee-for-service Medicare. Patterns of discretionary and non-discretionary service use suggest barriers in access to appropriate care among the oldest-old, racial minorities, those with only fee-for-service Medicare, and those dually eligible for Medicaid. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 Chapter 1 The Importance of Studying Older Adults’ Out-of-Pocket Costs: Although adults age 65 and older in the US are entitled to health care benefits through the Medicare program, coverage is incomplete. Medicare does not limit beneficiaries’ total out-of-pocket payments per year, unlike the great majority of employer-provided insurance plans. For outpatient services, beneficiaries in fee-for-service (FFS) Medicare must pay a $100 deductible, after which they are responsible for 20 percent of all covered expenses. The monthly premium for this coverage was $45.50 in 1999 (Health Care Financing Administration, 1999). Beneficiaries who visit physicians that do not accept Medicare reimbursement rates as payment in full can be responsible for more than 20% of the bill The Medicare program also covers little routine preventive care. While specific screening tests such as mammography and prostate exams have recently become covered, routine physical exams are not. For beneficiaries requiring hospital care, deductibles and co-payments are high, particularly for long stays. The deductible for the first 60 days in the hospital was $768 in 1999. If continued hospital care is needed, there is a daily co-payment ($192 in 1999) for days 61 to 90. After the 90th day, hospital care is not covered by Medicare unless the beneficiary elects to use up to 60 "lifetime reserve" days, which require a daily co-payment ($384 in 1999) (Health Care Financing Administration, 1999). Home health care is covered by Medicare Part A and B with no coinsurance, but only for those who meet strict disability criteria. The service must be prescribed by a doctor, and beneficiaries must be homebound and require skilled nursing care or physical Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. therapy. Also, there is a 20 percent coinsurance for medical equipment that may be needed to support beneficiaries at home, such as wheelchairs, hospital beds, oxygen and walkers. Such equipment can be expensive, and is often the largest source of out-of- pocket costs for those with the high spending (Medicare Payment Advisory Commission, 1999). Care in a nursing home (skilled nursing facility, or SNF) is covered only for rehabilitation after a hospitalization of 3 days or more, and only for 100 days. After the first 20 days, a co-payment is required ($96 per day in 1999) (Health Care Financing Administration, 1999). For the small portion of older adults who require long-term nursing home care, out-of-pocket costs for this care are much higher and more burdensome than those for any other type of health expense. Among those who incur the highest out-of- pocket health care costs, nursing home expenditures constitute a large portion of expenses (Rice, 1989; Rice & Gabel, 1986). Another important health cost not covered by Medicare is outpatient prescription drugs. Older adults’ out-of-pocket prescription costs are gaining increasing political attention as options for adding prescription coverage to Medicare are discussed (for example see Government Reform, 1999; Haddad, 1999; Waxman, 1999). Prescription drugs are increasingly becoming the preferred form of treatment for the diseases affecting older adults (Duka, 1999). Their use is widespread among the older population; 77% of Medicare beneficiaries reported regularly using prescription drugs in a recent survey (Schoen, Neuman, Kitchman, Davis, & Rowland, 1998). The cost of prescription drugs has also grown rapidly in the last decade (Duka, 1999), with prices for many drugs rising Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. faster than inflation (Haddad, 1999). A recent study by Families USA found that for the 50 prescription drugs most frequently used by the elderly, prices rose on average more than four times faster than inflation in 1998. Prescription drug costs are of particular concern because older adults who pay for their own prescription medications are typically charged much more than large insurance companies, government buyers, and HMOs, who can negociate low prices with drug companies because they buy in bulk. In a set of surveys recently conducted by the staff of the Congressional Committee Government Reform, individual purchasers paid two to three times more than big purchasers for the same drugs (Government Reform, 1999; Waxman, 1999). While it is common for manufacturers of any consumer item to offer lower prices to volume buyers, the price differential for consumer drugs is much higher for prescription drugs than for other retail items (Government Reform, 1999). Many beneficiaries purchase supplemental "Medigap" insurance to protect them from out-of-pocket health costs, in which case their out-of-pocket liabilities consist largely of insurance premiums. Beneficiaries who are provided with supplemental insurance through an employer, who are enrolled in Medicaid as well as Medicare, or who choose to obtain their health care through a pre-paid health plan have greater protection from out- of-pocket medical expenditures, but may still incur burdensome costs. Since the inception of the Medicare program in 1965, the cost of health care has been rising faster than inflation, which has contributed to a rise in out-of-pocket expenditures relative to income (Moon, 1992). As costs have risen, Medicare policy Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. changes have also increased the level of beneficiary cost-sharing (Moon, 1992). Thus, beneficiaries who are ill, disabled and/or require regular prescription medicines can incur very high out-of-pocket costs. For those with low incomes, even modest costs can be financially burdensome (Feder, Moon, & Scanlon, 1987; Wyszewianski, 1986a). This dissertation examines age, race, and insurance as predictors of out-of-pocket health expenditures by older adults in the United States, using two large survey samples of older adults. Descriptive analyses examine out-of-pocket costs in detail, and regression analysis is used to examine whether these variables are significant predictors of costs when controlling for other important factors. A conceptual framework is used to classify and identify links between other factors likely to affect costs. Rationale for Focussing on Age. Race, and Insurance: Concern regarding the risk of high and/or burdensome out-of-pocket health care costs among the older population has fuelled a desire to address this problem. However, in order to effectively intervene to help those most at risk, it is important to understand what factors are associated with the greatest and most burdensome costs. The factors chosen as the focus of this dissertation: age, race, and insurance, each have particular policy relevance, as described below. While these factors are known to be associated with health care costs, we lack a detailed understanding of how each is related to out-of-pocket expenditures for different health goods and services. Also important is an analysis of the independent contribution made by each factor, controlling for related variables. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5 Age Age has a history of being considered as a proxy for need in public policy. Because of this, it is important to determine the extent to which age is an accurate proxy, and to identify other factors that account in part for age-related variance in out-of-pocket costs. Due to age-specific reductions in mortality at the oldest ages, an increasing number of Americans are reaching very old age. A frequently cited statistic is that those 85 years and older constitute the fastest growing age group in the population, and the number of very old adults is projected to swell in the next century as the baby boomers continue to age (Schneider, 1999). Health care costs are generally found to be greater for older adults than for those who are younger. However, there are mixed findings as to whether costs increase steadily with age among the older population, and for which types of services. Given the projected growth in the number of older adults, it is important to examine the relationship between age, out-of-pocket costs for different types of care, and the burden of out-of-pocket health care costs among this population. Furthermore, it is not clear to what extent the relationship between age and out-of-pocket costs can be accounted for by health, gender, and facilitating factors such as income and insurance coverage. Race Race has been important in the history of the United States as a divisive social factor affecting life chances (Wallace, 1991). It is well known that both African Americans and Hispanics have lower mean income, are in poorer health, and face more barriers in access to health care than Whites. African Americans are more likely than White Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Americans to rate the performance of the nation’s institutions, including health services, as fair or poor (Blendon et aL, 1995). Among the older population, lower socioeconomic status among minority groups means that they are less likely than Whites to have private or employer-provided insurance as a supplement to Medicare. However, members of minority groups are more likely than Whites to be covered by Medicaid, which is designed to protect against high out-of-pocket health costs. Given racial differences in income, health, and insurance, the level of out-of-pocket costs and burden would be expected to vary by race. Examining racial differences in use and out-of-pocket costs for different health services will increase our understanding of the services for which minority races are at risk for high costs, and give clues as to whether costs are related to access. Also of policy relevance is to test whether race remains significant as a predictor of expenditures for different services when income, health and insurance coverage are controlled. This will provide insight into the factors underlying racial differences, and inform our efforts to address these differences. This is particularly important since the racial diversity of the older population is increasing, with minorities growing much faster than Whites as a share of the elderly population (Hooyman & Kiyak, 1999). Insurance Insurance coverage is of central concern when examining out-of-pocket costs, since the purpose of insurance is to reduce these costs. Although almost all older adults are covered by Medicare, out-of-pocket costs can be high without additional insurance to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. cover deductibles, co-pays, and goods and services not covered by Medicare, as discussed above. It is important to examine how different types of insurance coverage affect out-of- pocket costs and burden. The majority of older adults have insurance to supplement Medicare coverage, yet this may not decrease their out-of-pocket costs, particularly when the costs of insurance are considered as part of the out-of-pocket burden. A portion of older adults receive insurance through a previous or current employer, and examining the differences in out-of-pocket costs between those with employer-sponsored insurance and those without such coverage will give a better idea of the value of this benefit to older retirees. For those with low income and wealth, Medicaid is available as a form of supplemental insurance, and it is important to examine the extent to which Medicaid is reducing the burden of out-of-pocket costs. Analyses of out-of-pocket spending on specific services will yield insights regarding the types of goods and services for which Medicaid may not be providing adequate protection. Finally, some Medicare beneficiaries choose to receive their benefits through a pre-paid health plan. These plans have historically covered more goods and services and required lower deductibles and co-pays than traditional Medicare, but also tend to attract healthier beneficiaries. It is important to examine the extent to which differences in out-of-pocket spending by those in managed care are explained by favorable selection and by differences in the types of services used. Summary In summary, a deeper understanding of the variables related to having high out-of- pocket costs for different types of health goods and services will enable more focused and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. effective intervention to help those with the highest costs and the greatest out-of-pocket burden. The factors of focus in this dissertation each have particular policy relevance. While age and race are personal characteristics that cannot be changed, they continue to be characteristics along which society is divided. Assumptions are often made about the needs and deservingness of individuals and groups based on their race or age. Public policy has historically used age to determine eligibility for a range of programs and services (Neugarten, 1982), and services have also been either implicitly or explicitly targeted toward certain age and race groups. However, we have insufficient knowledge about how these characteristics relate to out-of-pocket costs to understand whether or how intervention based on these factors might be effective. With regard to insurance supplemental to Medicare, there is a growing movement focused on educating Medicare beneficiaries about their coverage options (Sofaer& Davidson, 1990; Stone, 1999). More detailed knowledge of the protection provided by different types of insurance can inform these efforts. Overview o f Remaining Chapters: Chapter 2 reviews the literature on older adults’ health care use and out-of-pocket spending, presents the conceptual framework used in this dissertation, and sets forth the research questions that will be addressed. Chapter 3 describes the data. The two samples used in this study are introduced, and weighting variables are described. The dependent variables and imputation procedures are presented, and the measurement of control variables in each sample is described. Chapter 4 describes the methodological approach for Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. both descriptive and multivariate analyses, and discusses several methodological issues. In Chapter 5, descriptive analyses of out-of-pocket costs in each sample are presented. The means and medians of spending variables before and after imputation are compared, and results are compared to those found in previous reports. Chapters 6, 7 and 8 focus respectively on age, race and insurance as predictors of out-of-pocket spending. Each chapter reviews the literature regarding health care use and costs with respect to the variable of interest, presents detailed descriptive data and hierarchical regression results that focus on that variable, and discusses the results and their implications. In Chapter 9, multivariate results are discussed with respect to the conceptual framework, the significance of the control variables as predictors of out-of-pocket costs is examined, and general conclusions are set forth. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 10 Chapter 2 Literature Review and Conceptual Framework Review of Literature on Out-of-Pocket Health Spending Data on Out-of-Pocket Costs Two main types of data are used to estimate older adults’ out-of-pocket health care costs. One type is macro-data: federal economic statistics and health industry data. The other is individual-level: survey data and Medicare records. Reports using macro-data are more consistently available than studies using survey data (e.g. Levit et aL, 1996; Ways and Means, 1996). However, out-of-pocket costs calculated from macro-data are aggregate estimates based on a number of assumptions and adjustments, and do not give an accurate or detailed picture of individual-level burden (Moon, 1991). At the individual level, Medicare records accurately reflect the health costs incurred by each beneficiary and the portion not covered by Medicare. However, Medicare records do not indicate which beneficiaries had these costs covered by supplemental insurance, and do not include out- of-pocket costs for services not covered by Medicare (Moon, 1991). In order to gain such information, older adults must be individually surveyed. Surveys that collect additional » information on respondents also provide the opportunity to examine the relationship between out-of-pocket costs and a number of sociodemographic and health variables. A main source of survey data on out-of-pocket costs has been a series of nationally representative surveys conducted by the Agency for Health Care Policy and Research (AHCPR) and the National Center for Health Statistics (NCHS). The first of these surveys was the 1977 National Medical Care and Expenditure Survey (NMCES). (In 1980, a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 11 similar study, the National Medical Care Utilization and Expenditure Survey (NMCUES) was conducted by the Health Care Financing Administration (HCFA) and NCHS.) The second was the 1987 National Medical Expenditure Survey (NMES). The third survey, the Medical Expenditure Panel Survey (MEPS) was begun in 1996, and data on out-of- pocket spending from this survey are not yet available. Data are also collected on out-of-pocket health spending as part of the annual Consumer Expenditure Survey (CEX), conducted by the Bureau of the Census under contract with U.S. Bureau of Labor Statistics. However, spending is reported in this survey at the family level, and reports cannot be disaggregated to the individual-level data needed to examine the correlates of individual expenditures (Waldo, Sonnefeld, McKusik, & Arnett, 1989). Due to increasing interest regarding older adults’ health costs, questions on out-of- pocket costs also have been included in surveys that were designed to study a broader range of social and economic characteristics. Two such surveys are used in this dissertation: the Panel Study of Income Dynamics (PSID) and the AHEAD study (Asset and Health Dynamics of the Oldest-Old), collected by the Survey Research Center at the University of Michigan. A significant new source of survey data on older adults’ out-of-pocket health care costs is the Medicare Current Beneficiary Survey (MCBS), funded by HCFA, which administers the Medicare program. The MCBS is an annual survey of a nationally representative sample of non-institutionalized Medicare beneficiaries, begun in 1991. In reviewing literature on out-of-pocket costs, it is important to distinguish Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 12 between estimates of out-of-pocket spending, and of ‘personal health care expenditures’ (PHCE). Many published estimates of health expenditures report PHCE, which is defined as “spending for the direct consumption of health care goods and services” (Waldo et aL, 1989, p. 111). However, PHCE refers to expenditures by ALL payers, including private insurance, public programs, and individuals. Also, PHCE are often reported at the aggregate level, as the amount (in billions) spent on health care for the whole population. While estimates of PHCE are regularly reported, there are relatively few individual-level reports and studies examining out-of-pocket costs, particularly among the older population. Studies Examining Out-of-Pocket Costs Two recent studies estimated the amount and burden of older adults’ out-of- pocket health costs by trending forward survey data collected earlier. The first trended forward data from the 1987 NMES to 1994, estimating total out-of-pocket costs including Medicare deductibles and coinsurance, balance billing by physicians, payments for medical goods and services not covered by Medicare (including home care but not nursing home care), and premiums for private insurance and Medicare Part B. Estimated out-of-pocket costs for 1994 were $2,803 per person, and the average burden of these costs was 23% of family income (American Association of Retired Persons (AARP) Public Policy Institute and The Urban Institute, 1994). A second study projected out-of-pocket costs from the 1993 MCBS to 1997, including the same costs described above with the exclusion of home care. A microsimulation technique was used to account for changes in factors likely to affect costs, such as Medicare cost-sharing rules, the use and costs of different types of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 13 services, insurance coverage (e.g. increased enrollment in HMOs), and demographic characteristics of the population (AARP Public Policy Institute and the Lewin Group, 1997). Mean out-of-pocket costs for 1997 were estimated to be $2,149 per beneficiary, and mean burden was 19% of individual income. Out-of-pocket health care spending by noninstitutionalized Medicare beneficiaries was also recently described as part of a report to Congress by the Medicare Payment Advisory Commission (1999). The Commission examined spending by those in the 1995 MCBS, as well as a sub-sample of beneficiaries who responded to the MCBS from 1992 through 1995. They found that Medicare paid about 62 percent of the health care costs of beneficiaries, and that this portion increased as total health costs increased. Cost-sharing for equipment and the lack of an annual limit on out-of-pocket spending were the most problematic factors leading to very high cost among a portion of beneficiaries. Another study examined out-of-pocket expenditures for prescription drugs in the 1995 MCBS by insurance type (Davis, Poisal, Chulis, Zarabozo, & Cooper, 1999). Mean expenditures per person were $303, and were lower for those with drug coverage ($232) than those with no coverage ($432). Gross & Brangan (1999) used the microsimulation technique discussed above to project these expenditures to 1999, yielding estimates of $320 for those with coverage and $590 for those without. While these studies provided valuable estimates of older adults’ out-of-pocket health costs, they consisted only of descriptive analyses, and did not use multivariate analyses to examine significant predictors of spending. Because many of the factors predicting spending are related to one another, multivariate analyses are necessary to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 14 identify independent predictors of costs. A small number of studies have used multivariate analyses to identify independent predictors of older adults’ out-of-pocket costs. However, most of these have either looked at only one type of health expenditure or at a select population, such as the disabled or low-income elderly. Rubin, Keolln, & Speas (1995) analyzed out-of-pocket health expenditures by elderly households in the 1980-81 and 1989-90 CEX. They found that older age, White race, and insurance coverage were predictors of higher expenditures, however these effects were not consistent across time periods and service types. Also, individual characteristics were those of the household reference person, which may not have accurately represented other household members. Using the 1977 NMCES, Cartwright, Hu and Huang (1992) examined insurance coverage as a predictor of older adults’ out-of-pocket expenditures for hospital and physician care combined. They found that the level of expenditures conditional on use was significantly lower for those with Medicaid coverage, those under 75, those with higher health ratings, females, and those with a lower education. Thomas & Kelman (1990) examined out-of-pocket health expenditures among low-to-moderate income Medicare beneficiaries living in the Bronx. Expenditures were significantly higher for those in the poorest health, those not covered by Medicaid, and those who obtained care primarily from a private physician. Stum, Bauer, & DeLaney (1996) examined the predictors of home care expenditures among 856 disabled older adults using data from the National Long Term Care Survey, and found that functional ability was the strongest predictor of out-of-pocket costs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 15 Predictors of out-of-pocket dental and prescription expenditures have been examined using data from the Elderly Health Supplement to the PSID (one of the survey samples used in this dissertation). Kington, Rogowski, & Lillard (1995) found that age and being married predicted lower out-of-pocket dental expenditures, whereas urban residence, wealth, and qualification for Medicaid predicted greater dental expenditures. For prescription medications, out-of-pocket expenditures were lower for Blacks, those in better health, those eligible for Medicaid, and those with private insurance that covered drugs (Lillard, Rogowski & Kington, 1999). Other studies have examined predictors of the burden of out-of-pocket costs relative to income. Coughlin, Korbin, and McBride (1992) examined the prevalence, sources, and predictors of catastrophic health expenditures among a sample of 1,783 severely disabled older adults. They found that prescription drugs were a source of catastrophic costs for the majority of their sample, and that nursing home costs were catastrophic for most who required nursing home care. Coughlin et aL (1992) also found that unique sets of personal characteristics predicted out-of-pocket costs for different services among their population of disabled adults. For example, age and level of cognitive and physical impairment predicted having catastrophic nursing home costs, whereas specific health conditions predicted out-of-pocket prescriptions expenses. The financial burden of out-of-pocket costs for prescription drugs has also been examined in the PSID (Rogowski, Lillard, & Kington, 1997). Those found to be most vulnerable to catastrophic costs were older adults with chronic conditions (particularly diabetes, angina, and hypertension), women, rural elderly and low income elderly. For the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 16 minority with insurance coverage for drugs, the fraction of household income spent on prescription drugs was 50% lower. Influences on Out-of-Pocket Expenditures A number of demographic, socioeconomic, and health variables have been associated with the level of out-of-pocket health expenditures on health goods and services. However, as discussed above, most studies have not analysed the statistical significance of these variables as predictors of out-of-pocket costs. Also, several variables that would be expected to independently predict expenditures have not previously been examined in studies of out-of-pocket costs. In contrast, a larger number of studies have used multivariate analyses to examine predictors of health care use. Thus, in reviewing the literature on factors affecting costs, studies that have found a particular factor to be a significant predictor of health care use may be cited. While there are important differences between health care use and out-of-pocket costs, factors affecting use would also be expected to predict costs in most cases. Age is often used as a predictor of health costs, and the elderly are know to incur higher costs than those who are younger. However, health costs may not rise uniformly with age among the older population. Also, the relationship between age and costs can be expected to vary for different types of health goods and services. The literature regarding age and out-of-pocket costs is reviewed in Chapter 6. Gender can be expected to affect out-of-pocket costs differently for different types of health services (for review see Mutran & Ferraro, 1988). This is due in part to gender Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 17 differences in the types and effects of medical conditions in women and men. It is well known that men are more likely to suffer from the types of diseases that require hospitalization, such as cardiovascular and respiratory disease (Verbrugge, 1995). In contrast, women are more likely to have non-fatal chronic conditions such as arthritis, or to survive more serious medical conditions to the point where maintenance care rather than critical care is required (Verbrugge, 1989). Consistent with these differences, Mutran and Ferraro (1988) found that when health was controlled, there was no gender difference in the use of physician services, but older men were more likely to be hospitalized than older women. The latter finding appeared to be due in part to the higher rates among men of diseases requiring hospitalization, however Mutran and Ferraro (1988) suggest that social differences in the way physicians perceive, interact with and treat women and men also help to explain differences in hospitalization. While Verbrugge (1995) found no gender differences in hospitalization for ischemic heart disease, Schulman et al. (1999) found that physicians are less likely to recommend cardiac catheterization for female patients, controlling for the physician’s assessment of the probability of coronary artery disease as well as for the patient’s age, level of coronary risk, type of chest pain, and the results of an exercise stress test. Such differences in treatment would be expected to lead to lower hospital costs for women. Consistent with this finding, Cartwright et al (1992) found that conditional on health care use, men had higher out-of-pocket expenditures than women for hospital and physician care combined. In contrast, women are likely to have higher costs for the types of care required for maintenance of disabling chronic diseases, such as long-term care and perhaps medi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 8 cations. Using MCBS projections of out-of-pocket costs for 1998 estimated by The Lewin Group, Gibson and Brangan (1998a) calculated that older women spent an average of $228 more out-of-pocket than men on total health costs when the costs of short-term nursing home care were included, and that women also spent more on prescription drugs. Regional variation in the cost of care can also account in part for variation in older Americans’ out-of-pocket health costs. Levels of health care use and average health care costs vary considerably across different areas of the country. For example, Moon (1992) found that in 1980 and 1989, Medicare beneficiaries living in the Northeast and the West had higher average use of physicians than those in the North Central and South parts of the U.S., controlling for other factors such as age, race, income, and insurance. Buczo (1989) found significant differences in the probability of hospitalization and in total hospital costs incurred by Medicaid beneficiaries in California, New York, Texas and Michigan, controlling for health and demographic factors. Similarly, living in a rural versus an urban area can affect health care use and costs. Those living in rural areas have been found to use less care and incur lower total Medicare costs than urban residents (Dor & Holahan, 1989; Hatten & Connerton, 1986; Krout, 1983; Nyman, Sen, Chan, & Commins, 1991). In an examination of urban-rural differences in physician expenditures under Medicare Part B, Dor & Holahan (1989) found that prices for care were lower in rural areas, but that the main factors contributing to the differences in costs were lower availability of hospital care and specialists in rural areas. Hospital admissions and length of stay were significantly lower in rural areas, and the only type of care for which urban and rural expenditures did not differ was visits to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 19 general and family physicians. Consistent with these findings, Braden & Beauregard, (1994) found that in the 1987 NMES, rural residents had lower use of hospital outpatient departments and emergency rooms than urban residents, and were more likely to rely on community doctors in private practice. Braden & Beauregard (1994) also found that among those in poor or fair health, rural residents had fewer ambulatory care visits on average than urban residents. Insurance is a key variable affecting the level of out-of-pocket health spending, because its purpose is to protect against out-of-pocket costs. Among the elderly, different types of insurance coverage in addition to Medicare have been shown to be related to health care use, total costs to the Medicare program, and out-of-pocket costs. Chapter 8 discusses the literature regarding four main types of coverage in addition to traditional FFS Medicare: private supplemental insurance, employer-sponsored insurance, membership in a Medicare HMO, and dual-eligibility for Medicaid. Education can increase health care use and the intensity and cost of services used in several ways. The improved literacy associated with higher levels of education can facilitate navigation of the health care system, and increase awareness of the importance of preventive and maintenance care. Among those with specific health conditions, those with higher levels of education are better able to access information regarding possible treatments. Those with more years of education are particularly likely to use more discretionary services such as dental care and preventive screening. For example, Conrad, Grembowski and Milgrom (1987) found that those with at least a high school education were more likely to use dental services and used a higher intensity of services than those Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 20 with less education. Consistent with this finding of greater service use, higher levels of education were found by Cartwright et aL (1992) to predict significantly higher out-of- pocket costs for hospital/physician care. Income and wealth affect both the ability to pay for health care and eligibility for public assistance to help with out-of-pocket health care costs. Both income and wealth are positively associated with health care use (Paringer, Bluck, Feder & Holahan, 1979) and out-of-pocket health costs, with expenditures increasing as income increases (Holahan, & Zedlewski, 1992). Because the elderly may draw on accumulated assets to pay for health care, it is important to examine wealth in addition to income as a predictor of out-of- pocket expenditures. Under the classic life cycle model of consumption and income, most families are expected to accumulate wealth during their working years, and gradually spend this wealth during their retirement years. Thus, resources available for consumption by older adults may be underestimated if only income is considered (Friedman, 1984; Hurd, 1989). It is unclear to what extent increased spending by those with higher incomes and wealth reflects over-use of medical care, and to what extent this pattern reflects inadequate access to care for those with fewer resources. There is evidence that those with lower incomes spend a higher proportion of their out-of-pocket health dollars on non- discretionary services, which suggests that they are less able to access primary preventive care (Wolinsky et aL, 1990). Illness, disability, and disease are strong predictors of health service use, and the level of illness, disability or disease would be expected to affect the intensity and cost of treatment. Need appears to be a more important predictor of health expenditures among Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 21 the elderly than ability to pay (Moon, 1992), particularly since almost all older adults are covered at least by Medicare. Self-rated health is a consistent predictor of health care costs, with those in worse health incurring higher Medicare costs and consuming a disproportionate share or Medicare resources (Eppig & Poisal, 1997). Those in poor health also have higher out-of- pocket costs and spend a higher proportion of their income on medical care (Cartwright et aL, 1992; Schoen et aL, 1998; Thomas & Kelman, 1990). The number of days spent in bed, a measure of acute illness, has been found to predict a greater number of physician visits, controlling for other factors such as age, race, income, and insurance (Moon, 1992). Chronic medical conditions are also important predictors of health care use and costs. In particular, Verbrugge (1995) reports that potentially fatal chronic conditions (cancer, ischemic heart disease, diabetes mellitus, and chronic obstructive pulmonary disease) lead to greater use of both ambulatory and hospital care than nonfatal conditions such as arthritis and sensory impairment. Buczko (1989) found that among a Medicaid population, many diseases were significant predictors of the number of hospital days, including cardiovascular, digestive, respiratory, and nervous system diseases, and musculoskeletal disorders. Chronic conditions have also been found to predict greater use of prescription medications (Linden, Horgas, Gilberg, & Steinhagen-Thiessen, 1997), and higher prescription drug expenditures (Mueller, Schur, & O’Connell, 1997). Functional ability, as measured by the ability to perform daily tasks, is also an important predictor of health care use and costs, with those who report difficulty with more tasks incurring higher Medicare costs and consuming a disproportionate share of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 22 Medicare resources (Eppig & Poisal, 1997; Liu, Wall, & Wissoker, 1997). Functional ability is particularly important as a predictor of long-term care expenditures. Two other measures of health expected to affect primarily long-term care costs are falling and cognitive impairment. Those who experience a fall show an increase in health service use and expenditures. Cognitively impaired older adults use more nursing home care, and thus have higher out-of-pocket nursing home costs than intact elderly (Linden et aL, 1997; LoGiudice, Waltrowicz, Ames, Brown, Burrows, & Flicker, 1997). However, findings regarding cognitive status as a predictor of health care use and costs have been mixed depending on the type of costs being predicted. In a study of older adults with dementia living at home, those with higher dementia symptom severity scores incurred higher out-of-pocket costs for acute and long-term care services combined (Weinberger et aL, 1993). However, Linden et aL (1997) found that poor cognitive status predicted lower use of medications, and LoGiudice et al. (1997) found that out-of-pocket costs for non- residential community services were similar for impaired and intact older adults. Finally, depression is a consistent predictor of increased health service utilization and higher health costs (Druss, Rohrbaugh, & Rosenheck, 1999; Simon & Katzelnick, 1997; Uniitzer et aL, 1997; Simon, Ormel, VonKorff, & Barlow, 1995; Simon, VonKorff, & Barlow, 1995). Among high users o f health care, a large proportion are found to have major depression and/or dysthymia (Katon et al., 1990). Depressed patients have longer hospital stays (Verbosky, Franco, & Zrull, 1993), and higher costs for every category of care (ie. primary care, medical specialty, medical inpatient, pharmacy, laboratory) (Simon, VonKorff and Barlow, 1995). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 23 Rationale for Current Study This dissertation addresses the limitations of previous research by examining out- of-pocket spending on a range of health goods and services in two large national survey samples of older adults. Both surveys include detailed measures of sociodemographic, health, and other variables likely to affect the level of out-of-pocket spending, enabling an analysis of age, race and insurance while controlling for potential confounding factors. The Elderly Health Supplement to the PSID ( n=1031) includes detailed reports of out-of-pocket spending on eight different health goods and services, which allows a more comprehensive analysis of the determinants of specific health costs than was previously available. The AHEAD survey (n=6237) includes reports of out-of-pocket spending in four categories, and is a much larger sample, which allows for more detailed analysis of out-of-pocket spending by sub-groups such as the oldest-old and those with different types of insurance. AHEAD is nationally representative of the population age 72 and over, which makes results generalizable to this population. Also, AHEAD allows examination of Hispanic elderly, and includes more detailed measures of health than PSID, including bed disability days, cognitive status and falls. Examining expenditures in two samples reveals the extent to which findings in the first sample are replicated in the second. Where findings are different, attempts to account for differences may provide additional insights. This dissertation also goes beyond previous descriptive reports of older adults’ out-of-pocket costs by examining in detail the amount spent on specific health goods and services, stratified by age, race, and types of insurance coverage. In addition to mean and median expenditures overall and by service type, the proportion of the sample using each Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 24 service and the proportion of users with uncovered costs is examined. This gives an indication of the prevalence of the problem of uncovered health costs in the elderly population, revealing which costs are concentrated among a small number and which are widespread. A breakdown of total out-of-pocket costs by service type indicates which services consume the largest portion of individual expenditures. The ratio of total out-of- pocket expenditures to income is also examined to explore the burden of costs for those in different groups. Specifically, the following descriptive research questions are addressed, in both survey samples. These questions are examined for the whole sample in Chapter 5, and by age, race, and insurance coverage in Chapters 6, 7, and 8. 1) What proportion of respondents report use of each health service? 2) What portion of those using each service report uncovered costs for that service? 3) What is the mean and median level of out-of-pocket spending overall, on specific services, and on insurance? 4) What is the mean proportion of total out-of-pocket costs spent on different health goods and services and on insurance? 5) What is the mean proportion of household income spent by individuals on total out-of-pocket health costs? 6) What percentage of the sample spent over 10%, over 30%, and over 50% of their income on out-of-pocket health expenses? In addition to these descriptive research questions, multivariate analyses are used in Chapters 6, 7, and 8 to examine whether age, race, and insurance types are significant independent predictors of out-of-pocket costs, overall and for different services. The next section describes the conceptual framework used to guide multivariate analyses. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 25 Conceptual Framework There is no established conceptual framework in the literature to guide the selection of independent variables when predicting out-of-pocket health care costs. However, studies examining total health care costs often refer to the the Andersen Behavioral Model of Health Services Use (Andersen, 1968; Andersen & Newman, 1973). This model has been widely used to predict the use o f health services, and is the most frequently cited theoretical framework for understanding health care utilization. The Andersen model, depicted in Figure 2.1, proposes that health care utilization is a function of predisposing characteristics, enabling factors, and need. Predisposing characteristics reflect the propensity of the individual to use services, enabling factors reflect the individual’s ability to secure services, and need reflects an individual’s level of illness. Variables in these categories were originally identified based on their relationship to health service use in other research and their demonstrated ability to be operationalized in social survey research. Predisposing characteristics include demographic variables such as age and gender, variables reflecting social structure, such as education, occupation, and race, and variables measuring health beliefs. Such variables exist prior to the onset of specific episodes of illness, and are not considered to be direct reasons for using services, however they affect the likelihood of service use. Enabling variables include family resources, such as income, savings, health insurance, and a regular source of care, as well as community resources, such as the ratio of physicians and hospital beds to the population in the area where the family lives. The level of such resources affects the availability of health services for those Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 26 who are predisposed to use services and/or have a need for service use. Need variables include perceived general health as well as the presence of health conditions or illness (called “evaluated health”, since awareness of particular conditions usually comes from physician diagnosis). Need is the most immediate determinant of health service use, and an individual must perceive illness or its possibility in order for use to take place. Andersen (1995) acknowledges that some variables in the model may act indirectly through others, but also feels that each factor can make an independent contribution to predicting use. The findings of research using the behavioral model are mixed, however need variables have traditionally been the strongest predictors of health service use (Andersen & Newman 1973). Figure 2.1: A Conceptualization of Andersen’s Behavioral Model of Health Utilization PREDISPOSING CHARACTERISTICS Demographic Social Structure Health Beliefs ENABLING NEED RESOURCES Perceived Personal/Family Evaluated Community HEALTH SERVICE USE The conceptual framework developed for this dissertation borrows from the Andersen model in several respects. However, since the Andersen model was developed Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 27 to predict and explain the use of health services, modification was required for the prediction of out-of-pocket health expenditures. Although many of the variables that predict increased health care use can also be assumed to increase health care costs, insurance coverage must be considered as a central modifying factor. Also, some variables (such as health behaviors) were excluded because they were not expected to be closely related to age, race or insurance, which were the main variables of interest in this study. Some variables were excluded because they were closely related to other variables in the model, which resulted in problems with multicollinearity. (For example, marital status was related to gender, and the ability to drive was correlated with health.) Modification of the model in this way is appropriate, as it was intended as a framework to guide the selection of relevant variables to include in analyses (Andersen & Newman, 1973). In developing the framework that would guide multivariate analyses, independent variables were categorized into groups according to the manner in which they were expected to affect health care costs. As seen in Figure 2.2, the variables fell into four main groups that resemble those in the Andersen model: demographic and geographic variables, variables reflecting socioeconomic status, insurance variables, and health variables. The first category— demographic and geographic variables— includes age, gender, and race, as well as variables reflecting regional Medicare spending and urban/rural residence. The demographic variables are expected to reflect social factors that differentially affect health care use and costs, such as preferences for health care use, the types and intensity of procedures prescribed, and access to care. If these variables remain significant when other variables are controlled, then such social factors, which are not Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 28 controlled in this study, can be discussed as possible explanations. Geographic variables control for aspects o f a respondent's geographic location that are likely to affect the availability and price of health care. Figure 2.2: Conceptual Framework for the Prediction of Out-of-Pocket Health Costs DEMOGRAPHIC Age Gender Race GEOGRAPHIC Regional Spending Urban / Rural SES Income Assets Education INSURANCE Medigap Employer-Sponsored Prepaid Plan Medicaid FFS Medicare Only NEED Self-Rated Health Disability Bed Days Health Conditions Cognitive Status Depression OUT-OF-POCKET The second category includes socioeconomic (SES) variables which indicate an individual’s ability to afford care, similar to the ‘enabling resources’ in the Andersen model. In this study, socioeconomic factors include income, wealth, and education. Education is included in this category rather than as a demographic variable because it is expected to affect health care costs primarily through its effects on SES. The third category includes insurance variables, which are modelled separately because of their central role in affecting out-of-pocket costs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 29 The final category includes measures of health, parallel to the ‘need’ factors in the Andersen modeL These variables are expected to affect out-of-pocket costs through their influence on the amount, types, and intensity of health care used. As seen in Figure 2.2, several measures of health are used in this study, including self-reported health, several chronic medical conditions, depression, functional ability, and days spent in bed due to illness. The variables used to predict expenditures for each service differ somewhat based on theoretical expectations. For example, two variables expected to affect primarily long term care expenditures— cognitive status and reporting a fall— are only used as predictors for nursing home, home care, and total expenditures. Sight and hearing difficulties are included as predictors of costs for equipment, since this category included expenditures for glasses and hearing aids. For dental expenditures, self-rated health and smoking status are the only measures of health used. Also, dental insurance, prescription insurance, and long-term care insurance are only used as predictors for the types of costs they would be expected to cover. Hierarchical regression is used to test the model developed from the conceptual framework. In the chapters examining age and race, demographic variables are entered into the model first, followed by health, SES, and insurance variables. In the insurance chapter, insurance and demographic variables are tested first, followed by health and SES variables. An important aspect of the conceptual framework is the distinction between more discretionary and less discretionary types of health services. While need variables are Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 30 generally found to be the strongest predictors of use in the Andersen model, the significance of predisposing and enabling factors has been found to vary depending on the types of service examined. Andersen (1968) originally hypothesized that the use of services considered ‘discretionary,’ such as dental visits, is more likely to be predicted by factors such as social structure, health beliefs, and enabling factors, whereas need and demographic characteristics are likely to be the primary predictors of less discretionary services, such as hospital visits. Physician visits, which fall more near the middle of the continuum between discretionary and non-discretionary use, are hypothesized to be predicted by all three categories of variables (Andersen, 1968; Andersen, 1995; Mitchell & Krout, 1998). For this study, the types of health goods and services for which out-of-pocket costs are examined can be categorized according to approximately where they fall along the discretionary continuum. Hospital stays and visits to a hospital emergency room (ER) are the least discretionary services (although costs for ER visits cannot be separated from physician costs in the data used here). The most discretionary services examined here are dental care, equipment, and “other special services” (such as home-delivered meals). Services in the middle of the discretionary continuum include outpatient surgery, visits to physicians and therapists, and prescription medications. Long-term care services also can be considered somewhat discretionary in that their use often depends on the availability and capacity of informal caregivers. It is hypothesized that demographic and SES variables are more likely to be independent predictors of out-of-pocket spending on services that are more discretionary. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 31 Chapter 3 Samples and Measures Samples The PSID Sample The Panel Study of Income Dynamics (PSID) is a longitudinal study begun in 1968 by the Survey Research Center at the University of Michigan which gathers data annually on a nationally representative sample of US individuals and their families. The survey focuses on the impact of economic and demographic dynamics, and includes a wide array of questions on income and demographic characteristics. Detailed reports of assets and debts were also included in a 1989 supplement to the regular interview. The rationale, measures, sampling, and other details of the study are presented in the PSID User’s Guide (Hill, 1992). This dissertation uses the 1990 Elderly Health Supplement to the PSID. In 1990, supplemental information on health service use, health spending, health status, and insurance coverage was collected for older adults in the PSID sample via a telephone interview and a mail-in questionnaire designed by the RAND Corporation. The telephone supplement contained detailed questions regarding use, out-of-pocket costs and insurance payments for eight different health goods and services in the past year. The mail-in questionnaire asked about health status and health insurance coverage, including the types of services covered by insurance. The telephone survey was completed by 1178 of the 1194 core households with heads and/or wives who were age 65 or older. For households in which both husband and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 32 wife were age 65 or over, information was collected separately for each person. The mail- in survey was given to individuals age 50 and over, with a response rate of 74%, yielding data on 2,429 individuals. The total number of respondents who had data on both the telephone and mail-in questionnaire was 1229. For these analyses, only respondents who were age 66 or over were selected, to ensure that Medicare beneficiaries had been covered for the entire year prior to the interview for which health expenditures were reported. This excluded 198 respondents, yielding a final sample size of 1031 for these analyses. Table 3.1 shows demographic, economic, and health characteristics of this sample. Analyses were weighted to compensate for unequal probabilities of selection, selective attrition from the panel from 1968-1990, nonresponse to the PSID in 1990, and nonresponse to the mail-in survey.1 The unweighted sample size was 1031 (weights were normed so that the weighted and unweighted sample sizes were the same). The AHEAD Sample The survey of Asset and Health Dynamics of the Oldest-Old (AHEAD) is a national, longitudinal survey of noninstitutionalized Americans age 70-103 that began in 1993. The rationale, measures, sampling, and other details of the study are presented elsewhere (Soldo, Hurd, Rodgers, & Wallace, 1997). Survey questions measured cognition, health, finances, and social network. Most respondents under 80 years of age provided data by phone, while most of those over 80 were given face to face interviews 1 Individual-level weights were created by PSID staff to reflect the original 1968 sampling weights for households initially included in the panel. I constructed an additional weights to compensate for nonresponse to the mail-in survey. These weights were the overall probability of nonresponse for each respondent, based on factors that significantly predicted nonresponse using logistic regression (self-rated health, education, and race). These probabilities were then added to the PSID weights. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 33 because of possible sensory and other problems. This dissertation examines data from the second wave of the study, which were collected in 1995-96. Wave 2 was used in this study because health spending is measured at the individual level in wave 2, whereas it is only measured at the household level in wave 1. Analyzing spending at the individual level allows examination of individual characteristics associated with spending. The data available for this dissertation were from a preliminary release of AHEAD wave 2, and thus could contain errors that will be corrected in the final public release of AHEAD wave 2. However, the portions of this data used here were carefully examined and cleaned by the author before analyses were run. Decisions regarding the treatment of missing data or inconsistent cases are noted in the relevant sections. Individual-level weighting was used in the analyses to account for oversampling of Mexican-American Hispanics, African Americans, and people living in Florida.2 When weighted, the wave 1 sample was nationally representative of the noninstitutionalized U.S. population age 70 and over. Wave 2 includes 246 respondents who were residing in a nursing home which they had entered between the first and second waves of the survey. In many households, a spouse of the primary respondent also completed the survey in both waves. However, spouses who were under age 70 (bom after 1923) in wave 1 were assigned zero weights, and are not included in the sample used for this study. This study examines the 6237 wave 2 respondents with nonzero weights; the weighted sample size is 6251. Table 3.1 shows demographic information for this sample. 2 The weights were created by AHEAD staff and are based on the wave 1 sample. Although these weights do not adjust for attrition in wave 2, they are the best weights available at this time, and their use is currently recommended by AHEAD staff. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 34 Measures Out-of-pocket costs in the PSID As part of the telephone questionnaire, respondents were asked to report their out-of- pocket costs for the previous year in eleven categories, shown in Table 3.1. These variables were combined to yield eight spending variables for these analyses, as shown in Table 3.1. For each service type, respondents were first asked about use. They were then asked to report payments for that service by payer type, beginning with out-of-pocket costs. The specific question on out-of-pocket costs for each service type was “How much did you (or your family living there) have to pay or expect to pay out-of-pocket for this, not counting what's covered by Medicare, Medicaid, insurance, or any other source?” Imputation of Out-of-Pocket Costs in PSID In each category, a small number of respondents who used services failed to report the exact amount spent (numbers shown in Table 3.1). These respondents were asked a series of unfolding questions to help determine a range in which their spending fell. Respondents who did not answer these unfolding questions were excluded from analyses on out-of-pocket costs for that health good or service. For respondents who did respond to at least some of the unfolding questions, out-of-pocket costs were imputed. Responses to unfolding questions were used to determine the range in which the respondent’s spending fell, and the respondent was assigned the mean amount spent by other respondents who reported spending a specific amount within this range. Covariates (such as age and gender) were not used in these imputations because there were no consistent predictors of the level of spending across spending variables. Also, using covariates for Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 35 imputation that will also be used as independent variables in regression analyses can artificially increase the likelihood that these variables will be significant predictors of costs. Figure 3.1 Questions on Health Care Utilization in PSID Since (date one year ago): Have you spent one night or more in a hospital? Have you spent one night or more in a nursing home, rehabilitation center, or other long term facility?1 Have you had surgery at an outpatient surgery facility, hospital, outpatient center, or an oral surgeon’ s office, that is, a place where you had surgery but you didn’ t stay overnight? Have you: - visited a general or family practitioner or general internist? - any other types of doctors or specialists, like cardiologists, surgeons, eye doctors, audiologists, or foot doctors physical, occupational, or speech therapists? - other health care providers for physical problems? - psychiatrists, psychologists, social workers, psychiatric nurses, or any other mental health care providers? - any other mental health care providers? - emergency rooms for physical or mental health care? Have you visited a dentist, an oral surgeon, or dental hygienist? Have you had one or more prescriptions filled for yourself? Have you - had to purchase prescription eyeglasses or a prescription hearing aid? - had to purchase or rent major medical items like crutches, wheelchairs, a special bed or chair, or remodel your home to accommodate any medical equipment or health condition? Has a paid professional, such as a nurse or trained attendant, come in and cared for your health or personal care needs (for example, changed dressings, given you medication, given you a bath, and things like that? Don’ t count people who only helped you with household chores). ■This question was prefaced with an description of facility types: A nursing home, sometimes called a skilled or intermediate care facility, is a place where people stay when they need nursing care, but don’ t need to be in a hospital. A rehabilitation hospital or unit is a place where people receive therapy, for example, after a stroke or accident. A residential care facility provides meals and assistance with daily living, but no nursing care. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission. Table 3.1 Percent with Missing and Imputed Data on Survey Questions Measuring Out-of-Pocket Costs in PSID, Summary Variables Created for Analyses Service type asked about in survey question Number Reporting Use Number Missing on Spending Number Imputed Number Excluded Combined spending variables used in analyses Hospital Stay 213 19 11 8 Hospital Care Physician treatment during hospital stay 213 33 16 17 Surgery with no overnight stay 135 16 8 8 Outpatient Surgery Physician treatment during surgery 135 8 3 5 Physician/Therapist/ER Visits 899 97 57 40 Ambulatory Visits Prescription Medications 858 113 77 36 Prescription Medications Dental care or oral surgery 395 11 8 3 Dental Care Medical Equipment 344 18 13 5 Equipment Stay in nursing home or other long term facility 49 4 4 0 Nursing Home Care Physician, therapist treatment in nursing home 49 5 1 4 Home Care 51 2 1 1 Home Care Insurance 984 58 58 - Insurance U ) ON 37 Out-of-pocket costs for insurance were reported on the mail-in questionnaire. Respondents missing on insurance were not asked unfolding questions to determine if their spending fell within a certain range. Imputation was done for all those missing on insurance using a hot-deck method with three covariates: gender, race, and whether or not insurance was obtained through an employer. Using this method, those missing on insurance expenditures were randomly assigned an amount that was reported by another respondent of the same gender and race and who had insurance from the same source. The first column in Table 3.1 shows the specific categories for which out-of- pocket costs were reported, and the second column shows the number of respondents who reported using that service. The third column shows the number of users who did not give a specific amount when asked about out-of-pocket costs. The fourth column shows the number of respondents who completed unfolding questions, and were assigned the mean for the spending category in which their spending fell. The fifth column shows the number who were excluded from analyses on that spending variable because they did respond to unfolding questions. (The numbers shown in Table 3.1 are not weighted, in order to show the actual number of respondents for whom responses were imputed.) As seen in the first column of Table 3.1, out-of-pocket costs for hospital, surgery, and nursing home care were reported separately from the costs of physician bills and therapy associated with this care. These costs were also imputed separately, and then combined to yield a single variable for each service, as seen in the last column of Table 3.1. This yielded variables reflecting out-of-pocket costs for eight types of services: 1) hospital care, 2) outpatient surgery, 3) physician/therapist/ER visits, 4) prescription Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 38 medications, 5) dental care, 6) equipment, 7) nursing home stays, and 8) home care. (Although not shown in Table 3.1, expenses for those who reported more than one hospital or nursing home stay were added to the out-of-pocket costs for the first stay.) Two summary variables measuring total out-of-pocket health spending were also created; one included expenditures on all eight services combined, and the other included these expenditures as well as out-of-pocket costs for private insurance or HMO premiums. Respondents were set to missing on the variables measuring total spending if they had been set to missing on any of the variables measuring spending for specific services. The sample sizes for each spending variable before and after imputation are shown in Chapter 5, where mean and median costs are reported. Out-of-Pocket Costs in AHEAD In AHEAD wave 2, respondents were asked about their use of eight different health goods and services in the two years since their first AHEAD interview. These are shown below in Figure 3.2. Respondents were asked to report their out-of-pocket health expenditures for these services in four categories, shown in Table 3.2: 1) nursing home and hospital stays, 2) physician visits, ER visits, outpatient surgery, and dental care3 , 3) home health care and other special services4 , and 4) prescription medications. 3 This category will be referred to as ‘ambulatory services’. 4 Those residing in a nursing home at the time of the wave 2 survey were not asked about use and costs for home care and other special services, and are thus excluded from analyses on this variable. However, these respondents are included in variables reflecting total out-of-pocket spending, and are given values of zero for spending on home care/special services. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 39 Figure 3.2 Questions Regarding Health Care Utilization in AHEAD Since wave 1 date: Have you been a patient in a hospital overnight? Have you been a patient overnight in a nursing home, convalescent home, or other long-term health facility? Have you seen or talked to a medical doctor about your health, including emergency room or clinic visits? Have you had outpatient surgery, not counting hospital stays? Seen a dentist for dental care, including dentures? Do you regularly take prescription medications? Has any medically trained person come to your home to help you (home medical care)? Have you used a special facility or service which we haven’t talked about, such as: an adult care center, a social worker, an outpatient rehabilitation program, or transportation or meals for the elderly or disabled? Table 3.2 Number Imputed on Questions on Out-of-Pocket Spending in AHEAD Question on Out-of-Pocket Spending Number Reporting Use Number Missing on Spending Number Imputed Number Excluded About how much did you pay out-of-pocket for nursing home and hospital stays since wave 1 date? 2238 384 349 35 About how much did you pay out-of-pocket for doctor, outpatient surgery, and dental bills since wave 1 date? 6024 1160 1113 47 On the average, about how much have you paid out-of-pocket per month for prescriptions since wave 1 date? 4935 677 638 39 About how much did you pay out-of-pocket for in- home medical care, special facilities or services since wave 1 date? 1194 148 130 18 Supplemental Insurance 3964 1386 1386 - HMO 805 106 106 - Long-Term Care Insurance 589 102 102 - Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 40 Respondents who reported use of a service were asked if the costs for this service had been completely covered, partially covered, or not covered at all by Medicare, Medicaid, or other health insurance. Those reporting no coverage or only partial coverage were asked to report their out-of-pocket costs in the relevant spending category. For the first three expenditure categories, respondents were asked to report their expenditures over the past TWO years, since the date of the AHEAD wave 1 survey. For prescription medications, respondents were asked to report their average monthly out-of-pocket expenditures over the past two years. In order to obtain a 2-year estimate of prescription costs, reported monthly costs were multiplied by 24. In addition to reporting out-of-pocket costs for health services, respondents were asked about the cost of premiums for three types of insurance: supplemental insurance, HMO premiums, and insurance covering long-term care. Respondents reported monthly, quarterly, or yearly premiums, and the amount was multiplied by the time frame to give the estimated amount spent over two years.5 All respondents were assumed to have held the insurance and paid the reported premium for the previous two years. This may have overestimated premiums for respondents who did not have coverage for the whole two-year period, or who experienced large premium increases over the survey period.6 5 Those who reported an amount on an insurance variable for a time period other than month, quarter or year were counted as missing and given imputed amounts using the hot-deck procedure, since information on time periods other than these was not yet available. 6 Also, for those who reported being enrolled in a pre-paid health plan for less than two years, it was estimated that the cost of previous coverage would have been approximately the same as the HMO premium. Because the mean reported insurance premium was higher than the mean reported HMO premium before imputation (as seen in Table 5.5 in Chapter 5), this likely led to a slight underestimation of insurance costs for some respondents in HMOs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 41 For respondents who reported holding more than one supplemental insurance policy and reported an amount spent on a second policy, this amount was added to the amount reported for the first policy. For long-term care insurance, respondents indicated how long they had had the insurance, and the reported amount was calculated for this time period. Imputation of Out-of-Pocket Health Costs in AHEAD For each of the health spending variables, a hot-deck procedure was used to impute values for those who did not give an exact spending amount, but indicated in subsequent unfolding questions that they had spent within a certain range. Actual responses of those who reported spending within each range were randomly chosen and assigned to those who did not give an amount but estimated spending to be within that range. For example, a respondent who did not give an exact amount and estimated his spending to be over $500 but under $5,000 could have been assigned any amount within this range that was reported by another respondent who was not missing on the spending variable. Those missing on the spending variable who reported in unfolding questions that they simply spent under a certain amount, such as “under $500" were never assigned zero, since it was assumed that those who spent nothing would not have failed to provide an amount. Covariates such as age and gender were not used for imputation, since none consistently predicted spending for those who responded, and using specific covariates for imputation might artificially increase the likelihood that these variables would be significant predictors of costs in analyses. The hot-deck method was chosen (rather than the category mean imputation Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 42 procedure used in PSID) because there was a larger number of people missing in several categories of each of the spending variables in AHEAD, and assigning the category mean would have reduced the variation in expenditures, leaving less variation to be accounted for in regression analyses. The percentage of responses imputed in this manner for each spending variable is shown in Chapter 5 in Table 5.2. Those who did not provide an amount and failed to answer the subsequent unfolding questions were set to missing for that spending category. Those with no use were assigned an out-of-pocket amount of zero. Imputation o f Out-of-Pocket Costs for Insurance Premiums in AHEAD Those who reported using an insurance type but did not give an amount spent on premiums were not asked subsequent questions about their range of spending. Thus, rather than imputing within unfolding categories for those who were missing, spending on each type of insurance was imputed within four race and gender categories using a hot- deck procedure. These covariates were chosen because they were significant predictors of the amount spent on HMO premiums and on supplemental insurance, among those who gave amounts for these insurance types. Those who were missing on the questions regarding use for both HMO and supplemental insurance (or who were missing on one and reported not having the other) were set to missing, since there was insufficient information on use to make an estimation of costs. Those who reported no insurance use or who indicated that their insurance was fully-covered by an employer were assigned zero. For those who said insurance was partly Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 43 covered by an employer but were missing on the amount spent out-of-pocket on premiums, donors in the hot-deck imputation procedure were restricted to those who had reported partial employer coverage . For long-term care insurance, amounts were assigned from respondents who indicated coverage for the same time period as the respondent with a missing amount. Those who reported more than one supplemental insurance plan were asked about the cost of the second plan. However, they were not asked about costs on plans beyond the second plan. Those who indicated that they held more than one supplemental plan and did not provide an amount for the second plan were not assigned an imputed amount for the that plan, since the variable indicating the number of plans appeared to be unreliable. For these reasons, the insurance variable may be an underestimate of total spending on insurance. Variables Reflecting Total Out-of-Pocket Costs in AHEAD After imputations for each of the spending variables, two variables reflecting total out-of-pocket costs were created: one included health expenditures in the four categories queried, and the other included these expenditures as well as expenditures on insurance premiums. Both variables reflected total spending over the previous two years. Respondents who had missing data on any of the four health spending variables (and who were not given imputed scores because they did not indicate that their spending fell within a certain range) were set to missing on the variables measuring total spending. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Measurement of Independent Variables in PSID and AHEAD This section describes the measurement of independent variables used in multivariate analyses, in the order shown in the conceptual framework in Chapter 2. Demographic variables: Age was measured continuously in years, and is discussed in more detail in Chapter 6. Race was measured using dummy variables, with Whites as the referent category. Minority race categories included African American (Black) in both PSID and AHEAD, as well as Hispanic in AHEAD. The measurement of race is discussed in more detail in Chapter 7. As explained in that chapter, a small portion of respondents in each sample reported themselves as a minority race other than Black or Hispanic. These respondents are set to missing for all analyses in Chapter 7. However, these respondents are included in analyses for other chapters, and are left with Whites as part of the referent group in regression analyses. Gender was self-reported, and females were coded as 1 in both survey samples. Socioeconomic variables: Because a main focus of both surveys was income and financial status, the surveys used in this study have extensive and thorough measures of income and assets. In PSID, income is a measure of total earnings, transfers, and asset income from the prior calendar year for the couple or individual if single. Wealth is a measure of the net value of assets (including respondent’s home) minus the value of debts (other than mortgages).7 Those who did not give an exact dollar amount on specific questions about assets were asked to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 45 indicate into which of several dollar ranges their answer would fall, and responses were imputed within this range by researchers at the University of Michigan who prepared the PSID data, using a hot deck method. Those who did not complete the wealth supplement to the PSID in 1989 were missing on wealth. In order to include these 80 respondents in analyses, a dummy variable was constructed indicating that for that observation, wealth data were missing and the sample mean was assigned. In AHEAD, income was based on self-reported income in wave 1 for the past year for the couple or individual if single. The measure of wealth used was based on a net worth summary variable from wave 1 of AHEAD, based on responses to many detailed questions regarding assets and debts.7 Wave 1 measures of income and wealth were used rather than wave 2 measures for two reasons.8 One reason is that economic measures at wave 1 could be considered a more accurate measure of the resources available to be spent on health care in the following two years. Another is that detailed imputation had been done by AHEAD staff for all of those missing on wave 1 wealth and income. Imputation was done by AHEAD staff using a hot-deck procedure. For those 7 In both PSID and AHEAD, assets included: net worth of real estate (including respondent’s home), vehicles, farm or business assets, stocks, mutual funds, IRA or Keogh accounts, money market funds, investment trusts, checking and savings accounts, certificates of deposit, savings bonds, Treasury bills, investments in trusts or estates, bond funds, life insurance policies, and special collections. Debts included money owed on such things as credit cards, medical or legal bills, or personal loans. 8 Results were similar when a wave 2 income measure was used, with some imputation by the author. A hot-deck procedure was used to impute an income amount for the 30.7% of the sample who did not provide an exact amount but indicated that their income fell within a certain range. In order to avoid losing the 12% who gave no income information in analyses, they ware assigned the mean score, and a variable indicating that they were missing on income was included. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 46 who provided an exact amount but indicated that their income fell within a certain range, an amount within this range was assigned. Respondent characteristics such as gender and race were also used as covariates. Whereas the measure of income used was a single self- reported amount, the wealth variable was the sum of many questions on different types of reported wealth, and was created by AHEAD staff. Individual components of wealth were imputed for those who were missing prior to the creation of the summary income and wealth measures. Because mean income and wealth are both skewed by a small number of high outliers, these measures were logged in analyses. Before taking the log, all negative wealth values were set to zero, and then those with values of zero for wealth were assigned a value of $1. Median weighted household wealth and income are shown in Table 3.1. Insurance coverage was measured using mutually exclusive dummy variables to represent those with employer-sponsored insurance, those in a prepaid health plan (HMO), and those dually-eligible for Medicaid. Those with privately purchased supplemental (medigap) insurance were used as the referent group. The creation of insurance variables is discussed in Chapter 8. Those without complete insurance information were excluded from analyses in Chapter 8. However, in order to include these respondents in analyses for other chapters where insurance was not the main focus, a dummy variable was created for those missing on insurance, and they were set to zero on other insurance dummy variables. Dental insurance, insurance covering prescription medications, and long-term care insurance were represented by dummy variables which were not mutually exclusive. Education level was self-reported and measured continuously in years in both Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 47 survey samples. In PSED, education was imputed for 25 respondents with missing data. Information from other questions on high-school completion and degree-level was used to determine approximate education level, and the mean number of years for respondents at that level was assigned. To control for regional differences in health costs, a variable was created for each individual equal to the mean Medicare expenditure per beneficiary in the area of the country where the respondent lived. In PSED, expenses were controlled at the state level, and spending figures for 1992 were used (this was the closest approximation possible, since spending figures for 1990 or 1991 were not available). In AHEAD, expenses were controlled for nine census divisions, using 1995 spending figures to correspond with the last part of the two-year period for which respondents were reporting health care costs. In PSED, urban residence was controlled using a dichotomous variable based on the Beale-Ross Rural-Urban Continuum Code; those living in areas classified as completely rural (either adjacent or not adjacent to a metropolitan area) were coded as zero, and others were coded as 1 on the urban variable. In AHEAD urban residence was controlled using a dichotomous variable where 1 indicated that the respondent lived in a Metropolitan Statistical Area (MSA), as defined in 1990 by the U.S. Census Bureau. Need: Self-rated health was measured in both surveys on a 5 point likert scale. In PSID, respondents were asked to compare themselves to those their own age, whereas AHEAD respondents were asked to rate their health in general. For analyses, responses were Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 48 reverse-coded so that 1 indicated poor health and 5 indicated excellent health. Functional ability was measured in both surveys based on respondents’ self- reported ability to perform Activities of Daily Living (ADLs). In PSID, an 8-point scale was used. Respondents were assigned one point for each of the following activities which they were limited in performing without help due to their health: eating, dressing, getting out of bed, getting out of a chair, walking inside, going outside, bathing, and using the toilet.9 In AHEAD, level of functional difficulty was measured on a six point scale, based on respondents’ ability to walk across a room, dress, bathe or shower, eat, get in or out of bed, and use the toilet. For each of these Activities of daily Living, one point was added to the scale if respondents reported having any difficulty performing the activity due to a health or memory problem, or if they reported that they used equipment or that someone helped them perform the activity. Two other measures of health used in the AHEAD survey were bed disability days and. falls. The AHEAD survey asked whether respondents spent one or more days in bed in the previous month due to illness or injury, and if so, how many days were spent in bed. Because 88% of respondents did not spend any days in bed, the variable was coded as ^In PSID, ADL limitations ware measured on the mail-in questionnaire using the question: Does your health limit you in each of the following activities? A yes/no response was required for each of the activities. This section was followed by a question on whether limitations had existed for three months or longer (with the response options: yes, no, or ‘not limited in any activity’). Several (about 55) of those who reported a limitation in one or more ADLs reported on the follow-up question that they were NOT limited in any activity. There were also several (about 75) who reported no limitations but answered yes or no to the follow-up question. Based on how the questions ware worded, it seems most likely that these individuals failed to carefully read the individual ADL questions and meant to give the opposite answers. Thus, for these respondents with discrepancies, the answers to the ADL questions were revarsed to create the summary variable. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 49 dichotomous rather than continuous, with those reporting more than 10 bed disability days coded as 1. Falling was measured in AHEAD using a dichotomous variable on which a value of 1 was assigned for those reporting a fall in the previous 2 years. In order to control for specific medical conditions in PSID, dummy variables were created for each of eight chronic conditions or illnesses: cancer, diabetes, high blood pressure, heart disease, neurological problems (including stroke), lung disease, digestive conditions, and arthritis. Specific medical conditions as they were queried in the mail-in questionnaire are shown in the Appendix. The ‘heart disease’ variable combined those reporting heart failure/enlarged heart and those using a pacemaker, and the ‘digestive conditions’ variable combined those reporting an ulcer and those reporting colitis/chronic inflamed bowel/enteritis. Respondents who reported ever having the disease were coded as 1. In order to include respondents who were missing on health conditions in regression analyses, they were assigned to the category of not having the disease. In AHEAD, dummy variables were created for each of nine chronic diseases: cancer, diabetes, high blood pressure, heart attack, congestive heart failure (CHF), other heart conditions, stroke, lung disease, and arthritis. Respondents who reported ever having the disease were coded as 1 (for heart attacks, respondents were only asked to report if one had occurred in the 5 years preceding the wave 1 interview or the two years between waves). The survey questions regarding medical conditions in AHEAD wave 2 can be seen in the Appendix. For heart conditions, respondents were first asked if they had any heart condition and then asked specifically about heart attack and CHF. Those who reported a heart condition but did not report heart attack or CHF were coded in this study Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 50 as having ‘other heart problems’. In order to include respondents who were missing on health conditions in regression analyses, they were assigned to the category of not having the disease. Depression was measured in AHEAD using a modified 8-item/8-point version of the Center for Epidemiological Studies— Depression Scale (CES-D) (Radloff, 1977), adapted especially for the AHEAD survey to assess depression. Higher scores indicate greater depression. Depression was not used as a predictor of costs in PSID. Although the PSID mail-in questionnaire asked a few depression questions, these did not constitute a depression scale, and preliminary analyses using a depression score created from these questions revealed that it was not a significant predictor of costs. Cognitive status was measured in AEEAD using a test containing 9 questions from the Telephone Interview for Cognitive Status (TICS) (Brandt, Spencer, & Folstein, 1988). Participants were asked to give the present date (month, day, year, day of the week), name the president and vice president, identify two items (scissors and cactus) from their descriptions, and count backward from 20. The maximum score was 10 points. The 768 (12.5%) who had a proxy responding to the interview for them did not complete the mental status test. (Proxies were used when the survey participant was too physically or mentally impaired to be interviewed). In order to include these those with proxies in the sample, they were assigned the mean mental status score, and a dichotomous variable indicating proxy interview was included. Cognitive status was not measured in PSID. Measures of vision and hearing problems were used in PSID for predictions of expenditures on medical equipment, including glasses and hearing aids. Those who Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 51 reported trouble seeing or deafness were coded as 1 on the sight and hearing variables (as shown in the Appendix), respectively. Table 3.3 shows the mean, standard deviation, and median for each of the continuous predictor variables, as well as the percent coded as 1 on each of the dichotomous variables. The significance of each of the control variables as a predictor of out-of-pocket costs is discussed in Chapter 9. Table 3.3 Means and distributions on predictor variables in PSED and AHEAD (weighted) PSED Elderly Health Supplement (n=1031) AHEAD Wave 2 (n=6251) Variables Percent Median Mean SD Percent Median Mean SD Demographic/ Geographic Age 73.0 74.7 6.9 78.0 79.1 5.7 Percent Female 62.5 62.8 African American 14.3 9.6 Hispanic - - - - 3.7 Regional Spending 4,753 4,760 417 Urban Resident 94.5 75.8 Health Self-Rated Health 3.0 2.7 1.0 3.0 3.0 1.2 Disability 0.0 1.6 2.8 0.0 0.9 1.5 10+ Days sick in Bed - 3.4 Cancer 6.0 16.5 Diabetes 11.7 14.4 Lung Disease 11.5 12.6 (Table continues) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 52 Table 3.3 continued PSID Elderly Health Supplement AHEAD Wave 2 Variables Percent Median Mean SD Percent Median Mean SD High Blood Pressure 40.8 54.6 Stroke 6.1 10.7 Digestive Problem 18.1 Heart Disease 12.5 24.1 Heart Attack 9.1 CHF 4.4 Arthritis 60.4 55.4 Fall in past 2 years 32.9 Depression 0.0 0.2 0.6 1.9 1.9 2.0 Cognitive Status - 9.0 9.0 1.4 Proxy Respondent - - 12.6 Sight problems 26.1 . . . . Deafness 22.5 . . . Insurance FFS Medicare Only 11.5 14.0 Medigap Insurance 37.4 47.8 Employer Insurance 24.5 14.0 Prepaid Health Plan 7.6 11.7 Medicaid 6.3 10.5 Missing on Insurance 12.7 2.0 Dental Insurance 8.4 18.0 Prescription Insurance 30.9 51.6 Socioeconomic Years of Education Household Income Household Wealth 12.0 10.8 3.5 12.0 11.1 3.6 16,400 23,844 26,470 15,101 23,256 29,764 62,000 163,578 378,999 86,000 192,803 400,328 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 53 Chapter 4 Methodological Approach Descriptive analyses In each Chapter, those in different age, race, and insurance groups are compared on each service type in terms of use, proportion of users with uncovered costs, mean and median out-of-pocket costs, and the proportion of household income spent on total costs. Mean vs. Median Spending Median spending is reported in addition to mean out-of-pocket costs, due to the skewed distribution of reported spending on each service type. For services where a small portion of users had very high costs that inflated the estimate of mean spending, median expenditures provide a more accurate estimate of out-of-pocket costs for the typical respondent. Out-of-Pocket Costs Across All Respondents vs. Users For each service type, mean and median out-of-pocket costs were calculated for the whole sample as well as among only those using each service. Measures of out-of- pocket spending across all respondents provide an estimate of the average out-of-pocket liability across the whole older adult population. This is of particular interest for services with high rates of use, and for estimates of total out-of-pocket expenditures. However, for services with low rates of use, mean spending across all respondents is less informative, because the amount spent by users is diluted by the large number of non-users with zero costs. For these services, measures of mean spending among only those using each service provide a better estimate of whether users typically face high out-of-pocket costs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 54 Estimates of Out-of-Pocket Burden In examining out-of-pocket costs, it is important to consider not only the magnitude of expenditures, but also individuals’ ability to pay for these costs. Even costs that might be seen as low or reasonable can be burdensome for those with few financial resources, and it is important to distinguish between high cost and financially catastrophic cases (Wyszewianski, 1986). To estimate the burden of out-of-pocket health spending on individuals in this study, total out-of-pocket costs for each respondent were divided by household income1 . Household income was used rather than couple or individual income because the questions regarding out-of-pocket costs in PSID specifically asked about expenditures by the respondent or his or her “family living there”. Although spending questions in AHEAD did not mention other family members, household income was used in AHEAD to maintain consistency in methods across the two survey samples. While this method underestimates the total out-of-pocket burden on the couple in cases where both members incurred costs and would have had to draw on the same income, it is an appropriate measure of individual burden. An alternative method would have been to use individual income for those living alone, and to divide the income of couples in half, calculating each member’s costs as a portion of his or her half of the income. However, this may overestimate individual burden, and would not be an accurate representation of the typical pooling of income by couples to pay for the needs of either lIn PSID, reported total out-of-pocket costs for the year were divided by yearly household income. In AHEAD, since out-of-pocket costs were reported for the previous two years, reported income for the previous year in wave 1 was multiplied by two before being used as the denominator in the ratio of costs to income. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 55 individual as they arise. In their estimates of out-of-pocket burden in the NMCUES, Feder et al (1987) used per capita income, but reported that results remained fundamentally unchanged when individuals and couples were analysed separately. For the small number of respondents who lived with family members other than a spouse, the use of household income assumes that the income of these other members would have been available to help pay for the health care costs of the respondent. Because there were likely households in which this was not the case, using household rather than couple income may have contributed to an underestimation of burden for these respondents. Some respondents reported spending 100 percent or more of their household income on total out-of-pocket health costs. This may reflect a situation in which respondents used savings in addition to income to pay for care, or may be the result of an overestimation of costs and/or underreporting of income. In order to avoid an upward bias in the mean proportion of income spent, spending as a proportion of income was capped at 100 percent for these analyses. The few respondents reporting zero household income and nonzero spending were also coded as spending 100 percent of income. Sensitivity analyses indicated that results would have been similar if this cap had not been applied, which is consistent with previous reports on this issue (Crystal, Johnson, & Kumar, 1998). Estimating Catastrophic Costs Additional descriptive analyses were performed in order to examine the proportion of respondents who faced out-of-pocket burdens that could be considered catastrophic. The proportion of household income spent was broken down to determine the percentage Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 56 spending over 10%, over 30%, and over 50% of income on out-of-pocket costs. While legislative discussions and private insurance policies often focus on ‘catastrophic costs’ as a problem of very large medical bills, it is more appropriate to focus on the burden of costs relative to ability to pay (Feder et aL, 1987). However, there is no agreement regarding the level of burden that should be considered catastrophic. Feder et al. (1987) point out that this decision is a matter of values; catastrophe can be defined as a level of burden that threatens one’s existing standard of living, or a standard of living defined as ‘reasonable’. Whether costs are catastrophic will vary by income and circumstances. Feder et aL (1987) report that expenses from 10 to 20 percent of income are typically considered catastrophic. Wyszewianski (1986a) considers 15 percent to be an appropriate cutoff for catastrophic costs, based on the use of this figure in other studies. Another potential cutoff is the threshold set by Congress after which out-of-pocket expenses are considered tax deductible, which was 7.5 percent of adjusted gross income in 1987 (Berki, 1986). This dissertation uses three thresholds of 10, 30, and 50 percent of income. Thresholds of 10 and 30 percent are used in an attempt to identify cases near the lower and higher ends of the range that might be considered catastrophic, and 50 percent is used to identify cases that would certainly be considered catastrophic. Individual vs. Aggregate Approach It is important to clarify that for calculations of out-of-pocket burden and of the distribution of spending by service type, an individual rather than an aggregate approach Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 57 was used. For estimates of burden, the spending ratio was computed for each individual, and then the mean of these ratios was calculated. In estimates of out-of-pocket burden using an aggregate approach, mean out-of-pocket costs are divided by mean income for the whole population. This approach is often necessary in reports that use macro-data, which does not contain information on individual incomes (e.g. Ways and Means, 1996). However, when data are available, the individual approach is preferable. This is because in aggregate estimates, the small portion of elderly with very high incomes tend to skew the mean income, leading to underestimates of individual burden (Moon, 1991; see AARP and The Urban Institute, 1994). An illustrative example of differences between the individual and aggregate approach is shown in Figure 4.1. For calculations of the distribution of total out-of-pocket costs by service type, there are also important differences between the individual and aggregate approaches. For the individual approach, used here, each respondent’s spending on each service type is computed as a portion of his or her total out-of-pocket costs. The mean of these portions is then calculated for each service type to obtain the mean portion of total costs spent on that service for the whole sample. In the aggregate approach, the breakdown of spending on each service type is calculated by dividing the mean total out-of-pocket costs for the whole sample by the mean amount spent on each service. When using the aggregate approach, the breakdown of spending on different services can be shown as a dollar amount spent on each service. However, when the individual approach is used, only the percentage of total costs spent on each service can be shown, since the amount is different for each respondent. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 58 Figure 4.1 Illustration of the Difference Between Individual and Aggregate Approaches* Individual Approach: For the group of three respondents below, mean individual spending as a proportion of household income is 14.3 percent. Respondent 1: $1.200 out-of-pocket costs = 3 % $40,000 income Respondent 2: $3.000 out-of-pocket costs = 30 % $ 10,000 income Respondent 3: $ 1.600 out-of-pocket costs = 10 % $ 16,000 income Calculation of mean burden: 3 % -+ 30 % +10% = 14.3 % 3 Aggregate Approach: For the same group of three respondents, total mean spending as a proportion of total mean income is 8.8 percent. Total out-of-pocket spending: $1,200 + $3,000 + $1,600 = $5,800 Total income: $40,000 + $10,000 + $16,000 = $66,000 Calculation of mean burden: $5.800 = 8.8 % $66,000 *format of illustration borrowed from AARP and The Urban Institute, 1994. Multivariate Analyses For each of the spending variables described in Chapter 3, hierarchical regression was used to examine the effects of independent variables on out-of-pocket costs. Entry of variables in each hierarchical step was guided by the conceptual framework described in Chapter 2. The order in which variables were entered was different in each chapter, depending on the variables of interest. Thus, each chapter contains a detailed description Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 59 of the specific hierarchical steps used. Because spending variables in both survey samples were skewed, with a small proportion of users reporting high costs, the natural logarithm of spending was used as the dependent variable for all service types. Ordinary Least Squares Regression and a Two-Part Model For most spending variables, a sizeable proportion of the sample had zero out-of- pocket costs because respondents did not use that type of health service in the period covered by the survey. The portion of respondents on each spending variable who had zero costs due to non-use is shown in Table 4.1. For variables with a large truncation at zero such as these, ordinary least squares (OLS) regression should not be used on the whole sample, since the assumption of a normal distribution on the dependent variable would be seriously violated, leading to inflated standard errors and biased coefficients. Also, the truncation at zero on these spending variables indicates that they are measuring two different things: health care use, and the level of costs dependent on use. There may be different patterns of predictors for these two parts of the variable. For example, Pohlmeier, Ulrich, Konstanz, & Mannheim (1995) concluded that the decision to see a physician and the decision regarding how often to contact the physician are made by two different decision makers and should be treated as two distinct processes. The decision to use a service is also likely to be somewhat separate from decisions regarding factors that affect the level of costs, such as frequency of use and the type and intensity of treatment. Thus, for analyses of these spending variables, a standard two-part model was used to examine costs conditional on use. First, probit analysis was used to estimate the probability of any use, and then OLS regression was used to estimate the total (log) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 60 expenditures among users. For some spending variables, such as those reflecting total out-of-pocket costs, the proportion of respondents with no use was small enough that a two-part model was not necessary. For these variables, OLS regression was used to predict out-of-pocket costs for all respondents. Table 4.1 indicates which spending variables were analyzed using a two- part model, and which only required OLS. Truncation at Zero Among Users Another problem with truncation at zero in this data is that even among users only, there were some spending variables on which a substantial proportion of users had zero out-of-pocket costs. For example, as seen in Table 4.1, 43.6% of those using hospital care in PSID reported zero out-of-pocket costs, and 66% of those using hospital/nursing home care in AHEAD reported no uncovered costs. Using OLS to predict spending on variables where there was truncation at zero among users could inflate standard errors and lead to biased OLS estimates. While taking the natural log of the spending variables helps to correct for this problem, an additional method of addressing this truncation is to use Tobit analysis rather than OLS regression to predict costs among users. The Tobit model was developed for censored variables, and is a hybrid between probit analysis and multiple regression (Tobin, 1958). In order to examine whether truncation at zero among users was an important problem with this data, additional analyses were done on a using Tobit analysis instead of OLS in the second part of the two-part model. Results (not shown) were almost identical to those obtained using OLS, indicating that the truncation at zero among users did not lead to biased estimates for these spending variables. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 61 Correction for Sample Selection Because the equations predicting spending in the two-part model are only done on the select group of self-reported users, the argument could be made that estimates are biased and do not reflect effects for the whole population. A method developed to correct for this type of selection bias is the self-selection or Heckman model (Heckman, 1976). This is a two-stage model as described above, which uses probit analysis followed by regression analysis among users. However, in order to correct for non-random selection into the group with data on the dependent variable, a correction term is included as a regressor in the second stage. This term, called the inverse of Mills ratio, reflects for each observation the inverse of the probability of that observation having data on the dependent measure. Including this term in regressions on users yields estimates describing expenditures that all respondents (including those without use) would have if they were all users. There is debate among economists as to whether it is appropriate to use a self selection model to examine health care expenditure data such as this. Researchers examining data from the Rand Health Insurance Study argue that the self-selection model is not appropriate for modeling health expenditure data because we know that non-users have zero expenses (Duan, Manning, Morris, & Newhouse, 1983; 1984). Unlike people in the usual self-selection problem, non-users with zero expenditures are not cases with missing data on the expense variable. Thus, rather than the unconditional equation given by the self-selection model, we are interested in the conditional equation obtained when the correction is not used (Duan et aL, 1983; Duan et aL, 1984). This conditional equation Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 62 gives estimates of actual expenditures among those who use services, and not 'potential’ expenses of the total sample. Based on this argument, a correction for sample selection is not used for the two-part models in this dissertation. However, in order to examine the effects of selection in this data, additional analyses were performed using the self-selection modeL Results were very similar to those from the main analyses, indicating that selection into the group of users was not an important problem that affected parameter estimates from two-part models in this study. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 63 Table 4.1 Percentage of Respondents with No Use and Percentage of Users with Zero Costs for Each Spending Variable in PSID and AHEAD Survey Samt>Ie Spending Variable % with no use % of users with zero costs Method Used Alternative Methods AHEAD Hospital, Nursing Home 65 66.6 2-part model 2-part with Tobit as second step 2-part with selection correction Ambulatory Services 3.4 28.4 OLS* - Prescription Medication 21 17.8 OLS - Home Health Care, Special Services 81 71.9 2-part model 2-part with Tobit as second step 2-part with selection correction Total excluding insurance 2.1 13.7 OLS - Total including insurance 0.9 7.8 OLS - PSID Hospital 78.5 43.6 2-part model 2-part with Tobit as second step 2-part with selection correction Outpatient Surgery 86.3 46.6 2-part model 2-part with Tobit as second step 2-part with selection correction Physician, ER Visits 11.2 34.2 OLS - Prescription Medication 15.3 18.7 OLS - Dental Care 59.8 9.7 2-part model 2-part with selection correction Equipment 66.4 13.7 2-part model 2-part with selection correction Nursing Home Care 95.2 30.2 2-part model 2-part with Tobit as second step 2-part with selection correction Home Care 94.8 85.7 2-part model 2-part with Tobit as second step 2-part with selection correction Total excluding insurance 4.1 7.4 OLS - Total including insurance 2.2 5.9 OLS - * Ordinary Least Squares Regression Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 64 Chapter 5 Out-of-Pocket Spending: Descriptive Analyses This chapter examines out-of-pocket costs in both PSID and AHEAD, overall and in each spending category. Whereas later chapters examine out-of-pocket costs stratified by the independent variable of interest, expenditures are examined here across all respondents in each survey sample. First, mean and median spending before and after imputation are compared. Then results are compared to other reports of out-of-pocket costs among older adults. Finally, other descriptive analyses are presented, including rates of reported use and uncovered costs among users of each service, the burden of total out- of-pocket spending as a portion of household income, and the distribution of out-of- pocket spending by service type. The first eight Tables in this Chapter compare non-imputed spending on each service type to spending estimates that include those whose costs were imputed, following the methods described in Chapter 3. For each service type, mean and median spending are presented, first for only respondents who initially provided an exact spending amount, and then for all respondents who either reported an amount or were assigned an imputed amount. The last two columns of each Table show the number and percent imputed on each spending variable. All means and medians are weighted, however the N’s shown in these Tables are not weighted, in order to give an exact report of the number of respondents imputed on each variable. Tables 5.1 to 5.3 compare non-imputed and imputed mean and median out-of- pocket expenditures in PSID. Table 5.1 compares imputed and non-imputed variables only among those who used each service and had uncovered costs. Table 5.2 does the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 65 comparison for all users c f each service, including those with zero out-of-pocket costs. Finally, Table 5.3 compares imputed and non-imputed spending for the whole sample, including non-users. Since only those with nonzero costs were imputed (or were used to provide category means for imputation), Table 5.1 depicts the most appropriate comparison of imputed and non-imputed spending. As seen in Table 5.1, for most service types, imputed costs were only slightly different than non-imputed costs. In the case of hospital care, imputed spending was $158 higher, indicating that those who did not report a specific out- of-pocket amount for hospital care had higher spending (as estimated by unfolding questions) than those who did provide an amount. For nursing home costs, the situation was reversed, with spending substantially lower on the imputed variable. As seen in Tables 5.2 and 5.3, these large differences between imputed and non-imputed spending were attenuated as those with zero costs were added. This occurs because most respondents did not use hospital or nursing home care, and thus had zero costs. The addition of a large number of zeros to both imputed and non-imputed spending washes out the difference that initially appeared when only those with nonzero costs were compared. For most other services, which were used by a larger portion of respondents, the difference between imputed and non-imputed spending remains similar as those with zero costs are added. Table 5.4 shows spending on the two summary variables created to indicate total out-of-pocket costs, both excluding and including insurance premiums. (These summary variables are shown for all respondents, rather than for only those using services or only those with nonzero costs). Though no respondents were directly assigned an imputed Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 66 Table 5.1 Comparison of Out-of-Pocket Costs Before and After Imputation for All Users of Each Service with Nonzero Costs in PSID Service Type Non-imputed Costs Imputed Costs Number Imputed Percent Imputed Hospital Stays Mean 737 895 (SD) (1,285) (1,368) (max = 9,525) Median 400 411 N 102 118 16 13.6 Nursing Home Stays Mean 9,438 8,699 (SD) (13,216) (12,675) (max = 60,044) Median 5,092 3,466 N 30 34 4 11.8 Outpatient Surgery Mean 597 589 (SD) (1,303) (1,261) (max = 8,000) Median 200 214 N 67 72 5 6.9 Physician/Therapist/ER Visits Mean 168 204 (SD) (309) (461) (max = 7,600) Median 100 100 N 529 586 57 9.7 Dental Care Mean 311 327 (SD) (658) (750) (max = 8,000) Median 105 110 N 349 357 8 2.2 Prescription Medications Mean 342 358 (SD) (480) (482) (max = 5,100) Median 150 180 N 611 688 77 11.2 Equipment Mean 251 250 (SD) (351) (354) (max = 2,125) Median 150 150 N 281 294 13 4.4 Home Care Mean 791 736 (SD) (798) (744) (max = 2,160) Median 523 400 N 6 7 1 14.3 Insurance Mean 884 867 (SD) (705) (688) (max = 5,496) Median 718 709 N 480 538 58 10.8 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 67 Table 5.2 Comparison of Out-of-Pocket Costs Before and After Imputation for All Users of Each Service in PSID Service Type Non-imputed Costs Imputed Costs Number Imputed Percent Imputed Hospital Stays Mean 416 539 (SD) (1,031) (1,146) (max = 9,525) Median 32 139 N 181 197 16 8.1 Nursing Home Stays Mean 6,658 6,323 (SD) (11,947) (11,530) (max = 60,044) Median 729 1,000 N 43 47 4 8.5 Outpatient Surgery Mean 332 338 (SD) (995) (980) (max = 8,000) Median 23 35 N 125 130 5 3.8 Physician/Therapist/ER Visits Mean 117 147 (SD) (265) (395) (max = 7,600) Median 50 63 N 784 841 57 6.8 Dental Care Mean 285 300 (SD) (634) (722) (max = 8,000) Median 100 100 N 383 391 8 2 Prescription Medication Mean 294 313 (SD) (456) (461) (max = 5,100) Median 100 120 N 727 804 77 9.6 Equipment Mean 224 224 (SD) (338) (342) (max = 2,125) Median 138 131 N 320 333 13 3.9 Home Care Mean 101 107 (SD) (378) (378) (max = 2,160) Median 0 0 N 48 49 1 2 Insurance Mean 699 700 (SD) (730) 714 (max = 5,496) Median 600 (600) N 608 666 58 8.7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 68 Table 5.3 Comparison of Out-of-Pocket Costs Before and After Imputation for All Respondents in PSID Service Type Non-imputed Costs Imputed Costs Number Imputed Percent Imputed Hospital Stays Mean 77 107 (SD) (468) (549) (max = 9,525) Median 0 0 N 994 1010 16 1.6 Nursing Home Stays Mean 327 334 (SD) (2,819) (2,823) (max = 60,044) Median 0 0 N 1022 1026 4 0.4 Outpatient Surgery Mean 40 42 (SD) (364) (365) (max = 8,000) Median 0 0 N 1019 1024 5 0.5 Physician/Therapist/ER Visits Mean 101 128 (SD) (249) (372) (max = 7,600) Median 30 40 N 910 967 57 5.9 Dental Care Mean 115 123 (SD) (413) (470) (max = 8,000) Median 0 N 1015 1023 8 0.8 Prescription Medication Mean 239 259 (SD) (427) (436) (max = 5,100) Median 50 66 N 895 972 77 7.9 Equipment Mean 74 76 (SD) (218) (222) (max = 2,125) Median 0 0 N 1003 1016 13 1.3 Home Care Mean 5 5 (SD) (84) (85) (max = 2,160) Median 0 0 N 1025 1026 1 0.1 Insurance Mean 500 513 (SD) (671) (664) (max = 5,496) Median 288 264 N 926 984 58 5.9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 69 spending total, the summary variables include respondents who have imputed costs on one or more of the specific services that make up the totaL The last two columns of Table 5.4 show the number and percent of respondents who were imputed on one or more components of the total spending variables. The first column shows the mean sum of non- imputed spending on all services, and the second column shows the mean sum of imputed spending variables. The differences between non-imputed and imputed total spending are larger than the differences for each of the service types. Total spending excluding insurance is $135 higher when imputed variables are used. This reflects the additive effect of slightly higher imputed versus non-imputed spending on several service types. Table 5.4 Comparison of Total Out-of-Pocket Costs as the Sum of Non-imputed and Imputed Cost Variables, for All Respondents in PSID Sum of Non- imputed Cost Variables Sum of Imputed Cost Variables Number imputed on one or more cost variables Percent imputed on one or more cost variables Total Costs excluding insurance (max = 10,667) Mean (SD) Median N total 808 (2,326) 795 755 943 (2,449) 348 890 135 15.2 Total Costs including insurance (max = 33,520) Mean (SD) Median N total 1,333 (2,440) 809 674 1,487 (2,600) 911 850 176 20.7 Tables 5.5 to 5.8 compare non-imputed and imputed two-year spending in AHEAD, in each of the four spending categories, and on premiums for the three insurance types. For prescription medications, average monthly costs are shown, since this is how Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 70 costs were reported by respondents. However, because this amount was multiplied by 24 before being included in the total cost variable, the estimated two-year amount is also shown. Table 5.5 compares imputed and non-imputed spending only among those who used each service and had uncovered costs. Table 5.6 does the comparison for all users of each service, including those with zero out-of-pocket costs. In Table 5.7, imputed and non-imputed spending on each variable is compared for the whole sample, including non users. As seen in Table 5.5, differences between imputed and non-imputed spending were small for most service types. However, for hospital/nursing home care, imputed hospital costs were quite a bit higher than non-imputed costs. This indicates that those who did not report a specific out-of-pocket amount on this variable had higher costs (as estimated by unfolding questions) than those who did provide an amount. When comparing the difference between imputed and non-imputed spending for each variable across Tables 5.5 to 5.7, in some cases the difference between imputed and non-imputed variables becomes larger as those with zero costs are added. For example, in Table 5.5, mean imputed spending on supplemental insurance is $41 higher than non- imputed spending when comparing only those with nonzero premiums. When those with zero premiums are added in Table 5.6, the difference increases to $246. In Table 5.7, when those who have zero premiums because they did not purchase supplemental insurance are added, imputed spending is $393 higher than non-imputed spending. This occurs because the imputed mean includes more people than the non-imputed mean. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 71 When those with zero spending are added, they lower the mean for both the imputed and non-imputed variables. However, because there are more respondents with nonzero costs in the imputed mean, the zeros do not lower this mean as much as they lower the non- imputed mean. As a hypothetical example, consider a case where there are 10 individuals contributing to the non-imputed mean, and each reports spending $100. In the imputed mean, there are 15 individuals, each with costs of $100. In this case, both the imputed and non-imputed means would be equal $100. However, if 10 individuals with zero costs are added to each mean, the zeros would have a greater impact on the non-imputed variable containing only 10 individuals. This mean would be lowered to $50, whereas the imputed mean would be $60. Table 5.8 shows summary variables for total spending on insurance, and total out- of-pocket costs, excluding and including insurance premiums. The insurance variable is the sum of premiums for supplemental insurance, HMO premiums, and long-term care insurance. Total spending is the sum of out-of-pocket costs for hospital/nursing home, ambulatory visits, two-year prescription costs, and home care/other special services. The last column shows that just over a quarter of respondents were imputed on one or more of the components of the total spending variable. When insurance is included in total spending, the proportion with at least one imputed component rises to almost half, reflecting the large number of respondents with imputed premiums for one or more types of insurance. Total spending excluding insurance is $777 higher when imputed variables are used. This reflects the additive effect of somewhat higher imputed versus non-imputed spending on several service types. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 72 Table 5.5 Comparison of Out-of-Pocket Costs Before and After Imputation for All Users of Each Service with Nonzero Expenses in AHEAD Service Type Non- imputed Spending Imputed Spending Number Imputed Percent Imputed Hospital, Nursing Home Mean 4,759 5,441 (SD) (12,820) (13,410) (max = 120,000) Median 700 1,000 N 377 726 349 48.1 Physician, ER, Outpatient Mean 660 791 Surgery, Dental (SD) (1,427) (1,802) Median 300 356 (max = 45,000) N 2965 4078 1113 27.3 Monthly Prescription Mean 102 101 Medication (SD) (268) (263) Median 40 (max = 4,800) N 3294 3932 638 16.2 Two-Year Prescription Mean 2,437 2,426 Medication (SD) (6,435) (6,303) Median 960 960 (max = 4,800) N 3294 3932 638 16.2 Home Health Care, Mean 2,238 2,301 Special Services (SD) (10,512) (10,114) Median 243 180 (max = 90,000) N 182 312 130 41.7 Supplemental Insurance Mean 2,863 2,904 (SD) (2,695) (2,796) (max = 58,608) Median 2,400 2,400 N 2017 3403 1386 40.7 HMO Premium Mean 1,647 1,650 (SD) (1,413) (1,391) (max = 11,592) Median 1,152 1,152 N 374 480 106 22.1 Long-term care insurance Mean 2,352 2,370 (SD) (1,964) (2,023) (max = 12,000) Median 1,984 1,997 N 391 493 102 20.7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 73 Table 5.6 Comparison of Out-of-Pocket Costs Before and After Imputation for All Users of Each Service in AHEAD Service Type Non-imputed variable Imputed Variable Number Imputed Percent Imputed Hospital/Nursing Home Mean 968 1,817 (SD) (6,111) (8,143) (max = 120,000) Median 0 0 N 1835 2184 349 16 Physician, Outpatient Mean 428 566 Surgery, ER, Dental (SD) (1,177) (1,552) Median 100 200 (max = 45,000) N 4720 5843 1113 19 Monthly Prescription Mean 81 83 Medication Costs (SD) (240) (239) Median 25 25 (max = 4,800) N 4222 4860 638 13.1 Two-Year Prescription Mean 1,938 1,995 Medication Costs (SD) (5,768) (5,745) Median 600 600 (max = 115,200) N 4225 4860 638 13.1 Home Health Care, Mean 432 647 Special Services (SD) (4,479) (5,322) Median 0 0 (max = 90,000) N 1038 1168 130 11.1 Supplemental Insurance Mean 2,261 2,507 (SD) (2,666) (2,784) (max = 58,608) Median 2,000 2,160 N 2578 3964 1386 35 HMO premium Mean 925 1,016 (SD) (1,317) (1,338) (max = 11,592) Median 360 720 N 699 805 106 13.2 Long-term care Mean 1,897 1,995 insurance (SD) (1,989) (2,044) Median 1,424 1,536 (max = 12,000) N 489 591 102 17.3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 74 Table 5.7 Comparison of Out-of-Pocket Costs Before and After Imputation for All Respondents in AHEAD Total Non-imputed variable Imputed Variable Number Imputed Percent Imputed Hospital/Nursing Home Mean 304 642 (SD) (3,459) (4,921) (max = 120,000) Median 0 0 N 5823 6172 349 5.7 Physician, Outpatient Mean 410 547 Surgery, ER, Dental (SD) (1,155) (1,529) Median 100 163 (max = 45,000) N 4899 6012 1113 18.5 Monthly Prescription Mean 62 66 Medication Costs (SD) (213) (215) Median 10 15 (max = 4,800) N 5513 6151 638 10.4 Two-Year Prescription Mean 1,485 1,578 Medication Costs (SD) (5,114) (5,171) Median 240 360 (max = 115,200) N 5513 6151 638 10.4 Home Care, Special Mean 73 120 Services (SD) (1,868) (2,322) Median 0 0 (max = 90,000) N 6072 6202 379 6.1 Supplemental Insurance Mean 1,289 1,682 (SD) (2,250) (2,523) (max = 58,608) Median 0 932 N 4790 6176 1386 22.4 HMO premium Mean 109 134 (SD) (544) (600) (max = 11,592) Median 0 0 N 5869 5975 106 1.8 Long-term care Mean 156 196 insurance (SD) (770) (870) Median 0 0 (max = 12,000) N 6007 6109 102 16.7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 75 Table 5.8 Comparison of Total Insurance Costs and Total Out-of-Pocket Costs as the Sum of Non-imputed and Imputed Cost Variables, for All Respondents in AHEAD Sum of Non- imputed Cost Variables Sum of Imputed Cost Variables Number imputed on one or more cost variables Percent imputed on one or more cost variables Total Insurance Costs (max = 58,608) Mean (SD) Median N 1,619 (2,569) 353 4491 2,000 (2,787) 1,296 6193 1702 27.5 Total costs excluding insurance (max = 183,600) Mean (SD) Median N 1,987 (5,546) 600 4229 2,764 (7,727) 880 5923 1694 28.6 Total costs including insurance (max = 183,600) Mean (SD) Median N total 3,586 (5,989) 2,043 3094 4,791 (8,262) 2,983 5904 2810 47.6 Comparison of Out-of-Pocket Costs in PSID versus AHEAD Out-of-pocket costs were higher in AHEAD than in PSID, as seen by comparing Tables 5.4 and 5.8. A main reason for this is that costs were reported for the previous two years in AHEAD, and for the previous year in PSID. Another main reason is that the time period for reported costs was six years later in AHEAD than in PSID, and reported costs are shown in the dollars for the year in which they were reported, unadjusted for inflation. Other factors that may explain the higher costs among AHEAD respondents are the older age range in this survey, and less precise measurement of out-of-pocket costs. In PSID, respondents were asked to provide detailed information on their out-of-pocket costs for each specific service type, as well as costs paid by Medicare, Medicaid, and private Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 76 insurance. In contrast, AHEAD respondents were asked to estimate their spending in four broad categories which were more loosely defined. The fact that respondents were asked to report combined expenditures on two or more services, and to estimate these expenditures over the previous two years, means that reported costs were likely more inaccurate in AHEAD. Table 5.9 compares mean out-of-pocket costs in PSID and AHEAD to other reports of older adults’ out-of-pocket spending. Reports are shown in chronological order, to facilitate comparison of different estimates over time. Each amount is shown first in the dollars of the year for which it was reported, and then (in parentheses), in 1995 dollars, adjusted for inflation.1 Because costs for some services likely increased less than average, and some increased more than average, estimates of costs adjusted for inflation for specific services may be inaccurate. The number of comparisons that can be made is limited, because out-of-pocket costs are given for different spending categories in each report. In order to make AHEAD results more comparable to other reports of one-year expenditures, two-year estimates out-of-pocket costs are divided by two. All estimates of 1 Amounts are given in dollars as of January 1995, because this represents approximately the middle of the period for which out-of-pocket costs were reported in AHEAD. Inflation in out-of- pocket payments was estimated by Braden et al. (1998, Figure 11, p.96) to be 4.4 percent in 1996, 1.6 percent in 1995, 0.9 percent in 1994, 3.0 percent in 1993, 5.4 percent in 1992, 5.8 percent in 1991, and 8.8 percent in 1990. These figures reflect average annual percent growth in direct spending by individuals of all ages on all health goods and services combined (excluding insurance premiums). For prescription medications, separate estimates of inflation from Braden et al. (1998) were used: 11.9 percent in 1991, 10.5 percent in 1992, 8.6 percent in 1993, and 9.0 percent in 1994 (Figure 16, p. 102). Estimates of out-of-pocket inflation were not available from 1987 to 1990, so the inflation adjustment for this period used a report of from Levit et al. (1997, Table 9, p. 187) of 11.3 percent pa- year for medical inflation (inflation in all private health expenditures, including payments by private insurance companies). For all inflation adjustments, the reference date used for each survey was the mid-point of the period for which expenses were reported. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 77 total out-of-pocket costs include cost-sharing under Medicare Part A and B and out-of- pocket costs for health services not covered by Medicare. When inflation-adjusted estimates from different surveys are compared, estimates from PSID and AHEAD are similar to one another, but often different than estimates from other surveys. Mean costs in PSID are particularly high compared to those in the Medicare Current Beneficiary Survey (MCBS) for hospital care, outpatient hospital, and physician/supplier services. However, median costs for these services in PSID were much lower, and, when adjusted for inflation, were more comparable to MCBS expenditures. For example, the median out-of- pocket cost for hospital care in PSID was $139 (as seen in Table 5.2 above), which equals $195 when adjusted for inflation. Thus, a main reason for the higher mean expenditures in PSID appears to be the stronger influence of high outliers in this sample than in the MCBS. Whereas the sample size in PSID was 1031, the MCBS included over 13,000 respondents, which provided a larger sub-sample of beneficiaries using each service, subduing the effect of outliers. The large mean expenditures for prescription medications in AHEAD may also be due to high outliers, since median prescription costs in AHEAD were $300 per user and $180 per capita. The lower expenditures in the MCBS may also be partially explained by the inclusion of disabled beneficiaries under age 65 in these out-of-pocket estimates. Mean out-of-pocket expenditures have been shown to be lower in the MCBS among disabled beneficiaries under 65 than among those 65 and older (due in part to greater coverage by Medicaid among the younger disabled) (Gibson & Brangan, 1998b). Thus, this difference in age range sampled may have resulted in an underestimation of out-of-pocket spending Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Table 5.9 Comparison of Means Spending in PSID and AHEAD to Other Reported Measures of Older Adults’ Out-of-Pocket Costs 1987 NMES1 1990 PSID 1992 MCBS2 1994 MCBS2 AHEAD 1994-953 1995 MCBS4 Hospital (per user) 539 (679) 227 (242) 331 (333) 1,817s Outpatient Hospital (per user) 338 (426) 88 (94) 95 (95) Physician/Supplier services (per user) 3716 (442) 303 (323) 354 (356) 5661 Dental (per user) 300 (438) 291 (363) 336 (351) Prescription Medications (per user) 313 (457) 315 (394) 326 (340) 998' Prescription Medications (per capita) 259 (326) 789* 304 (302) Total out-of-pocket costs - per capita 891 (1,720) 943 (1,188) 1,382 Total out-of-pocket costs including insurance premiums - per capita 1,487 (1,874) 2,396 3,150 (3,124) Total out-of-pocket costs including premiums for insurance and Medicare Part B - per capita 1,849 (2,330) 2.8504 (3,041) 2,866 'Expenditures by adults 65 and over. Total expenditures include home care but not nursing home costs, (Taylor & Banthin, 1994); T h e MCBS includes expenditures by all noninstitutionalized Medicare beneficiaries, including those who are under 65 and disabled. Numbers taken from tables of results prepared by Westat Inc. under contract to HCFA (Laschober & Olin, 1996; Olin & Liu, 1998); ’Two-year costs divided by 2; ^Medicare Payment Advisory Commission, 1999; ’Includes Nursing Home costs; 6 Expenditures for ambulatory visits and medical equipment are combined in PSID to compare with the MCBS spending category ‘Physician/supplier services’, which includes visits to health practitioners, diagnostic tests and services, medical and surgical services, medical equipment and supplies; ’Combined costs for outpatient surgery, physician visits, and dental care in AHEAD; ‘Monthly prescription amount multiplied by 24, -J 00 79 by older adults for these services. In contrast, inflation-adjusted estimates of total out-of-pocket expenditures including premiums for private insurance and Medicare Part B were higher in the MCBS than in PSID and AHEAD. One reason is that these estimates are from a different source than other reports of MCBS spending, and include institutionalized beneficiaries. The difference may also reflect the inclusion in these estimates of Medicare Part A premiums for the small number of beneficiaries who did not qualify for Medicare Part A but chose to purchase this coverage. Tables 5.10 to 5.14 present other descriptive analyses of out-of- pocket costs in PSID and AHEAD. Reported rates of use and uncovered costs among Table 5.10 Use and Uncovered Costs Among Users for Different Health Services in PSID and AHEAD PSID AHEAD Use Uncovered Costs Among Users Use Uncovered Costs Among Users* Any Services 97.8 94.1 98 86 Hospital Care 21 68.8 34.7 34 Outpatient Surgery 13 61.4 19.8 37.9 Physician/Therapist/ER 88.1 73.1 94.8 49.5 Prescription Medication 83.6 87.9 79.4 83.1 Dental Care 41.3 92 53.2 91.9 Equipment 34.8 89.6 - - Nursing Home Care 5.5 77.7 6.5 52 Home Care 5.2 15.3 14.8 17.9 Special Services - - 9.6 - * the small portion who did not say whether or not they had uncovered costs or who reported that their costs were not yet settled were counted among those reporting uncovered costs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 80 users are shown for each service type in Table 5.10. The percent reporting use of any service was very high in both PSID in AHEAD, which likely reflects the need for health services among the older population, as well as the role of the Medicare program in facilitating access to care. Rates of use were higher in AHEAD than in PSID for hospital care, outpatient surgery, physician visits, and home health care. Two likely explanations for this difference are the older age range in AHEAD, and the fact that use in AHEAD was reported for two years. The greater use of home health care in AHEAD may also reflect a trend toward increased use of this service during the 1990’ s (by those paying out- of-pocket because they do not meet Medicare eligibility criteria). The proportion of users with uncovered costs is surprisingly low in AHEAD compared to PSID for several services. This difference is difficult to explain, and may reflect incorrect responses by some users in AHEAD. As described in Chapter 3, the question regarding uncovered costs in AHEAD asked whether costs for each service were fully covered by Medicare, Medicaid, or other health insurance. Some beneficiaries may have responded that their costs were fully covered even if they were responsible for payments to satisfy their deductible. If this is the case, then out-of-pocket costs are likely underestimated in AHEAD, since those reporting that their costs were fully covered were assumed to have zero costs and were not asked to report an amount spent. Table 5.11 shows the burden of out-of-pocket health costs as a portion of household income in each survey sample. Mean estimates of burden were larger than median estimates, indicating that a portion of each sample had high levels of burden, which skewed the mean upward. Estimates of burden were larger in AHEAD than in PSID. This Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 8 1 Table 5.11 Proportion of Household Income Spent by Individuals on Total Out-of- Pocket Costs Excluding and Including Insurance Premiums in PSID and AHEAD Total Cost Variable* PSID AHEAD Total Costs excluding insurance premiums Mean 6.2 8.1 (SD) (13.5) (17.4) Median 1.8 2.1 Weighted N 894 5949 Total Costs including premiums for private Mean 9.5 13.9 insurance or a prepaid health plan (SD) 15.3 (20.1) Median 4.3 6.8 Weighted N 852 5929 Total Costs including premiums for private Mean 12.1 16.8 insurance or a pre-paid health plan, and (SD) (15.9) (20.7) Medicare Part B Median 6.7 9.5 Weighted N 852 5929 * includes only those nonmissing on all spending variables may partially reflect higher relative costs among AHEAD respondents, due to their older age range and greater use of several services. AHEAD respondents likely also had lower household income relative to those in PSID, due to their older age and their greater likelihood of living aicne. Also, in AHEAD, household income reported in 1994 (and multiplied by two) was used as a divisor for costs incurred in the following two years. This did not account for possible increases in income due to inflation over the time period, which may have offset inflation in medical prices. The difference in burden between the two survey samples was larger when the costs of insurance were considered, which may reflect high inflation in premiums for supplemental insurance in the 1990’ s. For example, Families USA (Dallek, 1996) reported double-digit inflation in the cost of supplemental policies between 1995 and 1996. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 82 Table 5.12 shows the percentage of respondents who spent over 10%, over 30% and over 50% of their household income on total health care costs. Results indicate that a small portion o f beneficiaries had very burdensome out-of-pocket costs that constituted over half of their household income. However, a sizeable portion of respondents spent over 10 percent of their household income on out-of-pocket costs; when expenditures for Medicare Part B are included as part of total costs, 36 percent of those in PSID and nearly 50 percent of those in AHEAD spent over 10 percent of income on health care. It is important to consider that these estimates of burden were calculated at the individual level, and that the actual burden of out-of-pocket spending for couples was greater, since both were drawing on the same income. However, these estimates do not consider household wealth, and thus may overestimate the burden of expenditures for those with high levels of savings or other assets that can be liquidated to pay for care. Table 5.12 Percentage Spending Over 10%, 30% and 50% of Household Income on Total Health Care Costs (Excluding and Including Insurance Premiums) in PSID and AHEAD % spending this % over 10% over 30% over 50% Total costs excluding premiums PSID 15.1 3.9 2.5 AHEAD 19.3 6.3 3.8 Total costs including private PSID 26.1 7.0 3.3 insurance premiums AHEAD 39.1 11.4 5.6 Total costs including private and PSID 35.8 9.4 3.6 Medicare Part B premiums AHEAD 48.6 14.6 6.7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 83 Tables 5.13 and 5.14 show the distribution of out-of-pocket spending on different services in PSID and AHEAD. In both survey samples, the largest portion of out-of- pocket spending was on insurance premiums. Insurance constituted a larger portion of Table 5.13 Distribution of Out-of-Pocket Spending on Different Services in PSID (n = 793) Percent of total out- of-pocket spending Hospital Stays 3.9 Outpatient Surgery 1.5 Physician/Therapist/ER 13.1 Dental Care 9.2 Prescription Medications 22.7 Equipment 8.0 Home Care 0.2 Nursing Home Care 2.4 Insurance 39.1 Total 100 Table 5.14 Distribution of Out-of-Pocket Spending on Different Services in AHEAD (N = 5464) Percent of total out-of- pocket spending Hospital/Nursing Home Stays 4.3 Physician/Outpatient Surgery/ Dental/ER 19.3 Prescription Medications 28.9 Home Health Care / Other Special Services 1.2 Insurance 46.4 Total 100 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. spending in AHEAD than in PSID. This may reflect greater spending on insurance in the mid-1990’ s compared to 1990, and may also reflect an overestimation of insurance expenditures in AHEAD. As discussed in Chapter 3, respondents were asked about the costs of current insurance coverage in 1996, and were assumed to have spent the same amount on insurance during the previous two years. The second largest portion of spending was on prescription medications in both survey samples. This reflects the widespread use and high costs of prescription drugs, and fact that they are not covered under FFS Medicare. Despite the fact that 65 percent of beneficiaries reported some form of prescription coverage in addition to Medicare in the 1995 MCBS (Davis et al., 1999), these results indicate that prescription drug expenditures constitute a large portion of the out-of-pocket health care burden of most older adults. The remaining Chapters examine differences in these descriptive results by age, race, and insurance coverage, to get a better picture of how out-of-pocket costs and out- of-pocket burden differ by these characteristics among older adults. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 85 Chapter 6 Age and Out-of-Pocket Health Costs It is well known that due to the higher prevalence of health conditions, illnesses, and disability among the elderly, the total costs of health care are higher for those over age 65 than for younger adults (for example see Waldo et aL, 1989). The higher health care costs among the older population are of particular policy concern because the U.S. population is aging. Schenieder & Guralnik (1990) warn that health costs will skyrocket as the oldest-old population continues to grow. Concern about the high public costs of caring for this group have led to suggestions as drastic as rationing health care by age (Callahan, 1987). However, there is mixed evidence as to whether health care costs rise with age among the older population. Those 65 and over are often considered as a single group (Torrey, 1989), or are stratified into groups that do not adequately distinguish the younger-old from the oldest-old (Peris, 1996). The few studies of out-of-pocket health spending among the elderly have found increased spending with age among the older population (AARP Public Policy Institute & The Lewin Group, 1997; AARP Public Policy Insititute & The Urban Institute, 1994; Moon, 1992). At each income level, those over age 75 report higher out-of-pocket expenditures than those age 65-74 (Moon, 1992). However, the relationship between age and expenditures may not be linear among the elderly: there is evidence that health expenditures plateau or begin to decline in the oldest age groups (Peris & Wood, 1996). Projections of spending by age in 1997 using the 1993 MCBS indicated that those 75-79 had the highest expenses ($2,360/yr), while those 85 and over spent almost the same amount as those 70-74 ($2,169 vs. $2,152) (AARP Public Policy Institute & The Lewin Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 86 Group, 1997). In regression analyses using dummy variables to represent successive age groups, Moon (1992) found that in 1980, those age 70-74 and age 75-79 each had significantly more physician visits than those 65-69, whereas those 80 and over did not, controlling for factors such as gender, race, income, insurance and bed disability days. Age has also been significant as a non-linear predictor of health care use and expenditures in studies using multivariate analyses. Freiman (1998) found that a squared age term was a significant positive predictor of the probability of hospitalization, and a significant negative predictor of out-of-pocket health expenditures among users of non hospital care. Logged age (of the household reference person) was a positive predictor of household medical expenditures in the 1980-81 and the 1989-90 CEX (Rubin & Koelln, 1993b). In a cohort analysis of health care utilization, Wolinsky, Mosely, & Coe (1986) found a J-curve relationship between age and the volume of physician utilization, with volume of use increasing until about age 80 and then beginning to decline. They concluded that the three most plausible explanations for this phenomenon were that the oldest old were substituting hospital for ambulatory services, substituting informal caregiver supports for physician services, or losing contact with physicians who have predeceased them. Another explanation for such non-linear effects of age on costs is selective survival of those who are healthier. Those who survive to very old age are a highly select group, representing the hardiest members of their birth cohort (Johansson, Zarit, & Berg, 1992; Peris, 1997; Peris & Wood, 1996). Manton, Woodbury, and Stallard (1995) found a plateau in functional scores at about age 95, and attributed this to the mortality at earlier ages of those who were more frail. Other explanations for a plateau or decrease in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 87 expenditures among the oldest old are that they are given less aggressive treatment, are too frail to undergo further treatment, or are given less expensive treatment in the last year of life (Meerding, Polder, Bonneux, Koopmanschap, & van der Maas, 1998; Peris, 1997; Peris & Wood, 1996). The oldest-old are also more likely to be treated in less expensive, nonteaching hospitals, which may lead to lower out-of-pocket hospital costs (Peris, 1997). Meerding et al. (1998) point out that the oldest old are more likely to be cared for in institutions, which reduces their use of outpatient care. When discussing the pattern of health care costs with age, it is also important to distinguish between expenditures for different types of health goods and services. While long-term care costs are consistendy found to increase with age, findings for acute care costs are mixed (Meerding et aL, 1998; Peris, 1996; Roos, Shapiro, & Tate, 1989). Schneider & Guralnik (1990) point out that those who survive to the oldest ages are more likely to suffer from diseases for which the prevalence increases exponentially with age, such as Alzheimer’s disease and other dementias, and are more likely to experience hip fracture associated with osteoporosis. These conditions are associated with higher costs for long-term care services in the oldest old. Similarly, the Government Accounting Office (1997) reports that chronic conditions may begin in middle age, but often progress in severity over time and require more costly and intensive treatment. However, the costs of treatment for many chronic diseases may not rise as sharply with age, and may in fact decline. Peris (1996) argues that it costs more to treat a disease the first time that an older adult is hospitalized for it; later hospitalizations due to a worsening of the condition do not require as many tests and procedures, as the patient is likely to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 88 have already undergone these. In a study of hospital costs by age, Peris (1996) found that the oldest-old were admitted to the hospital with less use-intensive diagnoses (DRG’s), and that DRG weights declined with age among the older population. Total hospitalization costs peaked at age 70 to 79, and then declined with age (Peris, 1996). This chapter helps to clarify the relationship between age and of out-of-pocket costs for different services by examining patterns of use and of spending by age across service types. Age is tested as both a linear and a non-linear predictor of out-of-pocket costs for different health goods and services, controlling for potential confounding factors. Age or Health? In the Andersen model of health service utilization, age is considered a “predisposing” variable, representing a “biological imperative” suggesting the likelihood of needing services (Andersen 1995, p. 2). Andersen (1968) postulates that health service use varies with age because older adults have “different types and amounts of illness” that require different patterns of medical care (p. 15). However, the fact that age was included in the model in addition to need variables suggests that it was meant to represent aspects of health not adequately captured by measures of perceived and evaluated health. Findings for the age variable in studies using the Andersen model to predict health care use have been mixed, with age often losing significance as an independent predictor of use when need variables are controlled. Wolinsky et aL (1983) found that among a sample of noninstitutionalized older adults, when controlling for other variables in the Andersen model, age was not a significant predictor of preventive dental contact, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 89 preventive medical contact, total doctor visits, emergency room visits, or hospital episodes. Linden et al (1997) found that age was not a significant predictor of physician use but was a significant predictor of prescription use when controlling for other variables in the Andersen modeL Evashwick, Rowe, Diehr & Branch (1984) found that age was a significant predictor of nursing home and home care use, but that age only remained significant for home care when enabling and need variables were controlled. An important focus of this chapter is to examine changes in the effects of age on out-of-pocket costs for different services when health is controlled. If age is a significant predictor of higher costs only until health is controlled, this will indicate that poorer health rather than older age is an explanation for higher costs. If age remains significant when health is controlled, this will indicate that other factors related to age but not controlled in analyses are affecting costs. There may also be significant non-linear changes in the effects of age when health is controlled, indicating different effects of health on costs among different age groups. Age and Socioeconomic Status An in-depth analysis of the economic resources of the oldest old performed by Atkins (1989) revealed that the oldest-old as a group have substantially lower economic status than the young-old. Lower income and wealth among many of the oldest-old could lead to lower out-of-pocket expenditures for discretionary services if they have insufficient resources to pay for care. It is also possible that the lower economic status of the oldest old is partially a result of having exhausted their resources on out-of-pocket payments for Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 90 health care. If this is so, then some older adults with low incomes may actually report higher out-of-pocket costs. Given the decrease in mean economic status with age among the elderly, the burden of out-of-pocket health spending as a portion of income would be expected to be greater for the oldest-old even if out-of-pocket costs do not increase with age. Lower levels of education among the oldest old will also likely affect their health service use and resultant out-of-pocket costs. The ability to read and understand medical information and instructions is important to obtain appropriate care and navigate the health care system. Gazmararian et aL (1999) found that among Medicare enrollees in a managed care organization, the likelihood of having inadequate or marginal levels of functional health literacy increased with age, and was markedly high in the 85+ age group. Enrollees age 85 and over had 8.62 times greater odds than those 65-69 of having inadequate or marginal functional health literacy versus adequate literacy. Although functional health literacy is related to education and cognitive status, these differences were significant even when these variables were controlled. The National Academy on an Aging Society (1998) reports that low health literacy is related to greater health care use and higher health care costs. When health is controlled, those with low literacy use fewer discretionary services, such as physician visits, and more hospital care. Thus, the lower health literacy among the oldest old may contribute to higher costs for less discretionary services, and lower costs for other service types. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 91 Age and Insurance In examining age as a predictor of out-of-pocket health costs, it is important to consider age differences in insurance coverage supplemental to Medicare. In their detailed study of supplemental insurance coverage, Chulis, Eppig, Poisal (1995) found that the share of beneficiaries with Medigap increased steadily with age, whereas the proportion with employer-sponsored insurance decreased with age. This is partially a cohort effect, since retiree health benefits were not common in the 1960’ s, when the oldest-old cohorts retired. It is also likely that beneficiaries purchase medigap insurance when they lose employer-coverage due to business closings or the death of a spouse. Because employer- coverage has historically been more generous than privately purchased coverage, the oldest-old might be liable for a greater portion of their health costs. Waldo & Lazenby (1984) report that the oldest-old pay a higher proportion of their total health costs out-of- pocket, and are more likely to be covered by Medicaid than the younger-old. Both of these effects reflect the greater use of nursing home care by the oldest-old; since Medicare and most supplemental insurance plans do not cover the costs of long-term nursing home stays, those living in a nursing home must pay for this care out-of-pocket, and then turn to Medicaid when their resources are exhausted. This study will examine whether insurance is an important determinant of age differences in out-of-pocket costs, by testing age as a predictor of costs both before and after controlling for different types of insurance coverage. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 92 Hypotheses In descriptive analyses, costs are expected to increase with age for most services, particularly nursing home and home care. However, there may be a plateau in costs among the oldest old, for reasons discussed above. Also, among users of each service, costs may be lower among the oldest-old if they receive a lower intensity of services. However, this may be offset by less generous insurance coverage among the oldest-old. The burden of total out-of-pocket costs is expected to increase with age, consistent with the age-related decline in income and wealth among the elderly. Consistent with known patterns of use, the oldest-old are expected to spend a larger portion of their total health care dollars on long-term care services than younger age groups. This is expected both because there is greater use of long-term care services among this group, and because the cost of these services (particularly nursing home care) is often much higher than the cost of other services. The oldest-old may also spend less on discretionary services, particularly if they face difficulties in access to care. In multivariate analyses, age is expected to be a predictor of higher nursing home and home care costs, but may not be a significant predictor of other costs, especially when other variables in the model are controlled. In particular, in cases where age is a significant positive predictor of costs, age is expected to become a non-significant or even a negative predictor of costs when health is controlled. Age is also expected to be a non-linear predictor of costs for some services. While health care use and costs are likely to be higher for the middle-old than the young-old, they may decline among the oldest-old, for reasons discussed above. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 93 Measurement of Age Age was measured continuously in years, and was self-reported in PSID1 and calculated from birth date in AHEAD2 . For some descriptive analyses (described below), three age categories were created in each survey sample. In PSID, those 66-74 were classified as the ‘young-old’, those 75-84 as ‘old-old’, and those 85 years and over as the ‘oldest-old’. In AHEAD, the age groups were created differently, to take advantage of the large number of very old respondents: those 72-79 were classified as the old-old, those 80-89 as the oldest-old, and those 90 and over as the very old. These categories were used in AHEAD because the youngest participants were age 72, and the sample was large enough to warrant examining those 90 and over as a separate group. For the testing of non-linear age effects in multivariate analyses, dummy variables were created using 5-year intervals up to age 90 in PSID and up to age 95 in AHEAD. Those older than this were included in the last dummy variable, representing the oldest old. Analyses Descriptive methods were chosen to facilitate detailed examination of use and spending across the age span sampled in each survey. Rather than reporting spending in a lPSID included measures of self-reported age as well as age calculated from reported birth date, and there were several cases where these two ages did not match. Reported age was chosen for these analyses since it is a reliable variable that was checked for consistency between waves of PSID (personal communication, PSID staff). 2 In AHEAD wave 2, two years were added to the age calculated in wave 1 (date of interview minus date of birth). Because the wave 1 and wave 2 interviews were not exactly two years apart for each respondent, this may have led to some small inaccuracies in age. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 94 small number of age groups, charts are used to depict descriptive results for respondents at each year of age. In order to get a smoother picture of results, a “moving average” is used in which the mean for each year of age actually represents the mean for all respondents that age as well as those one year younger and one year older. In addition to subduing the effect of outliers, this technique provides a larger sample size in situations where there are only a few respondents of a particular age. In order to show results for a sufficiendy large sample at the oldest plotted age, the oldest respondents were grouped together. The last point on moving average graphs is the mean of those 93 and over in PSID, and those age 98 and over in AHEAD. Charts are used to depict age patterns in 1) mean spending overall and for each service type, 2) the mean proportion reporting use of each service type, and 3) the mean burden of total out-of-pocket expenditures as a portion of household income. Tables are used to report the distribution of total out-of-pocket costs by service type for each of three age groups in both PSID and AHEAD. In hierarchical regression analyses, age and other demographic variables were entered first, followed by health variables in model 2, insurance in model 3, and SES variables in model 4. For all spending categories, age was tested first as a linear predictor of costs, using the variable measuring age continuously in years. Then, to test for non linear age affects, parallel models were run in which dummy variables representing five of the six 5-year age groups were entered instead of the continuous age variable. (The dummy variable for the youngest age group was left out as the referent age category.) In order to determine whether the age dummy variables improved the explanatory power of the model, the R2 for the model using dummy variables was compared to the R2 for the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 95 model using the continuous age variable. If using dummy variables increased the R2 , the significance of the increase was tested using the following F test (McClendon, 1994, p238): F = (R2-r*)/(g-2) , (1 - R2 ) / (n - g) where R2 is the variance explained by the equation using dummy variables, r2 is the variance explained by the equation with the continuous age variable, g is the number of dummy variables, and n is the sample size. Thus, the numerator is the increase in explained variance using the dummy variable equation, divided by the number of new variables when the dummy-variables are used. The denominator is the proportion of variance not explained by the dummy-variable equation, divided by its degrees of freedom. Non-linear age effects are shown and discussed in cases where the improvement in explanatory power using the dummy variables was significant, and one or more of the age dummy variables was significant in any of the hierarchical models. In these cases, the regression coefficients for each of the age dummy variables are shown to allow examination of the non-linear pattern. Results Table 6.1 show the distribution of both survey samples by age. In both PSID and AHEAD, there is a sufficient number of respondents of each age to provide a representative estimate of spending for those of each age. This is the case into the 90’ s in PSID, and into the late 90’ s in AHEAD. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 96 Patterns of use and out-of-pocket costs for each health service by age are shown using a moving average in Figures 6.1 to 6.8. Costs are shown across all respondents. Age patterns of spending among only those using each service are not shown, but were very similar to the patterns among all respondents for most service types. In each Figure, the first chart shows the percentage of those reporting service use in both PSID and AHEAD. Rates of service use vary more by age in PSID than in AHEAD, particularly among the oldest respondents. This likely reflects the greater effect of outliers on mean levels in PSID due to the smaller sample size. The remaining charts in each Figure show the pattern of mean out-of-pocket spending by age. Figure 6.1 shows mean use and expenditures for hospital care. As seen in chart a), hospital use did not rise dramatically with age in either AHEAD or PSID. There was a sharp rise among the oldest-old in PSID, which may be exaggerated due to the small sample size in this age group. In both samples, there was a small drop in expenditures among the oldest-old. As seen in chart b), out-of-pocket hospital costs in PSID did not show a linear increase with age, and in fact dropped among those in their early 80’ s. The peak in costs in the early 90’ s reflects high costs among a small number of respondents in this age group. Out-of-pocket costs are shown by age for AHEAD in chart c). Costs rise sharply in very old age, which reflects the fact that nursing home and hospital costs are combined in this spending category in AHEAD. As seen in Figure 6.2, nursing home use and costs rose dramatically among the oldest old in PSID. There was a similar increase in use with age AHEAD, although it was not as great for the oldest age group, and there was a dip around age 97. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 97 Table 6.1 Distribution of Respondents by Age in PSID and AHEAD (Weighted) Age Number this age in PSID Number this age in AHEAD 66 82 _ 67 70 - 68 76 - 69 75 - 70 50 - 71 44 15 72 55 443 73 65 554 74 46 552 75 39 531 76 49 434 77 54 456 78 31 382 79 40 346 80 35 345 81 21 284 82 29 296 83 33 274 84 25 232 85 22 192 86 13 182 87 12 150 88 17 133 89 9 109 90 9 81 91 4 59 92 3 59 93 1 36 94 4 29 95 3 19 96 1 17 97 3 11 98 2 12 99 - 7 100 - 6 101 - 3 102 - 2 103 - - 104 105 - 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 98 Figure 6.3 shows age patterns in use and costs for home care. Home care use rose with age in both PSID and AHEAD, with the greatest use among the oldest respondents. The distribution of out-of-pocket home care costs by age was bi-modal in PSID, with peaks in the early and late 80’ s. In AHEAD, costs for home care/other special services rose dramatically among those in their mid-90’ s, and then dropped off among the oldest old. Given the low overall use of home care in these samples, these age patterns were likely disproportionately affected by outliers. Figure 6.4a shows age patterns in physician use in both PSID and AHEAD. Rates of physician use were high among all age groups in AHEAD, whereas physician use declined among those in their 90’ s in PSID. Chart b) shows a gradual drop in out-of- pocket physician costs by age in PSID. As seen in chart c), results were similar for ambulatory service expenditures in AHEAD, with the exception of a sharp rise in the late 90’ s. This reflects one or more high outliers among this relatively small group. Figure 6.5a shows relatively low use of outpatient surgery in both survey samples, with small variations over most of the age range, and a dip followed by a rise in use at the oldest ages. Chart b) shows an uneven pattern of surgery costs by age, with the highest costs among those in their 80’ s. As seen in Figure 6.6a, a high percentage of respondents reported using prescription medication at all age groups in both PSID and AHEAD. The sharp drop in use at around age 96 in AHEAD likely reflects low or no use among the relatively few respondents in this group. As seen in charts b) and c), medication expenditures were fairly steady with age, and dropped among the oldest age groups in both survey samples. Figure Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 99 6.7 shows age patterns in equipment use and costs in PSID. While the use of equipment increased with age, equipment costs were more variable, and did not show a clear pattern. Figure 6.8 shows dental care use by age in both surveys, and dental costs for those in PSID. As seen in chart a), dental use declined with age in both surveys, although it increased among those in their 90’ s in PSID. Chart b) shows a decline in out-of-pocket dental costs with age, with a peak among those in their late 70’ s and early 80’ s. The pattern of total out-of-pocket costs by age is shown for PSID and AHEAD in Figure 6.9. (All three charts show total out-of-pocket costs excluding insurance premiums; very similar age patterns were found when premiums were included.) Chart a) reveals that total costs varied but did not generally increase with age when the costs of nursing home stays were excluded. In contrast, chart b) shows a sharp increase in total costs among those in their mid 90’ s when nursing home expenditures are included. A similar pattern is seen for total out-of-pocket costs including nursing home care in AHEAD, in chart c). (Total costs cannot be shown excluding nursing home costs in AHEAD, since hospital and nursing home expenses were reported together.) Age patterns in the level of out-of-pocket burden as a portion of income are shown in Figure 6.10. The proportion of household income spent on total out-of-pocket costs (excluding insurance) rose gradually with age until the early 90’ s in both survey samples. Mean burden rose more sharply among those in their 90’ s. This effect was more pronounced in PSID, which is may be due to the stronger effect of outliers. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 6.1: Hospital Use and Out-of-Pocket Costs by Age in PSID and AHEAD: Moving Average a) Percentage Reporting Hospital Use in PSID and AHEAD 100 AHEAD 67 70 73 76 79 82 85 88 91 94 97 Age b) Out-of-Pocket Hospital Costs in PSID o > a C D C L C O o o D- 700 600 500 400 300 200 V 100 O 0 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 Age c) Combined Out-of-Pocket Costs for Hospital and Nursing Home Care in AHEAD 10000 % 8000 c < D % 6000 o o o C L 4000 0 2 0 0 0 1 ■ S o 0 67 70 73 76 79 82 85 88 91 94 97 Age Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 6.2: Nursing Home Use by Age in PSID and AHEAD, and and Out-of-Pocket Nursing Home Costs by Age in PSED: Moving Average a) Percentage Reporting Nursing Home Use in PSID and AHEAD 100 < D CO 3 o > c 80 t: 60 0 Q _ £ 40 1 20 © a. 67 70 73 76 79 82 85 88 91 94 97 -PSID -AHEAD Age b) Out-of-Pocket Nursing Home Costs in PSID 5000 cn = 5 4000 c W 3000 o 20 0 0 o Q_ B 1000 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 102 Figure 6.3: Home Care Use and Out-of-Pocket Costs by Age in PSID and AHEAD: Moving Average a) Percentage Reporting Home Care Use in PSID and AHEAD 100 3 80 o> 70 PSID o O U (T 40 •£ 30 8 20 AHEAD Age b) Out-of-Pocket Home Care Costs in PSID 50 O ) =5 40 i n 30 o 20 o CL "5 1 0 Age c) Combined Out-of-Pocket Costs for Home Health Care and Other Special Services in AHEAD 2000 cn c c 1500 ( D Q . © 1000 500 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 6.4: Ambulatory Service Use and Out-of-Pocket Costs by Age in PSID and AHEAD: Moving Average a) Percentage Reporting Visits to Physician, Therapist, Emergency Room in PSID and AHEAD 100 a > CO Z ) O ) e 80 ^ 60 o a . £ 40 I 20 5 a. 67 70 73 76 79 82 85 88 91 94 97 -PSID -AHEAD Age b) Out-of-Pocket Costs for Physician, Therapist, and ER Visits in PSID 500 “ o 400 c o 200 o l a 's 100 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 Age c) Combined Out-of-Pocket Costs for Physician and ER Visits, Outpatient Surgery, and Dental Care in AHEAD © Q . C O 0 ) o o Q _ "o ■ 5 O 1250 1000 750 500 250 0 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 Age Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 104 Figure 6.5: Outpatient Surgery: Use by Age in PSID and AHEAD, and Out-of-Pocket Costs by Age in PSID: Moving Average a) Percentage Reporting Use of Outpatient Surgery in PSID and AHEAD 100 <D w 80 o > :§ 60 o Q. cc 40 | 20 © Q _ 0 67 70 73 76 79 82 85 88 91 94 97 Age b) Out-of-Pocket Outpatient Surgery Costs in PSID 180 j f 160 140 S. 120 “ 100 ■ § 80 £ 60 ■ 5 40 3 20 O 0 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 A ge / I 4 = / ■j— t 1 t T ' 7 — 1 PSID AHEAD Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 6.6: Use and Out-of-Pocket Costs for Prescription Medication by Age in PSID and AHEAD: Moving Average a) Percentage Reporting Use of Prescription Medication in PSID and AHEAD < D < 0 3 c n ■ E O a. a > a: C D o <D 0- 100 80 60 20 0 67 70 73 76 79 82 85 88 91 94 97 -PSID -AHEAD Age b) Out-of-Pocket Prescription Medication Costs in PSID (1 Year) o > C T 400 350 *o c 300 0 > Q . C O 250 a > 200 o o 150 Q _ o 100 ” 5 50 O 0 • V *1 — I 1 ----------1 ----------1 ---------- 1 ---------- 1 --------- 1 ---------- 1 ----------1 ---------- 1 ---------- 1 ----------1 1 1 1 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 Age c) Out-of-Pocket Prescription Medication Costs in AHEAD (Average per Month) 100 - o 80 tz C D W 60 <D 1 40 CL. 20 3 o 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 6.7: Use and Out-of-Pocket Costs for Equipment by Age in PSID: Moving Average a) Percentage Reporting Use of Equipment in PSID 100 a > c o 3 80 cn '-n 60 o a . ® 40 a a > o a > a. 20 67 70 73 76 79 82 85 88 91 94 97 Age b) Out-of-Pocket Equipment Costs in PSID 200 cn c £ 150 Q . CD 100 o o 5: so o 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 Age Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 107 Figure 6.8: Use and Out-of-Pocket Costs for Dental Care by Age in PSID: Moving Average a) Percentage Reporting Dental Care Use in PSID and AHEAD C D C O 0 3 C X I o Q _ CD G C CD O w ff l CL 100 90 80 70 60 50 40 30 20 10 0 67 70 73 76 79 82 85 88 91 94 97 -PSID •AHEAD Age b) Out-of-Pocket Dental Costs in PSID 500 0 3 =§ 400 c CD < » 300 g 200 CL 100 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 6.9: Total out-of-pocket costs by Age in PSID and AHEAD Moving Average a) Total Out-of-Pocket Costs in PSID excluding Nursing Home Costs 1600 e 1400 * o c <D 1200 C L C O 1000 a> 800 o o n 600 i o 400 3 200 o 0 € Age b) Total Out-of-Pocket Costs in PSID including Nursing Home Costs 16000 -|--------------------------------------------------------------------------------------- i* 14000 ---------------------------------------------------------------------------------------- E 12000 10000 jd 8000 o 6000 a. •4- 4000 o B 2000 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 Age c) Total Out-of-Pocket Costs in AHEAD including Nursing Home Costs „ 16000 -------------------------------------------------------------------------------------- .£ 1 4 0 0 0 --------------------------------------------------------------------------------------------------- § 12000 < £ - 10000 ffl 8000 ■g 6000 3 : 4000 ? 2000 3 Q 0 t 1 1 --------i--------1 i 1 1 i i 1 1 — i 1 1 r— 67 69 71 73 75 77 79 81 83 85 87 80 91 93 95 97 Age Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 109 Figure 6.10: Burden o f Total Out-of-Pocket Costs (excluding insurance) as a portion of Household Income by Age in PSID and AHEAD: Moving Average 100 90 ~ 80 r n " 70 60 PSID AHEAD 50 40 30 20 10 Age Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 110 Tables 6.2 and 6.3 show the distribution of total out-of-pocket spending by service type among the youngest, middle, and oldest age groups in PSID and AHEAD. In both survey samples, insurance premiums constituted the largest proportion of total out-of-pocket spending in all age groups. However, as expected, the mean portion of out-of-pocket costs spent on nursing home care in PSID and on home care/special services in AHEAD increased with age and was highest among the oldest-old. The proportion of total costs spent on hospital care in PSID was highest among the middle-old. Table 6.2 Distribution of Out-of-Pocket Spending on Different Services by Age Group in PSID Service Age 66-74 (N= 569) Age 75-84 (N = 359) Age 85+ (N= 103) Hospital Stays 3.4 5.1 2.7 Outpatient Surgery 1 1.5 4.2 Physician/Therapist/ER 13.9 12.1 11.6 Dental Care 10.6 8.4 3.4 Prescription Medications 21.5 25.5 21.4 Equipment 7.4 9.3 6.3 Nursing Home Care 0.2 2.9 15.3 Home Care 0.1 0.3 0 Insurance 41.9 34.8 35 Total 100 100 100 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I l l Table 6.3 Distribution of Out-of-Pocket Spending on Different Services by Age Group in AHEAD Service Age 72-79 Age 80-89 Age 90+ (N= 3713) (N = 2198) (N= 341) Hospital/Nursing Home Stays 2.9 6 9.2 Physician/Outpatient Surgery/ Dental/ER 19.9 19.1 13.2 Prescription Medications 27.4 31.3 31.6 Home Health Care / Other Special Services 0.6 1.8 3.7 Insurance 49.3 41.8 42.3 Total 100 100 100 Multivariate Results Hierarchical regression results for both PSID and AHEAD are shown in Tables 6.4 to 6.7. Table 6.4 shows results for spending on services where OLS regression was used on all respondents. Results for services for which a two-part model was used are shown in Tables 6.5 to 6.7. Non-linear effects were significant for dental care, home care, and nursing home costs in PSID. In AHEAD, non-linear effects were significant for hospital/nursing home costs, and for home care/special services. Thus, for these services, the regression coefficients for each dummy variable are shown. As seen in Table 6.4, age was not a significant predictor of physician costs in PSID. In AHEAD, age was a significant negative predictor of costs for ambulatory services in all models. Older age predicted greater spending on prescriptions in PSID until health was controlled, whereas age was not a significant predictor of prescription costs in AHEAD. For total out-of-pocket costs excluding insurance premiums, age was a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 112 significant positive predictor of spending until health was controlled, but did not significantly predict expenditures in AHEAD. For total out-of-pocket costs including insurance premiums, age was not significant in PSID, whereas age was a negative predictor of costs in AHEAD until socioeconomic variables were controlled. Table 6.5 shows results of the 2-part hierarchical models for outpatient surgery, dental care, and equipment in PSID. Age was not a significant predictor of use or costs for surgery and equipment. For dental care, there was a significant non-linear effect of age on use. The regression coefficients for the dummy variables indicate that successively older age groups were increasingly less likely to use dental care, with a plateau in the oldest age group. This age effect was attenuated when health was controlled, and only group aged 84-89 remained significantly less likely to use dental care after insurance was controlled. Hospital and nursing home results are seen in Table 6.6. For hospital care in PSID, age was a positive linear predictor of use until health was controlled in model 2. However, for hospital/nursing home care in AHEAD and for nursing home care in PSID, the effect of age use and out-of-pocket expenditures was nonlinear. The pattern of the age dummy variables shows an increasing likelihood of use in each successive age group. This pattern was most pronounced for nursing home use in PSID, as seen in Figure 6.2 above. For hospital/nursing home care in AHEAD, this non-linear effect on use became largely non significant when health was controlled in model 2. In contrast, the effect of age on nursing home use was attenuated when health was controlled, but remained significant in all models. A significant non-linear effect of age on nursing home expenditures in PSID emerged when insurance was controlled in model 3. As revealed by the pattern of the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 113 coefficients for the dummy variables, this effect on spending was in the opposite direction from the effect on use. Although all age groups spent significantly more than those 66 to 69, the magnitude of the age effect on expenditures declined in successive age groups. For users of hospital/nursing home care in AHEAD, costs were significantly higher among those in the oldest age group in all models. Table 6.7 shows similar results for home care in both survey samples. In PSID, there was a significant non-linear effect of age on use, with the greatest likelihood of use in the oldest age group. This effect was reduced when health was controlled in model 2, and became non-significant when insurance was controlled in model 3. In AHEAD, age was a significant non-linear predictor of use and of costs among users for home care and other special services. Each age group had greater likelihood of use than the youngest group, and the magnitude of the effect increased with age (with a dip for the 90-94 year- old group). When health and insurance were controlled in models 2 and 3, this effect remained significant for all except the oldest age group. Among those in the 85-89 and 90- 94 age groups who used home care, there were inconsistent effects of age group on out- of-pocket costs across models. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 114 Table 6.4 Hierarchical OLS Regression Results for Out-of-Pocket Spending by Age in PSID and AHEAD: Ambulatory Care, Prescription Medications, and Total Out-of-Pocket Costs (in Logged Dollars) Service Type Data Set Agef Model 1 Model 2 Model 3 Model 4 Physician, Therapist, PSID Age -0.02 -0.02 -0.02 -0.01 ER visits (N = 970) R2 0.026 0.046 0.111 0.139 Physician, Outpatient AHEAD Age -0.06*** -0.05*** -0.05*** -0.03*** Surgery, Dental Care, R2 0.040 0.056 0.108 0.147 ER visits (N = 6044) Prescription PSID Age 0.03* 0.01 0.00 0.01 Medications (N = 977) R2 0.036 0.153 0.216 0.223 AHEAD Age 0.01 0.00 0.00 0.01 (N = 6158) R2 0.012 0.136 0.214 0.223 Total Costs excluding PSID Age 0.03** 0.02 0.02 0.03* insurance premiums (N = 893) R2 0.084 0.12 0.203 0.239 AHEAD Age 0.00 -0.01 -0.01 0.01 (N = 5949) R2 0.031 0.079 0.167 0.198 Total Cost including PSID Age 0.01 0.01 0.01 0.02 insurance premiums (N = 852) R2 0.128 0.141 0.36 0.377 AHEAD Age -0.02*** -0.02*** -0.01** -0.00 (N = 5929) R2 0.082 0.095 0.346 0.365 Continuous age variable. * 2 < -05 ** £ < .01 *** 2 < -001 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Table 6.5 Hierarchical Two-Stage Regression Results for Outpatient Surgery, Dental Care, and Equipment Use and Out-of - Pocket Expenditures (in Logged Dollars) by Age in PSID Model 1 Model 2 Model 3 Model 4 Service type Data Set Age Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Surgery PSID Age (continuous) 0 .0 0 0.04 0 .0 0 0 .0 2 0 .0 0 0 .0 2 0 .0 0 0 .0 2 LLf / R2 -381 0.019 -378 0.074 -373 0 .1 0 0 -372 0.117 (N = 1025, 129 users) Dental PSID Age 70-74 -0.09 -0 .1 2 -0 .1 0 -0.13 -0.09 -0 .1 1 -0.04 -0.07 Age 74-79 -0.27* 0.32 -0 .2 2 0.31 -0.18 0.31 -0.05 0.36 (N = 1024, Age 80-84 -0.33* -0.14 -0.27* -0.14 -0.21 -0.09 -0.06 -0.03 420 users) Age 84-89 -0.75*** -0.60 -0 .6 6 *** -0.62 -0.60** -0.44 -0.47* -0.38 Age 90+ -0.70** -0.79 -0.50 -0.82 -0.44 -0.71 -0.16 -0.81 LL/R 2 -649 0.031 -634 0.032 -624 0.116 -680 0.134 Equipment PSID Age (continuous) 0 .01 -0 .0 1 0.01 -0 .0 2 0.01 -0 .0 1 0 .01 -0 .0 2 LL/R 2 -648 0.043 -638 0.109 -637 0.162 -635 0.190 (N = 1019, 348 users) + In cases where age dummy variables are shown, the youngest age group is the referent group (age 66-69 in PSID, age 72-74 in AHEAD), *LL = Log Liklihood * e < .05 ** p < .01 *** j> < ,001 Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Table 6.6 Hierarchical Two-Stage Regression Results for Hospital and Nursing Home Use and Out-of -Pocket Expenditures (in Logged Dollars) by Age in PSID and AHEAD Model 1 Model 2 Model 3 Model 4 Service type Data Set Age Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Hospital PSID Age (continuous) 0.03*** -0.05 0.01 -0.05 0 .01 -0.04 0.01 -0.03 Log Likelihood / R2 -489 0.060 -427 0.073 -425 0.098 -424 0.103 (N = 1011, 2 0 0 users) Nursing PSID Age 70-74 0.38 1.36 0.53 5.27 0.62 13.18** 0.65 16.6** Home Age 74-79 0.77* 6.33 0.71 8.13 0.76* 11.24** 0.76 13.5** (N = 1022, Age 80-84 1,07*** 6.29 0.97** 8.62 0.97** 10.03** 0.87* 10.16* 50 users) Age 84-89 1.58*** 7.58 1,45*** 9.77* 1.57*** 9.32* 1,74*** 10.55* Age 90+ 2.04*** 6.37 1.58*** 7.62 1.60*** 8.90* 1.57** 9.45 Log Likelihood / R2 -154 0.293 -127 0.551 -121 0.843 -1 1 0 0.889 Hospital, AHEAD Age 75-79 0.03 -0.30 -0 .0 0 -0.38 -0 .0 0 -0.37 0 .0 0 -0.34 Nursing Age 80-84 0.18*** 0 .0 0 0.01 -0 .21 0.01 -0 .2 0 0 .0 2 -0.13 Home (N = 6183, Age 85-89 0.42*** 0.45 0.17** 0.07 0.16** 0.01 0.18** 0.16 2185 users) Age 90-94 0.36*** 0.19 0.08 -0.30 0.07 -0.15 0 .1 0 0.04 Age 95+ 0.42** 2.90*** 0 .0 1 219*** 0 .01 2 ,io*** 0.04 2.25*** Log Likelihood / R2 -3958 0.025 -3452 0.059 -3436 0.098 -3434 0.104 + In cases where age dummy variables are shown, the youngest age group is the referent group (age 66-69 in PSID, age 72-74 in AHEAD). * j) < .05 ** j > < .01 *** g< .001 116 Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Table 6.7 Hierarchical Two-Stage Regression Results for Home Care Use and Out-of -Pocket Expenditures (in Logged Dollars) by Age in PSID and AHEAD Model 1 Model 2 Model 3 Model 4 Service type Data Set Age Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Home Care PSID Age 70-74 0 .1 2 -0.74 0.14 -1 .22 0.11 -0.62 0 .1 2 -0.81 Age 74-79 0.50* 0.16 0.43 -0.84 0.41 -0.79 0.40 -1.17 (N = 1028, Age 80-84 0.35 0 .2 0 0.21 -0.49 0.16 -0.31 0 .1 2 0.136 52 users) Age 84-89 0.62* 0.69 0.53 -1.23 0.48 -0.82 0.49 -2 .0 0 Age 90+ 1.05*** -0 .6 8 0.75* -1.98 0.64 -1.56 0.60 -2.42 LL* / R2 -194 0.083 -170 0.451 -167 0.502 -163 0.648 Home Health AHEAD Age 75-79 0.18*** 0.13 0.16** 0.18 0.16** 0.11 0.16** 0.14 Care, Other Age 80-84 0.36*** 0.44 0.19** 0.31 0 .2 0 ** 0.35 0.19** 0.41 Special (N = 5964, Age 85-89 0 .6 8 *** 0.57* 044*** 0.38 0 4 4 *** 0.39 0.42*** 0.45 Services 1154 users) Age 90-94 0.57*** 0.96* 0.24* 0.62 0.23* 0.69 0 .2 2 0.85* Age 95+ 0.76*** -0 .1 0 0.30 -0.57 0.29 -0.53 0.27 -0.26 LL/R 2 -2812 0.024 -2329 0.043 -2312 0.057 -2307 0.062 + For age dummy variables, the youngest age group is the referent group (age 66-69 in PSID, age 72-74 in AHEAD). *LL = Log Liklihood * E < ,05 ** e < .01 *** £ < .001 117 118 Discussion This Chapter was designed to examine whether out-of-pocket costs increase with age among the older population, whether these increases are linear, and whether the relationship between age and costs differs by service type. Results provided only some support for the widely held belief that health costs increase with age among the elderly. Use and out-of-pocket costs increased with age for long-term care services, which is consistent with other findings (Meerding et al, 1998; Roos et aL, 1989). Also, hospital use increased with age (until health was controlled), consistent with Frieman (1998), and age was a positive predictor of prescription costs in PSID. However, age was not a significant predictor of use or spending for the majority of services. Also, costs decreased linearly with age for ambulatory services in AHEAD, and use of dental care decreased with age. While age did predict higher total out-of-pocket costs in PSID, this likely reflects the inclusion of nursing home costs. Another hypothesis was that age would be a non-linear predictor of costs, with greater use and higher costs among the middle-old, and a plateau or decline among the oldest-old. There is some evidence of this effect in the charts showing patterns of use and spending with age; for example the drop in prescription costs among the oldest-old in both survey samples (Figure 6.6). However, the only non-linear effects that were significant in regression equations revealed an increasing effect of age into the oldest age groups. The pattern of these non-linear effects as shown by the coefficients for the dummy variables is informative, providing a more accurate picture of change with age than the coefficient for a linear effect. For example, Kington et aL (1995) found that age was a significant linear Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 119 predictor of lower dental use in the PSID, but their analysis did not reveal the increasing magnitude of this effect in successively older age groups. Another main goal of this Chapter was to determine whether the effects of age on spending can be explained by other factors, particularly health. As hypothesized, in most cases where age was a significant positive predictor of use or costs, controlling for health reduced or eliminated the effect of age. This occurred for prescription medication costs, hospital use, home care use, and total costs in PSID. These findings underscore the fact that poor health is a central determinant of health care use and costs, and highlight the importance of preventive health behaviors and disease management as ways to protect against high health costs as one ages. An interesting finding is that age continued to be a positive predictor of long-term care service use even after health was controlled. This indicates that other physiological and/or social factors related to advanced age (and not controlled in this study) are important determinants of long-term care service use. Wolinky & Johnson (1991) suggest that the oldest old have higher utilization because they are more physiologically and psychosocially frail, and are thus less able to rely on informal health services. The oldest- old are also least likely to have a living spouse who can provide informal care. Age effects were modified when insurance was controlled in some cases. For example, age emerged as a significant nonlinear predictor of out-of-pocket nursing home costs among users when insurance was controlled. This was likely a suppression effect reflecting the importance Medicaid coverage among this group; age was a positive predictor of out-of-pocket costs, but this relationship had been suppressed by the fact that Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 120 the older-old were more likely to be covered by Medicaid. Two hierarchical findings suggest that out-of-pocket spending was reduced due to lower SES among the oldest-old. For total costs including insurance premiums in AHEAD, the negative effect of age became non-significant when SES was controlled, suggesting that the lower spending with age was due to reduced ability to pay among this group. A similar conclusion can be drawn from evidence of a suppression effect for total costs excluding insurance premiums in PSID. For these costs, age reemerged as a positive predictor of total costs when SES was controlled, indicating that age was positively related to costs, but that this effect had been suppressed by the negative relationship between age and SES. As hypothesized, the burden of total out-of-pocket costs as a portion of income increased with age. This is likely due to decreasing income with age, as well as age-related increases in the use of nursing home and hospital care. Age remained significant as a negative predictor of out-of-pocket costs for ambulatory services in AHEAD in all hierarchical models. Thus, lower ambulatory costs with age do not appear to be due to better health, better insurance coverage, or lower SES. Lower costs may be explained by lower rates of use for dental care and outpatient surgery with age. The number of physician visits may also have declined with age, although the proportion of respondents reporting at least one visit remained high at all ages. Other possible explanations for lower costs with age are that the older-old received a lower intensity of services, received their care at lower cost institutions. It may be that the oldest old had less need for services due to a hardiness that helped them to survive to very old age. However, the pattern of service use and costs by Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 121 age provide some evidence for reduced access to care with age. The lower ambulatory costs, lower use of dental care and outpatient surgery, and greater hospital use with age may reflect barriers in access to appropriate care. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 122 Chapter 7 Race and Out-of-Pocket Health Costs This chapter examines the out-of-pocket health expenditures of African Americans compared to Whites in PSID, and African Americans and Hispanics compared to Whites in AHEAD. Four main policy questions are addressed: 1) Is the amount and burden of out-of-pocket spending greater among these racial/ethnic minority groups than among Whites? 2) How does the relationship between race and out-of-pocket costs differ by health service type? 3) Do race differences in out-of-pocket costs suggest inequities in access to health care? 4) Do race differences in spending appear to be due to differences in health, insurance coverage, or socioeconomic status? It is important to note that while membership in race categories is determined by genetic ancestry, race reflects primarily social rather than biological categories when it is used as a predictor in the social sciences (Williams, 1997). Because race is correlated with a variety of social factors, such factors will be the main focus of discussion in this examination of out-of-pocket spending by race. Review of the Literature Little is known about the amount and burden of out-of-pocket spending among older minority groups. Univariate comparisons of total out-of-pocket health spending by race in the 1977 NMCES and the 1987 NMES indicate that total out-of-pocket health care costs were lower for Blacks and Hispanics than for Whites, and these ethnic minorities were more likely than Whites to have no use o f services and no out-of-pocket costs (Rossiter & Wilensky, 1982; Taylor & Banthin, 1994). However, these surveys Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 123 included respondents of all ages, and race differences were not reported separately for older adults. Also, out-of-pocket costs were not reported separately for specific services. The few studies that have included race as a predictor of out-of-pocket costs in multivariate analyses have found that when race was a significant predictor of spending, the effect was in the direction of lower spending among minority groups. Rubin et aL (1995) found that race was an inconsistent predictor of total out-of-pocket health expenditures by elderly households in 1980-81 and 1989-90. When controlling for factors such as income, education, and housing tenure, Whites spent significantly more than non- Whites on medical services and on total health goods and services in 1980-81, and significantly more on prescription medications in 1989-90. Whites spent more out-of- pocket on insurance premiums than non-Whites at both time periods. Rubin & Koelln (1993b) found that in 1987, elderly White households spent more on health care than non- Whites even when the cost of insurance premiums was subtracted from out-of-pocket costs for services, controlling for education, income, assets, housing tenure, family size, age, and insurance. Stum et aL (1996) found that race was not a significant predictor of the amount spent out-of-pocket for home care among disabled older adults when controlling for other factors in the Andersen model However, only those with at least some out-of-pocket costs for home care were included in the study. Even fewer studies have examined racial differences in the burden of out-of-pocket spending among older adults. Coughlin et aL (1992) found that among the severely disabled population, race did not predict out-of-pocket burden for physician care, prescription medications, or total health spending. For hospital care, Whites had Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 124 significantly lower costs as a portion of income than non-Whites, whereas the out-of- pocket burden for nursing home care was higher for Whites than non-Whites. Stum, Bauer and Delaney, (1998) found that Blacks and Hispanics more likely to have catastrophic home care costs than Whites, but did not control for other factors. A review of the literature on race, health, and health care utilization provides insight into whether minorities would be expected to have higher or lower costs than Whites for different service types. African Americans and Hispanics are consistently found to be in poorer health than Whites. The prevalence of chronic diseases is higher among these groups (Markides & Black, 1996), and they have higher mortality rates at all ages until very old age (Hooyman & Kiyak, 1999). Blacks and Hispanics have higher level of functional disability than Whites, and they are more likely to rate their health as poor (Kaiser Family Foundation, 1999; Hooyman & Kiyak, 1999). The poorer health of Blacks and Hispanics relative to Whites might be expected to increase their health care costs. However, one factor that partially explains the poorer health status of minority groups is that they use less health care than Whites. Both Hispanics (Andersen, Giachello, & Aday, 1986) and African Americans (Blendon, Aiken, Freeman & Corey, 1989; Jones & Rene, 1994; Gomick et aL, 1996) proportionately underutilize health care services compared to Whites, and there is evidence that these groups are more likely to delay seeking care until their physical condition necessitates it (Bazargan, Bazargan, & Baker, 1998; Gomick et aL, 1996; Jones & Rene, 1994; Lannin, Mathews, Mitchell, Swanson, Swanson, & Edwards, 1998; Wolinsky etaL, 1990). Several factors related to race have been proposed as barriers in access to health Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 125 care among minority groups, including reduced resources in inner-city communities where racial minorities are concentrated, lower awareness of available providers, and institutional racism (Jones & Rene, 1994; Wallace, 1991; Weaver & Inui, 1975). Many inner-city hospitals have closed or relocated from neighbourhoods that have become heavily populated by Blacks, leaving residents with fewer health care resources (Wallace, 1991). Both Blacks and Hispanics are more likely than Whites and others to report having no usual source of care, such as “a particular clinic, health center, doctor’s office, or other place that he or she usually goes to if sick or in need of personal health advice” (Zuvekas & Weinick, 1999, p. 273). In the 1996 Medical Expenditure Panel survey, 29.6 percent of Hispanics and 20.2 percent of Blacks reported having no usual source of care, as compared to 15.5 percent of White and other respondents. Bazargan et aL (1998) found that among a sample of elderly African Americans, many reported a low level of perceived availability and accessibility of physicians, with a mean of 2.24 on a 4 point scale. The proportion of Hispanics lacking a ususal source of care rose substantially from 1977 to 1996, from 19.7 to 29.6 percent, which increased the existing gap between Hispanic Americans and Whites. This was partially explained by declines in health insurance coverage, but also occurred among those with insurance (Zuvekas & Weinick, 1999). Belonging to an HMO has been found to increase use for Blacks and Hispanics more than for Whites (Freiman, 1998), indicating that HMO membership may have a greater effect on increasing awareness among minority groups of a regular source of care. (However, this may also reflect a selection effect, where those more likely to use care are more likely to enroll in an HMO (Frieman, 1998).) Freiman (1998) also found that living in a high Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 126 income county increases use more for Whites than for Blacks or Hispanics, and that the health care utilization of Whites is more responsive to the physician fee schedule level than the use of Blacks and Hispanics. He saw this as evidence that minority groups may face more barriers in access to care that make them less able to respond to economic incentives in the health care market. Language barriers can be an important barrier in access to care and communication with health care providers among Hispanics elderly. It has been estimated that 40 percent of older Hispanics do not speak English (Hooyman & Kiyak, 1999). Members of minority groups may also have different attitudes toward health and the utilization of formal health services (Jones & Rene, 1994). There is evidence that minority elders have less trust in medical providers are more likely than Whites to relay on folk medicine and religious healing (Hooyman & Kiyak, 1999). Some suggest that minority groups have different cultural dispositions toward the use of health care than Whites (Wonlinsky,1982; Stahl & Gardner, 1976). Lannin et aL (1998) found that cultural beliefs and attitudes played an important role in explaining the tendency of African American women to delay seeking care for breast cancer. However, LaVeist (1994) argues that culture is ill-defined, and it is not clear that minority groups such as African Americans have a distinct culture. Racial differences in the use of long-term health care may also be related to cultural patterns of family caregiving. Blacks and Hispanics delay the use of formal long-term care services (home health and nursing home care) until later ages than Whites (despite the fact that they become disabled at earlier ages), because they are more likely to rely on informal caregivers, which include immediate and extended family members (Burton et aL, 1995; Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 127 Cagney & Agree, 1999). Reduced health care use by minority groups would be expected to result in lower health care costs in the short-run, however poorer health and high costs may eventually result from a lack of preventive care and disease management (Jones & Rene, 1994). For example, Bazargan et aL (1998) were concerned to find that elderly Blacks with heart problems or diabetes did not have more physician visits than those without these conditions, despite the fact that these are serious diseases which require frequent medical attention. As a result, Blacks are more likely to undergo procedures associated with poor disease management, such as amputation of the lower limb (Escarce, Epstein, Colby, & Schwartz, 1993; Gomick et aL, 1996). Blacks and Hispanics are less knowledgeable about cancer and its prevention than Whites (Baquet, 1988), and are less likely to undergo preventive screening such as mammography (Caplan, Wells, & Haynes, 1992; Gomick et aL, 1996). This decreases the likelihood of detecting cancer early enough to prevent metastasis (Hooyman & Kiyuk, 1994), resulting in the need for more intensive treatment. Consistent with this pattern of delaying care until use is no longer discretionary, older Blacks and Hispanics have fewer physician visits than Whites, but have greater emergency room use, and are found in some studies to have higher rates of hospitalization (Bazargan et al., 1998; Gomick et aL, 1996; Moon, 1992). However there is also evidence that minority groups are more likely to inappropriately use emergency rooms for less serious conditions (Cunningham, Clancy, Cohen, & Wilets, 1995). Improper use of emergency care may be due partly to lack of access to or awareness of other appropriate sites for care. Minority groups may also prefer to receive care in an institutional setting Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 128 than from an Anglo physician from whom they feel more distant (Weaver &Inui, 1975). Inappropriate use of health care by minority groups may also be explained by low functional health literacy, which is distinct from education level and overall literacy, and hinders navigation of the health care system and appropriate use of care. Gazmararian et aL (1999) found that among Medicare enrollees in a managed care organization, Blacks and Hispanics were significantly more likely than Whites to have inadequate or marginal levels of functional health literacy. Even when controlling for the socioeconomic factors of education and occupation, Blacks had 3.54 times greater odds than Whites of having inadequate or marginal functional health literacy versus adequate literacy, and Hispanics were more than twice as likely to have inadequate or marginal literacy. In a 1998 study, those with low health literacy were found to have fewer doctor visits but use substantially more hospital care when self-reported health was controlled (National Academy on an Aging Society, 1998). These racial differences in service use suggest that Blacks and Hispanics are likely to have lower mean out-of-pocket costs than Whites for services that are more discretionary, and similar or higher costs for less discretionary services such as hospital stays and ER visits. Also, if members of minority groups are more likely to delay care until health is poor, costs among those using services might be expected to be higher for Blacks and Hispanics. Univariate comparisons from the 1994 MCBS provide some support for this hypothesis; among users of each health service, Blacks and Hispanics had higher out- of-pocket costs than Whites for inpatient and outpatient hospital care (including ER), and lower out-of-pocket costs for dental care and prescription medications (Olin & Liu, 1998). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 129 Out-of-pocket costs in the category including visits to physicians and other health practitioners and medical equipment were lower for Hispanics than for Whites, but slightly higher for Blacks than for Whites. However, even among users, the relationship between health and the costs of care provided may be mediated by race. There is evidence that minorities who seek health care receive less intensive treatment than Whites for the same conditions. Escarce et aL (1993) found that among older Medicare beneficiaries, Blacks were less likely than Whites to receive 23 common procedures and tests, whereas Blacks were more likely than Whites to receive 7 such services. Whites were particularly more likely than Blacks to receive newer or higher technology services. There is also evidence that diseases are treated less aggressively among Blacks than Whites. For example, a recent analysis of two large national data bases revealed that elderly blacks diagnosed with early-stage lung cancer were less likely than white patients to undergo surgical resection, a life-saving technique for early-stage lung cancer (Bach, Cramer, Warren, & Begg, 1999). This difference remained after controlling for gender, coexisting illness, socioeconomic status, insurance coverage, and availability of care. Blacks also had a lower five-year survival rate than whites, which was at least partially explained by treatment differences (Bach et aL, 1999). Similarly, Dingham et aL (1999) found that in clinical trials on colon cancer, equal treatment resulted in equal outcomes for both races. They concluded that the worse prognosis among Blacks compared to Whites in the population at large was due to differences in treatment, even when accounting for diagnosis at later disease stage among Blacks. Even treatments which have higher use Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 130 rates among Blacks may be underused in relation to relative need. For example, Escarce et aL (1993) found that Blacks had higher use of three ophthalmologic procedures use to treat glaucoma. However, because the prevalence, severity, and morbidity of glaucoma is higher in Blacks, such procedures may still be under-used by elderly Blacks relative to Whites (Javitt, McBean, Nicholson, Babish, Warren, & Krakhauer, 1991). A recent study compared race differences in the use of medical procedures for different health conditions among older Medicare beneficiaries (Lee, Baker, Gehlbach, Hosmer, & Reti, 1998). While no race differences in procedure use were found for hip fracture and breast/colon cancer, substantial differences were found for stroke and coronary disease, conditions for which there is less widespread agreement on approaches to evaluation and treatment. This study confirmed previous findings that Blacks are less likely than Whites to undergo cardiac procedures such as angioplasty, cardiac catheterization, and Coronary Artery Bypass Grafting (CABG), even when controlling for insurance status and severity of disease (see Geiger, 1996). For example, Conigliaro (1999) found that Blacks in Vetran’s Administration (VA) hospitals received fewer cardiac procedures than Whites: White patients were 1.5 times more likely to receive cardiac catheterization, and 2.7 times more likely to undergo CABG. These differences persisted when controlling for many sociodemographic and health variables. Also, since all services are fully covered for VA patients, insurance coverage could not be used as an explanatory factor. While it is difficult to determine the relative contribution of different factors responsible for the lower use of medical procedures by minority groups, physician bias Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 131 appears to play a role. Schulraan et aL (1999), presented physicians with videotaped interviews of actors portraying patients, and found that physicians were less likely to recommend cardiac catheterization for black patients, controlling for the physician’s assessment of the probability of coronary artery disease as well as for the patient’s age, level of coronary risk, type of chest pain, and the results of an exercise stress test. In the case of breast cancer screening, Caplan et aL (1992) found that the most common reason for not having a mammogram among older Black women was that it was not recommended by a doctor. Richardson et al. (1987) found similar results in a study of older Hispanic women. While these results reveal a racial bias in physicians’ clinical decisions, there is also evidence that minorities are more likely to refuse recommended treatment. Conigliaro (1999) found that African Americans presented with a hypothetical decision for cardiac treatment refused angioplasty and bypass surgery more often than Whites presented with the same situations. This appeared to be at least partially due to lack of familiarity with these treatments. In this study, patients who were less familiar with the procedures were more likely to refuse them, and African Americans were less familiar with the procedures, rated themselves as less competent at estimating the risks of surgery, were more likely to overestimate risks, and were less likely to have a family member or friend who had undergone the procedure (Conigliaro, 1999). Higher likelihood of refusing treatment may also reflect a lack of trust in physicians and in the medical system among Blacks, who have experienced racial discrimination and may have reason to question whether recommended treatment is in their best interest (Conigliaro, 1999). Blacks are also less likely than Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 132 Whites to be satisfied with the way they are treated by physicians when they are ill, more likely to be dissatisfied with their care in the hospital, and more likely to feel that the duration of their hospital stay was too short (Blendon et aL, 1989). Two other important barriers in access to health care among minority groups are lower socioeconomic status (SES) and lower levels of insurance coverage. Older African Americans and Hispanics have lower mean incomes and accumulated wealth than Whites, resulting in part from lower levels of education, a lifetime of more limited employment opportunities, and concentration in low-paying jobs with fewer benefits (Hooyman & Kiyak, 1999). A higher proportion of African Americans live alone, and older adults who live alone are more likely to live in poverty (Angel & Hogan, 1994). African Americans and Hispanics are nearly three times more likely than Whites to be living below the poverty level (Hobbs & Damon, 1996; Kaiser Family Foundation, 1999). The lower SES among minority groups would be expected to result in lower health care use and lower costs, but perhaps higher burden among those with costs. Race differences in insurance coverage can affect out-of-pocket costs by influencing the use of care, as well the out-of-pocket liability for care that is used. The extent to which race differences in use are mirrored by race differences in out-of-pocket costs is modified by the level of insurance coverage. Blacks and Hispanics are more likely to be eligible for Medicaid coverage than Whites, but are much less likely to have supplemental insurance coverage (Hooyman & Kiyak, 1999; Kaiser Family Foundation, 1999). A recent analysis of three national data sets on service use suggest that insurance and income status are more important determinants of service use than ethnicity. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 133 However, while racial differences in health care use are generally reduced when SES and insurance are controlled, they often remain significant (Blendon et aL, 1989). Escarce et aL (1993) found that racial differences in the use of medical tests and procedures persisted even among those who were covered by Medicaid in addition to Medicare, suggesting that financial barriers alone could not fully account for race differences. The data used for the current study will allow a thorough investigation of race differences in use and out-of- pocket costs for several different health services, using a number of measures to control for race differences in health, insurance, and SES. The policy questions set forth at the beginning of the chapter will be addressed through four research questions, examined in both survey samples for Blacks and in AHEAD for Hispanics: 1) Do minority groups report higher or lower out-of-pocket health spending than Whites, overall and for different service types? 2) Are reported out-of-pocket costs more or less burdensome for minority groups? 3) How are race differences in out-of-pocket spending related to use, and to differences in costs among users? 4) Is race an independent predictor of out-of-pocket spending when controlling for health, insurance coverage, and socioeconomic status? Hypotheses for this Study In descriptive analyses, Blacks and Hispanics are expected to have lower use than Whites on services considered discretionary and on nursing home and home care. For the least discretionary services (ER visits and hospital stays), use for minority groups is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 134 expected to be higher than or similar to use for Whites. For the proportion of users with uncovered costs, it is difficult to predict which group will have a higher percentage of users with uncovered costs, since Whites are more likely to be covered by supplemental insurance, while Blacks are more likely to be eligible for Medicaid. Consistent with predictions of service use by minority groups, expenditures by Blacks and Hispanics are expected to be lower than those of Whites for more discretionary services, and similar or higher for less discretionary services. The extent to which race differences in use are mirrored by differences in out-of-pocket costs will be affected by the generosity o f insurance coverage among each race group (both supplemental insurance and Medicaid). When the distribution of out-of-pocket spending on different services is examined, minority groups are expected to spend a higher proportion of their out-of-pocket dollars on less discretionary services. Due to lower mean income among minority groups, the mean portion of income spent on health services is expected to be higher for Blacks and Hispanics. However, again this relationship will be mediated by differences in insurance coverage. For the cost of care among users, competing determinants make predictions for race differences difficult. Given the lower average health among minority groups, and the evidence that they are more likely to wait until health problems are more serious before seeking care, out-of-pocket costs among only those using each service might be expected to be higher for minority groups. However, this might be countered by racial bias or cultural beliefs that lead to the use of fewer or lower cost procedures among minority Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 135 users of care. Out-of-pocket costs among users will also be heavily influenced by the comprehensiveness of insurance coverage; be it supplemental insurance, a prepaid health plan, or Medicaid. In bivariate regression analyses, minority race is expected to be a negative predictor of spending on the moderately discretionary services of physician/therapist/ER visits and prescription medications. For nursing home and home care, where use and costs will be predicted separately, minority race may be a negative predictor of use if these groups are relaying on more informal long-term care. Costs among users may be higher for Blacks and Hispanics if they are in poorer health and require more care, but may be lower for minority groups since they are also more likely to be covered by Medicaid. Hypotheses are similar for outpatient surgery and equipment, where race may be a negative predictor of use, but a non-significant predictor o f costs among users. For hospital stays, race may not be a significant predictor of use if use is similar for minorities and Whites. It is difficult to predict whether or not race will be a significant predictor of out-of-pocket hospital expenditures among users, since Black and Hispanic hospital patients are likely to have poorer health, but may undergo less expensive procedures and have more comprehensive coverage under Medicaid. The effect of race as a predictor of total spending excluding insurance is difficult to predict given the mixed hypotheses for use and costs for different types of services. However, for total costs including insurance premiums, minority race is likely to be a negative predictor of spending, since Blacks and Hispanics are less likely to have supplemental insurance. In hierarchical regression, when health variables are added, race is expected to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 136 remain or become a negative predictor of use and costs. Because minorities have worse health on average, controlling for these differences may reveal or accentuate racial differences in use or intensity of treatment. If minority race remains a significant predictor when insurance and socioeconomic variables are controlled, then factors such as access to care and cultural practices, beliefs and attitudes may be explanations for differences in spending. Because the impact of such factors is likely to be stronger for the use of discretionary than non-discretionary services, it is hypothesized that race is more likely to be significant in the final models for more discretionary services. Measurement of Race In PSID, those reporting themselves as Black were contrasted with those reporting themselves as White. In AHEAD, two minority groups were contrasted with Whites: those reporting themselves as Black or African American and those who considered themselves as Hispanic or Latino. Those reporting themselves as both Black and Hispanic were coded as Black, and those who were both White and Hispanic were coded as Hispanic. Thus, ‘ White’ in AHEAD refers to non-Hispanic Whites. In both samples, there was an insufficient number of respondents in other racial or ethnic groups to allow meaningful analyses of these groups separately. In PSID, nine respondents reporting themselves as something other than Black or White were excluded from analyses in this Chapter, yielding a weighted sample size of 1024. In AHEAD, 109 respondents reporting themselves as something other than White, Black, or Hispanic were excluded from analyses in this Chapter, yielding a weighted sample size of 6175. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 137 Analyses Descriptive analyses were performed to examine race differences in mean and median expenditures, and in the burden of out-of-pocket costs as a portion of household income. In order to examine the relationship between costs and use, descriptive analyses were performed comparing race groups on rates of service utilization, the distribution of spending on different services, the proportion of users with uncovered costs, and mean and median expenditures per user of each service. Hierarchical regression analysis was performed for all of the spending variables in each data set, as discussed in Chapter 5. In the first hierarchical model, the dummy variables for race were entered, as well as age, gender, urban residence, and the variable reflecting regional/state Medicare spending. These demographic and geographic variables were included as control variables in case of possible relationships with race, however their correlation with race was not strong enough that they were expected to be confounding variables. In the second model, health variables were added to examine the effects of race independent from health. The third model included insurance variables, and in the fourth and final model, socioeconomic variables were controlled. As discussed in Chapter 5, for variables with a small portion of users, a two-part model predicting use and then costs among users was used for each hierarchical equation. For variables on which most respondents had nonzero costs, OLS alone was used for hierarchical equations. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 138 Results In PSID, 85.6 percent of the sample was White, and 14.4 percent was Black (weighted percentages). In AHEAD, 86.6 percent were White, 9.7 percent were Black, and 3.7 percent were Hispanic. Means and percentages on independent variables by race are shown in Table 7.1. Race differences on demographic variables were generally small and inconsistent across the two survey samples. Age was similar across race groups, and a higher percentage of Blacks were female in AHEAD. Blacks lived in states with slightly lower spending per beneficiary in PSID, whereas in AHEAD, minority groups lived in regions with higher spending per beneficiary. Slightly fewer Blacks lived in urban areas in PSID, whereas minority groups were more likely to live in urban areas in AHEAD. Race differences in health were consistent overall with well documented findings of poorer health among minority groups. Blacks and Hispanics had poorer self-rated health, higher depression scores, more functional difficulties, and were more likely to spend 10 or more days in bed than Whites. However, in many cases these differences were not as large as might be expected. This may be in part due to survival effects; because mortality is greater among minority groups, minorities who survive to their late 60’ s or 70’ s are more of a select group than Whites who reach these ages. Consistent with previous findings, Blacks and Hispanics reported substantially higher rates of high blood pressure and diabetes than Whites. However, minorities had lower cancer rates, and race differences for other conditions were inconsistent. Rates of insurance coverage by race were also consistent with previous reports, with minority groups much more likely to be dually eligible for Medicare and Medicaid, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 139 and much less likely to be covered by supplemental insurance. Blacks and Hispanics were also more likely than Whites to have no other coverage besides Medicare. Hispanics were more likely than Whites to be enrolled in a pre-paid health plan, reflecting concentration of Hispanics in areas with high penetration of Medicare HMOs (Kaiser Family Foundation, 1999). Race differences were large for all three measures of SES, with minority groups reporting fewer years of education and lower mean income and wealth. Table 7.1 Means and Percentages on Independent Variables by Race in PSID and AHEAD PSID AHEAD White N= 876 Black N = 148 White N = 5347 Black N = 599 Hispanic N = 229 Demographic Mean Age 74.9 73.7 79.1 79.4 78.7 Percent Female 62.6 61.9 62.4 67.6 61.4 Mean State Spending 3387 3351 4745 4802 4980 Percent Urban Residents 94.7 93.1 74.9 78.5 90.8 Health Self-Rated Health 2.80 2.32 3.02 2.58 2.68 Depression - - 1.85 2.19 2.23 Cognitive Status - - 9.12 7.84 8.49 Proxy Respondent - - 11.3 18.5 21.6 Percent with poor sight 24.0 38.3 - - - Percent with deafness 24.2 12.6 - - - Activities of Daily Living 1.45 2.31 0.8 1.0 1.0 Percent with 10+ sick Days - - 3.1 5.3 5.5 (Table Continues) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 140 Table 7.1 (continued) PSID AHEAD White Black White Black Hispanic Cancer 6.5 3.4 17.5 11.1 9.6 Diabetes 11.3 14.1 12.8 24.7 23.2 Lung Disease 11.5 11.6 13.4 6.4 10.4 High Blood Pressure 38.0 57.7 53.1 69.1 55.7 Neuro Problem / Stroke 6.3 5.1 10.6 12.9 7.8 Digestive Problem 16 15.4 - - - Heart Attack - - 92. 8.1 9.3 CHF - - 4.7 2.1 3.4 Heart Disease 13 19 24.8 20.9 17.8 Arthritis 59 69.8 53.9 66.4 61.3 Fall in past 2 years - - 33.7 29.2 28.8 Insurance Medicaid 3.5 22.6 6.6 30.4 45.7 Prepaid health Plan 8.6 5.5 12.1 8.7 18.2 Employer Insurance 28.8 16.9 15.4 7.4 3.5 Medigap Insurance 46.0 15.8 53.2 21.5 11.2 FFS Medicare Only 10.1 31.3 12.3 30.5 19.3 Missing on Insurance 11.8 18.8 1.5 4.9 4.1 SES Mean Education 11.28 7.77 11.62 8.37 6.15 Mean Household Income 25,275 15,572 27,243 17,837 16,195 Mean Household Wealth 196,516 58, 471 217,746 58,087 68,214 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Tables 7.2 and 7.3 show mean and median out-of-pocket spending across all respondents for each service type, and total expenditures excluding and including the costs of insurance. As seen in Table 7.2, Blacks had lower mean and median expenditures than Whites for each type of service and overall The race difference was largest for outpatient surgery, dental care, nursing home care and insurance. For several service types, median costs were zero for both races, reflecting the use of these services by only a small portion of the sample. For insurance, median costs were $450 for Whites and zero for Blacks, reflecting a large race difference in the proportion purchasing supplemental coverage. Table 7.2 Mean and Median On-Year Spending for All Respondents on Each Health Service by Race in PSID Service Whites Blacks Hospital Stays Mean 111 81 (SD) (607) (316) Median 0 0 N 864 144 Outpatient Surgery Mean 48 13 (SD) (421) (62) Median 0 0 N 871 148 Prescription Medication Mean 271 190 (SD) (460) (347) Median 90 18 N 838 135 Dental Care Mean 136 51 (SD) (523) (251) Median 0 0 N 873 146 Physician, Therapist, Mean 129 126 ER Visits (SD) (380) (347) Median 52 0 N 830 137 (Table continues) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 142 Table 7.2 continued Service Whites Blacks Nursing Home Stays Mean 388 23 (SD) (3,252) (351) Median 0 0 N 875 146 Home Care Mean 6 0 (SD) (98) (3) Median 0 0 N 874 147 Insurance Premiums Mean 556 252 (SD) (693) (467) Median 450 0 N 842 130 Total Costs Mean 1,026 470 (excluding insurance) (SD) (2,779) (703) Median 410 120 N 765 122 Total Costs Mean 1,606 770 (including insurance) (SD) (2,901) (1,033) Median 1,062 270 N 738 107 Race differences in spending were similar in AHEAD. As shown in Table 7.3, Blacks had lower expenditures for each service type and overall. The race difference was small for prescription medications and large for the categories of hospital/nursing home care and physician/ER/surgery/dental care. As seen in the last column of Table 7.3, Hispanics had substantially lower costs than both Blacks and Whites for all services with the exception of prescription medications, where mean spending for Hispanics was slightly higher but median spending was lower than in other groups. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 143 Table 7.3 Mean and Median Two-Year Spending for All Respondents on Each Health Service by Race in AHEAD Service Whites Blacks Hispanics Hospital/Nursing Home Mean 688 498 135 Stays (SD) (5,322) (3,304) (634) Median 0 0 0 N 5290 588 216 Physician, ER, Mean 579 393 258 Outpatient Surgery, (SD) (1,660) (799) (565) Dental Care Median 200 50 0 N 5199 554 227 Monthly Prescription Mean 66 64 78 Medication Costs (SD) (220) (193) (213) Median 16 7 0 N 5269 589 227 Home Health Care, Mean 130 85 22 Special Services (SD) (2,561) (952) (276) Median 0 0 0 N 5316 593 227 Insurance Premiums Mean 2,227 674 562 (SD) (2,989) (1,204) (1,043) Median 1,800 0 0 N 5321 586 225 Total Costs Mean 2,846 2,396 2,179 (excluding insurance) (SD) (8,174) (5,826) (4,944) Median 986 510 230 N 5117 545 216 Total Costs Mean 5,093 3,100 2,730 (including insurance) (SD) (8,726) (6,037) (5,044) Median 3,360 1,200 700 N 5105 538 215 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 144 The percentage of household income spent on health care (including and excluding insurance) is shown in Tables 7.4 and Tables 7.5. Race differences were small in both survey samples. In PSID, Blacks spent about one percent less of their household incomes than Whites on total costs excluding insurance premiums, and just over one-and-a-half percent less than Whites on total costs including premiums. In AHEAD, total out-of- pocket costs excluding insurance consumed a slightly higher mean proportion of house hold income for Blacks and Hispanics than for Whites, however the median portion of income spent was lower for Blacks than Whites, and lowest for Hispanics. When the costs of insurance were included, Blacks and Hispanics had lower mean and median spending as a proportion of income than Whites. Race differences were largest for the median proportion of income spent, with Blacks spending just over half as much of their house hold income as Whites, and Hispanics spending less than a third as much of their incomes. Table 7.4 Percentage of Household Income Spent on Total Out-of-Pocket Costs Excluding and Including Insurance Premiums in PSID Whites Blacks Total Costs Mean 6.6 4.2 (excluding insurance) (SD) (15.1) (6.3) Median 2.1 0.8 N 765 122 Total Costs Mean 10.0 6.3 (including insurance) (SD) (16.6) (9.2) Median 5.0 1.3 N 738 112 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 145 Table 7.5 Percentage of Household Income Spent on Total Out-of-Pocket Costs Excluding and Including Insurance Premiums in AHEAD Whites Blacks Hispanics Total Costs (excluding Mean 7.9 9.4 10.2 insurance) (SD) (17.3) (17.1) (19.9) Median 2.2 1.9 0.8 N 5117 545 216 Total Costs Mean 14.2 11.9 12.1 (including insurance) (SD) (20.4) (18.1) (19.9) Median 7.5 4.2 2.4 N 5105 538 215 Tables 7.6 and 7.7 examine the portion of each race group with mean spending over 10 percent, 30 percent, and 50 percent of household income on total out-of-pocket health costs excluding and including insurance. For total costs excluding insurance in PSID, a slightly higher proportion of Blacks than Whites spent over 10 percent of their incomes, whereas Whites were more likely to spend over 30 and over 50 percent of their incomes. In AHEAD, more Blacks than Whites spent over 10, over 30, and over 50 percent of their household income on out-of-pocket costs excluding insurance. A higher proportion of Hispanics than Whites also spent over 30 and over 50 percent of income, although the proportion of Hispanics spending over 10 percent of income was slightly lower than the proportion of Whites. When insurance premiums are included as part of total out-of-pocket costs, results are similar, with the largest increase being for the proportion of Whites spending over 10 percent of household income. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 146 Table 7.6 Percentage of Income Spent on Total Health Care (Excluding Insurance) by Race in PSID and AHEAD Percent spending this percentage over 10% over 30% over 50% PSID Whites 15.9 4.3 2.9 Blacks 11.0 1.2 0.5 AHEAD Whites 19.0 5.9 3.6 Blacks 22.7 7.9 5.1 Hispanics 18.3 10.1 6.4 Table 7.7 Percentage of Income Spent on total health care including insurance in PSID and AHEAD Percent spending this percentage over 10% over 30% over 50% PSID Whites 27.4 7.4 3.5 Blacks 16.9 4.9 2.7 AHEAD Whites 40.8 11.7 5.6 Blacks 30.5 10.2 6 Hispanics 26.1 10.7 6.7 Tables 7.8 and 7.9 show the percentage of all respondents in each sample reporting use of each service. Overall use of services was lower in PSID than in AHEAD, likely due to the two-year period of use in AHEAD. As seen in the columns for Whites and Blacks in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 147 each table, Blacks used fewer services than Whites overall, and the race difference was greater in PSID than in AHEAD. The race difference was greatest for dental care, with a much smaller portion of Blacks than Whites reporting dental care use. Blacks were less likely than Whites to report at least one visit to a physician or therapist. However, analyses not shown in Table 7.8 reveal that the mean number of physician visits was higher for Blacks than Whites in PSID (9.09 vs. 6.97), and higher for Blacks and Hispanics than for Whites in AHEAD (11.03 and 11.02 vs. 9.23). Three services for which Blacks had greater use than Whites were hospital care, emergency room visits, and home care. The race difference was smaller for hospital care and ER use than for home care. Blacks also had slightly greater use of nursing home care than Whites in AHEAD, whereas their rate of nursing home use was less than half as high as Whites in PSID. Differences in use for Hispanics compared to Whites and Blacks in the AHEAD sample are shown in Table 7.9. Overall service use by Hispanics was slightly lower than use by Blacks, and lower than use by Whites. In contrast to Blacks, Hispanics used slightly less hospital care than Whites. Use of physician visits was the same for Blacks and Hispanics, and use of prescription medication was similar for all races in AHEAD. Dental care use was slightly higher for Hispanics than for Blacks, but was still much lower than for Whites. Use of nursing home care was particularly low for Hispanics compared to Blacks and Whites, whereas home care use was the same for both minority groups. Both Blacks and Hispanics had lower use of ‘other special services’ than Whites, and fewer Hispanics than Blacks reported using such services. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 7.8 Percentage Reporting Use of each Health Service by Race in PSID Whites Blacks Use of Any Services 96.2 89.4 Hospital Use 20.3 24.7 Outpatient Surgery Use 13.5 9.3 Physician/Therapist Use 87.7 79.5 Emergency Room Use 8.7 10.1 Prescription Medication Use 83.9 81.9 Dental Use 45.9 15.7 Equipment Use 36.2 26.5 Nursing Home Use 5.9 2.3 Home Care Use 4.8 6.9 Table 7.9 Percentage Reporting Use of Each Health Service by Race in AHEAD Service Whites Blacks Hispanics Use of Any Services 98.1 96.8 96.5 Hospital Use 34.5 37.6 32.9 Use of Outpatient Surgery 20.9 15.4 11.8 Physician Use 95 93.3 93.3 Prescription Medication Use 79.5 79 80.6 Use of Dental Care 56.5 32.2 34.8 Nursing Home Use 6.6 7.2 2.4 Use of Home Health Care 14 19.2 19.2 Use of other Special Services 9 13.7 11.5 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 149 Tables 7.10 and 7.11 depict racial differences in the distribution of out-of-pocket costs among different health goods and services. (Results are presented only for those with complete data on spending for each service type.) As seen in Table 7.10 , Whites spent a greater proportion of their out-of-pocket health dollars on insurance and dental care compared to Blacks, and Whites spent a lower proportion than Blacks on hospital care, physician visits, prescription medications, and equipment. Findings for Blacks were similar in AHEAD. As seen in Table 7.11, Blacks spent less of their out-of-pocket dollars on insurance than Whites, and a greater proportion on other services, particularly prescription medications. Hispanics also spent less than Whites on insurance as a proportion of total spending, and more on prescription medications, physician visits and home health care / special services. However, Hispanics, unlike Blacks, spent a lower proportion of their out-of-pocket dollars on hospital/nursing home care than Whites. Table 7.10 Distribution of Out-of-Pocket Spending on Different Services by Race in PSID Service Whites Blacks Hospital Stays 3.7 5.6 Outpatient Surgery 1.5 1.6 Physician/Therapist/ER 12.6 16.1 Dental Care 9.8 4.7 Prescription Medications 21.3 36 Equipment 7.9 9 Home Care 0.2 0 Nursing Home Care 2.7 0 Insurance 40.3 27.1 Total 100 100 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 150 Table 7.11 Distribution of Out-of-Pocket Spending on Different Services by Race in AHEAD Service Whites Blacks Hispanics Hospital/Nursing Home Stays 4.1 6.3 3.6 Physician/Outpatient Surgery/ Dental/ER 18.5 23.9 26.1 Prescription Medications 27.1 43.1 41.1 Home Health Care / Other Special Services 1.1 1.4 2.1 Insurance 49.1 25.3 27.2 Total 100 100 100 The portion of those using each service who had uncovered costs in shown by race in Tables 7.12 and 7.13. As seen in the first row of each table, Blacks and Hispanics were more likely than Whites to have no uncovered costs at all. Blacks were also less likely than Whites to have uncovered costs for each specific service type used in PSID. Results for uncovered costs were similar in AHEAD, except that Blacks using hospital care were more likely than Whites to have uncovered costs, and had uncovered costs at a rate similar to Whites for outpatient surgery. Hispanics were less likely than Whites to have uncovered costs for every service type, and also had lower rates of uncovered costs than Blacks for most services. Rates of uncovered costs were particularly low among Hispanics for physician visits, prescription medications, and home care. Rates of uncovered costs were low for both Blacks and Hispanics compared to Whites for dental care and nursing home care. Percentages reporting uncovered costs were lower in AHEAD than in PSID overall and for several service types. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 151 Table 7.12 Percentage Of Users Reporting Uncovered Costs for Each Health Service by Race in PSID Service Whites Blacks Any Uncovered Costs 96.6 79.9 Uncovered Hospital Costs 70 67.3 Uncovered Surgery Costs 63.5 49.5 Uncovered Physician/Therapist/ER Costs 75.9 56.3 Uncovered Prescription Costs 90.7 73.3 Uncovered Dental Costs 92.7 79.6 Uncovered Equipment Costs 91.5 77.4 Uncovered Nursing Home Costs 81.4 70.7 Uncovered Home Care Costs 17.6 8.3 Table 7.13 Percentage Of Users Reporting Uncovered Costs for Each Health Service by Race in AHEAD* Service Whites Blacks Hispanics Any Uncovered Costs 88.5 75.7 59.5 Uncovered Hospital Costs 29.1 35 24.7 Uncovered Physician/ER Costs 49.9 45.6 28.9 Uncovered Outpatient Surgery Costs 34.1 34.5 30.1 Uncovered Dental Costs 92.9 73.9 80.6 Uncovered Prescription Costs 84.7 73.7 54.3 Uncovered Nursing Home Costs 47.5 27.3 27.3 Uncovered Home Health Care Costs 15.5 10.3 1.3 ♦percentages with uncovered costs excludes users who did not say whether or not they had uncovered costs or who reported that their costs were not yet settled Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 152 Mean and median out-of-pocket expenditures among only those using each service are shown in Tables 7.14 and 7.15. As seen in Table 7.14, in PSID Blacks had lower mean costs per user than Whites for most services, and these differences were largest for hospital and outpatient surgery costs. For physician/therapist/ER visits, median spending was lower for Blacks than Whites, while mean spending was higher for Blacks. This reveals that Blacks typically had lower costs for these ambulatory services, but that a portion of Blacks were high cost users. A similar pattern existed for dental care, where median spending was similar for both races, while mean spending was somewhat higher for Blacks. (This higher mean spending among Black users of dental care was found in this PSID data by Kington et aL (1995), however median spending was not presented in their study.) For nursing home care, Blacks had lower mean and higher median spending than Whites. This result is due to the fact that only two Blacks reported nursing home use, one with zero out-of-pocket costs and the other spending $9,120. For insurance, mean spending was lower and median spending much lower for Blacks than for Whites. Total mean and median spending was substantially lower for Blacks than Whites. In AHEAD, as shown in Table 7.15, Blacks had lower mean spending for all services, although the means for prescription drugs were almost equal. Median spending for users of hospital/nursing home care and home health care/special services was zero for all race groups, reflecting the fact that less than half of those using these services had uncovered costs. Median spending among users of physician/ER/surgery/dental care was much lower for Blacks than Whites, as was median spending on prescription medications and insurance. Differences in total out-of-pocket costs were large for mean spending and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. even larger for median spending, revealing that typical overall spending was much lower for Blacks than Whites. As seen in the last column of Table 7.15, Hispanic users had much lower out-of-pocket costs than Whites and Blacks for hospital/nursing home care, physician/ER/surgery/dental care, and home health care/special services. Spending on insurance and total out-of-pocket costs were also lowest for Hispanics. For prescription medications, median costs were much lower for Hispanics while mean costs were somewhat higher for Hispanics than for both Whites and Blacks, reflecting a portion of high-cost users in the Hispanic group. Table 7.14 Mean and median spending for users of each health service by race in PSID Service Whites Blacks Hospital Stays Mean 581 360 (SD) (1,266) (673) Median 200 8 N 165 33 Outpatient Surgery Mean 365 148 (SD) (1,085) (178) Median 50 0 N 114 13 Physician, Therapist, Mean 146 158 ER Visits (SD) (401) (384) Median 75 5 N 732 109 Prescription Medications Mean 326 239 (SD) (483) (382) Median 150 62 N 696 107 Dental Care Mean 298 354 (SD) (736) (595) Median 100 96 N 399 21 (Table continues) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 7.14 continued Service Whites Blacks Nursing Home Stays Mean 6,695 2,496 (SD) (11,933) (4,733) Median 2,416 4,560 N 51 2 Home Care Costs Mean 136 3 (SD) (456) (12) Median 0 0 N 41 10 Insurance Premiums Mean 699 637 (SD) (701) (679) Median 622 400 N 670 51 Total Costs Mean 1,072 541 (excluding insurance) (SD) (2,829) (737) Median 444 170 N 732 106 Total Costs Mean 1,649 856 (including insurance) (SD) (2,949) (1,077) Median 1,099 344 N 148 94 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 7.15 Mean and Median Spending for Users of Each Health Service by Race AHEAD Service Whites Blacks Hispanics Hospital/Nursing Home Mean 1,959 1,289 403 (SD) (8,815) (5,344) (1,061) Median 0 0 0 N 1858 227 77 Physician, ER, Mean 598 412 273 Outpatient Surgery, (SD) (1,683) (813) (575) Dental Care Median 200 60 0 N 5037 528 205 Prescription Medications Mean 83.9 81.5 97 (SD) (245) (215) (235) Median 30 18 7 N 4174 463 183 Costs for Home Health Mean 729 365 96 Care, Special Services (SD) (6,052) (1,956) (565) Median 0 0 0 N 947 138 52 Insurance Premiums Mean 2,682 1,534 1,308 (SD) (3,085) (1,520) (1,398) Median 2,208 1,100 984 N 4418 259 97 Total Costs Mean 2,902 2,482 2,257 (excluding insurance) (SD) (8,246) (5,908) (5,005) Median 3,000 600 290 N 5017 526 209 Total Costs Mean 5,126 3,194 2,744 (including insurance) (SD) (8,747) (6,103) (5,058) Median 3,380 1,284 700 N 5072 522 214 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 156 Multivariate Analyses Results of multivariate analyses are shown in Tables 7.16 to 7.19. Because the race variables are the main independent variables of interest in this chapter, Tables 7.16 and 7.17 show only the regression coefficients for race in each of the four hierarchical models. This allows examination of changes in the effect of race on out-of-pocket costs for each service as other control variables are added. Tables 7.18 and 7.19 show the full hierarchical model including all control variables for physician/therapist/ER visits in PSID and for ambulatory services in AHEAD. For hospital care and for outpatient surgery in PSID, race was not a significant bivariate predictor of use or of costs among users. Race remained non-significant in each of the hierarchical models for hospital care and outpatient surgery, with one exception: Black hospital users spent significantly less only in model 3, when insurance was controlled. For hospital care in AHEAD, neither Black nor Hispanic race predicted use or costs among users in any of the hierarchical models. Tables 7.17 and 7.19 show regression results for physician/therapist/ER visits in PSID, and for physician/ER/surgery/dental care in AHEAD. As seen in both Tables, race was a significant predictor of lower costs until socioeconomic variables were added in the fourth hierarchical model. When diseases were controlled in model 3 for PSID, the effect of race was reduced but remained significant at the .05 level. The results for prescription medications are shown in Table 7.16 for both PSID and AHEAD. In PSID, Blacks spent significantly less on prescription medications in all models, although the significance level dropped from .001 to .007 when insurance Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 157 variables were added, and to .03 when socioeconomic variables were added. In AHEAD, Blacks spent significantly less than Whites on prescription medications until socioeconomic variables were added in model 4. Hispanics in AHEAD had significantly lower out-of-pocket costs for prescription medications only in models 1 and 2, before insurance and socioeconomic variables were controlled. Regression results for nursing home care, dental care, and equipment in PSID, are shown in Table 7.17. For dental care, Blacks had significantly lower odds of using care than Whites, but there was no significant race difference in costs among users1 Blacks had significantly lower odds of using equipment (eyeglasses, hearing aids, medical equipment, home modifications) relative to Whites, but only in the first model, before health was controlled. Among users of equipment, Blacks had significantly lower costs in the first two models, until insurance was controlled. For nursing home care in PSID, Blacks had significantly lower odds of having a nursing home stay than Whites in all models except the second, where health was controlled but insurance and socioeconomic variables were not. Among those with a nursing home stay in the previous year, race was not a significant predictor of out-of- pocket costs. For home care in PSID, race was not a significant predictor of use or of out- of-pocket costs among users. In AHEAD, results for home health care / other special services differed for Blacks and Hispanics. Blacks were significantly more likely to use L The finding for costs among users is consistent with that found by Kington et al. (1995) using the PSID Elderly Health Supplement. However, the finding of lower odds of use for Blacks is not consistent with Kington et al. (1995), who did not find a significant race effect for use. This is likely due to differences in the exact variables, sample, and weights used. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 158 these services than Whites, but only in model 1, before health was controlled. Black race did not significantly predict the level of out-of-pocket costs among users. In contrast, Hispanics did not differ significantly from Whites in their use of home health care/special services, but costs among users were lower for Hispanics in the first two models, before insurance was controlled. For total out-of-pocket health care costs excluding the cost of insurance premiums, Blacks spent significantly less than Whites in all four models in PSID. In AHEAD, Blacks spent significantly less than Whites in the first three models, but this difference was not significant in the fourth model, when socioeconomic variables were controlled. For Hispanics in AHEAD, out-of-pocket costs were lower than for Whites in all four models. For out-of-pocket costs including the cost of insurance premiums, out-of- pocket costs were lower for Blacks in PSID and for Blacks and Hispanics in AHEAD in all four models. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 159 Table 7.16 Hierarchical OLS Regression Results for Out-of-Pocket Spending on Prescription Medications and Total Out-of-Pocket Spending (in Logged Dollars) by Race in PSID and AHEAD Service Type Race Model 1 Model 2 Model 3 Model 4 Prescription Medications PSID Black -0.87*** -1.40*** -0.76** -0.59* (N = 970) R2 0.036 0.158 0.219 0.226 AHEAD Black -0 39*** -0.61*** -0.25** -0.08 (N = 6085) Hispanic -0.71*** -0.81*** -0.13 0.14 R2 0.012 0.136 0.213 0.222 Total Costs excluding insurance premiums PSID (N = 886) Black R2 -1.78*** 0.087 -2.00*** 0.139 -1.33*** 0.205 -1.02*** 0.240 AHEAD Black -1.06*** -1.24*** -0.63*** -0.23 (N = 5878) Hispanic -2.05*** -2.12*** -1.08*** -0.43* R2 0.032 0.082 0.145 0.175 Total Costs including insurance premiums PSID (N = 845) Black R2 -2.05*** 0.124 -2.14*** 0.136 -1.06*** 0.352 -0.89*** 0.368 AHEAD Black -1.75*** -1.78*** -0.68*** -0.40*** (N = 5858) Hispanic -2.58*** -2.56*** -1.03*** -0.58*** R2 0.087 0.101 0.317 0.338 * E < .05 ** £ < .01 *** e < .001 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Table 7.17 Hierarchical Two-Stage Regression Results for Use and Out-of -Pocket Spending (in Logged Dollars) by Race in PSID and AHEAD: Hospital, Nursing Home, Outpatient Surgery, Dental, Equipment, and Home Care Model 1 Model 2 Model 3 Model 4 Service Data Set Race* Any Log costs Any Log costs Any Log costs Any Log costs Type use among users use among users use among users use among users probit (O L S) probit (O L S) probit (O L S) probit (O L S) Hospital PSID Black 0.16 -1.15 0.21 -1.09 0.03 -1.00 0.01 -0.93 (N = 1002,195 users) LL/R2 -484 0.066 -422 0.081 -416 0.145 -416 0.146 Nursing PSID Black -0.64 -2.76 -0.89* -4.83 -1.04* -4.48 -1.05* -5.06 Home (N = 1013,41 users) LL/R2 -145 0.101 -120 0.412 -115 0.764 -107.1 0.805 Hospital, AHEAD Black 0.10 -0.03 -0.03 -0.02 -0.05 -0.11 -0.02 0.08 Nursing Hispanic -0.00 -0.72 -0.09 -0.08 -0.18 i O oo -0.13 -0.05 Home (N = 6107, 2162 users) LL/R2 -3914 0.02 -3404 0.06 -3388 0.10 -3385 0.11 Surgery PSID Black -0.21 -0.48 -0.22 -0.45 -0.11 -0.35 i O -0.57 (N = 1015,128 users) LL/R2 -377 0.019 -374 0.080 -371 0.148 -369 0.155 Dental PSID Black -1.00*** -0.66 -0.89*** -0.67 -0.88*** 0.15 -0.40** 0.29 (N = 1014, 391 users) LL/R2 -645 0.017 -631 0.017 -621 0.105 -573 0.121 (Table continues) O s O Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission. (Table 7.17 continued) Model 1 Model 2 Model 3 Model 4 Service Dala Set Race+ Any Log costs Any Log costs Any Log costs Any Log costs Type use among users use among users use among users use among users probit (O L S) probit (O L S) probit (O LS) probit (O L S) Equipment PSID Black -0.26* -0.91** -0.22 -0.71* -0.16 -0.40 -0.10 -0.17 (N = 1007, 330 users) LL/R2 -642 0.043 -632 0.101 -630 0.149 -629 0.179 Home PSID Black 0.25 -0.99 0.16 -1.69 0.15 -1.42 0.23 -1.57 Care (N = 1017,48 users) LL/R2 -191 0.051 -168 0.428 -165 0.489 -161 0.627 Home AHEAD Black 0.19** -0.31 0.00 -0.02 -0.01 -0.16 -0.00 -0.03 Care Hispanic 0.18 -0.90* 0.05 -0.80* -0.03 -0.59 -0.01 -0.32 (N = 5891,1138 users) LL/R2 -2776 0.02 -2298 0.05 -2283 0.06 -2275 0.07 +White is the referent group *LL = Log Liklihood * E < .05 ** £ < .01 *** e < -001 162 Table 7.18 Hierarchical OLS Regression Results for Out-of-Pocket Expenditures on Physician/Therapist/ER Visits in PSID ( in Logged Dollars) (n = 963) Model 1 Model 2 Model 3 Model 4 Black -1.02*** -1 13*** -0.60* -0.27 Age -0.02 -0.02 -0.02 -0.01 Female 0.07 0.02 0.15 0.28 State Spending -0.00 -0.00 -0.00 -0.00 Urban 0.40 0.39 0.36 0.30 Self-Rated Health -0.09 -0.20* -0.28** Disability -0.03 -0.02 -0.02 Cancer 0.12 0.10 0.09 Diabetes -0.41 -0.44 -0.39 High Blood Pressure 0.51** 0.47** 0.45** Heart Problems 0.22 0.27 0.30 Stroke -0.29 -0.36 -0.30 Lung Disease 0.18 0.11 0.21 Digestive Problems 0.20 0.12 0.10 Arthritis 0.06 0.05 0.03 FFS Medicare Only -0.77** -0.39 Employer Insurance 0.15 0.07 Pre-Paid Health Plan -1 48*** -1.48*** Medicaid -2.22*** -1.25** Missing on Insurance -0.49* -0.26 Education 0.07* Household Income 0.21 Household Wealth 0.09** Missing on Wealth -0.44 R2 0.027 0.048 0.111 0.139 * E < -05 **e< .01 *** £ < .001 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 163 Table 7.19 Hierarchical OLS Regression Results for Out-of-Pocket Expenditures on Physician/Therapist/ER Visits in AHEAD (in Logged Dollars) (n = 5936) Model 1 Model 2 Model 3 Model 4 Black -1.07*** -0.99*** -0.48*** -0.01 Hispanic -1.82*** -1.73*** -0.82*** -0.01 Age -0.06*** -0.05*** -0.05*** -0.03*** Female -0.14 -0.13 -0.03 0.06 Regional Spending 0.00 0.00 0.00 0.00 Urban -0.05 -0.07 -0.1 -0.20* Self-Rated Health 0.11** 0.08* 0.02 Disability -0.15*** -0.08** -0.08** Bed Days -0.46* -0.4 -0.4 Depression 0.01 0.01 0.02 Cancer 0.30** 0.26** 0.17 Diabetes -0.03 0.04 0.06 High Blood Pressure 0.00 0 0.03 Heart Attack 0.04 0.12 0.15 CHF 0.1 0.15 0.17 Other Heart Problems 0.21* 0.21* 0.21* Stroke 0.07 0.15 0.17 Lung Disease -0.48*** -0.39*** -0.34** Arthritis 0.25** 0.30*** 0.35*** Medicaid -2.29*** -1.47*** Employer Insurance 0.32** 0.26* Pre-Paid Health Plan -0.48*** -0.45*** FFS Medicare Only -0.15 0.07 Missing on Insurance -0.84** -0.49 Education 0.12*** Household Income 0.10** Household Wealth 0.13*** R2 0.041 0.057 0.105 0.143 * E < -05 **£< -01 *** £ < -001 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 164 Discussion A main goal of this chapter was to examine whether minority groups report higher or lower out-of-pocket health spending than Whites, overall and for different service types. Descriptive analyses revealed that although race differences varied in magnitude across service type, Blacks and Hispanics had lower out-of-pocket costs overall and for each service type. With some exceptions, out-of-pocket expenditures among those using each service were also lower for Blacks and Hispanics than for Whites. A second main research question was whether the burden of out-of-pocket spending as a portion of household income was higher or lower for minority elders. Because of the lower average income among minority groups, it was hypothesized that even if Blacks and Hispanics had lower out-of-pocket costs, their level of burden might be higher. However, this hypothesis was generally not supported. There was some support for this hypothesis in AHEAD, where the mean burden for total out-of-pocket costs excluding insurance was slightly higher for Blacks and Hispanics, and minority groups were somewhat more likely to have overall costs that could be considered catastrophic. However, the median proportion of income spent on total out-of-pocket costs excluding insurance was lower for minority groups than Whites in both survey samples. For total costs including insurance premiums, these race differences were larger, with even lower mean and median burden levels for minorities compared to Whites. These results suggest that out-of-pocket liabilities for health care among older minority groups might not be an important policy concern. However, it is important to consider that measuring out-of-pocket costs as a portion of income may not accurately Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 165 reflect burden and ability to pay. African Americans report more problems than Whites in paying for basic needs such as clothing and housing, are more likely to report difficulties covering medical expenses. They also report more worries than White Americans that they will be unable to meet these needs in the future (Blendon et aL, 1995). The amount of money left after spending a certain proportion of income on health care will be less for those with low incomes than for those with high incomes, and may not be adequate to cover other necessities. Also, as discussed in Chapter 5, burden was calculated at the individual level using household rather than couple or individual income, which may have underestimated burden. This underestimation may have been greater for minority elderly, who are more likely than Whites to live in extended households, and might have appeared to have more income than was actually available for their care. However, in analyses of burden using couple income (not shown), the race difference was narrowed, but burden remained lower among Blacks. It is important to consider that low income can serve as a barrier to accessing appropriate health care, and that the lower out-of-pocket costs among minority groups in this study may reflect access barriers. Several of the results in this chapter suggest that older Blacks and Hispanics face more barriers in access to appropriate care than Whites. First, consistent with other studies, there was support for the hypothesis that minority groups would have lower use and costs for services considered more discretionary. Blacks and Hispanics in both survey samples were less likely then Whites to report seeing a dentist, having outpatient surgery, or seeing a physician. The race difference in physician use was smaller in AHEAD than in PSID, which likely reflects the fact that the question regarding physician visits in AHEAD Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 166 also included visits to the emergency room. In PSID, separate questions asked about physician, therapist and ER use, and Blacks were more likely than Whites to report an ER visit, which is a less discretionary type of care. Use of equipment, which also falls near the discretionary end of the continuum, was reported less frequently by Blacks than Whites in PSID. Use of prescription medications, which falls near the middle of the discretionary continuum, was similar across race in both survey samples. Use of hospital care, which is less discretionary, was higher for Blacks than Whites in both surveys samples, which is consistent with previous findings that Blacks delay seeking care until their condition necessitates it. Also consistent with this pattern is the finding that although a higher number of Blacks than Whites reported no physician contact, those who did use physician services appeared to have conditions that required more care, as evidenced by the higher mean number of reported physician/therapist visits was higher among Blacks in PSID. Another hypothesis for use by race was that Blacks and Hispanics would use less long-term care than Whites, reflecting greater reliance on family caregivers. This hypothesis was supported for nursing home use by Blacks in PSID, and by Hispanics in AHEAD. However, Blacks in AHEAD had slightly greater use of nursing home care than Whites. Also, home care use was greater for Blacks than Whites in PSID, and both Blacks and Hispanics used more home health care than Whites in AHEAD. This is likely due at least in part to poorer health among minority groups. When health was controlled in multivariate analyses, Blacks no longer had higher odds of using home health care in AHEAD. Minority race was not a significant predictor of home care use for Hispanics in AHEAD, or for Blacks in PSID in any of the hierarchical models. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 167 Patterns of mean out-of-pocket health expenditures for different services by race were not as consistent with the discretionary hypothesis as patterns of use. If out-of- pocket costs were directly proportional to use, then minority groups would be expected to have lower costs for discretionary services, and costs similar to or higher than Whites for less discretionary hospital care. However, mean spending over all respondents was in fact lower among minority groups for all service types. Race differences were generally larger for more discretionary services: in PSID, Blacks had much lower mean costs than Whites for outpatient surgery and for dental care, which are more discretionary. Also, median costs were lower among Blacks for physician visits, which are considered more discretionary, and for prescription medications, which are considered to be near the middle of the discretionary continuum. In PSID, the race difference in mean spending was smaller for hospital care, which is considered less discretionary. However, this finding was not replicated in AHEAD, where hospital costs were much lower for Blacks than Whites and even lower for Hispanics than for Whites, despite similar rates of use. The third research question asked about the extent to which race differences in expenditures appeared to be due to differences in use versus costs among users. As discussed above, Blacks and Hispanics reported lower use of most services than Whites, which explains at least in part the race differences in out-of-pocket costs across all respondents. As for costs among users, an additional hypothesis related to access was that if minority elders who used services had delayed seeking care and were in poorer health than Whites, they might require care that was more intensive and costly, leading to higher out-of-pocket costs. However, this hypothesis was generally not supported, since among Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 168 those using each service, Blacks and Hispanics had lower out-of-pocket costs than Whites for most services. This could have reflected superior coverage by Medicaid among minority groups. There is some support for this in both univariate and multivariate analyses. Blacks and Hispanics were more likely than Whites to be covered by Medicaid and more likely to have no uncovered costs overall and for most services used. Hispanics had the highest rates of Medicaid coverage, and the lowest rates of uncovered costs. Also, racial differences in the proportion of users with uncovered costs were smallest for hospital care, which is likely to be covered by supplemental plans. Race differences in the rate of uncovered costs among users were larger for other services such as prescription medications, dental care, and nursing home care, which are typically covered by Medicare but may not be covered by supplemental insurance. Multivariate analyses allowed examination of whether race was a significant predictor of out-of-pocket spending. In multivariate analyses, the race effects appeared to be explained by insurance differences for some services. Among users of prescription medication and home care in AHEAD, Hispanics no longer had significantly lower costs when insurance was controlled. In PSID, Blacks had lower equipment costs conditional on use only until insurance was controlled. However, since supplemental insurance was also controlled, this might also reflect a lower likelihood among minority groups of receiving the type and intensity of services associated with supplemental insurance coverage. It would be an important policy concern if the lower costs among minority users reflect inadequate treatment due to racial bias and other factors discussed in the review above. The lower use and related lower out-of-pocket costs among minority groups is of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 169 particular concern since Blacks and Hispanics have worse average health than Whites, and thus have a greater need for services. If treatment was distributed equally across race in proportion to health, then minority elders would be expected to use more services and have greater expenditures than Whites. However, in these analyses, Blacks and Hispanics had significantly lower odds of use for some services, and spent significantly less overall and on several services even before health was controlled. This means that minority groups used less care and spent less on care even before accounting for their worse health. The fourth research question, addressed using multivariate analyses, was concerned with whether race is an independent predictor of costs, or whether other factors can explain race-related differences in costs. Although findings were mixed, race did not appear to be a strong independent predictor of costs. In each regression model, the coefficients for race either attenuated or became non-significant as other variables were controlled. There were some equations in which race remained significant as a predictor of lower costs even when health, insurance coverage, and SES were controlled, including most equations for total spending, and equations for prescription medication costs and dental care use in PSID. This suggests that other factors related to race that were not controlled were responsible for race differences in costs. These factors, discussed above, include differences in access, cultural dispositions, and the intensity of treatment among users. There were two instances in which race effects became non-significant when health was controlled. One was greater use of home health care by Blacks in AHEAD, which indicates that greater use among this group was due to greater need. The other was lower Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 170 use of equipment by Blacks in PSID. This is difficult to explain, since it suggests that minorities had lower utilization of equipment due to better health. It may be that the measures used to control for health in the equations for equipment did not accurately reflect medical need for the services included in this variable: eyeglasses, hearing aids, medical equipment, and home modifications. As discussed above, controlling for insurance rendered race effects non significant for Hispanics in AHEAD on two health service types: prescription medication and home health care. This reflects greater coverage by Medicaid among Hispanics than Whites. In multiple regression equations for several services, race became non-significant as a predictor of lower costs when socioeconomic variables— education, and income, and wealth— were controlled. This occurred for physician care in both samples and both minority groups, and for prescription medications and total costs (excluding insurance premiums) for Blacks in AHEAD. This is consistent with previous findings that for some measures of health and health service utilization, race differences are eliminated when controlling for SES (Markides & Black, 1996; Mutchler & Burr, 1991). It is also consistent with Gomick et a l (1996), who found that older Black Medicare beneficiaries and low-incorae Whites had similar patterns of service utilization. Education likely had an indirect effect on expenditures through its influence the intensity of service use. Additional analyses (not shown) revealed that controlling for education alone explained race effects for physician care and prescription medications. For total out-of-pocket costs in AHEAD, income and wealth among Blacks likely had a direct influence on out-of-pocket costs by reducing the amount of money available for the purchase of health services after essential Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 171 needs were met. It is also likely that costs were influenced indirectly by a number of factors associated with low SES, such as low functional health literacy, type of neighbourhood, availability of providers, ease of transportation, fear of going out due to crime (LaVeist, 1994). The finding that SES explains some race effects has important policy implications. It suggests that the access problems that are more prevalent among minority groups can be addressed by interventions targeted toward groups with low SES, rather than toward specific minority groups. Of course, it is likely that the higher out-of-pocket costs among Whites reflects some over-use of unnecessary care, and such that it may not be an appropriate goal to raise the level of health care use among minority groups to level of Whites. For example, there is increasing evidence of inappropriate surgery among older Americans (see Escarce et aL, 1993). Also, the lower use of nursing home care among Hispanics is consistent with preferences among older adults to remain in the community and avoid institutionalization, and it may be positive that minority groups may be better able to delay or avoid nursing home placement through care in the community (Cagnee & Agree, 1999). However, with respect to overall service use, the poor health and the health care utilization patterns among minority groups suggests that there is considerable unmet need among these populations. Interventions encouraging the appropriate use of preventive and maintenance care are warranted, and have the potential to increase well being and decrease the financial burden of health costs in the long run by improving health. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 172 Chapter 8 Insurance and out-of-pocket health costs Insurance acts as central mediator between health care use and out-of-pocket liability, and thus warrants careful attention in any examination of out-of-pocket health care costs. Several studies have examined the effects of insurance on health care utilization, and on the total cost of care used, including costs covered by all payers. However, fewer studies have looked at the out-of-pocket costs incurred by those with different types of coverage in addition to Medicare. This section describes each type of insurance coverage, reviews the literature regarding health, health care use, and costs in relation to each insurance type, and puts forth policy questions to be addressed in analyses. Because the Medicare program does not provide full protection against out-of- pocket costs, the majority of Medicare beneficiaries have some form of additional insurance coverage. The most prevalent form of additional coverage is private supplemental insurance, either purchased individually or provided through an employer as part of a retirement plan. Another type of additional coverage is the Medicaid program, which is available to older adults with income.and wealth below specified levels. Other beneficiaries receive their Medicare benefits through a pre-paid health plan, which often covers more than traditional Medicare. Supplemental Insurance Two main types of supplementary insurance are available to Medicare beneficiaries: individually-purchased plans, and employer-sponsored plans that are Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 173 provided as a retirement benefit. These two types of coverage are described in detail by Chulis et aL (1995), and are briefly reviewed here. Individually-purchased plans, termed ‘medigap’ plans, have been designed to cover costs not covered by the Medicare program. Medigap plans vary premium costs, and in the extent and types of supplemental protection provided. Under federal law, all medigap plans sold since 1992 have had to offer one of 10 standardized benefit packages. Core benefits include coverage of the 20 percent coinsurance for Part B services, and the coinsurance for hospital stays greater than 60 days. Other optional benefits include nursing home coinsurance for stays over 20 days, the hospital and Part B deductibles, balance billing under Part B, preventive medical care, at- home recovery services, and prescription drugs. Different medigap plans have been found to vary widely in the extent of out-of-pocket protection they provide (Sofaer & Davidson, 1990). In addition to medigap plans, beneficiaries may purchase private insurance that covers a specific health good or service, such as prescription medications, dental care, or long-term care. Employer-sponsored supplemental insurance coverage is often a continuation of the health benefits provided prior to retirement. When a retiree becomes entitled to receive Medicare benefits, Medicare becomes the primary payer, and the employer decides how to coordinate its coverage with Medicare benefits. Employer-sponsored plans have traditionally provided more generous coverage than medigap plans, and employers often subsidize or pay the premium on behalf of the retiree. However there is evidence of reduced employer support for retiree health benefits since the early 1990's, with beneficiaries required to pay a larger portion of the premium (see Chulis et aL, 1995; Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 174 Lillard, Rogowski & Kington, 1997). In the 1992 MCBS, 37 percent of beneficiaries had medigap coverage, and 35 percent had employer-sponsored coverage. Seven percent of beneficiaries had private insurance in addition to employer-sponsored coverage, either to fill gaps in the employer- coverage, or because of uncertainties regarding continued employer coverage (Chulis et aL, 1995). About 20 percent of beneficiaries held more than one private supplemental policy, often to fill specific gaps in coverage that were not covered by their main medigap policy, such as prescription medications. Because the purpose of health insurance is to reduce out-of-pocket health expenditures, it might be assumed that those with supplemental insurance would have lower out-of-pocket costs. However, those with insurance tend to use more services than those without insurance. In the 1991 MCBS, higher levels of insurance coverage were associated with greater use and higher Medicare program payments on behalf of beneficiaries (Chulis, Eppig, Hogan, Waldo, & Arnett, 1993). In the 1994 National Health Interview Survey (NHIS), the use of health services by beneficiaries with medigap plans was 28 percent higher than for those with Medicare only, and the use of those with employer-sponsored insurance was 17 percent higher, controlling for health and sociodemographic factors (Christensen & Shinogle, 1997). In a study on health care use for arthritis in the 1992 MCBS, insurance coverage was a positive predictor of both initial use and the amount of physician care used for arthritis, controlling for other factors including severity of illness and comorbidities (Grana & Stuart, 1996). Blustein (1995) found that those with supplemental insurance coverage were much more likely to have Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 175 mammograms than those without insurance, even after Medicare began covering mammogram screenings. In the 1995 MCBS, those with supplemental prescription coverage were more likely to use prescription drugs and had a greater average number of prescriptions than those without such coverage (Davis et aL, 1999). Similarly, in the 1990 Elderly Health Supplement to the PSID (used here), Lillard et aL (1999) found that having insurance that covered prescription drugs was a significant positive predictor of the likelihood of any drug use. Consistent with this higher health care use, those with supplemental insurance incur higher total health costs. Among beneficiaries in the MCBS from 1991 to 1993, those with supplemental insurance incurred significantly higher costs to the Medicare program than those without supplemental or Medicaid coverage. Controlling for health, demographic, and socioeconomic variables, predicted Medicare spending for those with medigap was 15 percent higher than those with FFS Medicare only. Predicted spending for those with employer-sponsored insurance was 23 percent higher, and for those with both private and employer insurance was 32 percent higher than among those with FFS Medicare only (Khandker & McCormack, 1999). There is also evidence that more generous supplemental insurance increases use the most, with highest service use and total costs incurred for those with first-dollar coverage (McCaU, Rice, Boismier, & West, 1991). The greater use of services by those with insurance is partly due to moral hazard; demand for medical care is somewhat elastic, increasing when the price is reduced due to insurance (Manning et aL, 1988; Rubin & Koelln, 1993a). There is also evidence of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 176 adverse selection; those who anticipate needing more services or who have more favourable attitudes toward using health care are more likely to buy insurance (Rubin & Koelln, 1993a). Wolfe and Goddeeris (1991) found evidence of both moral hazard and adverse selection due to poor health among older adults purchasing medigap insurance. Hurd and McGarry (1997) found evidence of moral hazard but did not find support for adverse selection in the first wave of the AHEAD study. Those with supplemental insurance were more likely to see a physician, more likely to stay overnight in the hospital, and had longer hospital stays than those without insurance. However, those in poor health were less likely to be covered by supplemental insurance, which provides evidence of favorable rather than adverse selection. Lillard et aL (1997) also failed to find evidence of adverse selection for supplemental insurance in the PSID elderly health supplement. However, there may have been adverse selection based on preferences for service use that were not measured in these studies (Hurd & McGarry, 1997). For those with employer-provided coverage, adverse selection is less viable as an explanation for greater service use, because beneficiaries may not have chosen to purchase such coverage if it had not been provided for them. However, employer-provided plans tend to be more generous than medigap plans, which may increase health care use. While the term ‘moral hazard’ has negative connotations suggesting over-use of care, it is difficult to determine the extent to which those with insurance are over-using care, and the extent to which differences in use reflect barriers in access among those without supplemental coverage (Rowland & Lyons, 1996). An additional factor that may explain the greater service use among those with Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. supplemental insurance is its relation to socioeconomic status. There is a strong wealth effect in the purchase of supplemental insurance (Hurd & McGarry, 1997; Lillard et aL, 1997; Wolfe & Goddeeris, 1991), with those reporting higher income and wealth more likely to purchase medigap coverage. This suggests that Medicare beneficiaries purchase insurance primarily because they can afford to and wish to protect their wealth. Those with employer-provided insurance also tend to be better of economically, since retiree health benefits are usually provided in conjunction with a private pension, and both are associated with good jobs (Hurd & McGarry, 1997; Lillard et aL, 1997). Because those with supplemental insurance are better off financially, they may use more services because they can better afford the out-of-pocket costs for service not fully covered by their insurance. Also, education level is closely associated with income and wealth, and may increase health care use, as discussed in Chapter 2. Because there are competing determinants of out-of-pocket costs among those with supplemental insurance, it is difficult to predict the direction of the relationship between insurance coverage and out-of-pocket health spending. Those with supplemental insurance have better protection against out-of-pocket costs and may be healthier, but tend to use more services. Few studies have examined the out-of-pocket costs of those with supplemental insurance compared to other Medicare beneficiaries. The American Association of Retired Persons and The Lewin Group (1997) estimated that Medicare beneficiaries with supplemental coverage had slightly lower out-of-pocket costs (excluding insurance premiums) than those with Medicare only ($1,121 compared to $1,273). In the 1977 NMCES, Cartwright et aL (1992) found that having medigap coverage increased the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 178 probability of having out-of-pocket medical expenses, but did not significantly increase the amount spent among those with costs, controlling for health, demographic, and socioeconomic variables. In contrast, Rubin & Koelln (1993a, 1993b) found that elderly households with supplemental insurance had significantly higher out-of-pocket health expenditures (excluding insurance premiums) than those who were uninsured, controlling for age, race, education, income, assets, housing tenure, and family size. Lillard et aL (1999) found that among users of prescription medications in the PSID, those with insurance that covered medications had significantly lower out-of-pocket drug costs, controlling for other factors. When the cost of insurance premiums are counted as part of older adults’ out-of-pocket costs, the out-of-pocket liability of those with insurance is very likely to be higher than for those with no supplemental insurance or with employer- provided coverage. When out-of-pocket costs included insurance premiums in the 1997 study by the AARP Public Policy Institute and The Lewin Group, those with supplemental insurance were projected to incur higher out-of-pocket costs ($2,610 compared to $1,735 for those without supplemental coverage). In order to improve our understanding of the relationship between supplemental coverage and out-of-pocket costs, this chapter gives a more detailed and comprehensive picture than previous studies of out-of-pocket health spending by those with medigap insurance and those with employer-provided insurance compared to other insurance groups. In addition to comparing total out-of-pocket costs excluding and including insurance premiums, expenditures for eight different health goods and services are examined. This gives a picture of how insurance coverage affects the use and costs of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 179 different service types, and for which service types differences are greatest. Additional comparisons among only those who used each service indicate the effects of insurance coverage on the intensity of services used. The burden of out-of-pocket costs as a portion of income is also examined for those with and without supplemental coverage, both excluding and including the cost of insurance premiums. Variables indicating insurance coverage specifically for dental care, prescription medications, or long-term care are also included as predictors of out-of-pocket spending on these services and on total out-of- pocket costs. Prepaid Health Plan Membership The majority of Medicare beneficiaries are in a fee-for-service (FFS) system, under which physicians, hospitals and other providers submit bills to the Medicare program and are reimbursed for each service they provide. Beneficiaries are billed by providers for the portion of costs not covered by Medicare, and may in turn be reimbursed for these costs if they are covered by supplemental insurance. However, an alternative option available to beneficiaries is to receive their benefits through a prepaid health plan, or Medicare HMO. Those who live in the service area of one or more plans that accept Medicare risk- contracts can enroll in a plan, and must receive care from providers authorized by the plan. In 1997, 12% of beneficiaries were enrolled in Medicare HMOs (Schoen et aL, 1998). For each beneficiary in the plan, the Medicare program pays the plan a set monthly fee equal to 95% of the average cost of care for FFS beneficiaries in that region (adjusted for demographic factors). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 180 Prepaid plans are required to provide beneficiaries with all health services covered by Medicare, and may charge premiums in addition to Medicare Part B, and deductibles, and co-pays equal to those in the FFS program. However, in order to attract beneficiaries, many plans set these fees at a lower level than in ITS Medicare or waive them completely. Also, plans often provide coverage for goods and services not covered by FFS Medicare, such as preventive care, outpatient prescription drug coverage, mental health benefits, extended hospital stays, and dental care (Sofaer & Kenney, 1989). As a result, beneficiaries in prepaid health plans have been found to have lower out-of-pocket costs than those in FFS Medicare (AARP Public Policy Institute & The Lewin Group, 1997; Crystal et aL, 1998; Rubin & Koelln, 1993b; Sofaer and Kenney, 1989). There is also evidence that enrollees in prepaid plans use fewer health services (see Sofaer & Kenney, 1989). In the 1994 NHIS, FIMO enrollees used fewer services than those with supplemental insurance and those with Medicare only (Christensen & Shinogle, 1997). Goldzweig et aL (1997) found that older Medicare beneficiaries in a fee-for-service setting were twice as likely to have their cataracts removed as beneficiaries in two prepaid health plans, controlling for age, gender, and diabetes. The lower use of care by enrollees in prepaid plans has fueled concerns about access and appropriate care in prepaid settings. Because prepaid plans receive a set fee regardless of the amount of care they provide, there may be an incentive to underprovide care. However, it is important to consider the health of beneficiaries when examining differences in use. There is general agreement that those who join plans are healthier than those who remain in FFS Medicare. Cox & Hogan (1997) found that those who enrolled in Medicare HMOs used 37 percent fewer services Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 181 in the six months prior to their enrollment than those who remain in FFS Medicare, and that those who disenrolled from Medicare HMOs used 60 percent more services in the 6 months after disenrollment than others in FFS Medicare. Cox & Hogan (1997) also found that new enrollees had lower expenditures for chronic conditions prior to enrollment, and that disenrollees had higher expenditures compared to those who remained in FFS. Similarly, a recent GAO study (Government Accounting Office, 1997) found that elderly Medicare beneficiaries enrolled in HMOs had fewer and less costly chronic conditions than those who remained in fee-for-service Medicare, and that those who disenrolled had more chronic conditions and higher costs than those who remained. Better health among HMO enrollees means that they have less need for expensive health services, and health differences appear to at least partially explain their reduced service use and lower costs.1 However, most studies finding lower out-of-pocket costs among Medicare beneficiaries in prepaid plans have not controlled for health differences. The use of multivariate analyses in this study allows an examination of whether membership in a prepaid plan is a significant predictor of out-of-pocket costs when controlling for beneficiaries’ health. Another important consideration is that while overall service use is lower in prepaid plans, use of some types of care may be greater while use of other types may be lower than in FFS. Because prepaid plans are at risk for the health expenses of their 1 This favorable selection in HMOS contributes to plans’ ability to offer more generous coverage. Many agree that as a result of favorable selection into prepaid health plans, the reimbursement rate for prepaid plans is too high, allowing plans to profit more than intended (Brown, Bergeron, Clement, Hill, & Retchin, 1993; Dowd, Christianson, Feldman, Wisner, & Klien, 1992). A government response to this apparent problem of overpayment was to require plans that receive payments in excess of their projected costs to use a portion of profits to increase benefits. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 182 members and cannot disenroll those who become ill, plans have an incentive to provide preventive care and disease management in order to keep beneficiaries healthy. Thus, ideally, those in prepaid plans should have greater use of services that are more discretionary, and lower use of less discretionary services. This treatment pattern was seen among those enrolled in HMOs in the 1994 NHIS; compared to those with FFS Medicare only, HMO enrollees had more outpatient but fewer inpatient hospital days (Christensen & Shinogle, 1997). Prepaid plans may also have greater use of cost-effective treatments for beneficiaries who do become ilL For example, Merril et aL (1999) recently found superior overall survival among Medicare beneficiaries in HMOs compared to those in FFS Medicare, and concluded that this may have been partially due to increased cancer screening and greater use of adjuvant therapies in HMOs. The current study allows in- depth analysis of differences in service use and associated out-of-pocket costs for those in prepaid plans compared to those with other types of coverage. Two studies examining the out-of-pocket costs of those in prepaid plans costs concluded that their reduced out-of-pocket costs appear to be primarily due to lower premiums and lower costs for care rather than lower use of services. AARP & The Lewin Group (1997) estimated that most of the difference in out-of-pocket costs between HMO enrollees and FFS beneficiaries not enrolled in Medicaid was due to lower premiums for those in prepaid plans. HMO enrollees were projected to spend about $550 less in private insurance premiums and $145 less on health goods and services than FFS beneficiaries. Sofaer and Kenney (1989) estimated that out-of-pocket costs for treatment of 13 different illnesses would be much lower for enrollees in two Los Angeles HMOs than for those in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 183 traditional Medicaid (with no supplemental coverage). Premium costs were not considered, and lower costs for enrollees were attributed primarily to the elimination of hospital deductibles, the reduction or elimination of co-payments, and coverage of prescription medications. The chapter examines out-of-pocket costs for those in prepaid plans overall and for users of each service, indicating the extent to which differences in costs are related to differences in use. Medicaid Older adults with income and wealth below a specified low level qualify for Medicaid coverage. Medicaid is a joint federal-state program that covers the health care costs of certain groups of poor persons, including mothers and children, and the elderly. For older adults who are covered by both Medicare and Medicaid (called dual-eligibles), Medicare remains the primary payer. Medicaid pays for Medicare premiums, co-payments, and deductibles, and for health goods and services not covered by Medicare, including outpatient prescription medications and extended nursing home care. The exact package of benefits covered by Medicaid and the generosity of coverage differs from state to state, with some states implementing more rigid prescription limits, and some states covering additional services such as dental care and community-based alternatives to nursing home care. Because Medicaid provides first-dollar coverage, dual-eligibles would be expected to have lower out-of-pocket costs than other beneficiaries. Cartwright et aL (1992) found that dual-eligibles who used hospital and/or physician care spent significantly less than Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 184 others on these services combined, controlling for health and sociodemographic variables. Total out-of-pocket health costs measured in the MCBS were projected to be lower in 1997 for dual eligibles than for other insurance groups (AARP & The Lewin Group, 1997). However, this comparison was for all goods and services combined, including the cost of premiums for private insurance and Medicare Part B. This may have masked differences in out-of-pocket costs for specific health goods and services that are not covered in all state Medicaid programs. For example, coverage for dental care under Medicaid is uneven across states, and among users of dental care in the Elderly Health Supplement to the PSID, out-of-pocket dental expenditures among users were higher for those estimated to be eligible for Medicaid (Kington et aL, 1995). Two other factors affecting out-of-pocket costs for Medicaid beneficiaries are health and access to care. Those dually-eligible for Medicare and Medicaid have poorer average health than other Medicare beneficiaries (Rowland & Lyons, 1996; Schoen et aL, 1998), and incur higher costs to the Medicare program than all other insurance groups when health status is not controlled (Khandker & McCormack, 1999). Dual eligibles also report more difficulties in obtaining needed care. Schoen et aL (1998) found that older Medicaid beneficiaries were nearly three times more likely to report difficulties getting care or problems obtaining specific services than those with private supplemental insurance. As discussed in the chapter on race, there is evidence that those with low socioeconomic status delay or have difficulty getting care until their physical condition necessitates it. Thus, Medicaid beneficiaries may have lower use of more discretionary types of care, and higher use of less discretionary services. Those who do use care are Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 185 likely to be in poorer health than non-Medicaid users, and require more expensive care. However, because of the protection provided by Medicaid, differences in out-of-pocket costs as a function of use and of health may only occur for goods and services for which Medicaid coverage is incomplete. Even when the level of out-of-pocket spending is lower for Medicaid beneficiaries than for other groups, the burden of these costs relative to income may be greater for dual-eligibles given their low levels of income. The data used in this study allows comparison of the out-of-pocket costs and out- of-pocket burden of dual-eligibles compared to other insurance groups. Comparisons are made for eight different health goods and services as well as for total health costs, including and excluding premiums for private insurance. This will indicate the extent to which Medicaid is protecting beneficiaries as intended, and identify the specific areas in which Medicaid beneficiaries are vulnerable to high or burdensome out-of-pocket costs. Analyses that may help to identify access difficulties among dual-eligibles are a comparison of rates of use for each service type, and examination of out-of-pocket costs for all beneficiaries as compared to only those using each service. Multivariate analyses will examine whether Medicaid coverage is a significant predictor of out-of-pocket costs independent of health, income, and demographic factors. Hypotheses for this study This chapter examines out-of-pocket costs overall and for different health goods and services for five different insurance groups: those with individually-purchased supplemental insurance, those with employment-based supplemental insurance, those Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 186 enrolled in prepaid health plans, those dually eligible for Medicaid and Medicare, and those with Medicare coverage only. Consistent with past findings, those with supplemental insurance are expected to use more services than other groups. When the cost of insurance premiums is included, those with supplemental insurance are expected to have higher out-of-pocket costs than other groups. Because premiums are often subsidized for those with employer-sponsored insurance, those with medigap are expected to have the highest insurance costs. However, whether out-of-pocket costs excluding insurance premiums are higher for those with medigap and employer-sponsored insurance will depend on the competing effects of superior coverage and greater use among these groups. If out-of-pocket costs are found to be greater for those with supplemental insurance, then multivariate analyses will allow examination of the extent to which this difference is due to averse selection or to wealth effects. Differences that remain after controlling for demographic variables, health, and SES may be attributed to moral hazzard or reduced access to care among other groups. Those enrolled in prepaid plans are expected to have lower out-of-pocket spending than those with supplemental insurance, even when the cost of premiums is considered. However, due to favorable selection into plans, this effect is likely to be attenuated when health is controlled in multivariate analyses. HMO enrollees are also likely to spend less out-of-pocket cost due to lower service use; those in prepaid plans are expected to have lower use of hospital care, but perhaps greater use of more preventive or discretionary services. Medicaid eligibles are expected to have lower out-of-pocket costs than those with Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 187 other types of insurance (although their costs may be similar to those in HMOs). However, their costs for services not fully covered by Medicaid, such as dental care and prescription medications, may be similar to or higher than those in other insurance groups. Due to the low economic status of dual-eligibles, their out-of-pocket costs may be more burdensome as a portion of income than those of other insurance groups. Use of discretionary services may be lower for Medicaid beneficiaries, reflecting possible barriers in access to care. Measurement of Insurance in PSID Information on insurance coverage was taken from questions in the mail-in survey. Each respondent was categorized as belonging to one of five health coverage groups: Medicaid eligible, enrolled in a prepaid health plan, covered by supplemental (Medigap) insurance, covered by employer-sponsored insurance or no coverage beyond FFS Medicare.2 All respondents were assumed to be covered by Medicare, since 99% of the population 65+ is covered by Part A and 97% is covered by Part B. The mail-in survey included a question on Medicaid eligibility, however this appeared unreliable; many of those who reported Medicaid eligibility as well as insurance had high incomes, which suggests that they were mistaken about their Medicaid eligibility. Thus, an exogenous variable reflecting Medicaid eligibility was created by applying federal eligibility criteria for Supplemental Security Income (SSI) to each individual or couple’s reported income and 2 Sensitivity analyses revealed that results were similar when insurance variables ware coded differently. (For example, when those missing on the question regarding membership in a prepaid plan were set to missing on all types of insurance and excluded from analyses.) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 188 wealth.3 The eligibility criteria for full benefits were used, so that those coded as Medicaid eligible did not include those eligible to receive only partial benefits through the Qualified Medicare Beneficiary (QMB) program. This program requires state Medicaid programs to cover the Medicare Part B premium, as well as deductibles and copayments for Medicare-covered services for those who do meet Medicaid eligibility criteria, but have incomes below the Federal Poverty Level (FPL), and asset levels below twice the eligibility criteria for SSI. This program had just begun in 1989, the year before these data were collected (Carpenter, 1998). Those who reported currently being a member of a pre-paid health plan such as an HMO (and who were not coded as Medicaid eligible) were coded as belonging to a prepaid health plan. Those who reported having insurance that paid “hospital, doctor, or other medical bills not covered by Medicare (such as Medigap),” and who did not report 3 In 1989, 37 states (including the District of Columbia) chose to offer Medicaid eligibility to all recipients of Supplemental Security Income (SSI), and 14 States chose to use more restrictive Medicaid eligibility criteria for low-income aged (Health Care Financing Administration, 1990). In 1989, the federal SSI benefit standard was an income below $4,632 ($386/month) for individuals and $6,948 ($579/month) for couples, and countable resources below $2,000 for individuals and $3,000 for couples (Social Security Administration, 1997). In the determination of SSI eligibility, there are certain exclusions from countable income, such as the first $20/month of Social Security benefits, and exclusions from countable resources, such as a primary residence, a business, or vehicle used for transportation to the doctor, etc. (Most beneficiaries who own a vehicle exclude it for this purpose (personal communication, Pasadena Social Security office, July 1999). The PSID couple income variable was used to calculate income eligibility, after excluding the first $20/month of Social Security benefits for each spouse who received Social Security, and excluding the reported value of SSI benefits for those who received SSI. To calculate wealth eligibility, household wealth information from the 1989 wealth supplement was used. The value of a business and of any transportation was excluded. Wealth was not available at the couple-level, and applying the level of assets allowed for a couple may have led to some respondents being classified as ineligible for Medicaid when they would have in fact been eligible if the wealth of other family members had not been counted. Those who were missing on wealth were classified as though their wealth fell below the level for eligibility. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 189 being eligible for Medicaid or belonging to a prepaid health plan were classified as having supplemental insurance. Those who reported that they obtained this insurance through a current, previous, or spouse’s employer were coded as having employer-sponsored insurance. (Information was not available on the portion of the premium paid by the employer.) Those who reported not being eligible for Medicaid and not being a member of a prepaid health plan or having supplemental insurance were coded as having no insurance (FFS Medicare only).Those who did not report supplemental insurance or membership in a pre-paid health plan (N = 131) and were not eligible for Medicaid were excluded from analyses due to insufficient insurance information. Those who reported that their private insurance or HMO covered prescription medications were coded as having prescription insurance. Those missing on this question were coded as not having prescription insurance. The variables reflecting dental insurance and long- term care insurance were created in the same way. The weighted number of respondents who were classified as having each type of insurance, as well as the proportion missing on insurance, is shown in Table 3.3 in Chapter 3. Measurement of Insurance in AHEAD AHEAD respondents were categorized into the same insurance groups as PSED respondents. However, rather than creating an exogenous Medicaid eligibility variable using wealth and income criteria, survey questions regarding Medicaid coverage were used. Those reporting currently being covered by Medicaid or having Medicaid coverage in the past two years were coded as Medicaid beneficiaries. This variable appeared to be Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 190 reliable since the interviewer described the Medicaid program before asking about coverage, and respondents who indicated that they were covered were asked to provide the number from their Medicaid cards. This variable was self-reported and was not created by applying eligibility criteria beneficiaries’ income and wealth. Thus, in addition to those eligible for full Medicaid benefits, it likely includes those who received only partial Medicaid benefits through the QMB program, described above. It may also include those who had incomes between 100 and 120 percent of the Federal poverty level and had their Medicare Part B premiums paid for by Medicaid through the Specific Low-Income Medicare Beneficiaries (SLMB) program. Those who reported receiving their Medicare benefits through an HMO (and who were not coded as Medicaid eligible) were coded as belonging to a prepaid health plan. Respondents were asked if they had health insurance that paid “any part of hospital or doctor bills (sometimes called a medigap policy)”, and were asked how many insurance plans they had. Those who reported that part or all of the premium for this type of insurance was paid by an employer, union, or spouse’s employer or union, and who were not in a prepaid plan, were coded as having employer-sponsored insurance. All others who reported such health insurance and were not in a prepaid plan were coded as having individually-purchased supplemental insurance. Those who reported not being eligible for Medicaid and not being a member of an HMO or having supplemental insurance were coded as having no insurance (FFS Medicare only). Those who reported more than one supplemental insurance plan were coded as having the type of plan that they mentioned as their primary plan. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 191 Those with insufficient information on insurance coverage were excluded from analyses (N = 124); this included those who were missing on the questions regarding Medicaid coverage and/or supplemental insurance coverage, or who were missing on the HMO question and reported not having Medicaid or supplemental insurance coverage. Those who reported that one of their private insurance plans covered the cost of prescription medications were coded as having private prescription insurance. Those missing on this question were coded as not having prescription insurance. The same procedure was followed for the creation of a dental insurance variable. Those coded as having long-term care insurance were those who reported having, “insurance that specifically covers any part of long-term care, such as personal or medical care in the home of in a nursing home”, aside from government programs. The weighted number of respondents with each type of insurance is shown in Table 3.3 in Chapter 3. Descriptive analyses The five insurance groups were compared on each service type in terms of use, proportion of users with uncovered costs, mean and median spending overall and among users. Insurance groups were also compared on the mean and median proportion of income spent on overall out-of-pocket costs, and on the proportion spending over 10%, over 30%, and over 50% on out-of-pocket costs. Multivariate Analyses Hierarchical regression analysis was performed for out-of-pocket spending on each service type in each survey sample. As discussed in Chapter 5, for variables with a small Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 192 portion of users, a two-part model predicting use and then costs among users was used for each hierarchical equation. For variables on which most respondents had nonzero costs, OLS alone was used for hierarchical equations. In regression analyses, dummy variables are used to represent employer-sponsored insurance, membership in a prepaid health plan, dual-eligibility for Medicaid, and coverage by FFS Medicare only. Since those with privately purchased supplemental insurance were the majority, this group was left out of regression models as the referent group. Thus, multivariate hypotheses and findings are discussed for each insurance group relative to those with supplemental insurance. Three hierarchical models were tested for each service type. In the first hierarchical model, the dummy variables insurance were entered, as well as age, gender, urban residence, and the variable reflecting regional/state Medicare spending. These demographic and geographic variables were included as control variables in case of possible relationships with insurance type, however they were not expected to be important confounding variables. In the second hierarchical model, health variables were added to examine the effects of insurance independent from adverse or favorable selection. In the third model, socioeconomic variables were controlled, to examine the effects of insurance on out-of- pocket costs independent from ability to pay, and from other factors associated with SES that are likely to affect use and costs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 193 Results Descriptive Results Table 8.1 shows the distribution of respondents on independent variables by insurance type in both PSID and AHEAD. In both survey samples, those in prepaid plans and those with employer-sponsored insurance had the youngest mean age, whereas those dually-eligible for Medicaid and those with only FFS Medicare were the oldest. Females were much less likely than males to have Medigap insurance in PSID, and employer- sponsored insurance in AHEAD. Racial minorities were much more likely to be covered by Medicaid or to have only FFS Medicare than to have private insurance. Rural residents were most likely to be covered by Medicaid or medigap, and least likely to be in prepaid plans. Health measures indicated that Medicaid beneficiaries and those with only FFS Medicare had the worst self-rated health and the highest prevalence of medical conditions. This is consistent with previous findings, as is the better health among those in HMOs and those with employer-sponsored insurance. Finally, on socioeconomic measures, mean education and income were highest among those with employer-sponsored insurance, and those with medigap had the highest wealth in both survey samples. Mean income and wealth were lowest among those dually-eligible for Medicaid, reflecting the eligibility criteria for this coverage. The proportion of those using each health service by insurance type is shown in Table 8.2 for PSID and Table 8.3 for AHEAD. In both survey samples, those with only FFS Medicare coverage were least likely to report use of at least one service, although overall rates of using at least one service were quite high among all groups. In PSID, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission. Table 8.1 Means and Percentages on Independent Variables by Insurance Type in PSID and AHEAD Medicare Only Employer Insurance Medigap Insurance Pre-Paid Health Plan Medicaid PSID AHEAD PSID AHEAD PSID AHEAD PSID AHEAD PSED AHEAD (N = 119) (N = 875) (N = 252) (N = 876) (N = 386) (N = 2985) (N = 78) (N = 733) (N = 65) (N = 659) Demoeranhic Mean Age 77.0 80.1 72.7 78.3 75.0 78.9 72.2 78.2 76.7 80.7 Percent Female 66.3 63.7 53.7 55.4 37.2 62.5 60.0 59.0 79.2 76.8 Percent Black 32.0 19.9 8.6 4.9 5.3 4.2 9.1 6.8 51.5 27.1 Percent Hispanic - 4.9 - 0.9 - 0.9 - 5.4 - 15.8 Mean State Spending 3,382 4,804 3,449 4,677 3,375 4,722 3,325 4,831 3,497 4,894 Percent Urban Residents 96.9 77.5 95.9 75.3 92.6 71.6 100 97.1 87.8 70.5 Health Self-Rated Health 2.19 2.93 3 3.07 2.87 3.06 2.9 3.17 2.05 2.31 Activities of Daily Living 2.88 0.92 0.98 0.67 1.23 0.70 0.69 0.69 3.03 1.93 Percent with 10+ sick Days - 4 - 2.3 - 2.4 - 3.7 - 8.1 Cancer 7.9 15.3 4.4 17.5 6.5 17.1 2.3 17.9 5.9 11.3 Diabetes 12.1 16.5 10.9 12.4 11.7 12.1 10.1 15.1 17.6 23.2 Lung Disease 16.6 11.7 13.5 10.6 10.6 12,9 6.7 9.8 9.2 17.7 (Table Continues) 194 Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Table 8.1 (continued) Medicare Only Employer Insurance Medigap Insurance Pre-Paid Health Plan Medicaid PSID AHEAD PSID AHEAD PSID AHEAD PSID AHEAD PSID AHEAD High Blood Pressure 47.7 52.2 39.6 54.3 38.3 53.8 46.5 52.6 42.7 65.7 Neuro Problem / Stroke 11.6 10.2 5.0 9.5 5.9 9.4 3.6 9.2 8.4 20.1 Digestive Problem 19.6 - 14.4 - 17.5 - 17.3 - 13.4 - Heart Attack 27.2 8.4 8.6 8 10.7 8,3 12.4 9 14.4 14.5 CHF - 3.8 - 3.4 - 3.9 - 4.5 - 8.8 Heart Disease - 23.7 - 27 - 23.7 - 22.9 - 24.3 Arthritis 76 52.5 59.9 51.3 60.8 54 55.2 54.4 77.3 72.1 Fall in past 2 years - 31 - 32.4 - 31.4 - 31.9 - 43.2 Depression - 2.04 - 1.8 - 1.74 - 1.73 - 2.61 SES Mean Education 8.7 10.2 12.2 12.1 11.5 11.8 11.8 11.7 6.59 7.4 Mean Household Income 15,167 22,510 29,271 30,619 26,303 28,845 24,241 24,954 6,996 13,865 Mean Household Wealth 75,968 153,095 205,500 220,593 209,180 242,473 162,784 200,228 38,098 31,888 V O U \ 196 those with Medicaid were more likely than those with no insurance but less likely than those with other insurance types to use at least one service. Hospital use was substantially higher for those with Medicaid than for other insurance groups in both PSID and AHEAD. In PSID, hospital use was also higher among those with only FFS Medicare than among those with insurance. Use of outpatient surgery was highest for those with employer-sponsored and medigap insurance in both survey samples, and was particularly low for Medicaid eligibles in PSID. Physician use was lowest for those with only FFS Medicare in both survey samples, and was also lower for Medicaid eligibles than for those with insurance in PSID. Prescription medication use was highest for those in prepaid plans in PSID, and for those covered by Medicaid in AHEAD. Use of dental care was much lower for those with Medicaid and those with only FFS Medicare than for those with insurance in both PSID and AHEAD. Equipment use in PSID was also lowest for Medicaid eligibles and those with no insurance. In both survey samples, use of nursing home and home care was highest for those with Medicaid and lowest for those in prepaid plans. The proportion of those using each service who had uncovered costs is shown by insurance type in Table 8.4 for PSID and Table 8.5 for AHEAD. Those with Medicaid were least likely to have uncovered costs for all services used except hospital and home care, in both survey samples. Those in prepaid plans were least likely to report uncovered hospital costs. A surprising proportion of those reporting no coverage besides FFS Medicare reported no uncovered costs for the services they used, particularly in AHEAD. Possible explanations for this are discussed below. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 197 Table 8.2 Percentage Reporting Use of Each Health Service by Insurance Type in PSID Medicare Only Employer Insurance Medigap Insurance Pre-Paid Health Plan Medicaid Use of Any Services 90.6 96.9 96.3 95.0 93.4 Hospital Use 26.3 16.3 22.0 19.3 35.5 Outpatient Surgery Use 10.2 16.2 14.1 7.2 5.7 Physician/Therapist/ER Use 80.8 90.0 90.4 92.5 81.1 Prescription Medication Use 81.6 82.8 83.8 92.2 87.5 Dental Use 14.6 54.1 45.5 59.9 19.5 Equipment Use 30.5 36.3 36.0 38.0 26.8 Nursing Home Use 14.2 4.3 2.6 1.1 18.0 Home Care Use 10.4 2.6 5.9 1.3 11.2 Table 8.3 Percentage Reporting Use of Each Health Service by Insurance Type in AHEAD Service Type Medicare Only Employer Insurance Medigap Insurance Pre-Paid Health Plan Medicaid Use of Any Services 95.2 99.3 98.2 98.9 98.2 Hospital Use 30.1 33.3 33.9 29.0 51.8 Use of Outpatient Surgery 15.4 23.5 21.7 16.9 15.8 Physician Use 90.7 96.6 95.2 96.6 95.1 Prescription Medication Use 74.0 80.9 79.5 77.2 87.0 Use of Dental Care 41.3 60.3 59.2 57.3 28.7 Nursing Home Use 6.2 4.5 5.0 3.9 18.3 Use of Home Care 11.9 12.4 14.3 10.7 29.4 Use of other Special Services 7.2 8.0 8.5 8.8 21.6 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 198 Table 8.4 Percentage Of Users Reporting Uncovered Costs for Each Health Service by Insurance Type in PSID Medicare Only Employer Insurance Medigap Insurance Pre-Paid Health Plan Medicai d Any Uncovered Costs 85.7 96.4 98.7 89.0 71.5 Uncovered Hospital Costs 58.2 80.0 75.1 22.7 56.4 Uncovered Physician Costs 68.5 80.6 78.3 57.2 35.6 Uncovered Outpatient Surgery Costs 83.3 61.7 64.1 20.7 29.4 Uncovered Prescription Costs 74.8 88.4 96.9 82.6 59.3 Uncovered Dental Costs 79.7 92.9 96.5 85.4 46.9 Uncovered Equipment Costs 66.5 95.3 97.3 85.8 79.0 Uncovered Nursing Home Costs 62.8 100.0 100.0 0.0 52.4 Uncovered Home Care Costs 22.0 42.3 10.0 0.0 5.5 Table 8.5 Percentage Of Users Reporting Uncovered Costs for Each Health Service by Insurance Type in AHEAD* Service Type Medicare Only Employer Insurance Medigap Insurance Pre-Paid Health Plan Medicaid Any Uncovered Costs 82.6 91.4 91.5 87.3 59.7 Uncovered Hospital Costs 46.7 43.8 27.3 16.6 31.5 Uncovered Physician Costs 58.0 63.2 48.7 44.6 19.8 Uncovered Outpatient Surgery Costs 48.4 43.8 32.3 25 17.3 Uncovered Prescription Costs 82.8 82.8 90 83.3 51.7 Uncovered Dental Costs 90.9 89.8 95.6 91.5 58.3 Uncovered Nursing Home Costs 45.3 72.8 52.4 39.3 20.6 Uncovered Home Care Costs 16.0 26.3 13.4 5.1 8.5 *users missing on costs or with costs not settled are not counted as having uncovered costs Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 199 Mean and median out-of-pocket costs for all respondents by insurance type are shown in Table 8.6 for PSID and Table 8.7 for AHEAD. Costs are lower in PSID than in AHEAD, reflecting the fact that reported costs were for one year in PSID and the previous two years in AHEAD. (Other reasons for higher costs in AHEAD are discussed in Chapter 5). For hospital care, those in prepaid plans had the lowest out-of-pocket costs while those with only FFS Medicare had the highest costs in both survey samples. Those with only FFS Medicare had the highest costs for outpatient surgery in PSID, and the highest costs for ambulatory services in AHEAD. Dual-eligibles had the lowest costs for physician/therapist/ER visits in PSID, and for ambulatory services in AHEAD. For physician/therapist/ER visits in PSID, out-of-pocket costs were highest for both supplemental insurance groups. Prescription medication expenditures were lowest for Medicaid eligibles in PSID, and for those with Medicaid and those with employer- sponsored insurance in AHEAD. Consistent with patterns of use, out-of-pocket dental and equipment costs were lowest for Medicaid eligibles and those with only FFS Medicare coverage in PSID. Home care costs in PSID were low for all respondents, reflecting good coverage by Medicare for this service. In AHEAD, costs for home health care/other special services were highest for those with employer-insurance, fairly high for those in Medicaid, and lowest for those in prepaid plans. Those with FFS Medicare coverage only had the highest nursing home costs in PSID. Insurance premiums were highest for those with individually-purchased supplemental insurance in both PSID and AHEAD. As seen in the last two rows of Tables 8.6 and 8.7, total out-of-pocket costs excluding insurance premiums were highest for those with individually-purchased supplemental insurance in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 200 PSID, and for those with only FFS Medicare in AHEAD. Those with the lowest total out- of-pocket costs excluding insurance premiums in PSID were HMO enrollees, although spending among Medicaid eligibles was only slightly higher, and Medicaid eligibles had the lowest total costs when insurance premiums were included. In AHEAD, those covered by Medicaid had the lowest total out-of-pocket costs, both excluding and including insurance premiums. In both survey samples, those with individually-purchased supplemental insurance had much higher costs than other groups when insurance premiums were included. An examination of median out-of-pocket costs shows a similar pattern, with some exceptions. For services with low rates of use, median costs were zero for all insurance groups. These included hospital, nursing home, and home care in both survey samples, as well as outpatient surgery, dental, and equipment in PSID. Medicaid eligibles in PSID had the lowest median for total costs excluding insurance, whereas mean costs in this category were slightly lower for those in prepaid plans. In AHEAD, median costs for ambulatory services were zero for Medicaid recipients, and were also much lower for those in prepaid plans than for other insurance groups. For total costs excluding insurance premiums in AHEAD, median costs were highest for those with medigap insurance, whereas mean costs were highest for those with FFS Medicare only. This reflects a portion of high cost users among those with FFS Medicare only. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 201 Table 8.6 Mean and Median One Year Out-of-Pocket Spending on Each Service and Overall by Insurance Type, for All Respondents in PSID Service Type Medicare Only Employer Insurance Medigap Insurance Pre-Paid Health Plan Medicaid Hospital Stays Mean 181 67 131 19 55 (SD) (644) (345) (706) (104) (157) Median 0 0 0 0 0 N 117 245 381 78 61 Outpatient Surgery Mean 87 31 37 5 10 (SD) (483) (151) (263) (45) (71) Median 0 0 0 0 0 N 118 251 385 78 65 Physician, Mean 127 149 150 32 17 Therapist, (SD) (439) (281) (487) (53) (46) ER visits Median 28 75 75 3 0 N 112 238 371 72 56 Prescription Mean 264 172 351 224 76 Medications (SD) (407) (375) (537) (403) (127) Median 42 32 160 55 0 N 110 243 371 75 55 Dental Care Mean 14 128 142 200 18 (SD) (52) (278) (509) (421) (83) Median 0 0 0 0 0 N 116 252 382 78 65 Equipment Mean 49 82 96 74 36 (SD) (171) (224) (283) (200) (84) Median 0 0 0 0 0 N 118 250 382 78 63 Nursing Home Mean 677 442 314 0 314 Stays (SD) (4,755) (3,081) (2,703) - (1,556) Median 0 0 0 0 0 N 118 252 387 78 64 Home Care Mean 13 5 7 0 0 (SD) (102) (54) (123) - (5) Median 0 0 0 0 0 N 119 252 385 78 64 Insurance Mean 0 545 846 494 78 Prem ium s (SD) - (766) (671) (605) (218) Median 0 312 720 365 0 N 119 252 385 78 65 (Table continues) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 202 (Table 8.6 continued) Service Type Medicare Only Employer Insurance Medigap Insurance Pre-Paid Health Plan Medicaid Total Mean 782 937 1,190 564 583 (excluding (SD) (1,279) (2,758) (3,191) (614) (1,766) insurance Median 325 362 500 300 20 premiums) N 101 227 342 69 47 Total Mean 782 1,466 2,042 1,072 664 (including (SD) (1,279) (2,812) (3,247) (854) (1,961) insurance Median 325 853 1384 775 30 premiums) N 101 227 342 69 47 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 203 Table 8.7 Mean and Median Two Year Out-of-Pocket Spending on Each Service and Overallby Insurance Type, for All Respondents in AHEAD Service Type Medicare Only Employer Insurance Medigap Insurance Pre-Paid Health Plan Medicaid Hospital, Mean 926 780 531 287 925 Nursing Home (SD) (5,665) (5, 801) (4,080) (3,383) (6,522) Median 0 0 0 0 0 N 868 873 2959 730 649 Physician, Mean 653 597 581 535 194 Outpatient Surgery, (SD) (1,458) (1,140) (1,649) (2,068) (627) Dental, ER Median 200 250 200 110 0 N 841 855 2918 723 610 Prescription Mean 1,870 1,082 1,734 1,623 1,083 Medications' (SD) (6,375) (4,976) (4,594) (6,050) (4,677) Median 360 240 600 240 0 N 863 864 2952 728 647 Home Care, Mean 78 228 99 62 153 Special Services (SD) (1,260) (3,515) (2,285) (796) (2,749) Median 0 0 0 0 0 N 868 876 2977 729 654 Insurance Mean 963 1,182 3,261 1,772 350 Premiums2 (SD) (552) (2,288) (3,137) (2,385) (936) Median 0 0 2,568 1080 0 N 875 876 2985 733 656 Total Mean 3,228 2,685 2,864 2,483 2,147 (excluding (SD) (8,080) (9,542) (6,957) (7,404) (8,096) insurance Median 882 750 1,200 560 129 premiums) N 829 843 2873 716 596 Total Mean 3,324 3,892 6,132 4,252 2,485 (including (SD) (8,132) (9,728) (7,697) (7,777) (8,216) insurance Median 960 1,896 4,416 2180 196 premiums) N 829 843 2873 716 596 1 Monthly costs multiplied by 24 2 Reported premium multiplied to give premium costs for two years 3 43 respondents reported only long-term care insurance Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 204 Mean and median out-of-pocket costs for only those using each service are shown by insurance type in Table 8.8 for PSID and Table 8.9 for AHEAD. In terms of which insurance groups had the highest and lowest costs for each service type, expenditures among users reflect the same pattern as seen across all respondents, above. A notable exception is for nursing home care in PSID. Among users, expenditures were highest for those with employer-sponsored and medigap insurance, whereas across all respondents, costs were highest for those with FFS Medicare only. This reflects the fact that those with only FFS Medicare were more likely to use nursing home care than those with insurance, but were less likely to have uncovered costs for this care. For total costs excluding and including insurance premiums, the rank ordering of insurance groups by out-of-pocket spending level was the same among those using at least one service as among all respondents, above. The pattern of median costs by insurance and service type was similar to the pattern of mean costs, with some notable exceptions. Median costs for hospital stays in PSID were substantially higher for those with employer-sponsored and medigap insurance than for other groups, whereas mean costs were highest for those with FFS Medicare only. Median costs for users of hospital/nursing home care and home care in AHEAD were zero for all insurance groups, whereas mean costs were quite high. This is consistent with the finding in Table 8.3 that less than half of hospital users and home care users in AHEAD reported any uncovered costs, and indicates that a portion of each group had high costs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 205 Table 8.8 Mean and Median One Year Out-of-Pocket Spending on Each Service and Overall by Insurance Type, for Users of Each Service in PSID Service Type Medicare Only Employer Insurance Medigap Insurance Pre-Paid Health Plan Medicaii Hospital Stays Mean 711 459 615 97 175 (SD) (1,181) (808) (1,450) (226) (283) Median 30 171 212 0 0 N 30 35 81 15 19 Outpatient Mean 850 198 266 76 181 Surgery (SD) (1,333) (316) (669) (177) (215) Median 140 51 50 0 0 N 12 40 53 6 4 Physician. Mean 158 167 166 35 23 Therapist, (SD) (481) (294) (510) (55) (50) ER visits Median 50 79 78 10 0 N 90 213 334 66 45 Prescription Mean 329 210 422 244 89 Medications (SD) (427) (403) (560) (421) (134) Median 80 61 250 68 11 N 89 199 309 69 47 Dental Care Mean 109 236 311 333 92 (SD) (116) (346) (710) (515) (181) Median 75 100 104 150 20 N 15 137 175 47 13 Equipment Mean 161 231 271 194 143 (SD) (288) (321) (423) (294) (139) Median 68 150 146 111 145 N 36 90 135 30 16 Nursing Home Mean 4,766 10, 238 12,016 _ 1,864 Stays (SD) (14,236) (11,723) (12,586) - (4,464) Median 1,000 5,000 11,682 - 0 N 17 11 10 0 11 Home Care Mean 125 200 130 4 (SD) (315) (292) (553) - (17) Median 0 0 0 - 0 N 12 6 21 0 7 Insurance Mean - 545 846 494 609 Premiums (SD) - (766) (671) (605) (430) Median - 312 720 365 432 N 0 252 386 78 8 (Table continues) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 206 (Table 8.8 continued) Service Type Medicare Only Employer Insurance Medigap Insurance Pre-Paid Health Plan Medicaid Total Mean 879 970 1,241 598 641 (excluding (SD) (1,329) (2,807) (3,239) (621) (1,828) insurance Median 358 393 527 316 33 premiums) N 90 219 328 65 43 Total Mean 879 1,484 2,090 1,109 729 (including (SD) (1,329) (2,864) (3,290) (867) (2,029) insurance Median 358 853 1,414 791 30 premiums) N 90 219 328 65 43 Median insurance premiums for those with employer-sponsored insurance were zero, indicating that at least half of those with this type of insurance had the premium completely covered by an employer. For total costs excluding insurance premiums in PSID, median costs were substantially lower for Medicaid eligibles than for other groups, whereas the mean costs of Medicaid eligibles were closer to those of other groups. The burden of total out-of-pocket costs as a portion of household income is shown for PSID in Table 8.10 and for AHEAD in Table 8.11. (These calculations were possible only for respondents with complete spending data on each service.) In each table, mean and median burden is shown first for total costs excluding insurance premiums, then for total costs including premiums, and finally for total costs including insurance premiums and premiums for Medicare Part B. To obtain this estimate, the cost of Medicare Part B for the survey period in each survey sample was added to total reported spending. The costs of Medicare Part B premiums was not added to spending for dual-eligibles, since Medicaid was assumed to have paid the premium for these respondents. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 207 Table 8.9 Mean and Median Two Year Out-of-Pocket Spending on Each Service and Overall by Insurance Type, for Users of Each Service in AHEAD Service Type Medicare Only Employer Insurance Medigap Insurance Pre-Paid Health Plan Medicaid Hospital, Mean 2,948 2,294 1,557 981 1,679 Nursing Home (SD) (9,854) (9,740) (6,862) (6,270) (8,872) Median 0 0 0 0 0 N 272 297 1010 213 358 Physician, Mean 705 605 599 544 201 Outpatient Surgery, (SD) (1,504) (1,146) (1,669) (2,088) (637) Dental Median 200 200 200 125 0 N 780 843 2830 711 590 Prescription Mean 2,537 1,341 2,186 2,112 1,248 Medications1 (SD) (7,351) (5,500) (5,047) (6,858) (5,027) Median 840 360 960 480 0 N 636 697 2657 561 562 Home Care, Mean 521 1,409 549 390 464 Special Services (SD) (3,256) (8,526) (5,372) (1,988) (4,770) Median 0 0 0 0 0 N 130 141 534 115 216 Insurance Mean 1.9593 1,182 3,261 1,772 1,462 Premiums2 (SD) (1,702) (2,288) (3,137) (2,385) (1,548) Median 1,600 0 2,568 1080 984 N 43 876 2985 733 157 Total Mean 3,400 2,706 2,920 2,512 2,188 (excluding (SD) (8,250) (9,581) (7,010) (7,452) (8,169) insurance Median 980 750 1,228 578 144 premiums) N 787 837 2818 708 585 Total Mean 3,495 3,892 6,132 4,252 2,531 (including (SD) (8,304) (9,728) (7,697) (7,777) (8,289) insurance Median 1007 1,896 4,416 2180 200 premiums) N 787 843 2873 716 585 1 Monthly costs multiplied by 24 2 Reported premium multiplied to give premium costs for two years 3 Long-term care insurance only Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 208 When insurance premiums were not included, mean burden was highest in both survey samples for those with only FFS Medicare coverage. When the cost of HMO or insurance premiums was included, mean burden was highest for those with individually purchased insurance in both survey samples. Mean burden for total costs both excluding and including insurance premiums was lowest for those in prepaid plans in PSID, and for those with employer-sponsored insurance in AHEAD. As seen in the last row of Tables 8.10 and 8.11, when the cost of premiums for Medicare Part B was added to total spending, the mean burden of spending as a portion of income was lowest for those with Medicaid in both survey samples, and highest for those with medigap insurance. Table 8.10 Percentage of Household Income Spent on Total Out-of-Pocket Costs by Insurance Type in PSID Medicare Only (N = 101) Employer Insurance ( N = 227) Medigap Insurance (N = 342) Pre-Paid Health Plan (N = 69) Medicaid (N = 47) Total out-of-pocket Mean 8.7 4.9 6.8 3.4 7.6 costs excluding (SD) 13.6 (13.6) (14.6) (4.9) (18.1) insurance premiums Median 2.1 1.4 2.4 1.5 0.3 Total out-of-pocket Mean 8.7 7.4 12.6 6.3 8.2 costs including (SD) 13.6 (14.5) (17.3) (7.8) (18.6) premiums for insurance or a prepaid health plan Median 2.1 3.2 6.4 3.6 0.3 Total out-of-pocket Mean 13.3 9.2 15.2 8.6 8.2 costs including (SD) (14.2) (15.1) (18.0) (8.7) (18.6) premiums for insurance and Medicare Part B Median 7.7 4.8 8.7 6.2 0.3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 209 Table 8.11 Percentage of Household Income Spent on Total Out-of-Pocket Costs by Insurance Type in AHEAD Medicare Only (N = 829) Employer Insurance (N = 843) Medigap Insurance (N = 2873) Pre-Paid Health Plan (N = 716) Medicaid (N = 596) Total out-of-pocket Mean 9.9 6.0 8.5 7.0 8.5 costs excluding (SD) (19.2) (15.0) (17.5) (16.5) (17.5) insurance premiums Median 2.7 1.6 2.8 1.4 0.6 Total out-of-pocket Mean 10.1 8.6 17.7 12.3 10.3 costs including (SD) (19.3) (16.3) (21.0) (19.3) (18.9) premiums for insurance or a prepaid health plan Median 2.8 3.6 10.9 5.4 1.1 Total out-of-pocket Mean 14.1 11.4 20.8 15.6 10.3 costs including (SD) (19.7) (16.7) (21.7) (19.9) (18.9) premiums for Median 7.3 6.5 13.7 4.3 1.1 insurance and Medicare Part B Examining median burden paints a slightly different picture. Median spending as a portion of income was highest for those with medigap insurance, even when insurance premiums were not included as part of out-of-pocket spending. Those dually-eligible for Medicaid had the lowest median burden on all measures of total costs in both survey samples, whereas the mean burden among dual-eligibles was similar to that of other groups before Medicare Part B Premiums were considered. This difference between mean and median burden reflects a portion of high-burden users in the Medicaid group. Tables 8.12 and 8.13 show the portion of those in each insurance group with total out-of-pocket costs (including insurance and Medicare Part B premiums) over 10%, 30%, and 50% of income in both PSID and AHEAD. In both survey samples, those with medigap insurance were most likely to spend 10% and 30% of household income on Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 210 health care, whereas those with employer insurance were least likely to have very burdensome costs at 50% of income. In both survey samples, those eligible for Medicaid were most likely to spend over 50% of income, but were least likely to spend over 10% of income, indicating that a few dual-eligibles had very high costs. Table 8.12 Percent Spending Over 10%, 30%, and 50% of Household Income on Total Out-of-Pocket Costs (Including Premiums for Insurance and Medicare Part B) by Insurance Type in PSID Percent spending this percentage over 10% over 30% over 50% Medicare Only 42.0 9.9 5.1 Employer-sponsored Insurance 22.0 5.5 3.2 Medigap Insurance 47.1 13.5 4.2 Pre-Paid Health Plan 28.0 4.3 0.0 Medicaid 12.1 8.5 7.5 Table 8.13 Percent Spending Over 10%, 30%, and 50% of Household Income on Total Out-of-Pocket Costs (Including Premiums for Insurance and Medicare Part B) by Insurance Type in AHEAD Percent spending this percentage over 10% over 30% over 50% Medicare Only 35.2 10.8 5.8 Employer Sponsored Insurance 31.2 6.2 3.7 Medigap Insurance 63.7 19.4 8.3 Pre-Paid Health Plan 44.4 13.3 5.9 Medicaid 25.8 10.2 5.9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 211 Multivariate Results Regression results for both PSID and AHEAD are shown in Tables 8.14 to 8.17. Table 8.14 shows results for spending on services where OLS regression was used on all respondents. Results for services for which a two-part model was used are shown in Tables 8.15 to 8.17. Those with medigap insurance were the referent group in all models. Results are described here by insurance type across tables. Fee-for-Service Medicare Only Those with no insurance coverage besides FFS Medicare spent significantly less than those with medigap insurance overall and on most services. An exception is for hospital/nursing home care in AHEAD. As seen in Table 8.15, those with only FFS Medicare had lower odds of using hospital/nursing home care in Models 2 and 3, when health was controlled. However, among users, costs were significantly higher for those with only FFS Medicare than for those with supplemental insurance. Having only FFS Medicare was not a significant predictor of costs for ambulatory services in AHEAD, or of use or costs for hospital or home care in PSID. As seen in Table 8.14, costs for physician/therapist/ER visits in PSID were only lower for those with FFS Medicare coverage until socioeconomic variables were controlled in Model 3. Those with only FFS Medicare had significantly lower odds of using dental care in PSID, and home care/special services in AHEAD, as seen in Tables 8.15 and 8.16. As shown in Table 8.15, those with only EFS Medicare had greater odds of using nursing home care in PSID, but users had significantly lower costs in Model 2, which controlled for health but not SES. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 212 Employer Sponsored Insurance Multivariate findings for employer insurance were mixed; for some services, out- of-pocket expenditures were lower among those with employer insurance, for some services costs were higher, and for several services employer insurance was not a significant predictor of costs. As seen in Table 8.14, having employer-sponsored insurance was not a significant predictor of physician expenditures in PSID, whereas those with employer-insurance in AHEAD spent significantly more on ambulatory services than those with medigap. For prescription medications, those with employer-sponsored insurance spent significantly less than those with medigap insurance in both survey samples. The two-part models for hospital and nursing home costs are shown in Table 8.15. In PSID, employer insurance did not predict hospital use or expenditures. Those with employer insurance had greater odds of using nursing home care, but only in Model 1, before health was controlled. Among those who used nursing home care in PSID, out-of- pocket costs were significantly higher for those with employer insurance than for those with medigap, but only in Model 2, where health was controlled but socioeconomic variables were not. For hospital/nursing home expenditures in AHEAD, having employer insurance did not significantly predict use, but costs among users were significantly higher for those with employer insurance than those with medigap. As seen in Table 8.16, employer insurance was not a significant predictor of use or costs for outpatient surgery, dental care or equipment. Table 8.17 shows results of the two-part models for home care. Employer insurance was not a significant predictor of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 213 home care use in either sample, or of home care expenditures in PSID. In AHEAD, those with employer insurance had significantly higher home care costs than those with medigap, but only in models 1 and 2, before socioeconomic variables were controlled. For total expenditures, as seen in Table 8.14, employer-insurance was not a significant predictor in either sample when the cost of insurance premiums were excluded. However, when insurance premiums were included in total spending, those with employer- insurance spent significantly less than those with medigap coverage in both PSID and AHEAD. Prescription, Dental, and Long-Term Care Insurance This section reports results for insurance specifically covering prescription medications, dental care, or long-term care, although these results are not shown in the Tables above. Having insurance that covered prescription medications predicted significantly lower medication costs in AHEAD, but not PSID.4 Those with dental insurance were more likely to use dental care in PSID, but did not have significantly lower dental costs or total costs.5 In AHEAD, those with dental insurance spent significantly less on physician/surgery/dental care and on total out-of-pocket costs in all hierarchical models. Having long-term care insurance was only a significant predictor of lower home care costs in the fourth hierarchical model in PSID, when socioeconomic variables were added. 4 This is the result when using a OLS on the whole sample to predict prescription expenditures. However, when a two-part model is used, having prescription insurance does predict significantly Iowa: drug costs among users in the PSID Elderly Health Supplement (Lillard et al.,1999). 5 This replicates findings by Kington et al. (1995) using this data. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 214 Prepaid Health Plan Those in Medicare HMOs spent significantly less than those with medigap insurance, overall and for most services. The exceptions to this pattern are described here. As seen in Table 8.15, for nursing home costs in PSED, being in a prepaid plan was not a significant predictor of use, and those in prepaid plans who used nursing home care had significantly lower expenditures than those with medigap only in Model 2, where health but not socioeconomic variables were controlled. As seen in Table 8.16 and 8.17, being in a prepaid plan was not a significant predictor of use or costs for surgery, dental care, or equipment in PSID, and did not predict home care use or costs in either survey sample. Medicaid Those reporting Medicaid coverage in AHEAD and those estimated to be eligible for Medicaid in PSID spent significantly less than those with medigap insurance, overall and for most services. Those with Medicaid were significantly more likely to use hospital/nursing home care in AHEAD and nursing home care in PSID, as seen in Table 8.15. However, in PSID, Medicaid coverage no longer predicted nursing home use or costs in Model 3, when socioeconomic variables were controlled. As seen in Table 8.17, Medicaid eligibility did not predict home care use or costs in PSID. In AHEAD, those with Medicaid had greater odds of using home care/special services than those with medigap, but Medicaid coverage did not predict costs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 215 Table 8.14 Hierarchical OLS Regression Results for Out-of-Pocket Spending by Insurance Type in PSID and AHEAD: Ambulatory Services, Prescription Medications, and Total Out-of-Pocket Spending (in Logged Dollars) Service Type Data Set Insurance Type + Model 1 Model 2 Model 3 Physician, Therapist, PSID Medicare only -0.65* -0.79** -0.42 ER visits Employer insurance 0.10 0.14 0.05 (N = 850) Prepaid plan -1.45*** -1.50*** -1.49*** Medicaid -2.14*** -2.17*** -1.24** R2 0.104 0.123 0.145 Physician, AHEAD Medicare only -0.20 -0.18 0.06 Outpatient Surgery, Employer insurance 0.37*** 0.35** 0.30** Dental Care, (N = 5948) Prepaid plan -0.38** -0.40*** -0.35** ER visits Medicaid -2.37*** -2.27*** -1.38*** R2 0.101 0.109 0.149 Prescription PSID Medicare only -1.18*** -1.62*** -1.48*** Medications Employer insurance -0.84*** -0.73*** -0.77*** (N = 855) Prepaid plan -0.68* -0.78* -0.79** Medicaid -2.07*** -0.26 -2.13*** R2 0.093 0.230 0.234 AHEAD Medicare only -0.41*** -0.41*** -0.34*** Employer insurance -0.29*** -0.27*** -0.28*** (N = 6055) Prepaid plan -0.21* -0.19* -0.17* Medicaid -1.01*** -1.37*** -1.07*** Prescription insurance -0.67*** -0.76*** -0.78* R2 0.068 0.214 0.224 Total Costs excluding PSID Medicare only -1.11*** -1.39*** -1.01*** insurance premiums Employer insurance -0.24 -0.21 -0.32 (N = 787) Prepaid plan -0.80** -0.88** -0.89** Medicaid -2.29*** -2.56*** -1.64*** R2 0.142 0.206 0.238 AHEAD Medicare only -0.53*** -0.55*** -0.34*** Employer insurance 0.03 0.05 0.02 (N = 5858) Prepaid plan -0.32** -0.32** -0.27* Medicaid -2.02*** -2.42*** -1.61*** R2 0.100 0.166 0.200 Total Cost PSID Medicare only -2.38*** -2.58*** -2.35*** including insurance Employer insurance -0.52** -0.50** -0.56*** premiums (N = 780) Prepaid plan -1.09*** -1.15*** -1.15*** Medicaid -3.39*** -3.65*** -3.07*** R2 0.343 0.377 0.393 AHEAD Medicare only -0.78*** -2.28*** -2.13*** Employer insurance -1.01*** -1.00*** -1.02*** (N = 5858) Prepaid plan -0.78*** -0.79*** -0.76*** Medicaid -3.47*** -3.69*** -3.09*** R2 0.313 0.336 0.359 + Medigap insurance is the referent group * E < .05 **e < .01 *** £ < .001 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission. Table 8.15 Hierarchical Two-Stage Regression Results for Hospital and Nursing Home Use and Out-of -Pocket Expenditures (in Logged Dollars) by Insurance Type in PSID and AHEAD Model 1 Model 2 Model 3 Service Type Data Set Insurance Type+ Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Hospital PSID Medicare only 0.09 -0.76 -0.18 -0.67 -0 .2 2 -0.56 Employer insurance -0.24 -0.01 -0.24 -0.06 -0.23 -0.07 II O O oo to Prepaid plan -0.04 -2.78** 0 .0 2 -2 .8 6 ** 0.0 1 -2.87** 181 users) Medicaid 0.38 -1.40 0.21 -1.49 0 .1 0 -1.23 LL* / R2 -424 0.125 -371 0,138 -370 0.142 Nursing PSID Medicare only 1.09*** -2.76 0.80** -6 .1 0** 0.70* -5.47 Home Employer insurance 0.50* -0.41 0.52 -4.66* 0.53 -4.38 (N = 893, Prepaid plan 0.06 -5.59 -0.01 -9.86* -0.04 -9.33 45 users) Medicaid 1 1 9*** -5.19* 0 .8 6 * -8.06** 0.55 -5.47 LL/R 2 -119 0.323 -1 0 2 0.721 -93 0.756 Hospital, AHEAD Medicare only -0.08 1.34*** -0.14* I -0.13* Nursing Employer insurance 0 .0 0 1.19*** -0 .0 2 1.0 2 *** -0.03 1,0 1 *** Home (N = 6079, Prepaid plan -0.08 -0.67* -0 .1 2 -0.67** -0.11 -0 .6 6 * 2150 users) Medicaid 0.57*** -0.46* 0 .2 1 ** -0.92*** 0,28*** -0.55* LL/R 2 -3831 0.053 -3368 0 .1 0 1 -3364 0.108 + Medigap insurance is the referent group *LL = Log Liklihood * E < .05 ** e < .01 *** e < .001 216 Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Table 8.16 Hierarchical Two-Stage Regression Results for Outpatient Surgery, Dental Care, and Equipment Use and Out-of - Pocket Expenditures (in Logged Dollars) by Insurance Type in PSID Model 1 Model 2 Model 3 Service type Data Set Insurance Type Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Surgery PSID Medicare only -0.19 1.80 -0 .2 2 1.35 -0.17 0.65 Employer insurance 0.06 -0.04 0.07 0.13 0.05 0.26 (N = 897, Prepaid plan -0.36 -1.83 -0.33 -2 .2 0 -0.33 -2.43 114 users) Medicaid -0.47 -1.09 -0.50 -0.81 -0.40 -2.24 Log Likelihood / R2 -328 0.094 -325 0.149 -323 0.181 Dental PSID Medicare only -0.84*** -1.27** -0.77*** -1.33** -0.45* -1.16* Employer insurance 0 .1 0 -0 .1 2 0 .1 0 -0.13 0 .0 2 -0 .2 2 (N = 897, Prepaid plan 0.23 -0.01 0.24 -0.03 0.26 -0 .0 0 387 users) Medicaid -0.43* -2.92*** -0.29 -3.04*** 0.76** -2 .1 0** Dental Insurance 0.47** -0.71** 0.43** -0 .6 8 * 0.36* -0.74** Log Likelihood / R2 -543 0.119 -534 0 .1 2 1 -500 0.165 Equipment PSID Medicare only -0.13 -1.36*** -0 .1 2 -1.31*** -0.06 -1.27 Employer insurance 0.04 0 .0 1 0.19 0 .0 2 0 .0 1 0.03 (N = 892, Prepaid plan 0 .1 0 -0.56 0 .1 2 -0.55 0.13 -0.51 306 users) Medicaid -0.25 -0.73 -0.25 -0.82 -0.05 -0.40 Log Likelihood / R2 -559 0.126 -551 0.147 -550 0.166 + Medigap insurance is the referent group * E < .05 ** E < -01 * * * E < *001 to I — * -4 Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Table 8.17 Hierarchical Two-Stage Regression Results for Home Care Use and Out-of -Pocket Expenditures (in Logged Dollars) by Insurance Type in PSID and AHEAD Model 1 Model 2 Model 3 Service type Data Set Insurance Type* Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Any use probit Log costs among users (OLS) Home Care PSID Medicare only 0.19 0.28 -0.10 -0.08 -0.21 -0.79 Employer insurance -0.32 1.89 -0.36 0.64 -0.37 -0.11 (N = 898, Prepaid plan -0.65 -0.05 -0.80 -1.60 -0.87 -1.44 48 users) Medicaid 0.30 -0.49 0.10 -1.68 -0.15 -1.71 Log Likelihood / R2 -165 0.125 -144 0.589 -143 0.752 Home Health Care, AHEAD Medicare only -0.16* 0.41 -0.22** 0,39 -0.22** 0.44 Other Special Services Employer insurance -0.03 0.55* -0.03 0.50* -0.02 0.47 (N = 5869, Prepaid plan -0.05 0.25 -0.10 0.18 -0.10 0.21 1136 users) Medicaid 0.54*** -0.31 0.22** -0.35 0.20* -0.07 Log Likelihood / R2 -2718 0.029 -2292 0.046 -2287 0.051 + Medigap insurance is the referent group * E < .05 ** e < -01 *** E < -001 218 219 Discussion The goal of this chapter was to examine the out-of-pocket protection provided by different types of insurance coverage supplemental to Medicare. Multivariate analyses were used to examine whether differences in out-of-pocket spending by insurance type could be explained by related differences in health or socioeconomic status. Results are discussed below for each insurance type. Supplemental Insurance versus FFS Medicare Findings for out-of-pocket costs among those with medigap or employer- sponsored insurance compared to those with only FFS Medicare were mixed. Those with only FFS Medicare had higher mean expenditures on hospital care in both survey samples, on outpatient surgery in PSID, and on ambulatory services in AHEAD. Mean total expenditures excluding insurance premiums were also highest for those with only Medicare in AHEAD. These findings indicate that medigap and employer-sponsored insurance provided important out-of-pocket protection. However, in regression analyses, only costs for hospital/nursing home care were significantly higher for those with no insurance compared to those with medigap. Several other categories of out-of-pocket costs were higher for those with supplemental insurance. Those with medigap and those with employer insurance had higher mean expenditures for physician visits, prescription medications, dental care, and equipment in PSED, and for home care/special services in AHEAD. As a result, even when insurance premiums were not included in total costs, those with medigap and employer- sponsored insurance had the highest mean total out-of-pocket costs in PSID and highest Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 220 median costs in AHEAD. Several of these differences were significant in multivariate analyses. Compared to the referent group of those with medigap, those with only FFS Medicare had significantly lower total expenditures and prescription medication costs in both survey samples. In PSID, dental, physician and equipment costs were also significantly lower for those with only FFS Medicare than for those with medigap. These findings are consistent with Rubin & Koelln (1993a), who found that families with insurance spent more out-of-pocket than others, net of insurance premiums. Potential explanations hypothesized to account for higher out-of-pocket costs among those with insurance are adverse selection, wealth effects, and moral hazard or improved access to care. Adverse selection does not appear to explain these finding, since the health of those with medigap and employer-insurance was similar to that of HMO enrollees and better than those with Medicaid or with FFS Medicare only. Also, spending differences were significant even when multiple measures of health were controlled. This is consistent with Hurd & McGarry (1997), who found favorable selection among those purchasing medigap coverage in wave 1 of the AHEAD survey. However, although there does not appear to have been adverse selection of sicker beneficiaries into plans, there may have been unmeasured adverse selection based on health attitudes that predisposed individuals to a greater intensity of service use. Hierarchical results provide limited evidence for wealth effects. When SES was controlled, not having insurance only lost significance as a predictor of lower spending for physician visits and equipment in PSID. Patterns of service use and costs among users provide support for moral hazard and/or increased access to services as explanations for higher expenditures among those Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 221 with supplemental insurance. Those with medigap and employer-insurance were more likely than those with only FFS Medicare to report use of most services, with the exception of hospital and home care in PSID, and nursing home care in both samples. However, these are measures of whether each service was used at all, and do not reflect volume or intensity of use. Additional analyses (not shown here) revealed that insurance groups did not differ significantly in the number o f physician visits reported in AHEAD. Among only those using each service, out-of-pocket costs were higher for those with supplemental insurance for several discretionary services, which may indicate use of more and/or higher cost tests and treatments. It is difficult to determine the extent to which the greater use of some services among those with insurance reflects moral hazard or over-use of services, and to what extent it reflects greater access to appropriate care among the insured. Patterns of service use suggest that those with only FFS Medicare coverage may not have received adequate maintenance and preventive care, since they had lower use of discretionary services such as physician visits, outpatient surgery, and dental care, but had higher rates of hospital use. Another potential explanation for the higher costs among those with supplemental insurance is that they actually faced higher costs for the same procedures than those with only FFS Medicare. It is well known that there is price discrimination and cost-shifting in the health care market; in order to offer lower prices to bulk payers such as employers and prepaid health plans, service providers typically charge more to those with individually- purchased insurance. Those with supplemental insurance may have been more likely to see physicians who did not accept assignment for their care, and thus face balance billing Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 222 charges (although these charges are covered under some supplemental plans as discussed above). Physicians and hospitals may also have been more likely to aggressively collect Medicare deductibles and coinsurance from those with insurance. This is one possible explanation for the large portion of those with only FFS Medicare coverage who reported no uncovered costs for services used. These respondents should have been responsible for deductibles and coinsurance if they did not have any additional coverage. However, physicians and hospitals may have been more likely to accept reimbursement by Medicare as payment in full for these patients. It is difficult to find up-to-date information on the extent of this practice, since it is illegal. However, there is evidence that it may have been widespread in the past. In a 1976 survey of physicians, 40 percent reported that they did not try to collect the coinsurance and deductibles from their Medicare patients, but would bill their insurance companies (Mitchell, Cromwell & Schurman, 1981, in Cromwell & Burstein, 1985). Also, in a 1988 interview, the inspector general of the Department of Health and Human Services reported that over 25 percent of hospitals regularly waived Medicare deductibles and coinsurance as a marketing technique (Anonymous, 1988). Employer-sponsored insurance did not appear to provide significantly greater out- of-pocket coverage than medigap, as measured by total spending excluding premiums. While those with employer insurance did have lower prescription medication costs, they spent significantly more on hospital/nursing home, ambulatory care, and home care/special services in AHEAD. These higher costs may reflect moral hazard; although rates of reported use were similar for both insurance types, higher costs among users of these Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 223 services with employer insurance may reflect a greater intensity of service use. When the cost of premiums is considered, overall out-of-pocket costs and burden were much lower for those with employer-insurance than those with medigap, reflecting the fact that employers were subsidizing and often fully covering premium costs. As expected, those with individually-purchased medigap coverage had the highest mean and median expenditures for insurance premiums. The cost of premiums among this group is of concern, because for total out-of-pocket costs including insurance premiums, those with medigap spent more out-of-pocket as a portion of household income than all other insurance groups. Those with medigap insurance were also more likely to have costs that could be considered catastrophic, with the greatest likelihood of spending over 30 percent of household income on health care. The finding that those with supplemental insurance face more burdensome costs than other Medicare beneficiaries is of concern because it indicates that medigap coverage increases out-of-pocket costs, rather than protecting beneficiaries as intended. However, this may be acceptable if the insurance is increasing access to appropriate care and keeping beneficiaries healthier. The fact that dental coverage predicted lower out-of-pocket costs for ambulatory services and lower total costs in AHEAD likely reflects the fact that insurance plans covering dental care were more generous overall than other plans. The lower prescription costs among those with prescription coverage in AHEAD is consistent with Lillard et aL (1999). Although this finding was not replicated in PSID, additional analyses (not shown) indicate that when a two-part model is used to predict costs, prescription insurance does significantly lower drug costs among users. The difference in findings when a two-part Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 224 model is used is likely due to the relationship between prescription insurance and medication use. Those not using medications were less likely to have prescription insurance, and they had zero costs due to non-use. When a two-part model is not used, the inclusion of these respondents with no insurance and zero costs dilutes the effect of insurance as a predictor of lower costs among users. The fact that having long-term care insurance did not predict lower nursing home costs reflects that fact that very few of those who used nursing home care had this insurance. In PSID, the finding that long-term care insurance predicted lower costs only when socioeconomic variables were controlled reflects a suppression effect. Those with higher income and wealth were more likely to have long-term care insurance, and this positive relationship suppressed the negative effect of long-term care insurance on out-of- pocket costs until wealth and income were controlled. Prepaid Health Plan As expected, those in Medicare HMOs had lower overall out-of-pocket costs than those with supplemental insurance. Enrollees paid less out-of-pocket on insurance than those with Medigap or employer-sponsored benefits, and had lower expenditures for several service types. Costs of care were also less burdensome for those in HMOs than for those in other insurance groups. Those enrolled in prepaid plans were in better health than others, and had the lowest use of nursing home and home care, which reflects previously documented favorable selection into plans. However, out-of-pocket spending among enrollees was Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 225 significantly lower even when health was controlled, indicating that lower costs were not due to favorable selection. Patterns of health care use did not provide strong support for concerns regarding reduced access to care among those in HMOs. As hypothesized, enrollees had lower rates of hospital use than other respondents, but were more likely to use discretionary services such as physician and dental care. This reflects the desired pattern of preventive, maintenance care, and is consistent with findings suggesting that lower out-of-pocket costs among enrollees are primarily due to better coverage rather than lower use (Sofaer & Kenney, 1989) Those in prepaid plans had the highest use of prescription medications, which likely reflects adverse selection of beneficiaries requiring expensive medications. One finding that may indicate under-provision of care among HMO enrollees is their lower rates of outpatient surgery use. This finding is consistent with Goldzweig et aL (1997), and further scrutiny is required to determine whether it indicates barriers in access among enrollees or appropriate provision of care (Obstbaum, 1997). Medicaid As expected, those dually-eligible for Medicaid and Medicare had lower out-of- pocket costs than other groups, and were least likely to have uncovered costs for all services used except hospital and home care. Out-of-pocket costs were significantly lower for dual-eligibles than for those with medigap overall and for the majority of services. These spending differences were significant before controlling for the worse health of Medicaid beneficiaries, and most remained significant even when controlling for socioeconomic status, indicating that reduced costs were primarily due to Medicaid Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 226 coverage rather than to reduced ability to pay. These results indicate that Medicaid was protecting most beneficiaries from high out-of-pocket costs as intended. However, as hypothesized, patterns of service use suggest barriers in access among this population. Dual-eligibles were less likely than other groups to report use of any services, and less likely to use more discretionary services such as physician care, outpatient surgery, equipment, and dental care. This lower use is particularly disconcerting given the poorer health and greater need for care among Medicaid beneficiaries. Dual-eligibles were substantially more likely than those in other insurance groups to use hospital care, which may be an indication that they were less likely to seek care until their conditions had become very serious. The greater use of home care and ‘other special services’ by dual-eligibles in AHEAD may be explained by coverage of services such as adult day care and meal programs by some state Medicaid programs and by other programs targeted toward low- incorae elders, such as the Older Americans Act. The greater use of long-term care services among Medicaid beneficiaries is consistent with their poorer health, and with the fact that Medicaid is the major public payer for nursing home care. It is also important to note that the burden of out-of-pocket health spending was quite high for some dual-eligibles. Although median out-of-pocket costs and median burden were much lower for Medicaid beneficiaries than for other groups, mean differences were smaller, indicating that a portion of dual-eligibles had high, burdensome costs. It will be important in future studies to examine the main sources of out-of-pocket costs for dual-eligibles with catastrophic expenditures. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 227 Chapter 9 General Discussion Significance of Control Variables The primary focus of this dissertation was on age, race, and insurance as predictors of out-of-pocket health expenditures. Because of this focus, the above Chapters did not discuss the significance of other variables in the conceptual framework that were controlled in hierarchical analyses. This section describes and discusses findings for these demographic, geographic, health and SES variables as predictors of out-of-pocket costs. (Because these variables were included as control variables and were not of primary interest, these results are not shown in Tables.) Variables are discussed in their hierarchical order of entry, as described in Chapters 6 and 7: demographic and geographic variables were tested alone first, then health variables were added, followed by insurance variables, and finally socioeconomic variables. Gender Gender is the one demographic variable controlled in analyses that was not discussed above. The effect of gender on out-of-pocket costs was expected to differ by service type, and to be related to differences in types of health conditions, as discussed in Chapter 2. Results of hierarchical regression models revealed that women spent significantly less on ambulatory services in AHEAD, but only until health variables were controlled. This supports the hypothesis in Chapter 2 that men would have higher costs because they are more likely to have acute health conditions that are expensive to treat (Verbrugge, 1995). Also, consistent with previous findings of gender differences in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 228 hospital use (Mutran & Ferraro, 1988), women in both PSID and AHEAD were significantly less likely to use hospital care. These differences remained significant in 2- part hierarchical models even when controlling for health conditions and other variables in the conceptual modeL Factors not controlled in analyses that might explain these differences are greater disease severity in men, and a possible bias against hospitalization of women. There were no significant gender differences in out-of-pocket costs among those using hospital care, which suggests that disease severity and types of treatment were similar for women and men among those admitted. Women were more likely than men to use nursing home care in PSID and home care/other special services in AHEAD. Women also spent significantly more on prescription medications in both survey samples. This is consistent with the greater prevalence of disabling chronic health conditions among women. However, with the exception of prescription costs in PSID, these gender effects are not fully explained by health differences, because they remained significant when controlling for health and other variables in the conceptual modeL These results may reflect gender differences in selective survivaL Since men have higher mortality rates at all ages than women (Heeren, Lagaay, Hijmans & Rooymans, 1991), men who survive with chronic conditions are a more hearty, select group, whose conditions may require less care. Gender differences in long term care use reflect the fact that women are more likely to live alone, whereas men are more likely to have a spouse who can provide informal care in the community (Hooyman & Kiyak, 1999). Marital status was not controlled in this study because living with a partner was negatively correlated with being female (r = -.035 in PSID, r = -0.4 in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 229 AHEAD). The magnitude of these relationships indicated that multicollinearity may have been a problem in regression analyses, inflating the standard errors of both variables and reducing their likelihood of significance. For total costs excluding insurance premiums, gender did not predict spending in initial hierarchical models. However, when SES was controlled in PSID, women had significantly higher costs. This reveals that although men and women had similar total costs, women had higher costs than men relative to their lower economic status. In AHEAD, total costs were significantly higher for women only after controlling for Medicaid, which reflects womens’ greater reliance on public assistance to pay for care. Gender was not a significant predictor of spending in PSID for physician care, dental care, equipment, home care, or total costs including insurance. In AHEAD, women spent less on total costs including insurance premiums only until insurance coverage was controlled, indicating that their lower expenditures were due to lower likelihood of purchasing private insurance. Regional Differences in Medicare Spending Mean Medicare spending per beneficiary in the area where the respondent lived was used to control for regional variation in health care costs. This variable was a significant predictor of out-of-pocket costs in some cases. For hospital, physician, and prescription expenditures in PSID, higher regional Medicare spending predicted significantly lower out-of-pocket costs. While this seems counterintuitive, it may be that hospitals and physicians in areas with high Medicare reimbursement rates are less likely to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 230 engage in balance billing, or to collect deductibles and coinsurance from patients. Lower prescription expenditures in areas with high Medicare spending may reflect greater penetration of Medicare managed care in these areas. There were also several suppression effects, where the regional spending variable became significant in later hierarchical models, due to its complex relationships with other variables, such as insurance coverage and SES. Urban vs. Rural Residence Those living in urban areas were expected to use more hospital care and have higher hospital expenditures than rural elders. However, in AHEAD, those living in urban areas used less hospital care, and users had lower combined costs for hospital/nursing home care in all hierarchical models. This is not consistent with previous findings of urban/rural differences in hospital use (Braden & Beauregard, 1994; Dor & Holahan, 1989), and is difficult to explain. Since the AHEAD survey includes only those age 72 and over, these findings may reflect a reversal of urban/rural patterns of hospital use at the oldest ages. Under-use of hospital care at younger ages may result in greater need for hospital care at older ages. Also, since the AHEAD survey took place at a later date than previous studies on urban/rural hospital use, these findings may reflect a disproportionate shift from inpatient to outpatient care in urban vs. rural areas in the 1990’ s. For ambulatory services in AHEAD, urban residents spent less than those in rural areas only when SES variables were controlled, indicating that rural residents had higher expenditures relative to their lower economic status. Urban residents were less likely to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 231 use home care/speciaL services in AHEAD, but only until health was controlled. This indicates that greater use of these services by rural residents was associated with poorer health among this group. A combination of these SES and health effects is seen for total costs excluding insurance in AHEAD. These costs were lower for urban residents until health was controlled, and became significantly lower again when SES was controlled. In PSID, there were no significant urban/rural differences in out-of-pocket costs. This may reflect a lack of power due to small sample size, since only 5.5 percent of the sample (N = 57) was classified as rural. Health As expected, those with higher self-rated health (SRH) were less likely to use hospital, nursing home or home care. Better SRH also predicted lower expenditures on prescription medications and lower total out-of-pocket costs (excluding insurance) in both surveys. In PSID, better SRH predicted lower spending on physician care only after medicaid coverage was controlled. This reflects the fact that those covered by Medicaid had worse health ratings but also had lower costs, which had suppressed the otherwise negative relationship between SRH and costs. In contrast to the findings above, better SRH predicted higher spending on ambulatory services in AHEAD. While at first this effect appears counterintuitive, it may reflect better health that resulted from greater use of ambulatory services in the past. Also, this appears to be a wealth effect, since SRH became non-significant when SES was controlled in the last hierarchical model. Better SRH also predicted greater use of dental Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 232 care in PSID until SES was controlled. These results highlight the importance of financial status in facilitating access to care, and help to disentangle the complex positive relationships between health, education, financial status, and discretionary health care use. Those reporting more difficulties with Activities of Daily Living (ADL) were more likely to use hospital and nursing home care in both survey samples, and costs among those using hospital/nursing home care also increased with functional disability in AHEAD. ADL disability lost significance as a predictor of nursing home use when insurance was controlled, reflecting the importance of Medicaid in facilitating access to nursing home care among the functionally disabled. Greater ADL difficulties predicted greater use of home health care/special services in AHEAD, and higher home care expenditures among users in both survey samples. In PSID, functional disability became non-significant as a predictor of higher home care expenditures among users when SES was controlled. This revealed that income and wealth increased access to home care services, and had a greater effect on the amount spent than need for services. Total out-of- pocket costs (both including and excluding insurance) in PSID only emerged as significantly higher for those with greater ADL difficulty when insurance was controlled, reflecting the greater likelihood of Medicaid coverage among this group. Specific medical conditions predicted use and out-of-pocket costs for several service types, and selected results are reported and discussed here. Those reporting a stroke had significantly greater odds of using hospital and home care in both PSID and AHEAD, whereas stoke did not predict use and costs for ambulatory services. The greater use of hospital care among those with a stoke likely reflects the need for acute Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 233 intervention at the time of the stroke, while the use of home care reflects the need for support and therapy to treat remaining problems. Those with heart conditions and those with cancer were more likely to use hospital care in both survey samples, and had higher prescription costs and higher total expenditures in AHEAD. Those with cancer in AHEAD also had higher out-of-pocket costs for ambulatory services, but this effect became non significant when SES variables were controlled. This finding is of concern because it suggests that cancer patients with more money may have had better access to care, over used care, or faced higher prices for the same treatments. While arthritis did not predict out-of-pocket costs for any services in PSID, those with arthritis in AHEAD had higher costs for ambulatory services and prescription medications, used more hospital and home care, and had higher total out-of-pocket costs (both excluding and including insurance) than those without arthritis. Those reporting high blood pressure had significantly higher prescription medication expenditures and higher total out-of-pocket costs than those without the condition in both PSID and AHEAD. Those with high blood pressure also had higher physician costs in PSID and greater use of hospital care in AHEAD. Given the high prevalence of both arthritis and high blood pressure among the older population, the increased costs associated with these diseases are an important concern. Those with diabetes were more likely to use home care in both survey samples. In PSID, those with diabetes were also more likely to use hospital care. In AHEAD, diabetes also predicted higher medication expenditures and higher total out-of-pocket costs. Because the complications of diabetes are often associated with poor diseases Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 234 management, the higher out-of-pocket costs associated with diabetes highlight the importance of promoting prevention and proper disease management among those with this condition. For equipment in PSID, those with lung disease had greater odds of use, and those reporting deafness had higher costs. This likely reflects the need for oxygen among those with lung conditions, and the cost of hearing aids among those with hearing problems. Because health is such an immediate determinant of the need for health care, it may be seen as surprising that several medical conditions did not predict use or costs for several service types. One likely explanation for this is that the dummy variables reflecting health conditions included all respondents who reported ever having the condition. Many respondents likely reported conditions that they had had for several years, which did not require expensive treatment during the survey period. Also, the dummy variables for health conditions did not take into account disease severity, which would affect the amount of care needed and its cost. Two other measures of heath in AHEAD were depression, and the number of days spent in bed due to illness. Higher depression scores were associated with greater prescription expenditures. This may reflect drug therapy for depression, although such therapy would be expected to reduce depression scores. Higher depression scores also predicted greater use of hospital/nursing home care and home care/special services. While this is consistent with the literature documenting greater use and higher costs among those who are depressed, the direction of the causal relationship is not clear. Particularly in the case of long-term care, it is less likely that respondents were seeking care because of their Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 235 depression, and more likely that they were depressed because of their disability and need for care. Those who spent ten or more days in bed due to illness had greater use of hospital/nursing home and home care/special services in AHEAD. These effects remained significant in all hierarchical models, and were significant even when controlling for other measures of health, reflecting that days spent in bed represents a dimension of need not captured by other health measures. Education, Income, and Wealth Because coverage of health care costs under Medicare is incomplete, income and wealth were expected to be significant determinants of the amount spent out-of-pocket on health care. In particular, income and wealth were expected to be stronger predictors of out-of-pocket costs for more discretionary services. Similarly, higher levels of education were expected to predict greater use and higher costs for discretionary services. In hierarchical analyses for the chapters above, income, wealth, and education were all included in the same regression model, and each was a significant predictor of out-of- pocket costs in several cases. However, these three variables were closely correlated with one another in both survey samples, with Pearson correlations and between each pair of variables greater than .4 in PSID and greater than .3 in AHEAD. Because of concerns about multicollinearity, additional analyses were run to test education, income, and wealth individually as predictors of costs in hierarchical analyses. Results of these analyses are not shown in tables, but are described below. For these analyses, the first model included only Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 236 the variable of interest (education, income, or wealth) and demographic/geographic variables. In the second model, health variables were added, and in the third model, insurance variables were controlled. In several cases, education, income, and wealth were each significant independent predictors of higher total out-of-pocket costs in all hierarchical models. This was true for total out-of-pocket costs in both survey samples, and for physician and dental care in PSID. In AHEAD, SES variables also independendy predicted higher out-of-pocket costs in all models for ambulatory services, prescription medications and hospital/nursing home care (conditional on use). These findings partially reflect coverage by Medicaid among those with the lowest SES, since this coverage reduces out-of-pocket costs among this group, as seen in Chapter 8. These results are also consistent with the discretionary hypothesis, and suggest that higher SES increased access to care among those above the poverty level, particularly for services at the discretionary end of the continuum. The higher costs among users of hospital/nursing home care among those with higher SES may indicate use of more expensive treatments, and may even serve as evidence of over-use. For hospital use in both survey samples, higher education, income, and wealth predicted lower use until health was controlled. This pattern of findings also occurred in some hierarchical models for nursing home and home care. These findings reveal that lower use of these services was due to superior health among those with higher SES. In some hierarchical models, SES variables predicted higher costs only until insurance was controlled. This occurred for wealth as a predictor of prescription costs and dental costs among users in PSID, which likely reflects the greater coverage of these Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 237 services by Medicaid for those with low levels of wealth. For home health care and other special services in AHEAD, those with higher income and wealth were less likely to be users even when health was controlled. This may reflect participation in means-tested programs covered by Medicaid or the Older American’s Act. Results with Respect to the Conceptual Framework Variance Accounted for by Variables in the Conceptual Framework This section presents and discusses the proportion of variance in out-of-pocket costs accounted for by the four groups of variables in the conceptual framework. As discussed in Chapter 2, need variables have typically been found to be the strongest predictors of health care use in the Andersen model. However, this study primarily examined out-of-pocket costs rather than health care use. Because