University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

Download

Open document

Flip pages

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 UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will 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. Each original is also photographed in one exposure and is included in reduced form at the back of the book. 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 UMI directly to order. UMI A Bell & Howell Information Company 300 North Zeeb Road, Ann Arbor MI 48106-1346 USA 313/761-4700 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. 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 By Freddi I. Segal-Gidan 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 Polity) May 1997 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 9733132 Copyright 1997 by Segal-Gidan, Freddi I. All rights reserved. UMI Microform 9733132 Copyright 1997, by UMI Company. AH rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES. CALIFORNIA 90007 This dissertation, written by F.ns.cW.L.S8flAJ..T&j.fl3J!i..................................... under the direction of hax. 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 Date . ..? .9 .:.. .1 9.9.7. DISSERTATION. COMMITTEE Chairperson Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. DEDICATION DEDICATED WITH LOVE TO MY GRANDMOTHER, PAULINE POLINSKY SEGAL BORN IN ODESSA, RUSSIA, MAY 19, 1899. SHE HAS TAUGHT ME ABOUT AGING FIRSTHAND. SHE IS MY HISTORY AND INSPIRATION, AND I HER LEGACY. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. FOREWORD No work, no struggle, is completed by an individual acting alone. This research is the culmination of a journey that began many years ago and the route has been influenced by many along the way. I could not have traveled the road and reached this place without the love, support, advice, and friendship of many. Initially I must acknowledge my parents, Colette and Harold Segal, who have always, through word and action, encouraged me to strive for and be whatever it was I dreamed of. And now my father will have, at long last, his daughter the doctorl I must acknowledge the help, guidance, and criticism of my doctoral committee; Edward Schneider, Jeffrey McCombs, and Martin Levine. They encouraged and supported me throughout the many iterations of the research design, execution, and final printed form. Also, to my adviser, Phoebe Leibig, who has helped me to believe in myself by believing in me. My co-cohort students, now good friends and gerontologic colleagues, Linda Wray Valentine Villa, and Carl Reynolds paved the way and continued to assure me there was plenty of room on the path for me to follow. Without them I know that I would have been diverted many times. Many physicians with whom I have had the pleasure of working over the past 14 years as a physician assistant (PA) have encouraged me in whatever direction I have wanted to go and supported my further education and training. From my first physician supervisor/colleague, Cranford Scott, I learned not just medicine, but the art of patient care and the business of private practice. Robert Blackman, Marshall Byrnes, Gary iii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Sutter, and the other physicians at California Primary Physicians (now Healthcare Partners) shared their innovative practice and believed in my abilities to become a geriatric specialist. Ken Brummel-Smith, who was my first geriatric teacher, remains the epitome of a caring provider and teacher to both students and patients. And ultimately, Helena C. Chui, whose support has been unfailing. She remains my inspiration and role model of that unique individual who can successfully combine clinical expertise, research, and teaching. I would be remiss not to mention those other physicians from my training as a health associate, Archie Golden and Dennis Carlson in particular, who have influenced me directly and from afar over time and geography. Many physician assistant colleagues have encouraged me to help blaze the trail of PAs as health service researchers, teachers, and policy makers. Jim Cawley, who was my adviser so many years ago and whose pursuit of a Ph.D. has remained an inspiration, though unknown to him; Jack Runyan, PA, Ph.D., who began and achieved the doctoral journey before me; and my dear friend, Lauren Geisler, whose love and friendship has become a centering part of my life. Staff at the Leonard Davis School, especially Pauline Abbott, have throughout the years offered support, encouragement, and a ready shoulder when needed; my classmate from the Master's Program and into our doctoral training, Jeff Hyde, whose data programing expertise made this work possible; and my friend, brought into my life by fete and timing, Sandy Reynolds, who showed me personally that a career continues to build and unfold in ways chartered and unanticipated, all have my complete appreciation. My co-workers at the Rancho Los Amigos Geriatric Neurobehavior Center have encouraged, cheered, and assisted in so many ways over the years, it is hard to imagine how I could have continued to work and complete my degree without them. Thank yous iv Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. to Barbara Smith, Marilyn Carmack, Andele Ireland, Maureen Wimderlick, Lee Willis, Sue Collins, Bryan Kemp, Edith Marquez, and Scott Linus. Many thanks to the individuals at the HMO who were eager to help an optimistic and struggling graduate student several years ago. They advocated for me and without their support I would never have gained access to the data that served as the body of this dissertation. And my gratitude to the others in the HMO over the last several years who have assisted me both directly and indirectly. This research, in its early stages, was partially funded by the AARP-Andrus Foundation. Their early support of my research proposal gave me not only the necessary seed money, but also the confidence, to pursue an idea and the data to study it. Although it has been difficult for them, without the love, hugs, and adoration of my husband Jon (no matter how grudgingly given) and my two wonderful sons, Jared and Hayden, who entered my life during graduate school, I could not have continued. I know they will be glad to have me back from the computer and more available for their lives. Finally, I must recognize the patients and families I have had the privilege to care for as a PA the past 18 years. They remain my inspiration and keep me focused on what really matters as they struggle with a health care system that at times produces miracles, but more often frustration. I can only hope my efforts in some way make a difference both in their lives, and over time in a more ethical and equitable system for us all. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. TABLE OF CONTENTS DEDICATION ............................................................................................... ii FOREWORD ............................................................................................... iii LIST OF TABLES............................................................................................ vii ABSTRACT ............................................................................................... ix INTRODUCTION............................................................................................ 1 CHAPTER 1 REVIEW OF THE LITERATURE....................................... 5 CHAPTER 2 METHODOLOGY ............................................................... 37 CHAPTER 3 DESCRIPTIVE RESULTS ................................................... 53 CHAPTER 4 PREDICTORS OF UTILIZATION OF HEALTH CARE SERVICE.................................................. 74 CHAPTER 5 PREDICTORS OF COST OF CA RE................................... 98 CHAPTER 6 PREDICTING DEATH AND HIGH COST DYING ____ 126 CHAPTER 7 DISCUSSION OF RESULTS AND THEIR IMPLICATIONS.................................................... 137 REFERENCES .............................................................................................. 156 APPENDIX A Construction of the Non-Respondent Population......................... 170 B New Enrollee Questionnaire......................................................... 172 C Simple Logistic Regression Last 3 Months of L ife....................... 177 D Simple Logistic Regression Last Month of L ife............................ 186 vi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. LIST OF TABLES Table 2.1 Demographic Variables................................................................ 42 Table 2.2 Part A Claims D ata........................................................................ 44 Table 2.3 Health Care Questionnaire Variables .......................................... 48 Table 3.1 Age, Gender, and Survival Characterisitics of the Sample Population.............................................................. 54 Table 3.2 Five Year Age Cohort Distribtution of the Total and Respondent Population........................................................... 55 Table 3.3 Death Rate in the Total Sample and Respondent Populations Compared to National Census D a ta ............................................ 57 Table 3.4 Proportion of the Population that Died by Five Year Cohort. . . . 58 Table 3.5 Cost Distribution by Type of Service for Total Sample and Respondent Populations......................................................... 59 Table 3.6 Total Costs by Five Year Age Cohort ......................................... 62 Table 3.7 Average Cost by Five Year Age Cohort of Decedents and Survivors with Part A Use .................................................... 65 Table 3.8 Cost Distribution of Decedents and Survivors among the Total and Respondent Populations.............................. 67 Table 3.9 Diagnoses by Frequency of Occurrence in the Total Sample and Respondent Populations with Part A C laim s......................... 68 Table 3.10 Questionnaire Responses as a Percentage of the Population .... 70 Table 4.1 Logistic Regression Odds Ratio of Having Part A Costs During the 12 Month Study Period ............................................ 76 Table 4.2 Normalized Odds Ratios for Total Costs of M ales...................... 78 Table 4.3 Logistic Regression Odds Ratio of Having Hospital Costs for 12 Month Study P e rio d ................................................................ 81 Table 4.4 Normalized Odds Ratios for Hospital Costs of M ales................. 84 vii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.5 Logistic Regression Odds Ratio of Having LTC Costs for the 12 Month S tudy.......................................................................... 87 Table 4.6 Normalized Odds Ratios for Long-term Care Costs of Males . . . 89 Table 4.7 Logistic Regression Odds Ratio on Having Part A Costs for 12 or 3 Month Periods for Respondents................................. 93 Table 4.8 Logistic Regression Odds Ratio on Having Hospital Costs for 12 or 3 Month Period for Respondents................................... 94 Table 4.9 Logistic Regression Odds Ratio on Having Long-term Care Costs for 12 or 3 Month Period for Respondents................................... 95 Table 5.1 OLS Regression Simple Models on Total Costs for Claimants During die 12 Month Study ......................................................... 100 Table 5.2 Computed OLS Regression Results in Dollars for Total Costs of Males with Claims for Part A Service U s e .............................. 102 Table 5.3 OLS Regression Simple Models on Hospital Costs for Claimants During the 12 Month Study ......................................................... 104 Table 5.4 Computed OLS Regression Results in Dollars for Hospital Costs of Males with Claims for Hospital Services................................. 107 Table 5.5 OLS Regression Simple Models on Long-term Care Costs for Claimants During the 12 Month Study................................... 108 Table 5.6 Computed OLS Regression Results in Dollars for Long-term Care Costs of Males with Claims for Long-term Care Services I l l Table 5.7 Complex OLS Regression Models for Total Costs of Those with Claims for Phrt A Services .................................................. 115 Table 5.8 Complex OLS Regression Models for Hospital Costs for Those with Part A Claims ..................................................................... 119 Table 5.9 Complex OLS Regression Models of Long-term Care Costs for TTiose with Part A C laim s....................................................... 123 Table 6.1 Odds Ratio on the Probability of Dying During the 12-month Study ....................................................... 127 Table 6.2 Odds Ratio on the Probability of Dying During the 12-month Study for Those Who Used Part A Services............... 129 Table 6.3 Odds Ratio on the Probability of Dying with Costs in Excess of $10,000 During the 12-month Study........................................ 133 viii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ABSTRACT The elderly consume a disproportionate share of health care resources and health care costs are higher for those elderly who die compared to those who survive. Thus, many researchers, policy makers, and health professionals conclude that excessive resources are being expended on elderly persons who will die anyway. However, studies of Medicare payments show a consistent pattern of decreasing costs with increasing age of death (Lubitz & Riley 1993; Riley et al., 1986) leading others to suggest that implicit age-based rationing may already be occurring. Is there evidence for age-based rationing of health care? This thesis explores the effect age and death have on health care resource utilization by elderly enrollees in an HMO. It is hypothesized that if age-based rationing occurs, it might more readily be detected when utilization patterns are examined within a managed care system where incentives are to avoid costly care that might be considered excessive or unnecessary. Utilization of Medicare Part A services by new elderly enrollees (N = 28,536) in an HMO was examined for one year from time of enrollment. Logistic regression models were used to examine the probability of service utilization and ordinary least squares (OLS) regression models were then estimated for the costs incurred by those who utilized services. Logistic regression models were also developed to predict death and high-cost death. Age and death were found to have an independent effect on the use and cost, and to jointly affect the probability, of utilization of services. Specifically, death has a much stronger effect than age on utilization, but the effect of death decreases with age. Costs also increase with both age and death, yet costs are highest for younger aged elderly decedents compared to oldest-old decedents. Age alone is a weak significant predictor of death, but loses both ix Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. significance and strength when health status is controlled for by using data for discharge diagnoses and self-reported health status. Therefore, the results indicate that age is not being used to limit access to care, but rather is being used to limit aggressive care for the dying. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. INTRODUCTION The growing aged population and their increased use of health care services has heightened concerns about the escalating cost of providing health care, now and into the foreseeable future. As a result, decisions regarding the utilization of health care resources at the end of life are increasingly being viewed as within the province of policy makers. Considerable research documents the high cost of health care associated with both increasing age and death (Gomick, McMillan, & Lubitz, 1994; N. Roos, Shapiro, & L. Roos, 1986; Scitovsky, 1984, 1994). Death at any age is costly, but it is more costly for the elderly compared with those of younger ages (Waldo & Lazenby, 1984). The high cost of health care for many individuals in the last period of life has led to calls for health care rationing by age (Callahan, 1987; Daniels, 1986, 1988; Lamm, 1987; Veatch, 1988). Whether we should continue current policies that implicitly ration access to care, or develop explicit policies, based on age or other criteria, has become a topic of debate among providers and policy makers. Most of the data on the use and cost of health care by the elderly are derived from care received in the traditional fee-for-service environment where there are incentives to provide additional, and often more costly, care. In an effort to contain escalating costs, more and more Americans, including elderly beneficiaries in the Medicare program, are receiving their health care from alternative delivery systems. Under these variations of managed care the incentives are to provide necessary, but not excessive, care. This raises questions about what impact such systems will have on 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. utilization and costs among the elderly, particularly regarding the potential rationing of health care services for the very old or sick. This research examines the effect that increasing age and death have on the use of health care services during a 1-year period of time for a cohort of elderly enrollees in a Medicare Health Maintenance Organization (HMO). The study addresses five research questions: 1 . What is the pattern of utilization of health care services, especially for high cost health care, by elderly enrollees in an managed health care system? 2. What is the effect of age on utilization of health care services among the elderly enrollees in a managed health care system? 3. Does the relationship between age and the use of high cost health care among elderly enrollees differ for survivors and decedents? 4. Is there any evidence that the influence of age on the amount or mix of health care services provided to elderly enrollees changes when controlling for health status or the type of service provided? 5. Does the existence of an advanced directive effect utilization of services for elderly enrollees in a managed care system? The first chapter presents a review of the literature on the utilization of health care resources by the elderly. Studies on the relationship between age and the cost of care and the cost of dying are reviewed. The applicability of these studies to managed care and the changing health care market place are discussed. Rationing within the health care system is then reviewed with attention to proposals calling for age-based limitations on care. The second chapter describes the methodology employed to answer the research questions posited. The methods utilized to construct the sample populations are 2 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. described. The construction of the data set, utilizing claims-based data and variables derived directly from a mailed health interview survey conducted by the HMO on new enrollees, are discussed. The multivariate statistical techniques used to model predictors of health care utilization, the cost of care for those who used services, and death are then reviewed. Chapter 3 presents a description of the sample population of new elderly enrollees in the HMO used for this study. The demographic characteristics of the 5,926 respondents to the health interview survey and the 22,610 other new elderly enrollees in the health plan who comprised the sample population are compared. The frequency of diagnoses and procedures derived from discharge data on those members of both population groups who used institutional (Medicare Part A) services are presented as measures of the population's health status. Answers to the health interview questionnaire by the respondents prior to entering the health plan provide a more detailed description of this sub-sample population's health status and life-style characteristics. Logistic regression models were developed to examine predictors of health care service utilization. These results are presented in Chapter 4. Particular attention is paid to the effect of age and death on the probability of use of any institutional service (hospital and long-term care) over three different time periods (1 year, 3 months, and 1 month). These models were then used to compute normalized odds ratios for men ages 65 and 85 who died or survived. This was done to concretely illustrate the differential effect of age and death on the probability of utilization of different types of health care services over different periods of time for elderly survivors and decedents of different ages. 3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ordinary least square (OLS) regression models were then developed in order to examine predictors of cost for those who used institutional services. These results are presented in Chapter 5. Again, the impact of age and death on costs is highlighted for total, hospital, and long-term care costs over 12, 3, and 1 months. To demonstrate the impact of age and death on cost, the models were used to calculate costs for young-old and oldest-old survivors and decedents. Chapter 6 examines predictors of death and high cost death using logistic regression models. The focus is on the effect of age on predicting death and how this is impacted by indicators health and health status derived from discharge diagnoses and respondents' questionnaire data. A discussion of the results with emphasis on the policy implications is contained in Chapter 7. The patterns of utilization of health care services among elderly enrollees in an HMO are very similar to those reported in previous research from the fee-for-service sector. A small proportion of the population was responsible for a disproportionate amount of health care use and cost. Age among survivors was found to be associated with increasing costs. Decedents had higher costs than survivors, but those who died at younger old ages had higher costs for all types of care than more aged elderly who died. Whether these findings represent age-based rationing of care and by whom is discussed. Finally, areas for future research resulting from this study are suggested. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER 1 REVIEW OF THE LITERATURE This chapter presents the conceptual framework on which this dissertation is based. The policy issues that provide the motivation for this work are first briefly presented. The literature on health care utilization is reviewed as it relates to age and death is then reviewed, with particular attention to cost of care at the end of life and high cost care for the elderly. Patterns of health care use in health maintenance organizations (HMOs) are then presented, again with emphasis on studies involving the elderly. Finally, a discussion of rationing of health care is undertaken in an effort to understand how age and death figure into the discussion of this controversial subject. Policy Issues The impact of our aging society on the health care system has become a major topic of debate and a focus of public policy discussions during the past decade. Considerable attention by both researchers and policy makers has focused on the disproportionate use of health care services by the elderly and the increased costs associated with dying. From this research many have concluded that excessive amounts of health care resources are expended to keep old people alive, even when the prognosis is poor and they will die anyway, despite intensive care. This has led to concerns that precious health care resources are being "wasted" on the elderly that could better be spent on others (the uninsured and children). Calls for cost containment 5 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. strategies that target high users, particularly the elderly, cite this research and * reasoning as justification for limiting health care past some pre-determined age (Callahan, 1989; Daniels, 1988; Smeeding, 1991; Veatch, 1988). However, it is this author’ s experience as a clinician that the number of very old people being maintained on ventilators or receiving other expensive life-saving measures is quite small. In fact, aggressive, heroic care that might be undertaken for younger individuals is not as common an occurrence for the very old as many believe. This disparity between clinical practice and the conclusions drawn from the population-based research raises questions about the role of age and death in resource utilization within the elderly population. Demographic characteristics and health status together have been shown to influence health service utilization. However, the role that advancing age and impending death play in the utilization of health care by older persons, and the costs of these services, is not clear. For example, among elderly decedents age has been shown to be inversely related to health care costs (Lubitz & Riley, 1993; Riley, Lubitz, Prihoba, & Stevenson, 1986). Elderly individuals who die before age 75 in the fee- for-service sector have been found to incur more costly health care services in the period preceding death than those elderly over age 85. However, in survivors the pattern is reversed: increasing age is associated with increasing costs. These results suggest that for individuals of equal health status (i.e., impending death) age has a different effect. Recent policy changes encourage Medicare enrollees to join managed health care plans in an attempt to contain health care costs. However, the effect of managed health care plans on patterns of utilization and cost among the elderly is not well 6 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. understood. Specifically, the relationship between age, death, and health care costs among the elderly has not been examined in the managed care setting. These issues call for an examination of the policy proposal for age-based rationing of health care and to determine if such a policy can be used as a method of cost containment. It has been suggested that age-based rationing may already be occurring implicitly, but evidence to support such a contention is not definitive. If trends in the fee-for-service sector represent over-utilization due to economic incentives, then the move to managed care should eliminate, or at least minimize, these patterns and possibly make age-based rationing more apparent. To understand the findings that have influenced this policy issue and to determine if claims that medical resources are being misused for the dying elderly are accurate, a careful review and evaluation of the research related to this issue was undertaken, with particular attention to use in the last period of life. Age and Death Effects Considered Jointly A number of studies have described the joint effect of increasing age and death on health care services utilization and costs. In a seminal piece of research on the relationship between health care costs, age, and death, Lubitz and Prihoda (1984) found that expenditures were higher for Medicare beneficiaries who died than for survivors. However, medical care expenditures declined with age for elderly decedents, but increased with age among elderly survivors. Lubitz and Riley (1993) examined Medicare payments per person year for decedents and survivors by age. Their findings also demonstrated a declining cost pattern for decedents, while payments for survivors increased with increasing age, thus narrowing the gap between the average payment for decedents and survivors. 7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Scitovsky (1989) studied medical care in the last 12 months of life of members of a multi-specialty medical group. She found that total expenditures for those who died after age 80 were about 80% of decedents in the younger old age groups. Different patterns of medical care services utilization accounted for these age group differences. The oldest old who died received more supportive services such as nursing home and home health care while hospital care comprised a larger proportion of the costs for younger old age persons who died. A review of Medicare expenditures for elderly enrollees in 1967, 1975, 1979 and 1982 by analysts at the Health Care Financing Administration (Riley et al., 1986) found that average Medicare expenditures for persons who die have increased at about the same rate as Medicare expenditures for those who survive. They concluded that the data indicated that "expensive methods of prolonging the lives of terminally ill patients are not the culprit behind increasing Medicare program expenditures" (p. 62). Another study from the Health Care Financing Administration (HCFA) (Gomick, McMillan, & Lubitz, 1993) examined patterns of Medicare payments for three cohorts over a 16-year period. In contrast to survivors, per capita payments for decedents decreased consistently as the age of the cohort increased. Advancing age, in and of itself, was not found to be the cause of rising Medicare payments over an older person's later life span. Payments rose rapidly as death approached for the elderly at all ages in each of the three cohorts, with higher Medicare payments for decedents in younger cohorts. A recent study that estimated total lifetime Medicare expenditures at age of death found that, among Medicare beneficiaries, lifetime payments increased with age at death up to age 89 and then leveled off (Lubitz, Beebe, & Baker, 1995). However, they also found that Medicare payments actually decreased with increasing age in the 8 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. years prior to death. Average payments per beneficiary in the 3 years prior to death were not related to age at death. This led them to conclude "that any limitation of care related to age takes place in the t>me near death" (p. 1002). A study of health care utilization prior to death among Canadians found that those dying at older ages had higher total costs (Roos et al., 1987). This was due primarily to nursing home utilization by the very old, a component of care that is not captured in studies utilizing Medicare data, because most nursing home care is custodial in nature and falls outside the payment parameters of Medicare. It is important to note that the studies cited above, and the vast majority of those reviewed below, whether cross-sectional or longitudinal, are solely descriptive. Age and Health Care Costs The high medical expenses of the elderly have received attention due to their disproportionate contribution to the growing national expenditures devoted to health. In 1990 national health care expenditures comprised 12.4% of the GNP, a marked increase from 5.3% in 1960. Health care costs for the elderly accounted for about one- third of the total annual health expenditures, although the elderly constitute just under 13% of the population. This focus on the elderly has been due to the fact that a large portion of the medical bills of the elderly are paid for through publicly funded programs (Medicare and Medicaid). Yet, it should be noted that only about 60% of total health care costs for those over age 65 are paid for by public funds (Medicare pays 40%, Medicaid and other government programs pay about 18%) (Davis & Rowland, 1986; Waldo etal., 1987). It is well-documented that use of health care services increases with age (Fisher, 1980; Waldo & Lazenby, 1984). The elderly have the highest average expenditures 9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. for all types of health care except dental services (Hahn & Lefkowitz, 1992). They are hospitalized more often and have longer lengths of stay, have more physician visits a year than the rest of the population, and occupy 90% of all long term care beds (Evashwick, Rowe, Diehr, & Branch, 1984; Gagnon, 1978; Polliak & Shavitt, 1977). Further, as a cohort ages, its rate of hospital and physician use has been shown to increase (Wolinsky, Mosely, & Coe, 1986). Increasing life expectancy, especially at the oldest ages, along with increased prevalence of chronic diseases, has led to estimates of escalating health care costs for the elderly (Schneider & Guralnik, 1990). A small proportion of the older population is responsible for the majority of health care costs of the elderly. The Congressional Budget Office (CBO) of the United States estimated that average total costs for acute care services were 12 times higher for those who required hospitalization during the year than those who did not. They found that health care expenditures for the elderly were higher and concentrated in the 22% of Medicare enrollees who required hospitalization each year (CBO, 1987). In a longitudinal study of patterns of utilization of health care resources in Canada, Roos and Shapiro (1981) found hospital usage more related to age than ambulatory care. The very old had a greater risk of hospitalization, but only marginally higher usage of physician visits, than younger elderly. In a study of hospital use among the elderly, charges and length of stay increased with increasing age (Rosenthal & Landefeld, 1993). However, multi-variate analyses demonstrated that these differences were explained by severity of illness. In an even more recently published study of acute hospital costs by the elderly, those over age 90 were found to have lower costs per stay even though they have longer lengths of stay, than younger elderly (Peris & Wood, 1996). 10 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Nursing home costs account for about one-fifth (21%) of total health care expenditures of older persons. Scitovsky (1989) found that average expenses for nursing home care increased with age among decedents: $326 on average for decedents under age 65, $1,262 for those 65 to 79, and $5,407 on average for those over age 80. A study of the British Columbia Long Term Care program utilizing structural equation modeling demonstrated a differential effect of age on health services utilization by gender (Ellencweig & Pagliccia, 1994). Older males had increased hospital use and decreased procedures. Females, regardless of age, had a marked decrease in both hospital use and procedures. Using Medicare expenditure data, Fuchs (1984) demonstrated that adjusting for age-sex differences in survival status eliminates much of the age-related increase in health care expenditures. In fact, for those over age 80, Fuchs found a decline rather than an increase in expenditures. His data, however, were limited to Medicare reimbursement and did not include non-reimbursable expenditures for pharmaceuticals and nursing home care, the latter being heaviest for the oldest age group. Effect of Dying on Cost and Utilization Much of the elderly's health care utilization has been shown to be associated with dying, rather than simply aging. Approximately 70% of all deaths in the United States occur among people age 65 and older (Schick & Schick, 1994). Studies in both the U.S. and Canada have demonstrated that a small minority of the population accounts for a majority of health care expenditures and that use of health care services intensifies as death approaches. This has been shown to be true across all population age groups, including the young (Starfield, van den Berg, Steinwachs, Katz, & Horn, 11 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1979) and aged (Garfinkle, Riley, & Iannachione, 1988; Lubitz & Prihoda, 1984; Riley et al., 1986; Roos & Shapiro, 1981). Studies over the past two decades have shown the proportion of elderly decedents who used costly services prior to death to be fairly constant. Lubitz and Prihoda (1984) reported that 5.9% of Medicare beneficiaries who died in 1978 accounted for 28% of Medicare expenditures. Similar findings were reported in a later study by Riley et al. (1986) who found that 5% of decedents accounted for 27.3% of Medicare reimbursements. A study of the total medical care expenditures of older persons living in the community in 1980 conducted by Kovar (1983) found that 5% of the elderly who either died or were institutionalized during the year accounted for 22% of total expenditures while they lived in the community. An unpublished study by the Health Care Financing Administration, Office of Research and Demonstration in 1991 was conducted to update the findings reported above. They examined Medicare-covered services and payments of decedents in 1985 and again reported similar results. During 1986 decedents comprised 5.3% of the study population and accounted for 26.7% of expenditures, excluding hospice care. Several other studies have examined medical costs of Medicare beneficiaries who died in relation to total expenditures during the calendar year of study. An early study of Medicare beneficiaries who died in 1967 reported that the 5% of beneficiaries who died that year accounted for 22% of total Medicare expenditures for the year (Pira & Lutins, 1973). Another similar study examined deaths of Medicare beneficiaries during 1979 and found that the 5% of beneficiaries who died accounted for 21% of total Medicare expenditures for that year (Riley & Lubitz, 1989). Recent published data from the Medicare program (Helbing, 1992) demonstrate that the effect of death on health care costs persists even in this era of cost 12 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. containment. In 1990, the 6.6% of Medicare beneficiaries who died accounted for 21.5% of all Medicare payments. The average payment per decedent was four times greater than the average survivor payment. Another approach has been to examine the cost of care of elderly persons for different periods of time prior to death. Regardless of the time period selected research has repeatedly concluded that Medicare expenses in the last year of life are heavily concentrated in the last weeks of life. Lubitz and Prihoda (1984) found in their study that 77% of all expenses of decedents occurred in the last 6 months of life. Just under one-half (46%) of the costs in the last year of life were spent in the last 60 days and 30% in the last month. McCall (1984) also found that a large portion of the health care expenditures for those who die occurred in the last month of life. Riley et al. (1987) found that 36% of decedents' costs were incurred in the last 30 days of life. While it is true that elderly people who die incur greater expenditures than those who do not die, most elderly people who die do not incur high expenditures. In Lubitz and Prihoda’ s (1984) study of Medicare enrollees, they found that 69% of enrollees who died incurred expenditures less than $5,000 and 45% incurred less than $2,000 in Medicare expenditures in the year before death. Furthermore, as previously discussed, the average reimbursement for persons who died decreased with increasing age. Reimbursement for persons over age 85 who died was about one-half of the average reimbursement for persons age 67 to 69 who died. Medicare payments per enrollee have repeatedly been shown to be substantially higher for decedents than survivors (Lubitz & Riley, 1993; Riley et al., 1987). The most recent research published on the subject confirmed earlier studies that decedents do indeed comprise a disproportionately expensive group. In a 16-year longitudinal study that followed three cohorts of Medicare beneficiaries (Gomick, McMillan, 13 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Lubitz, 1993), per capita Medicare payments for beneficiaries who died were double those for survivors. Further, estimated costs for the final 12 months of life accounted for one-third (33%) of a cohort's lifetime Medicare payments. Different cost patterns have been reported for elderly decedents and survivors. Lubitz and Prihoda (1984) found that 92% of Medicare enrollees who died in 1978 used some Medicare-covered services in their last year of life compared to 58% of enrollees who survived. Further, Medicare payments for persons in their last year of life were six times as great as payments for survivors. In a study of Medicare decedents during 4 separate years, Lubitz and Riley (1993) reported that a very small proportion (3.1%) of decedents incurred no Medicare payments in the last year of life compared with almost one-quarter (22.9%) of survivors. In contrast, in a study of medical care utilization for 4 years before death in Manitoba, Canada, N. Roos, Montgomery, and L. Roos (1987) found death at older ages to be more, rather than less, costly. Decedents over age 85 had total health expenditures 29% higher than those ages 75 to 84 and 74% higher than those age 65 to 74. Use of hospital services was somewhat lower for those over age 85, but offset by higher utilization rates of nursing home and home health care services. The impression exists among many that a greater share of health care resources is being devoted to care for dying patients than in the past. A study by Lubitz and Riley (1993), using Medicare data over 12 years, soundly refutes this notion. They demonstrated that "Medicare payments for decedents as a percentage of the total Medicare budget changed little" (p. 1092). In a study of the last year of life of Medicare beneficiaries, Riley et al. (1986) found that the cause of death was an important determinant of reimbursement level and therefore directly affected total costs. They found considerable variation in 14 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. patterns of expenditures for decedents with different diseases. Those with malignant neoplasms had high levels of reimbursement for the 2 years preceding death. In comparison, older persons with more chronic conditions (diabetes, COPD, nephritis, and liver disease) demonstrated patterns of consistently high use over 5 to 6 years preceding death. Scitovsky and Capron (1987), in reviewing studies of medical care at the end of life, concluded "it appears unlikely that more than a relatively small part of the high medical expenses at the end of life ... are due to excessively aggressive care of terminally ill patients. Most of these expenses seem to be for the care of very ill, but not necessarily dying patients, care that... is relatively conservative, yet expensive" (p. 73). In reviewing utilization and costs in the period before death, it is important to keep in mind that among the elderly who incur significant health care costs, most have multiple chronic conditions and have ill health. Therefore, it is not surprising that they consume a large proportion of health care resources. Health care expenses are higher for those whose conditions require on-going medical care and should be expected to be higher than for those who are not ill. Most studies of health care costs for the elderly in the United States have been limited to examining Medicare beneficiaries and those services reimbursable under Medicare. A report to Congress in 1991 on this topic noted that the "data on health care spending at the end of life is rather limited and out of date" (King, 1991). Most of the studies available used data that is over a decade old, prior to the widespread use of some newer technologies and before the institution of many cost containment mechanisms, particularly Medicare's prospective payment system and the enrollment of Medicare enrollees in managed health care plans. 15 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. High Cost Illness Studies of high cost illness consistently demonstrate that a considerable portion of these costs are incurred by individuals in the period just prior to their death. High cost illness has been defined as "instances where expenditures exceed an amount considered to be large, without regard to source of payment or ability to pay" (Wyszewianski, 1986, p. 383). However, no consistent definition of high cost illness has been developed. The term has been used to reflect a portion of expenditures based on either total costs or income. A commonly employed approach has been to examine the total distribution of cases and focus on the top 5% or 10% (Berki, Lepowski, Wyszewianski et al., 1985; Lubitz & Prihoda, 1984; Schroeder, Showstack, & Roberts, 1979). Recent Medicare data support previous findings that a small proportion of enrollees account for a majority of program expenditures (Helbing, 1992). In 1990 8.4% of all Medicare enrollees accounted for 64.4% of all payments. High average payments per beneficiary were related to those who died, had end stage renal disease (ESRD), were inpatient hospital users, and had inpatient surgery. Terminal illnesses tend to be costly for individuals at all ages (Gibbs & Newman, 1982). When examined as a population, the dying elderly were found to generate high expenses, but the number of decedents with high cost illnesses in any given year is actually quite small. Lubitz and Prihoda (1984) found only 3% of decedents in 1978 had Medicare reimbursements in excess of $20,000, and only 1% in excess of $30,000. Among the approximately 10,000 Medicare enrollees in the study by Lubitz and Prihoda (1984) who received more than $30,000 in Medicare reimbursement, similar numbers (5,000) survived and died during the year of study. 16 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A skewed distribution of health care expenditures with increasing care among the top few percent, and predominance of the elderly among those with high costs, was reported from the National Medical Expenditure Survey (NMES) in 1987 (Hahn & Lefkowitz, 1992). One percent of the population accounted for 30% of all expenditures, and 2% accounted for 41% of expenditures. Further, almost one-half (47%) of those in the top 1% of expenditures were over age 65. High cost expenditures reflect frequent use and need of medical care. Utilizing data from the National Medical Care Utilization and Expenditure Survey (NMCUES) conducted in 1980-1981, Berki et al. (1985) reported that 54.2% of high users of hospital inpatient services died within the year of study. Almost two-thirds (71.7%) of decedents in their study were in the intermediate- or high-use categories for hospital services. In two studies of high cost patients in San Francisco Bay area hospitals, Schroeder et al. (1979) and Schroeder, Showstack, and Schwartz (1981) found that 15% of these patients died in hospital and 34% of those discharged had died within 2 years. Certain technologies and medical interventions are in themselves so costly that any individual requiring the service would, by definition, experience a high cost illness. Examples of such technology include a stay in the intensive care unit (ICU), cardiac surgeries (bypass or other open heart surgery), and organ transplantation. Lubitz and Riley (1993) reported that only a small proportion of decedents (4.9%) had the kind of high expenses that would suggest aggressive, high technology medical services. Further, in their study a similar number of decedents and survivors were among the 1% who had highest cost care. The aged and the chronically ill are the primary consumers of intensive medical care services. Intensive care unit (ICU) services are extraordinarily costly (Chassin, 17 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1982). It has been estimated that ICU expenditures account for 15% of all hospital charges in the United States (Knaus & Thibault, 1982). Persons admitted to ICUs who do not survive the hospitalization have been shown to incur higher expenditures than those who survive. Civetta (1973), in a study of the surgical ICU at the Massachusetts General Hospital, found that costs were inversely related to survival. In Thibault's (1980) study at the same facility, he found that 23% of ICU admissions accounted for 37% of total charges. Studies of cancer patients have also reported that expenditures of decedents exceeded those for survivors by, on average, twice as much (Scotto & Chiazze, 1974). Surgery is another high cost intervention. Even among the very old, age 80 or 90 and above, surgical outcomes have been found to vary based on factors other than age such as functional status and co-morbidities (Edmunds, Stephenson, & Ratcliffe, et al., 1988; Hosking, Warner, Lobell et al., 1989). Other illnesses can be considered high cost due to their chronic nature and the need for continuous medical intervention in which costs accumulate. Examples of these types of high cost cases include AIDs, end stage renal disease with dialysis, and institutional care such as a nursing home. Slightly over one-half the patients receiving dialysis in the United States are over age 55. A recent study of elderly dialysis patients (Byrne, Phillip, & Cohen, 1994) showed a marked rise in mortality among the elderly with increasing age of initiation of dialysis. No patients who began dialysis treatment after age 80 were alive 5 years later. Recently a rigorous multi-center study designed to prevent costly high technology-driven death was completed (SUPPORT, 1995). They were unable to demonstrate any improvement in care or patient outcomes and therefore unable to affect cost of care. Physicians caring for hospitalized individuals in the advanced 18 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. stages of one of nine costly diseases were provided frequent data on the patient's likelihood of survival. In addition, specially trained nurses were provided to enhance communication between patients, their families, and the physicians. Over one-third of the patients who died spent 10 or more days in the ICU. The number who received mechanical ventilation or were comatose prior to death was not affected by the interventions, resulting in no impact on symptoms or costs associated with these expensive deaths. The argument is frequently made that increased use of expensive life-sustaining interventions for terminally ill patients is responsible, at least in part, for escalating health care costs of the elderly. Current research, however, provides only indirect evidence that inappropriate care is being provided to the elderly, some of whom do not survive. In her review of the literature through 1983, much of which has been cited above, Scitovsky (1984) argued that the data do not support this contention because acute hospital costs actually decrease with increasing age. In a subsequent review and update on this same topic, Scitovsky (1994) again came to the same conclusion: "Studies in the past decade do not support the hypothesis that it is high cost, high technological treatment of patients who die that has driven up medical costs" (p. 580). The greatest single potential for financial savings among the elderly lies in prospectively identifying high-cost decedents, those who receive costly, aggressive medical interventions in the months prior to death. Previous research has consistently shown that only a small percentage (3.5%) of Medicare decedents had high cost health care expenditures in the last year of life. Using 1987 data it has been estimated that if these high cost decedents could be identified and were not treated, the total financial savings would equal $2.8 billion, a small, but significant, savings of the total Medicare expenditures for the year (Jahnigen & Binstock, 1991). Denial of treatment to all high 19 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. cost users, not just decedents, would potentially save considerably more funds, but have the additional consequence of costing lives. Hospital Use The hospital is the most cost-intensive component of the American health care system. The majority of health care expenditures for the elderly population are for services rendered in the hospital (Waldo & Lazenby, 1984). Therefore, it is not surprising that hospital costs have repeatedly been found to comprise a disproportionate amount of total health care costs and that hospital care has been the focus of many of the studies on high cost care and the use and cost of health care services at the end of life. The earliest studies on medical care use at the end of life were published in the mid-1960s and focused solely on hospital services (Sutton, 1965; Wunderlich & Sutton, 1966). These studies were based on data in the National Mortality Survey, a probability sample of persons who died in the United States in 1961. It was found that 48% of all deaths, and 45% of deaths of persons over age 65, occurred in short-stay hospitals (Sutton, 1965). In addition, they found that approximately one-third of decedents of all ages used some hospital care in the last year of life. About 72% of decedents in 1961 had one or more episodes of hospitalization in the last year of life (Wunderlich & Sutton, 1964). In this same study, the number of days of hospitalization during the last 12 months of life was found to increase markedly with advancing age, rising from an average of 6 days for those under 1 year, to a high of 118 days for decedents 85 years and older. With the implementation of the Prospective Payment System under Medicare, length of hospital stay for the elderly has declined. If these studies were duplicated today a different pattern of hospital length of stay would no doubt be found. 20 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Hospital use has been shown to be "a clear and unequivocal function of age " (Barer, Evans, Hertaman, & Lomas, 1987, p. 857). The elderly consume a large number of hospital days both in absolute number of days and on a per capita basis; however, fewer than one-quarter of the elderly are hospitalized during any year. Further, a small proportion (5%) of the elderly population was found to consume almost two-thirds (59%) of the total hospital days used by the elderly during a 1-year period (Roos & Shapiro, 1981). Timmer and Kovar (1966), in a study of adult decedents age 25 and older from 1964-1965 found a similar percentage (73%) had hospital or institutional care in the 12 months prior to death. In contrast, they found only 13% of the survivor population used hospital or institutional care during the same year. In a later study (1971) of adult health care utilization in 1964 and 1965, Timmer and Kovar found that the median hospital bill of decedents was almost three times higher than that of hospital patients who did not die. Mushkin (1974) examined data 10 years later and found that 20% of non-psychiatric hospital and nursing home expenses were for persons who died. In a retrospective study of medical care expenses of all persons who died cared for by a group of physicians during 1983 and 1984, Scitovsky (1984) found that 60% of the average cost of care during the last year of life was for hospital care. Similarly, Lubitz and Riley (1993) found that inpatient hospital care accounted for 70% of payments for Medicare decedents compared with 53% for survivors. In a study of Medicare enrollees in Colorado in 1978, McCall (1984) found that having an inpatient hospital stay was the "single variable influencing the mean use and cost of services in the last year..." (p. 332). She found that in the last year of life enrollees averaged 20.4 inpatient hospital days, with 54% of the utilization 21 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. concentrated in the last quarter year of life. Garfinkle et al. (1988), in a study of "high cost users," found that 97% of this population had at least one hospital stay. The cost of care is substantially greater for those who use hospital services. Christensen, Long, and Rodgers (1987) reported Medicare expenditures for beneficiaries who required hospitalization during a year were four times the expenditures of those who did not require hospital care. The trends in hospital use and cost associated with death have also been found in similar studies in Canada. A study of residents of Manitoba, Canada (Roos et al. 1987) found hospital utilization increased in the period prior to death with a dramatic increase in overall hospital utilization among adult decedents at all ages compared to survivors. Among individuals aged 45 to 74 years old, death was significantly associated with the number of days spent in hospital during each of 4 years prior to death. For those age 85 and older, however, the prior year's hospitalization was not associated with death. The vast majority of the studies on hospital use and cost among the elderly have been primarily descriptive. In one of the few studies to use multivariate statistical techniques, Culler, Callahan, and Wolinsky (1995) developed predictive models of hospital costs among decedents from the Longitudinal Study on Aging (LSOA). Lower total hospital charges were associated with older age at baseline. Higher hospital charges were associated with female gender, being on Medicaid, higher population density, poor perceived health status, and more frequent physician and hospital utilization in the period prior to the study. Hospitals are not only a frequent source of care during the last year of life but have been also been found to be a frequent, but declining, place of death for America's elderly. Presently only 17% of deaths occur outside of institutional settings. In the 22 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. early 1980s, the proportion of elderly dying in hospital remained fairly constant about 55%. This declined slightly to 51% in 1986 after implementation of the Medicare Prospective Payment System (McMillan, Lubitz, McBean, & Russell, 1990). A study based on a stratified sample of death certificates of residents age 65 and over in Fairfield County, Connecticut (Brock, Foley & Losonczy, 1987) found 45% of deaths occurred in hospital, 30% at home, and 25% in nursing homes. In a more recent study of Medicare-covered services during the last 90 days of life, Gaumer and Stavens (1991) found that older Americans are likely to die somewhere other than a hospital. They reported that hospital deaths had decreased from 51.1% in 1982 to 45.4% in 1986, while out-of-hospital deaths in the 15 days post discharge increased. McMillan et al. (1990) also reported a shift in the place of death away from the hospitals to an older person's own home or nursing home. Non-hospital Utilization Information on the effect of dying on the use of physician services, nursing home, home health care and pharmaceuticals is much less well understood than that of the hospital. In the United States the proportion of the population residing in a nursing home has remained relatively constant at approximately 5%. As the numbers of elderly persons in the population have risen, the total number of older persons residing in a nursing home has also grown. It has been estimated that one of every 11 American who turned 65 in 1990 will spend at least 5 of their remaining years of life in a nursing home (Kemper & Murtaugh, 1991). Roos et al. (1987) demonstrated a strong association between nursing home use and age for residents of Manitoba, Canada. They reported a pattern of regular increases in the number of nursing home days used in each of the 4 years prior to death 23 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. for the elderly at all ages. Time spent in a nursing home increased significantly for those 85 years old for years prior to death and for 8 years prior to death for those between age 65 and 84. Nursing home care increased dramatically by age for decedents from a mean of 22.6 days for those age 45 to 64 to 55.7 days for those aged 65 to 74, to 181 days for those aged 75 to 84 and 419.3 days for decedents over age 85. In their study of elderly deaths in Connecticut during the early 1980s, Gaumer and Stavins (1991) found an increase in the use of home care, durable medical equipment, and physicians' services during the last 90 days of life. They reported that about 6% of Medicare decedents used a skilled nursing facility during their last 90 days of life. The effect of disability among the elderly on health services utilization remains an area of debate. McCall (1984) reported that disabled Medicare beneficiaries have on average higher overall charges and utilization rates of all health care services, and it is the very old who are the most likely age group to be disabled. Guralnik, LaCroix, Branch, Kasl, and Wallace (1991) found that disability rates increased with age at death. However, Scitovsky (1988) found that mean total expenses for those over age 65 did not differ significantly by functional status. Hospital expenses declined with decreased functional status, though, and nursing and home health expenses rose sharply with declining functional status. It can also be argued that the disabled have more impaired health and therefore are more likely to be selected out by mortality. The removal of the sicker, higher users of care should then leave an increasingly healthier older population. This has led some to hypothesize that future cohorts of the very old may actually be healthier very old 24 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. populations who will have decreasing need for health care services compared to younger old survivors (Fries, 1980). Managed Care The enrollment of significant numbers of elderly Medicare beneficiaries in managed health care plans is a recent occurrence. In 1985, 1.72 million Medicare beneficiaries belonged to Medicare Health Maintenance Organizations (HMOs) (Lawlor, 1995). This had grown to 3.2 million people, 8.7% of all Medicare beneficiaries, by early 1995. In California almost one in five Medicare beneficiaries (19.6%) belong to a Medicare Health Maintenance Organization. While managed care, and HMOs in particular, have been well studied for younger individuals, the effect of this type of health care delivery system on utilization, cost, and the health care of the elderly has not been extensively studied. The expectation is that any health care delivery system will provide basic and essential services (Eddy, 1991). What constitutes such services, how such definitions are arrived at, and how they impact on the care provided is unclear within the Medicare system, especially in the risk- contracting between the Health Care Financing Administration and HMOs. The most recent and exhaustive study published to date was a 4-year study of Medicare HMOs conducted by Mathematical Policy Research (Brown, Clement, Hill et al., 1993). They concluded that Medicare HMOs appeared to increase the volume of services but reduce their intensity. Enrollees in these managed care plans had comparable hospital admission rates, but shorter lengths of stay and more frequent use of skilled nursing facilities, and no increase in the total number of skilled nursing days. McCombs, Kasper, and Riley (1990) studied whether cost savings could be demonstrated in two of the Medicare HMO demonstration health plans. In one of the 25 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. plans, they found differences in the savings over time for both survivors and decedents. There were no savings for survivors during the first year of enrollment compared to the previous year, but significant savings during the second year. Savings for decedents were found, but to a lesser extent. In the second plan, no savings could be demonstrated for either survivors and there were increased expenditures for decedents compared to Medicare beneficiaries in the fee-for-service sector. In a study of elderly Medicare enrollees in a health maintenance organization, differences in health care services utilization rates between those 80 years of age and older were compared to those ages 65 to 79 (Johnson, Mullooly, & Greenlick, 1990). Those over age 80 were more likely to be admitted to the hospital and to see physicians more frequently than the younger old age group. Thomas and Kelman (1990) compared patterns of health care utilization among elderly residents in four different types of health care plans over 3 years. Enrollees in the preferred provider plan (PPO) had mean and total lengths of hospital stay significantly shorter than those of other plan members. Similar patterns of high use among a small percentage of enrollees has been documented in HMOs for several decades (Densen, 1959). Levkoff, Welte, Coakley et al. (1988), in examining a group of elderly enrollees in an HMO, found similar trends to that reported in the fee-for-service sector. A small percentage of members accounted for a disproportionately large amount of all types of services. The majority (86%) of enrollees had no acute hospitalizations. This is one of the few studies to use multivariate techniques and showed that age was a statistically significant predictor of average annual per capita expenditures. 26 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Freeborn. Pope, McFarland, and Siegenthaler (1988) compared utilization among elderly enrollees in a health maintenance organization. They found that consistently high users of medical care had multiple physical medical problems and high levels of psychological distress. A study of long-term adult (elderly and non- elderly) enrollees in a prepaid group practice HMO found 13% were consistently high users of outpatient services (McFarland et al., 1985). High use was more common the older the age group, but age was not a predictor of frequent outpatient care. Research on quality of care among different service delivery systems has shown that care delivered by HMOs is generally comparable to or better than that in the fee- for-service sector (Cunningham & Williamson, 1980). However, most of this research concentrated on young adults and only limited data existed on the quality of care received by elderly enrollees in HMOs. Carlisle, Siu, Keeler et al. (1992), colleagues at RAND, studied the quality of medical care received by elderly individuals in an HMO or the fee-for-service sector hospitalized with an acute myocardial infarction. They found no difference in sickness-adjusted mortality and hospital care that was better based on an algorithm for process of care among those over 65 years in the HMOs compared to a national sample of fee-for-service Medicare beneficiaries. Rationing In every society allocation decisions must be made with respect to a wide variety of resources. The allocation of health care resources on both the macro and micro level in the United States and many other industrialized nations has become among the most problematic and controversial. Increasingly it has been recognized that resources available to meet health care needs are limited and that decisions about the allocation and rationing of health care resources are inevitable. Many believe that in order to 27 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. limit expenditures we will, as a society, be forced to restrict access to medical care for some or certain services, even those that are potentially beneficial. These discussions have arisen not from utilitarian or egalitarian concerns, but rather as a direct response to the increasing financial cost of health care. It is the increasing pressure to control costs that has pushed the question of who will get how much care into the spotlight. Hence, the problem is one of what benefit is to be gained at what cost. Widespread concern has arisen that in order to control escalating health care costs we will, as a nation, have to develop public policy to ration health care on a larger scale than ever before (Aaron & Schwartz, 1990). The question for many is not whether, but how, we will ration health care services (Churchill, 1987). Others have argued that economics alone does not provide sufficient justification for more explicit public policies to ration medical care. It has been suggested that inflationary and wasteful aspects of our current health care system need to be corrected in order to control costs and then "we should be able to afford all the services we really need, provided we use our resources wisely" (Reiman, 1990, p. 913). E. Emanuel and L. Emanuel (1994) examined the potential cost savings if advanced directives, hospice care, and refusal of aggressive treatment at the end of life was instituted. They proposed that $18.1 billion, 3.3% of total health expenditures, or $5.4 billion from Medicare (6.1% of Medicare expenditures) would be saved at best. It can be argued that although a small amount overall, such a savings would be significant and allow increased access to care for the uninsured. On the other hand, in the overall scheme of things, even if such savings could be achieved they would not amount to much. Further, their calculations did not take into account the expense of alternative therapy or supportive care for those who refuse aggressive life-sustaining interventions. 28 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Decisions regarding the distribution of health care resources and access to these services have always been made, but until recently they were largely implicit and generally occurred as the unintended results of other policy choices. Recent discussion has begun to focus on the need for the explicit rationing of health care resources where "someone will consciously and publicly define what activities should be rationed and how" (Eddy, 1991, p. 106). The change will be in the way rationing occurs: who does it and who is affected by it. Price rationing, the dominant form of health care rationing in America has largely affected the poor. Kapp (1989, 1991) argued that rationing should take the form of explicit restrictions on third party payments for services. This has already begun to occur in Oregon's Medicaid system and in the denial by HMOs of certain treatments which they consider experimental (e.g., bone marrow transplantation). More explicit categorical rationing may potentially affect a broader and more diverse segment of the population. During the past two decades the American health care system has been in the midst of a transformation from an individualistic model in which the doctor-patient relationship is primary, toward a bureaucratic model in which decisions by both doctors and patients are heavily influenced by external considerations. Witness the embrace of managed care as the solution and the increasing attention to outcomes research which will set enforceable standards of care. The issue of cost has become an increasingly important factor in medical decision making. Concern about rising costs has led policymakers to promote enrollment in HMOs and other managed care plans for Medicare beneficiaries specifically as a means to control costs. Managed care is one approach to allocation, or rationing, of health care that relies on a set of economic incentives that differ significantly from those of traditional fee-for-service medicine (Luft, 1982). The goal 29 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of managed care is to provide access to care while containing costs by providing care in a cost-efficient method and eliminating unnecessary care. What role is played by cost, quality, and need in treatment decisions within such a system of care remains debatable. Cost controls, not quality, are the driving force in managed care. Siegler (1989) suggested that consideration of cost, availability of care, and quality of life are increasingly taking precedent in treatment decisions. Further, he argued that if patient selection decisions are going to be made, these should "begin with the strongest, most intelligent and most articulate patients" (p. 27) and not focus on the poor, aged or uneducated as targets of such plans. Clinicians and ethicists (as well as economists) would agree that expending resources on someone who is going to die, regardless of the intervention, is useless, wasteful, and unnecessary. The difficulty arises in trying to predict death with some degree of certainty on an individual basis. Hertzman and Hayes (1985) asked "Which clinical indicators predict 'natural death' before expensive interventions are initiated?" (p. 374). The answer for the vast majority of medical practice in the late twentieth century is "We don't know." It has been argued that the most obvious area to impose limits on health care is for the terminally ill whose care is often intensive and expensive (Bayer, Callahan, Fletcher et al., 1983; Veatch, 1988). However, prospectively identifying the terminally ill remains a problem. Terminal illness has been defined as "an illness in which, on the basis of the best available diagnostic criteria and in light of available therapies, a reasonable estimation can be made prospectively and with a high probability that a person will die within a relatively short time" (Bayer et al., 1983, p. 1491). Differentiating those known to be dying, the terminally ill, from those who may die, the critically ill, remains problematic. The 30 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. discussion becomes even more muddled when data from retrospective analysis on decedents are interpreted as representing the terminally ill. Medical decisions of whether or how to intervene for any condition are rarely based solely on scientific evidence, but increasingly involve ethical choices. Many decisions in clinical medicine are more fundamentally ethical than scientific in nature (Furrow, Wartman, & Brock, 1988). The fact that strict scientific evidence to support many medical decisions is often lacking is one of the key elements behind the recent move toward outcomes research. The Office of Technology Assessment (U.S. Congress) has estimated that only 20% to 30% of all clinical decisions are supported by hard scientific evidence. Clinical decisions regarding different diagnostic and treatment options involve the weighing of multiple costs and benefits by and for each individual patient. Issues Concerning Age-based Rationing The rapid rise in technological interventions and the consequent costs, coupled with the increasing numbers of older persons, have led in the past decade to a proliferation of books and articles about rationing and the role that age should play in this discussion (Barry, 1991; Binstock & Post, 1991; Blank, 1986; Evans, 1991, 1983; Haber, 1986). The proposals for age-based rationing of medical care often assert that older persons are being kept alive against their will through the use of expensive and costly technological interventions. However, research findings do not support this conclusion (Scitovsky, 1984, 1994). Implicit in the argument for age-based rationing is the assumption that outcome, particularly death, can be accurately predicted and anticipated, in persons of advanced age. In fact, the dying, regardless of age, are not an easily discernible population when viewed prospectively. 31 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Age is an obvious, statistically significant predictor of death. Yet, as a single factor old age is a poor predictor of clinical outcome. This is largely because the older population is extremely diverse and much more heterogeneous as a population than younger persons. Variability in function and physiologic and psychologic measures makes "advanced age, by itself, an inadequate criterion to use for categorical exclusion from medical diagnosis or intervention" (Jahnigen & Binstock, 1991, p. 26). A frequent assumption in proposals for age-based rationing is that older patients are poor candidates for complicated, and often costly, medical interventions. However, a review of the literature on successful outcomes for a wide variety of interventions (Jahnigen & Binstock, 1991) has demonstrated that old age in itself is not a good predictor of success or failure. Rather, it is the individual's underlying clinical condition and functional status, regardless of age, that are known to predict outcome. Age often influences medical decisions, but it remains unknown to what extent it is utilized as a determinant of what is or is not done. In medical practice age may be employed indirectly, perhaps as a proxy for functional ability, mental status, life expectancy, or possibly economic status in decisions of whether to intervene and to what extent. In a policy paper published by the American Geriatrics Society (Barondess, Kalb, Weil et al., 1988), it is stated "We behave, as clinicians, as though age has strong relevance to such decisions, but are substantially less clear concerning the ethical or moral implications of such a stance" (p. 920). If this is indeed the case, then implicit age-based rationing might be viewed as already occurring, and perhaps sanctioned, within the existing health care system. Recent discussions have focused on whether age should be adopted as an explicit categorical criterion for patient selection. A number of ethicists, philosophers, 32 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and politicians have written extensively advocating and justifying the use of age as a worthwhile criterion for patient selection (Daniels, 1988; Veatch 1988). Callahan (1987) provided one of the most provocative and widely discussed proposals for age- based rationing of health care services. He argued that the goal of medical science to lengthen the natural life span must be abandoned in order to both save costs and make old age a meaningful time of life. Underlying his argument to limit care at the end of the natural life span (somewhere between 70 and 80 years) is a presumption that life- extending technologies are frequently used on older people who neither want them nor can benefit from them. However, as was discussed earlier, the data do not support his conclusions. Arguments against the use of age as a categorical characteristic for access to fundamental services, such as health care, have focused on justice, equity, utility, and fairness. Chronological age provides an easy target and this is part of its appeal. However, the heterogeneity of the older population raises questions about the use of such an arbitrary criterion as age. The postponement of death to older and older ages leaves an increasingly larger and older population with possibly more chronic disease and disability during their last years of life (Schneider & Brody, 1983) that make them a population vulnerable to policies that limit access to care. Childress (1984) argued that it is unfair to treat a person differently when that person has no responsibility for the differences. He wrote "Ageism, like racism and sexism, involves a set of beliefs, attitudes and practices that unjustly and unjustifiably discriminate against a group. We have no responsibility for our aging; if we live long enough, we will age. . . . ageism is comparable to racism and sexism and should be rejected for similar reasons, which also appear to exclude the use of age as a criterion for the distribution of medical care" (p. 1116). 33 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Congressional reports have on several occasions alluded to the potential for rationing of new technologies and health care, especially for the elderly (U.S. Congress, 1984,1987). This may already be occurring. With the increasing prevalence of managed health care plans, it is possible that de facto rationing may be institutionalized with age as one of many potential factors to determine access to certain types of health care (Schwartz & Mendelson, 1992). Older persons rarely choose longer life as a goal in and of itself. Most individuals, when questioned, freely state that they prefer independence and, if unable to care for themselves, accept the inevitability of death. Public attitudes about the use of chronological age as a criterion for allocating health care resources were recently studied (Zweibel, Cassel, & Karrison, 1993). A majority were willing to withhold life-prolonging medical care to critically ill older persons near death and unlikely to recover. However, few were willing to withhold care based on age and most felt that decisions regarding life-extending interventions should be left to the individual. In reviewing whether age is the best criterion available for patient selection and whether there is any moral justification for age-based rationing, Olson (1989) pointed out that "To deny an older person potentially beneficial treatment as a panacea for a younger society's inability to cope with dying does not seem particularly just" (p. 606). To deny the elderly beneficial treatment because of society's inability to cope with the costs of health care for the dying, old and young alike, seems equally unjust. There are some indications that age-based rationing of health care may already occur, implicitly, and may be the norm for certain technologies. Aaron and Schwartz (1984) provided evidence of implicit age-based rationing in the United Kingdom for kidney dialysis. Jahnigen and Binstock (1991) asked "Suppose it were possible . . . to identify those Medicare patients who were going to die within the year, and whose 34 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. treatment would be comparatively costly; to choose to not undertake aggressive treatment of them; and thereby to save unnecessary health care costs. How much would be saved ...?" (p 29). Using clinical data to prospectively identify those elderly persons who are going to die remains a complicated and inaccurate undertaking. However, as clinical acumen increases in predicting favorable outcome categories it may become more plausible to establish rules of exclusion. Establishing criteria for final selection will be more difficult. If the goal is cost savings, then it is imperative that costs, along with clinical data, be included in the first decision step. Research to date has documented some of the clinical characteristics of high cost illness patients and others of those most likely to die in the fee-for-service delivery system. The patterns of utilization and health care expenditures among these two groups and within different delivery systems however still remains to be established. The changes in the health care system, especially the introduction of the Prospective Payment System and trend toward managed care for Medicare beneficiaries raise questions about the general applicability of our knowledge on the use and cost of health care services by the elderly derived from the fee-for-service Medicare system. The incentives in this changed health care environment could potentially alter both the level and intensity of care provided to the elderly. Patterns of care may more easily become institutionalized and be the fodder for, or product of policies, both formal and informal. It is hypothesized that age-based rationing of health care services already occurs, without explicit policy. Evidence of this should be more easily apparent in a managed care environment. Therefore this study was undertaken to explore the role of age and death in patterns of utilization and cost among elderly enrollees in an HMO. This is timely since there are increasing numbers of elderly American Medicare beneficiaries 35 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. entering into a health care system that is rapidly evolving and has little experience caring for the unique needs of an increasingly elderly population. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER 2 METHODOLOGY This study was designed to examine the role that age and death play in the older population's utilization of health care services in a managed care environment. To do so, this research focused on whether any evidence of implicit age-based rationing of health care services can be detected from aggregate claims-based data. To detect age- based rationing of health care for the elderly in a managed care setting, a series of specific research questions was addressed: 1. What is the pattern of health care services utilization, especially for high cost care, by elderly enrollees in a managed health care system? 2. What is the effect of age on utilization of health care services among elderly enrollees in a managed health care system? 3. Does the relationship between age and the cost of care among elderly enrollees differ for survivors and decedents? 4. Is there any evidence that the influence of age on the amount or mix of health care services provided to elderly enrollees is related to their health status or type of service provided? 5. Does the existence of an advanced directive affect utilization of services for elderly enrollees in a managed health care system? 37 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Sample Construction The study population was a convenience sample of 32,470 elderly enrollees in a senior health plan that is a Medicare-qualified health maintenance organization (HMO) in the Southern California region. The population was drawn from all new enrollees in the health plan from October 1991 through May 1992. The study sample was constructed from an initial sample of 21,158 new enrollees in the Southern California region who were mailed a health interview survey between October 1, 1991 and February 28, 1992. Of those enrollees surveyed, 11,870 persons over age 65 returned a completed survey with useable data. This is referred to as the respondent population. Respondents who had incomplete questionnaire responses due to missing responses to one or more pertinent questionnaire items were deleted from the study. As a result, the respondent population for this study consisted of 5,926 individuals over age 65 with answers to all pertinent questionnaire items. In order to assess whether the behavior of the respondents differed in any significant way from that of all enrollees, it was necessary to identify the non respondent population, those who failed to return the mailed questionnaire. This proved impossible because no master list of new enrollees sent the initial survey was maintained by the health plan. Therefore, the non-respondent sample was created from all other new elderly enrollees in the senior health plan between October 1991 and May 1992 from the same five-county region in Southern California. Only individuals who remained continuously enrolled for 12 months or died during their first 12 months in the plan were included in this non-respondent sample (N = 22,610). This population of other new enrollees may have included the non-respondents to the health assessment questionnaire, but this could not be confirmed. (For a more detailed explanation of how the non-respondent sample was created see Appendix A.) 38 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The survey was designed by the HMO to identify "high utilizers" among new enrollees. The original mailing included all new enrollees in the health plan between the designated dates, regardless of