of the central role of insurance as a mediator between use and out-of-pocket costs, insurance was expected to be a major predictor of expenditures, perhaps accounting for more variance than other variables. Consistent with this hypothesis, insurance emerged as the most important predictor of total out-of-pocket costs, and expenditures for several services. Table 9.1 shows the percentage of variance accounted for by each group of variables as they were successively entered in hierarchical regression models for each type of out-of-pocket costs. These numbers are obtained from hierarchical models as tested in Chapters 6 and 7: demographic variables were entered first, followed by health, insurance, and finally SES variables. The increase in the R2 was calculated for each hierarchical model in Tables 6.4 to 6.7 in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 238 Chapter 6, and is shown below as the percentage of variance explained by each group of variables. For each service type, bold font is used in Table 9.1 to highlight the percentage of variance for the group of variables that accounted for the most variance. As seen in Table 9.1, insurance variables accounted for the largest proportion of variance in total costs (both including and excluding insurance) in both PSID and AHEAD. Insurance also accounted for the largest proportion of variance in out-of-pocket expenditures for ambulatory services and hospital/nursing home care (among users) in AHEAD, and for physician care and dental care (among users) in PSID. Health variables accounted for the largest percentage of the explained variance in out-of-pocket costs for prescription medications in both survey samples, and for outpatient surgery, equipment, and home care among users in PSID. Demographic and health variables accounted for approximately equal portions of variance in nursing home costs among users in PSID. Demographic variables accounted for the largest portion of variance in costs among users of hospital care in PSID, and among users of home care/special services in AHEAD. However, this may reflect the fact that demographic variables were entered first and were correlated with other variables in the model, with a portion of the variance that they account for being shared by other variables. To test for this, separate competing analyses were run in which health variables, insurance variables, and SES variables were each entered into the model first. In almost all cases, these analyses yielded the same rank- ordering of variable groups, with the same group of variables accounting for the most variance in costs for each service. An exception was for hospital costs among users in PSID, where demographic variables accounted for the most variance when they were Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. entered first, but insurance variables accounted for the largest portion of the variance when they were entered first. This suggests that the correlation between demographic variables and insurance coverage was stronger among hospital users than among the whole sample, and that a portion of the variance in costs among this group was shared by these correlated variables. Table 9.1 Percentage of Variance in Out-of-Pocket Costs Accounted for by Each Group of Variables in Conceptual Framework, for Each Service Type in PSID and AHEAD Service Type Data Set Demo graphic variables Health variables Insurance variables Socio economic variables Physician, Therapist, ER Visits PSID 2.6 2.0 6.5 2.8 Ambulatory Services AHEAD 4.0 1.6 5.2 3.9 Prescription Medications PSID 3.6 11.7 6.3 0.7 Prescription Medications AHEAD 1.2 12.4 7.8 0.9 Total Costs (excluding insurance) PSID 8.4 3.6 8 3 3.6 Total Costs (excluding insurance) AHEAD 3.1 4.8 8.8 3.1 Total Costs including insurance PSID 12.8 1.3 21.9 1.7 Total Costs including insurance AHEAD 8.2 1.3 25.1 1.9 Outpatient Surgery (among users) PSID 1.9 5.5 2.6 1.7 Dental Care (among users) PSID 3.1 1.0 8.4 1.8 Equipment (among users) PSID 4.3 6.6 5.3 2.8 Hospital Care (among users) PSED 6.0 1.3 2.5 0.5 Nursing Home Care (among users) PSID 29.3 25.8 293 4.6 Hospital, Nursing Home Care (among users) AHEAD 2.5 3.4 3.9 0.6 Home Care (among users) PSID 8.3 36.8 5.1 14.6 Home Health Care, Special Services (among users) AHEAD 2.4 1.9 1.4 0.5 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 240 The finding that insurance variables accounted for the largest proportion of variance for most services confirms the importance of insurance as a mediator between use and out-of-pocket costs. However, as seen for medigap insurance in Chapter 8, insurance may not always serve to decrease out-of-pocket expenditures. It is important to consider that insurance may increase health care use in addition to reducing costs, and that the net result may be for insurance to increase costs. The implications of these complex insurance findings are further discussed below. The fact that health variables accounted for more of the variance in prescription costs than insurance variables reflects the inconsistent coverage of prescription medications by traditional supplemental insurance. This may also explain the dominance of health in accounting for equipment and home care costs, which may not always have been covered by Medicare or supplemental insurance. For example, insurance may have been less likely to cover equipment such as glasses, hearing aids, and home modifications, or to cover home care for those who did not meet strict Medicare eligibility criteria. The Discretionary Hypothesis A main hypothesis with respect to the conceptual model was that demographic and SES variables would be more likely to independently predict costs for services considered more discretionary. A review of results from final hierarchical models reveals that this hypothesis was partially supported. When controlling for other variables in the model, age and race were not significant predictors of hospital care, but did predict costs for several of the more discretionary services. (Age predicted lower ambulatory costs in AHEAD and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 241 higher long-term care expenditures, and race predicted lower prescription, dental, and total costs.) Gender predicted higher long-term care and prescription costs, but also remained significant as a predictor of lower hospital use, which is not consistent with the discretionary hypothesis. Socioeconomic variables independently predicted higher costs for several ambulatory services, and higher hospital/nursing home costs among users, but did not predict use of hospital care. These results reflect the importance of socioeconomic factors in increasing access to discretionary services. Andersen (1995) comments that the types of factors that significantly predict costs give an indication of whether access to health care is equitable. Andersen contends that access to services could be considered equitable if demographic and need variables accounted for most of the variance in service use, and inequitable if social structure and enabling resources accounted for more of the variance. Consistent with this idea, the results of this study suggest that access to discretionary services among Medicare beneficiaries is somewhat inequitable. As discussed below, efforts to increase access to appropriate care among those with poor patterns of use are necessary to address this problem. Limitations of This Study While this study provided valuable insights on the predictors of older adults’ out- of-pocket costs for different service types, it had several limitations. One limitation is that these data reflect out-of-pocket spending in 1990 and 1994-95, and may not accurately represent current costs. In particular, the cost of prescription medications has been Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 242 growing much faster than the costs of other health services in recent years, partially reflecting higher drug costs, and partially reflecting increases in use (Braden et aL, 1998). Accuracy of Self-Reports Another important limitation of this study was the use of self-reported data. Studies comparing reported self-reports to health records typically find that survey respondents underreport health care use (Eppig & Chulis, 1997; Glandon, Counte and Tancredi 1992; Jobe, White, Kelly, Mingay, Sanchez & Loftus, 1990; McKinlay, 1972; Moeller & Mathiowetz, 1991). Those in worse health and with lower levels of utilization may be more accurate in their reports or even overestimate their use (Glandon et aL, 1992). Fewer studies have examined the accuracy of self-reported medical costs. Eppig & Chulis (1997) comment that major, more expensive medical events are more likely to be remembered and reported at the interview, which may mean that those in better health underestimate their costs. Marquis, Marquis & Newhouse (1976) found that out-of- pocket physician costs were overestimated, but that dental costs were not. The Medicare claims files of respondents in both PSID and AHEAD are now available to be matched with this data, which will allow verification of the accuracy of self- reported use for Medicare-covered services in future studies. Information from Medicare claims files will also allow an examination of the proportion of total health care costs paid by Medicare and other payers (Eppig & Chulis, 1997). However, linking survey data with Medicare claims files may not increase the accuracy of estimated out-of-pocket costs. Those linking the MCBS survey data with Medicare claims files gave precedence to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 243 survey reports of out-of-pocket costs, assuming that they were more accurate than Medicare records (Eppig & Chulis, 1997). Cross-Sectional vs. Longitudinal Data Finally, these analyses were limited by the cross-sectional nature of the data. While they provided a valuable snapshot of the level and burden of out-of-pocket costs over a year (or over two years, in AHEAD), they did not provide information on patterns of out- of-pocket spending over time. High out-of-pocket costs in one or two years will not present the same financial hardship to an individual as persistent high costs over time (Medicare Payment Advisory Commission, 1999). Continued collection of information on out-of-pocket costs in future waves of AHEAD and in the MCBS will allow analyses of trends over time. Longitudinal data will also allow more accurate examination of the relationship between age and out-of-pocket costs. While some interesting non-linear age effects were observed in Chapter 6 of this study, these could not be distinguished from cohort and period effects. Longitudinal data will also yield a clearer picture of the causal order in the relationship between health care use and self-rated health. Conclusions Mean total out-of-pocket health costs including insurance premiums were $1,487 in PSID (1990), and constituted 9.5 percent of household income. Mean reported expenditures for two years (1994-95) in AHEAD were $4,791, and constituted an average of 14 percent of household income over that period. It is important to consider that Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 244 spending was not distributed equally among all persons in each group. Some beneficiaries incurred costs much higher than average, and many incurred costs lower than the average (Coughlin et aL, 1992). Analyses of out-of-pocket costs and burden revealed that a small portion of beneficiaries faced very high and/or burdensome costs. These findings reflect the fact that Medicare does not place a cap on out-of-pocket costs to protect beneficiaries from the risk of catastrophic costs. A cap on out-of-pocket expenses at a specified dollar amount has been suggested as an addition to Medicare to address this problem. However, Feder et al. (1987) caution that an out-of-pocket cap at a fixed dollar amount would not help those with low incomes, who would be burdened by costs below the cap. In fact, the higher the cap, the more likely it would be to help mostly high-income elderly, who could more easily afford to spend up to the cap. To address these problems, Feder et al. (1987) suggest having a cap proportionate to income. They also recommend expansion of the cap to cover things not covered by Medicare (such as prescription medications) and conclude that both the cap and the expansion would be necessary to alleviate the problem of catastrophic expenses among low-income elderly (Feder et al., 1987). A breakdown of spending by service type in both PSID and AHEAD revealed that on average, insurance premiums constituted the largest portion of out-of-pocket costs, and prescription drug costs were the second most expensive item. Estimates of median spending indicate that for some services, costs were not exceptionally high, but were faced by the majority of beneficiaries. Costs which affect a large number of persons are more likely to become a policy concern than those affecting relatively few individuals. As noted Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 245 by the Medicare Payment Advisory Commission (1999), this may be the case with the current debate regarding drug coverage under Medicare. While some beneficiaries face very high prescription medication costs, the debate is as much about the fact that most beneficiaries face out-of-pocket costs as it is about the magnitude of these costs. It is important to consider that the cost of care can be a barrier to access, with potential adverse health consequences. For example, analyses (not shown above) revealed that 4.7 percent of AHEAD respondents (N= 292) reported that because of cost, they had ended up taking less medication than was prescribed for them at some time during the previous two years. Out-of-pocket costs may also have harmful effects on the mental health of older adults and those helping to pay for their care. The level of out-of-pocket health expenditures have been found to predict symptoms of anxiety and depression among those caring for an older spouse (Burton, Schulz, Beach, Jackson, & Hirsch, 1998). This effect was significant controlling for age, race, gender, education, number of care needs, proportion of care for which help was given, and reported caregiving strain. As has occurred several times since the inception of the Medicare program, older adult’s out-of-pocket costs are gaining increasing attention in the research and policy arenas, with a growing consensus that intervention is needed. As discussed in Chapter 1, a main goal of this dissertation was to identify the factors that lead to high out-of-pocket costs, in order to inform the design of focused and effective interventions. The sections below discuss the implications of the main findings with respect to this goal. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 246 Age as a Proxy for Need Analyses of age as predictor of out-of-pocket costs revealed that among the older population, age was an accurate proxy for the need for long-term care services. However, age was not a significant predictor o f use or costs for the majority of other services, even before controlling for health. Thus, these findings suggest that it would be appropriate to use age as a basis for intervention to reduce long-term care expenditures, but not other types of out-of-pocket health costs. In contrast, health variables were often significant predictors of use and costs, and accounted for the largest portion of variance in costs for several services. As seen in Chapter 6, in most cases where age was initially a significant predictor of use or costs, age-related increases in out-of-pocket costs were explained by poor health. These findings underline the importance of research on the prevention and control of age-related health conditions. Schneider (1989; 1999) points out that investment in such research currently constitutes a tiny fraction of total Medicare spending. He stresses the need for more intense research efforts, arguing that increased expenditures on research to help prevent and cure the chronic diseases of old age will be an essential element of reducing projected medical costs. Similarly, Hodgson and Cohen (1999) examine the costs of circulatory diseases and emphasize the importance of lifestyle and medical interventions to prevent and control these conditions. Verbrugge (1995) also stresses that more research is needed on the chronic health conditions of old age. However, she argues that research on nonfatal disabling conditions is particularly underfunded in comparison to biomedical research on fatal diseases. An increase in research on the prevention and appropriate treatment for Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 247 diseases that disable people for many years will also be important in reducing health care costs among the aging population. While the focus on improving health is important, some results in this study point to the need to consider social factors as determinants of health care use and costs, particularly for long-term care. For example, age and gender remained significant as predictors of nursing home use even when health, insurance, and SES were controlled. These results suggest that females and the oldest-old are more likely to end up in nursing homes because they have no one to care for them in the community. Thus, increased and innovative efforts to support older adults in the community would help to decrease nursing home costs, as well as enabling older adults to remain in their preferred setting. As mentioned in Chapter 7, reduced nursing home use among minority groups may indicate that they are more skilled at this type of informal community support. Race, SES and Access to Care Analyses in Chapter 7 revealed that minority race was a significant predictor of lower out-of-pocket costs, particularly for discretionary services. While this meant that minority elders faced less burdensome out-of-pocket costs than Whites, several findings suggested that elderly Blacks and Hispanics faced barriers in access to appropriate care. Access to care was a theme that surfaced several times in the above chapters. Patterns of service use across the discretionary continuum provided evidence for barriers in access to care among the oldest-old, racial minorities, those eligible for Medicaid, and those with Medicare only. Despite their worse health, these groups had lower rates of use for Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 248 discretionary services, but had similar or higher rates of use for hospital care. As noted by Obstbaum (1997), the labelling of services as discretionary does not mean that they are not needed, important types of care. While lower use of preventive and maintenance care may lead to lower out-of-pocket costs in the short run, this pattern of use eventually results in worse health, and may result in higher out-of-pocket costs. These findings underline the need for interventions to improve access to care among groups at risk. Hooyman & Kiyuk (1999) outline a number of cultural, economic, and structural barriers to service utilization among minority elders, and put forth recommendations for addressing these barriers. However, while it is appropriate to design interventions that are culturally sensitive, an important finding in hierarchical analyses was that low SES often explained race differences in out-of-pocket costs. This finding, combined with the finding of poor access among those eligible for Medicaid, suggests that interventions to improve access to care should be targeted toward groups with low SES rather than specific minority groups. Many of the suggestions outlined by Hooyman & Kiyuk (1999) can be applied to low income communities in general. For example, they stress the importance of building on community strengths, and using existing organizational structures such as churches and community centers to link people to health services (Mayers & Souflee, 1990-91; Wood & Wan, 1993). Bringing mobile services into low-income areas may also be effective in increasing utilization of preventive screening and care. For example, Schweitzer, French, Ullman & McCoy (1998) tested the cost-effectiveness of a mobile mammography service, and concluded that it offered an important tool for fighting breast cancer in poor women. They Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 249 stressed the need for more studies on the cost-effectiveness of such interventions. Findings on the effects of poor functional health literacy also point to a need to address this problem. Interventions to improve the functional health literacy of groups at risk, including the oldest-old, minority elders, and those with low SES will help to increase appropriate use of care, resulting in improved health and lower costs in the long run. The Central Role of Insurance Overall, insurance emerged as the most important independent predictor of out-of- pocket expenditures among the older population, most often accounting for the highest proportion of variance in out-of-pocket costs. However, as seen in Chapter 8, the relationship between insurance coverage and out-of-pocket costs was complex, with different types of insurance exerting opposite effects on the level of expenditures. Medicaid coverage consistendy exerted the strongest effect on lowering out-of- pocket costs, indicating that it was providing needed support to vulnerable beneficiaries as intended. Through the Medicaid program, the administradve structure is already in place to use income and wealth as a basis for intervening to help more of those with the highest out-of-pocket burden. Over the past decade, a number of programs have incrementally expanded Medicaid coverage to those who are poor but do not meet stringent Medicaid eligibility criteria (For review see Carpenter, 1998). Future efforts to better meet the needs of those with high health care costs are likely to continue to build on the Medicaid program rather than on Medicare (Feder et aL, 1987), particularly since the recent attempt to increase protection against catastrophic costs through the Medicare program was Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 250 politically unsuccessful (Torres-Gil, 1990). However, the incremental addition of coverage for specific preventive services under Medicare has also been an important policy trend. In addition to making policy changes to extend coverage to those who need it most, it is important to increase participation in existing programs. The high and burdensome costs among a small portion of those eligible for Medicaid in PSID suggests that these respondents, although eligible, were not enrolled in the program. Studies also reveal that a large portion of beneficiaries eligible for partial Medicaid support do not enroll. Neumann, Bemardin, Evans, & Bayer (1995) found that only 41 percent of those eligible for support under the QMB program were enrolled. In response to this concern, Congress recently appropriated $6 Million to the Social Security Administration (SSA) to evaluate approaches to increase SSA participation in outreach and enrollment (Carpenter, 1998). Such efforts to increase beneficiary awareness of available assistance will help to reduce out-of-pocket costs among those most likely to experience burdensome expenditures. Results for medigap and retiree health benefits suggest that they serve primarily to increase access to care rather than to decrease out-of-pocket health costs. As seen in Chapter 8, those with medigap had the highest use of most services, and the highest out- of-pocket costs for most discretionary services. The median burden of total out-of-pocket costs was highest for those with medigap insurance even before the cost of premiums was considered. These results suggest that increasing the rates of private supplemental insurance coverage among Medicare beneficiaries would not lead to lower out-of-pocket burden. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 251 Supplemental insurance does appear beneficial to the extent that it increases access to appropriate care. However, as discussed earlier, it is very difficult to distinguish appropriate use from over-use of care. There is evidence that certain medical procedures are overused among Medicare beneficiaries, with adverse health consequences (see Escarce et aL, 1993). In addition to these adverse effects on health, over-use of care also drives inflation in premium costs, further increasing beneficiaries’ out-of-pocket burden. Continued research on the effectiveness and outcomes of medical procedures will help to identify the scope of the problem of over-use, and hopefully lead to more appropriate treatment of beneficiaries independent from insurance coverage. The possibility that those with supplemental insurance are charged a higher price than those without insurance for the same services also warrants further investigation. Further regulation may be required to limit the practice of price shifting by Medicare providers. Patterns of use and costs for those with employer-sponsored insurance were similar to patterns among those with medigap. The main difference was that insurance premiums were lower for those with eraployer-coverage. However, recent analyses reveal that the number of firms offering retiree health benefits is declining, the portion of employers who require retirees to pay premiums is increasing, and benefits have become less generous (Hewitt Associates LLC, 1997). Analyses in Chapter 8 revealed that membership in a Medicare HMO provides out- of-pocket protection second only to Medicaid. However, HMOs are not available to all beneficiaries, particularly those in rural areas (Haddad & Slass, 1999). Also, a recent trend has been for prepaid plans to withdraw from Medicare or reduce their service areas, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 252 leaving beneficiaries to seek other insurance arrangements (Laschober, Neuman, Kitchman, Meyer & Langwell, 1999). In a recent survey of those involuntarily disenrolled from Medicare HMOs, Laschober et aL (1999) report that 39 percent reported higher monthly premiums, and one-third experienced a decline in benefits. Although two-thirds enrolled in another HMO, one-quarter of these reported paying higher premiums than in their previous HMO and said that they expect to have higher hospital and doctor expenses. Those hardest-hit by the disenrollment were the near-poor, the oldest-old, those in poor health, and racial minorities. With the inception of the Medicare+Choice program under the 1997 Balanced Budget Act, beneficiaries face an array of new insurance options. Analysis of the effects these types of coverage on out-of-pocket costs will provide important insight into the types of coverage most likely to protect against high out-of-pocket costs while preserving appropriate access to care. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 253 References Cited AARP Public Policy Institute & The Lewin Group. (1997, December). Out-of- pocket spending bv Medicare beneficiaries age 65 and older: 1997 projections. (AARP Report No. 9705). Washington, DC: Public Policy Institute, American Association of Retired Persons. AARP Public Policy Institute & The Urban Institute. (1994, April). Coming up short: Increasing out-of-pocket health spending bv older Americans. Washington, DC: Public Policy Institute, American Association of Retired Persons. Andersen, R. M. (1968). A behavioral model of families’ use of health services. (Research Series No. 25) Chicago: Center for Health Administration Studies, University of Chicago. Andersen, R. M. (1995). Revisiting the behavioral model and access to medical care: Does it matter? Journal of Health and Social Behavior. 36. I-10. Andersen, R. M., Giachello, A. L., & Aday, L. A. (1986). Access of Hispanics to health care and cuts in services: A state-of-the-art overview. Public Health Reports. 101. 238-252. Andersen, R. M., & Newman, J. F. (1973). Societal and individual determinants of medical care utilization in the United States. Milbank Memorial Fund Ouarterlv/Health and Society. 51. 95-124. Angel, J. L., & Hogan, D. P. (1994). The demography of minority aging populations. In Minority elders: Five goals toward building a public policy base (2n d ed.) Washington, DC: The Gerontological Society of America. Anonymous. (1988). IG Seeks Clearer Regulations On Fraud And Abuse. Hospitals. 62H0Y 79-80. Atkins, G. L. (1989). The economic status of the oldest old. Milbank Memorial Fund Quarterly. 63(21. 395-419. Bach, P. B., Cramer, L. D„ Warren, J. L„ & Begg, C. B. (1999). Racial Differences in the Treatment of Early-Stage Lung Cancer. New England Journal of Medicine. 341. 1198-205. Baquet, C. R. (1988). Cancer prevention and control in the black population. In J. S. Jackson (Ed.) The Black American elderly. New York: Springer. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 254 Bazargan, M., Bazargan, S., & Baker, R. S. (1998). Emergency department utilization, hospital admissions, and physician visits among elderly African American persons. The Gerontologist. 38. 25-36. Berki, S. E. (1986). A look at catastrophic medical expenses and the poor. Health Affairs. Winter. 138-145. Blendon, R. J., Aiken, L. H., Freeman, H. E„ & Corey, C. R. (1989). Access to medical care for black and white Americans: A matter of continuing concern. JAMA. 261(2), 278-285 Blendon, R. J., Scheck, A. C., Donelan, K., Hill, C. A., Smith, M., Beatrice, D., & Altman, D. (1995). How white and African Americans view their health and social problems: Different experiences, different expectations. JAMA. 273. 341-346. Blustein, J. (1995). Medicare coverage, supplemental insurance, and the use of mammography by older women. New England Journal of Medicine. 332(17). 1138-1143. Braden, J. & Beauregard, K. (1994, February). Health status and access to care of rural and urban populations (AHCPR Publication No. 94-0031). National Medical Expenditure Survey Research Findings 18, Agency for Health Care Policy and Research. Rockville, MD: Public Health Service. Braden, B. R., Cowan, C. A., Lazenby, H. C., Martin, A. B., McDonnell, P. A., Sensenig, A. L., Stiller, J. M., Whittle, L. S., Donham, C. S., Long, A. M., & Stewart, M. W. (1998). National health expenditures, 1997. Health Care Financing Review. 20(1). 83- 126. Brandt, J., Spencer, M., & Folstein, M. (1988). The telephone interview for cognitive status. Neuropsychiatry. Neuropsychology, and Behavioral Neurology. 1. 111- 117. Brown, R. S., Bergeron, J. W., Clement, D. G., Hill, J. W., & Retchin, S. M. (1993). The Medicare risk program for HMOs: Final summary report on findings from the evaluation. (Final report under HCFA Contract no. 500-88-0006). Princeton, NJ: Mathematica Policy Research. Buczko, W. (1989). Hospital utilization and expenditures in a Medicaid population. Health Care Financing Review. 11(1). 35-47. Burton, L. C., Schulz, R., Beach, S., Jackson, S., & Hirsch, C. (1998). High out-of-pocket health costs are associated with mental morbidity among elderly spousal caregivers. Abstract Book/Association for Health Services Research. 15. 258-259. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 255 Cagney, K. A., & Agree, E. M. (1999). Racial differences in skilled nursing care and home health use: The mediating effects of family structure and social class. Journal of Gerontology: Social Sciences. 54B(4). S223-S236. Callahan, D. (1987). Setting limits: Medical goals in an aging society. New York: Simon & Schuster. Caplan, L. S., Wells, B. L., & Haynes, S. (1992). Breast cancer screening among older racial/ethnic minorities and Whites: Barriers to early detection. The Journals of Gerontology. 47fSpecial Issue). 101-110. Carpenter, L. (1998). Evolution of Medicaid coverage of Medicare cost sharing. Health Care Financing Review. 20(2). 11-18. Cartwright, W. S., Hu, T. W„ & Huang, L. F. (1992). Impact of varying Medigap insurance coverage on the use of medical services of the elderly. Applied Economics. 24(5), 529-539. Christensen, S., Long, S. H., & Rodgers, J. (1987). Acute health care costs of the aged Medicare population: Overview and policy options. The Milbank Quarterly. 65(3). 397-425. Christensen, S., & Shinogle, J. (1997). Effects of supplemental coverage on use of services by Medicare enrollees. Health Care Financing Review. 19. 5-17. Chulis, G. S., Eppig, F. J., Hogan, M. O., Waldo, D. R„ Amett, R. H. m . (1993). Health insurance and the elderly: Data from MCBS. Health Care Financing Review. 14(3). 163-181. Chulis, G. S., Eppig, F. J., Poisal, J. A. (1995). Ownership and average premiums for Medicare supplementary insurance policies. Health Care Financing Review. 17Q). 255-275. Conigliaro, J. (1999 June). Racial variation in the use of cardiovascular procedures in the VA. Paper presented at the annual meeting of the Association for Health Services Research, Chicago, IL. Conrad, D. A., Grembowski, D., & Milgrom, P. (1987). Dental care demand: insurance effects and plan design. Health Services Research. 22(3). 341-367. Coughlin, T. A., Liu, K., & McBride, T. D. (1992). Severely disabled elderly persons with financially catastrophic health care expenses: Sources and determinants. The Gerontologist. 32f3L 391-403. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 256 Cox, D. F. & Hogan, C. (1997). Biased selection and Medicare HMOs: Analysis of the 1989-1994 experience. Medical Care Research and Review. 54(3). 259-274. Cromwell, J. & Burstein, P. (1985). Physician losses from Medicare and Medicaid discounts: How real are they? Health Care Financing Review. 6(4). 51-68. Crystal, S., Johnson, R. W., & Kumar, R. (1998). Out-of-pocket health care costs among elderly Medicare beneficiaries. Paper presented at the annual meeting of the Gerontological Society of America, Philadelphia, PA. Cunningham, P. J., Clancy, C. M., Cohen, J. W., & Wilets, M. (1995). The use of hospital emergency departments for nonurgent health problems: A national perspective. Medical Care Research and Review. 52(4). 453-474. Dallek, G. (1996, October). The crushing costs of Medicare supplemental policies. Washington, DC: Families USA. Davis, M., Poisal, J., Chulis, G., Zara.bozo, C., & Cooper, B. (1999). Prescription drug coverage, utilization, and spending among Medicare beneficiaries. Health Affairs. 18(1), 231-243. Dingham, J. J., Colangelo, L., Tian, W., Jones, J., Smith, R., Wickerham, D. L., & Wolmark, N. (1999). Outcomes Among African-Americans and Caucasians in Colon Cancer Adjuvant Therapy Trials: Findings From the National Surgical Adjuvant Breast and Bowel Project. Journal of the National Cancer Institute. 91(22). 1933-1940. Dor, A., & Holahan, J. (1989). Urban-rural differences in Medicare physician expenditures. Washington, DC: Urban Institute. Dowd, B., Christianson, J., Feldman, R., Wisner, C., & Klien, J. (1992). Issues regarding health plan payments under Medicare and recommendations for reform. The Milbank Quarterly. 70(31. 423-453. Druss, B. G„ Rohrbaugh, R. M., & Rosenheck, R. A. (1999). Depressive symptoms and health costs in older medical patients. American Journal of Psychiatry. 156(3). 477-479. Duan, N., Manning, W. G. Jr., Morris, C. N., & Newhouse, J. P. (1983). A comparison of alternative models for the demand for medical care. Journal of Business & Economic Statistics. 1(2). 115-126. Duka, W. (1999, September). Drugs stir up Medicare debate: Backing for new benefit grows, though ‘99 outlook is uncertain. AARP Bulletin. 40(8). 3, 24-25, 32. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 257 Eppig, F. & Chulis, G. S. (1997). Matching MCBS and Medicare Data: The Best of Both Worlds. Health Care Financing Review. 18(3). 211-229. Eppig, F. J., & Poisah J. A. (1997). Measuring the health status of Medicare beneficiaries: 1995. Health Care Financing Review. 18(4). 125-131. Escarce, J. J., Epstein, K. R„ Colby, D. C., & Schwartz, J. S. (1993). Racial differences in the elderly’s use of medical procedures and tests. American Journal of Public Health. 83. 948-954. Evashwick, C., Rowe, G., Diehr, P., & Branch, L. (1984). Factors explaining the use of health care services by the elderly. Health Services Research. 19f3). 357-382. Feder, J., Moon, M., & Scanlon, W. (1987). Nibbling at catastrophic costs. Health Affairs. Winter. 5-19. Freiman, M. P. (1998). The demand for healthcare among racial/ethnic sub populations. Health Services Research. 33(41. Friedman, L. S. (1984). Microeconomic Policy Analysis. New York: McGraw-HilL Gazmararian, J. A., Baker, D. W., Williams, M. V., Parker, R. M., Scott, T. L., Green, D. C., Fehrenbach, S. N., Ren, J., & Koplan, J. P. (1999). Health literacy among Medicare enrollees in a managed care organization. JAMA. 281(6). 545-551. Gibson, M., & Brangan, N. (1998a). Out-of-pocket spending on health care bv women age 65 and over in fee-for-service Medicare: 1998 projections. Publication No. FS72. Washington, DC: AARP Public Policy Institute. Gibson, M., & Brangan, N. (1998b, November). Out-of-pocket spending and financial barriers to care by disabled Medicare beneficiaries under age 65. Paper presented at the annual meeting of the Gerontological Society of America, Philadelphia, PA. Geiger, H. J. (1996). Race and health care— an American dilemma? New England Journal of Medicine. 335Q P . 815-816. Glandon, G. L., Counte, M. A., & Tancredi, D. (1992). An analysis of physician utilization by elderly persons: Systematic differences between self-report and archival information. Journal of Gerontology: Social Sciences. 5. S245-S252. Goldzweig, C. L., Mittman, B. S., Carter, G. M., Donyo, T., Brook, R. H., Lee, P., & Mangione, C. M. (1997). Variations in cataract extraction rates in Medicare prepaid and fee-for-service settings. JAMA. 277(221. 1765-1768. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 258 Gomick, M. E., Eggers, P. W., Reilly, T. W., Mentnech, R. M., fitterman, L. K., Kucken, L. E„ & Vladek, B. C. (1996). Effects of race and income on mortality and use of services among Medicare beneficiaries. New Eneland Journal of Medicine. 335. 791- 799. Government Accounting Office. (1997, August). Medicare: Fewer and lower cost beneficiaries with chronic conditions enroll in HMOs. (GAO/HEHS Publication No. 97- 160). Washington: GAO. Government Reform, Committee on. (1999). Prescription drug pricing in the 1s t Congressional District in Oregon: Drug companies profit at the expense of older Americans. U.S. House of Representatives. Washington, DC: Minority Staff Report Prepared for Representative David Wu. Grana, J., & Stuart, B. (1996-97). The impact of insurance on access to physician services for elderly people with arthritis. Inquiry. 33(4). 326-338. Gross, D., & Brangan, N. (1999). Medicare beneficiaries and prescription drug coverage: Gaps and barriers. (Issue Brief No. 39). Washington, DC: AARP Public Policy Institute. Haddad, K. (1999, November). Hard to swallow: Rising drug prices for America’ s seniors (Families USA Publication No. 99-107). Washington, DC: Families USA Haddad, K., & Slass, L. (1999, September). Rural neglect: Medicare HMOs ignore rural communities (Families USA Pub. # 99-105). Washington, DC: Families USA. Hatten, J. M., & Connerton, R. E. (1986). Urban and rural hospitals: How do they differ? Health Care Financing Review. 8(2). 77-78. Health Care Financing Administration. (1990). Medicare and Medicaid Data Book. 1990. (HCFA Publication No. 03314). Baltimore, MD: HCFA. Health Care Financing Administration (1999). Medicare and vou 2000. (HCFA Publication No. 10050). Baltimore, MD: HCFA. Heckman, J. J. (1976). The common structure of statistical models of truncation, sample selection and limited dependent variables and a simple estimator of such models. Annals of Economic and Social Measurement. 5. 475-492. Heeren, T. J., Lagaay, A. M., Hijmans & Rooymans, H. G. M. (1991). Prevalence of dementia in the ’oldest old’ of a Dutch community. Journal of the American Geriatrics Society. 39. 755-759. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 259 Hewitt Associates LLC. (1997). Retiree health trends and implications of possible Medicare reforms. Washington, DC: Kaiser Family Foundation. Hill, M. (1992). The panel study of income dynamics: A user guide. Newbury Park, CA: Sage. Hobbs, F. & Damon, B. L. (1996). 65+ in the United States. Washington, DC: Department of Commerce, Bureau of the Census. Hodgson, T. A. & Cohen, A. J. (1999). Medical care expenditures for selected circulatory diseases: Opportunities for reducing national health expenditures. Medical Care. 37(10). 994-1012. Holahan, J., & Zedlewski, S. (1992). Who pays for health care in the United States? Implications for health system reform. Inquiry. 29. 231-248. Hooyman, N. & Kiyak, H. A. (1999). Social gerontology: A multidisciplinary perspective (5th Ed.). Boston: Allyn & Bacon. Hurd, M. D. (1989). The economic status of the elderly. Science. 244. 659-664. Hurd, M. D., & McGarry, K. (1997). Medical insurance and the use of health care services by the elderly. Journal of Health Economics. 16. 129-154. Javitt, J. C„ McBean, A. M., Nicholson, G. A., Babish, J. D., Warren, J. L., & Krakhauer, H. (1991). Undertreatraent of glaucoma among Black Americans. New England Journal of Medicine. 325. 1418-1422. Jobe, J. B., White, A. A., Kelley, C. L., Mingay, D. J. & Sanchez, M. J. (1990). Recall strategies and memory for health-care visits. The Milbank Quarterly. 68(21. 171- 189. Johansson, B., Zarit, S. H., & Berg, S. (1992). Changes in Cognitive Functioning of the oldest old. Journal of Gerontology: Psychological Sciences. 47. P75-80. Jones, W. Jr., & Rene, A. A. (1994). Barriers to health service utilization and African Americans. In I. L. Livingston (Ed.) Handbook of Black American health: The mosaic of conditions, issues, policies, and prospects. Westport, CT: Greenwood. Kaiser Family Foundation, The Henry J. (1999, April). Medicare’s role for Latinos. The Medicare Program: The Kaiser Medicare Policy Project. Washington, DC: Kaiser Family Foundation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 260 Katon, W., VonKorff, M., Lin, E., Lipscomb, P., Russo, J., Wagner, E., & Polk, E. (1990). Distressed high utilizers of medical care: DSM-IH-R diagnoses and treatment needs. General Hospital Psychiatry. 12. 355-362. Khandker, R. K L , & McCormack, L. A. (1999). Medicare spending by beneficiaries with various types of supplemental insurance. Medical Care Research and Review. 56(2): 137-55. Kington, R., Rogowski, J., & Lillard, L. (1995). Dental expenditures and insurance coverage among older adults. The Gerontologist. 35(41. 436-443. Krout, J. A. (1983). Correlates of service utilization among the rural elderly. The Gerontologist. 23(51. 500-504. Lannin, D. R., Mathews, H. F., Mitchell, J., Swanson, M. S., Swanson, F. H., Edwards, M. S. (1998). Influence of socioeconomic and cultural factors on racial differences in late-stage presentation of breast cancer. JAMA. 279. 1801-1807. Laschober, M. A., & Olin, G. L. (1996). Health and health care of the Medicare population: Data from the 1992 Medicare Current Beneficiary Survey. Rockville, MD: Westat, Inc. Laschober, M. A., Neuman, P., Kitchman, M. S., Meyer, L. & Langwell, K. M. (1999). Medicare HMO withdrawls: What happens to beneficiaries? Health Affairs. 18f6L 150-157. LaVeist, T. A. (1994). Beyond dummy variables and sample selection: What health services researchers ought to know about race as a variable. Health Services Research. 29(1), 1-16. Lee, J. A., Baker, C. S., Gehlbach, S., Hosmer, D. W. & Reti, M. (1998). Do Black elderly Medicare patients receive fewer services? An analysis of procedure use for selected patient conditions. Medical Care Research and Review. 55('3L 314-333. Levit, K. R., Lazenby, H. C., Braden, B. R., Cowan, C. A., McDonnell, P. A., Sivarajan, L., Stiller, J. M., Won, D. K., Donham, C. S., Long, A. M., & Stewart, M. W. (1996). National health expenditures, 1995. Health Care Financing Review. 18(1). 175- 214. Levit, K. R., Lazenby, H. C., Braden, B. R., Cowan, C. A., Sensenig, A. L., McDonnell, P. A., Stiller, J. M., Won, D. K., Martin, A. B„ Sivarajan, L., Donham, C. S., Long, A. M., & Stewart, M. W. (1997). National health expenditures, 1996. Health Care Financing Review. 19fD. 161-200. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 261 Lillard, L. A., Rogowski, J., & Kington, R. (1997). Long-term determinants of patterns of health insurance coverage in the Medicare population. The Gerontologist, 37(3), 314-323. Lillard, L. A., Rogowski, J., & Kington, R. (1999). Insurance coverage for prescription drugs: effects on use and expenditures in the Medicare population. Medical Care. 37f9L 926-36. Linden, M., Horgas, A. L., Gilberg, R., & Steinhagen-Thiessen, E. (1997). Predicting health care utilization in the very old: the role of physical health, mental health, attitudinal and social factors. Journal of Aging and Health. 9(11. 3-27. Liu, K., Wall, S., & Wissoker, D. (1997). Disability and Medicare costs. The Milbank Quarterly. 7564). 461-493. LoGiudice, D., Waltrowicz, W., Ames, D„ Brown, K., Burrows, C., & Flicker, L. (1997). Health care costs of people referred to an aged care assessment team: The effect of cognitive impairment. Australian and New Zealand Journal of Public Health. 21. 311-316 Manning, W. G., Newhouse, J. P., Duan, N., Keeler, E. B., Benjamin, B., Leibowitz, B. A., Marquis, M. S., & Zwanziger, J. (1988). Health insurance and the demand for medical care: Evidence from a randomized experiment. American Economic Review. 77(31. 251-277. Manton, K. G., Woodbury, M. A., & Stallard, E. (1995). Sex differences in human mortality and aging at late ages: The effect of mortality selection and state dynamics. The Gerontologist. 35. 597-608. Markides, K. S., & Black, S. A. (1996). Race, ethnicity and aging. In R. H. Binstock and L. K. George (Eds.), Handbook of aging and the social sciences (4th ed., pp. 153-170). San Diego, CA: Academic Press. Marquis, K. H„ Marquis, M. S., & Newhouse, J. P. (1976). The measurement of expenditures for outpatient physician and dental services: Methodological findings from the Health Insurance Study. Medical Care. 14(11), 913-931 Mayers, R. S. & Souflee, L. (1990-91). Utilizing social support systems in the delivery of social services to the Mexican-American elderly. Journal of Applied Social Sciences. 15. 31-50. McCall, N., Rice, T., Boismier, J., & West, R. (1991). Private health insurance and medical care utilization: Evidence from the Medicare population. Inquiry. 28. 276-287. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 262 McClendon, M. J. (1994). Nonlinear relationships. Chapter 6 in Multiple regression and causal analysis. Itasca, IL: F. E. Peacock. McKinlay, J. B. (1972). Some approaches and problems in the study of the use of services— An overview. Journal of Health and Social Behavior. 13. 115-396. Medicare Payment Advisory Commission. (1999, June). Beneficiaries’ financial liability and Medicare’s effectiveness in reducing personal spending. Chapter 1 in Report to Congress: Selected Medicare issues. Washington, DC: MedPAC. Meerding, W. J., Polder, J., Bonneux, L., Koopmanschap, M., & van der Maas, P. (1998). Health-care costs of ageing. The Lancet. 351. 140. Merrill, R. M., Brown, M. L., Potosky, A. L., Riley, G., Taplin, S. H.; Barlow, W., Fireman, B. H. (1999). Survival and treatment for colorectal cancer Medicare patients in two group/staff health maintenance organizations and the fee-for-service setting. Medical Care Research and Review. 56(2). 177-96. Mitchell, J., & Krout, J. A. (1998). Discretion and service use among older adults: The behavioral model revisited. The Gerontologist. 38(2). 159-168. Moeller, J. & Mathiowetz, N. (1991). Correcting errors in prescription drug reporting: A critique. Health Affairs. Spring. 210-211. Moon, M. (1991). Measures of health care spending and the elderly. (Urban Institute Publication No. 6132-01). Washington, DC: The Urban Institute. Moon, M. (1992). Increasing burdens of health care spending on the elderly. (Urban Institute Publication No. 6132-02). Washington, DC: The Urban Institute. Moon, M. (1999). Growth in Medicare spending: What will beneficiaries pav? New York: The Commonwealth Fund. Moon M., Kuntz, C., & Pounder, L. (1996, December). Protecting low income medicare beneficiaries. The Commonwealth Fund. Mueller, C., Schur, C., & O'Connell, J. (1997). Prescription drug spending: The impact of age and chronic disease status. American Journal of Public Health. 87. 1626-1629. Mutchler, J. E., & Burr, J. A. (1991). Racial differences in health and health care service utilization in later life: The effect of socioeconomic status. Journal of Health and Social Behavior. 32. 342-356. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 263 Mutran, E., & Ferraro, K. F. (1988). Medical need and use o f services among older men and women. The Gerontologist. 43(5). S162-S171. National Academy on an Aging Society. (1998, October). Understanding health literacy: New estimates of costs of inadequate health literacy. Presented at the Pfizer Conference on Health Literacy “Promoting health literacy: A call to action.” Washington, DC. Neugarten, B. L. (1982). Policy for the 1980’ s: Age or need entitlement? In B. L. Neugarten (Ed.) Age or need? Public policies for older people (pp. 19-32). Beverly Hills: Sage. Neuman, P. J., Bemardin, M. D., Evans, W. N., & Bayer, E. J. (1995). Participation in the Qualified Medicare Beneficiary Program. Health Care Financing Review. \1(TS. 169-178. Nyman, J. A., Sen, A., Chan, B. Y., & Commins, P. P. (1991). Urban/rural differences in home health patients and services. The Gerontologist. 31(41. 457-466. Obstbaum, S. A. (1997). Should rates of cataract surgery vary by insurance status? JAMA. 277I22L 1807-1808. Olin, G. L., & Liu, H. (1998). Health and health care of the Medicare population: Data from the 1994 Medicare Current Beneficiary Survey. Rockville, MD: Westat, Inc. Paringer, L„ Bluck, J., Feder, J., & Holahan, J. (1979). Health status and use of medical services: Evidence on the poor, the black, and the rural elderly. Washington, DC: The Urban Institute. Peris, T. T. (1997). Acute care costs of the oldest old. Hospital Practice. 32. 123-4, 129-32, 137. Peris, T. T., & Wood, E. R. (1996). Acute care costs of the oldest old: they cost less, their care intensity is less, and they go to nonteaching hospitals. Archives of Internal Medicine. 156. 754-760. Pohlmeier, W„ Ulrich, V., Konstanz, U., & Mannheim, U. (1995). An econometric model of the two-part decisionmaking process in the demand for health care. Journal of Human Resources. 30(2L 339-361. Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement. 1. 385-401. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 264 Rice, T. (1989). The use, cost, and economic burden of nursing home care in 1985. Medical Care. 271121. 1133-1147. Rice, T. & Gabel, J. (1986). Protecting the elderly against high health care costs. Health AfFairs. Fall. 5-21. Rogowski, J., Lillard, L. A., & Kington, R. (1997). The financial burden of prescription drug use among elderly persons. The Gerontologist. 37(4). 475-482. Roos, N. P., Shapiro, E., & Tate, R. (1989). Does a small minority of elderly account for a majority of health care expenditures?: A sixteen-year perspective. The Milbank Quarterly. 6713/4’ ). 347-369. Rossiter, L. F. & Wilensky, G. R. (1982). Out-of-pocket expenditures for personal health services. (DHHS Publication No. (PHS)82-3332). National Health Care Expenditures Study Data Preview 13. National Center for Health Services Research. Hyattsville, MD: DHHS. Rowland, D., & Lyons, B. (1996). Medicare, Medicaid, and the elderly poor. Health Care Financing Review. 18(2). 61-69. Rubin, R. M„ & Koelln, K. (1993a). Determinants of household out-of-pocket health expenditures. Social Sciences Quarterly. 74('4’ > . 721-735. Rubin, R. M„ & Koelln, K. (1993b). Out-of-pocket health expenditure differentials between elderly and non-elderly households. The Gerontologist. SStD . 595-602. Rubin, R. M., Koelln, K., & Speas, R. K (1995). Out-of-pocket health expenditures by elderly households: Change over the 1980s. Journal of Gerontology: Social Sciences. 50B. S291-S300. Schneider, E. L. (1989). Options to control the rising health care costs of older Americans. JAMA. 261(6). 907-908. Schneider, E. L. (1999). Aging in the third Millennium. Science. 283. 796-797 Schneider, E. L., & Guralnik, J. M. (1990). The aging of America: Impact on health care costs. JAMA. 263071. 2335-2340. Schoen, C., Neuman, P., Kitchman, M., Davis, K., & Rowland, D. (1998). Medicare beneficiaries: A population at risk. Menlo Park, CA: The Henry J. Kaiser Family Foundation and The Commonwealth Fund. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 265 Schulman, K. A., Berlin, J. A., Harless, W., Kemer, J. F., Sistrunk, S., Gersh, B. J., Dube, R., Taleghani, C. K., Burke, J. E„ Williams, S., Eisenberg, J. M., & Escarce, J. J. (1999). The effect of race and sex on physicians’recommendations for cardiac catheterization. New England Journal of Medicine. 340(8). 618-26. Schweitzer, M. E., French, M. T., Ullmann, S. G., & McCoy, C. B. (1998). Cost-effectiveness of detecting breast cancer in lower socioeconomic status African American and Hispanic women through mobile mammography services. Medical Care Research and Review. 55Q):99-115. Simon, G. E., & Katzelnick, D. J. (1997). Depression, use of medical services and cost-offset effects. Journal of Psychosomatic Research. 42(4). 333-344. Simon, G. E., Ormel, J., VonKorff, M., & Barlow, W. (1995). Health care costs associated with depressive and anxiety disorders in primary care. American Journal of Psychiatry. 152f3). 352-357. Simon, G. E., VonKorff, M., & Barlow, W. (1995). Health care costs of primary care patients with recognized depression. Archives of General Psychiatry. 52. 850-856. Sofaer, S., & Davidson, B. N. (1990). Mness-episode approach: Costs and benefits of Medigap insurance. Health Care Financing Review. 11(4). 121-131. Sofaer, S., & Kenney, E. (1989). Financial consequences of joining a Medicare HMO: An application of the illness episode approach to estimating out-of-pocket costs. Journal of Health Politics. Policy and Law, 14. 565-585. Soldo, B. J., Hurd, M. D., Rodgers, W. L., & Wallace, R. B. (1997). Asset and health dynamics among the oldest-old: An overview of the AHEAD study. Journal of Gerontology: Psychological and Social Sciences. 52B (Special Issue), 1-20. Social Security Administration. (1997). Social security handbook (13th Ed.). (SSA Publication No. 65-008) Washington, DC: SSA. Stahl, S., & Gardner, G. (1976). A contradiction in the health care delivery system: Problems of access. The Sociological Quarterly. 17. 121-130. Stone, R. I. (1999, November). Helping Professionals help seniors: The role of the Center for Medicare Education. Symposium presented at the annual meeting of the Gerontological Society of America, San Francisco, CA. Stum, M. S., Bauer, J. W., & DeLaney, P. J. (1996). Out-of-pocket home care expenditures for disabled elderly. Journal of Consumer Affairs. 30H V 24-47. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 266 Stum, M. S., Bauer, J. W„ & Delaney, P. J. (1998). Disabled elders’out-of-pocket home care expenses: Examining financial burden. The Journal of Consumer Affairs, 32(1), 82-105. Taylor, A., & Banthin, J. (1994). Changes in out-of-pocket expenditures for personal health services: 1977 and 1987. (AHCPR Publication No. 94-0065). National Medical Expenditure Survey Research Findings 21, Agency for Health Care Policy and Research. Rockville, MD: Public Health Service. Thomas, C., & Kelman, H. R. (1990). Unreimbursed expenses for medical care among urban elderly people. Journal of Community Health. 15C2). 137-149. Tobin, J. (1958). Estimation of relationships for limited dependent variables. Econometrica. 26. 24-36. Torres-Gil, F. (1990). Seniors react to the Medicare catastrophic bill: Equity or selfishness? Journal of Aging and Social Policy. 2(1). 1-8 Torrey, B. B. (1989). Sharing increasing costs on declining income: The visible dilemma of the invisible aged. Milbank Memorial Fund Quarterly. 63(2). 377-394. Uniitzer, J., Patrick, D. L., Simon, G., Grembowski, D., Walker, E., Rutter, C., & Katon, W. (1997). Depressive symptoms and the cost of health services in HMO patients aged 65 years and older. JAMA. 277(20). 1618-1623. Verbosky, L. A., Franco, K. N., & Zrull, J. P. (1993). The relationship between depression and length of stay in the general hospital patient. Journal of Clinical Psychiatry. 54(5), 177-181. Verbrugge, L. M. (1989). The twain meet: empirical explanations of sex differences in health and mortality. Journal of Health and Social Behavior. 30( 3), 282- 305. Verbrugge, L. M. (1995). Seven chronic conditions: Their impact on US adults' activity levels and use of medical services. American Journal of Public Health. 85(2). 173-182. Waldo, D. R., & Lazenby, H. C. (1984). Demographic characteristics and health care use and expenditures by the aged in the United States: 1977-1984. Health Care Financing Review. 6(1). 1-29. Waldo, D. R., Sonnefeld, S. T., McKusik, D. R., & Arnett, R. H. m . (1989). Health expenditures by age group, 1977 and 1987. Health Care Financing Review. 10(4). 111 - 120. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 267 Wallace, S. P. (1991). The political economy of health care for elderly blacks. In M. Minkler & C. L. Estes (Eds.) Critical perspectives on aging: The political and moral economy of growing old (pp. 253-270). Amityville, N.Y.: Baywood. Waxman, H. A. (1999, July-August). Should a prescription drug benefit be added to Medicare as a federal government program? Yes: Let’s keep Medicare’s promise. AARP Bulletin. 40(71. 29, 31. Ways and Means, Committee on. (1996). 1996 Green Book: Background material and data on the programs within the jurisdiction of the Committee on Wavs and Means. U.S. House of Representatives. Washington, DC: USGPO. Weaver, J. L., & Inui, L. T. (1975). Information about health care providers among urban low-income minorities. Inquiry. 12. 330-343. Weinberger, M., Gold, D. T., Divine, G. W., Cowper, P. A., Hodgson, L. G., Schreiner, M. S., & George, L .K. (1993). Expenditures in caring for patients with dementia who live at home. American Journal of Public Health. 83. 338-341. Williams, D. R. (1997). Race and health: Basic questions, emerging directions. Annals of Epidemiology. 7(51. 322-333. Wolfe, J. R., & Goddeeris, J. H. (1991). Adverse selection, moral hazard, and wealth effects in the Medigap insurance market. Journal of Health Economics. 10. 433- 459. Wolinsky, F. D., Aguire, B. E„ Fann, L. J., Keith, V. M., Arnold, C. L„ Niederhauer, J. C., & Dietrich, K. (1990). Ethnic differences in the demand for physician and hospital utilization among older adults in major American cities: Conspicuous evidence of considerable inequalities. The Milbank Quarterly. 67. 412-449. Wolinsky, F. D., Coe, R. M„ Miller, D. K., Prendergast, J. M., Creel, M. J., & Chavez, M. N. (1983). Health services utilization among the noninstitutionalized elderly. Journal of Health and Social Behavior. 24. 325-337. Wolinsky, F. D., Mosely, R. R. H, & Coe, R. M. (1986). A cohort analysis of the use of health services by elderly Americans. Journal of Health and Social Behavior. 27. 209-219. Wood, J. B. & Wan, T. (1993). Ethnicity and minority issues in family caregiving to rural black elders. In C. Barresi and D. Stull (Eds.), Ethnic elderly and long-term care (pp. 39-56). New York: Springer. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 268 Wyszewianski, L. (1986a). Families with catastrophic health care expenditures. Health Services Research. 21(51. 617-633. Wyszewianski, L. (1986b). Financially catastrophic health care expenditures. Health services Research. 25(51. 617-634. Zuvekas, S. H., & Weinick, R. M. (1999). Changes in access to health care, 1977- 1996: The role of health insurance. Health Services Research. 34. 271-279. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 269 Appendix A Questions regarding medical conditions in PSID and AHEAD 1) PSID Elderly Health Supplement: The mail-in questionnaire contained the question below (only the conditions used in this study are listed). Respondents circled all the conditions that they had. As far as you know, do you have any of the following conditions/problems? Arthritis or rheumatism Cancer (except skin cancer) Major paralysis or neurologic problems (including stroke) Using a cardiac pacemaker Heart failure or enlarged heart Asthma or other severe lung problems, such as chronic bronchitis, emphysema Ulcer (duodenal, stomach or peptic) Chronic inflamed bowel, enteritis, colitis Trouble seeing Diabetes High blood pressure or hypertension Deafness Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 270 2) AHEAD wave 2: In wave 1 of AHEAD, respondents were asked about medical conditions using the questions below. In wave 2, similar questions were asked. Respondents who had reported the condition in wave 1 were asked to verify this report, and those who had not reported the condition in wave 1 were asked: Since we last talked to you, that is (Wave 1 date), has a doctor told you that you have (condition)? Conditions were described in the same way as wave 1, with exceptions noted below. Cancer: Has a doctor ever told you that you have cancer or a malignant tumor, excluding minor skin cancers? Diabetes: Wave 1: Do you have diabetes? Wave 2: Has a doctor told you that you have diabetes or high blood sugar? Lung Disease: Wave 1: Not including asthma, has a doctor ever told you that you have chronic bronchitis or emphysema? Wave 2: Has a doctor told you that you have chronic lung disease such as chronic bronchitis or emphysema? High Blood Pressure: Has a doctor ever told you that you have high blood pressure or hypertension? Has a doctor ever told you that you had a stroke? Has a doctor ever told you that you had a heart attack, coronary heart disease, angina, congestive heart failure, or other heart problems? Wave 1: Have you had a heart attack in the last five years? Wave 2: Have you had a heart attack in the last two years? Congestive Heart Failure: Wave 2 only: In the last two years (since we last talked to you) has a doctor told you that you have congestive heart failure? Arthritis: Wave 1: During the last 12 months, have you seen a doctor specifically for arthritis or rheumatism? Wave 2: Have you had or has a doctor told you that you had arthritis or rheumatism? Stroke: Heart Problems: Heart Attack: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Using network perspective to examine the organization of community -based elder care systems across four communities

PDF

Marriage and health: The effect of marital status on health and the mechanisms by which marital status affects health among elderly people

PDF

Evidence of age-based rationing of health care to the elderly: The effect of age and death on the use of health care services by elderly enrollees in an HMO

PDF

The construct validity of the Health Utilities Index in patients with chronic respiratory disease in a managed care population

PDF

Variations in physician practice patterns for eye care under the National Health Insurance of Taiwan

PDF

The extent and effectiveness of nursing home regulation in the 50 states

PDF

Service evaluation and quality improvement of a geriatric evaluation and management program

PDF

Assessment of prognostic comorbidity in hospital outcomes research: Is there a role for outpatient pharmacy data?

PDF

A new paradigm to evaluate quality-adjusted life years (QALY) from secondary database: Transforming health status instrument scores to health preference

PDF

Effects of a formulary expansion of the use of SSRIs and health care services by depressed patients in the California Medicaid program

PDF

Heart disease among middle -aged and elderly persons in the United States: Trends and a multistate model

PDF

The 1996 Health Insurance Portability and Accountability Act (HIPAA) and the 1997 Balanced Budget Act (BBA): Implications for organizational change and delivery of care in cardiology group practices

PDF

The measurement of physical functioning in a longitudinal study of Americans 70 years of age and older

PDF

Assessing the cost implications of combined pharmacotherapy in the long term management of asthma: Theory and application of methods to control selection bias

PDF

Legislative intent and aging policy in Los Angeles County: the relative effects of age versus need in public conservatorship

PDF

Physician adherence to national hypertension guidelines in an elderly Medicaid population

PDF

The perceived credibility of health-related Web sites within an information -seeking context

PDF

Performance management in health care in Iceland

PDF

Leading change initiatives: communication and bounded agency in a health care organization

PDF

Controlling for biases from measurement errors in health outcomes research: A structural equation modeling approach

Asset Metadata

Creator Stewart, Susan Tracy (author)

Core Title Out -of -pocket health expenditures by older adults in relation to age, race, and insurance

Degree Doctor of Philosophy

Degree Program Gerontology/Public Policy

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Gerontology,health sciences, health care management,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-69555

Unique identifier UC11338273

Identifier 3018035.pdf (filename),usctheses-c16-69555 (legacy record id)

Legacy Identifier 3018035.pdf

Dmrecord 69555

Document Type Dissertation

Rights Stewart, Susan Tracy

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA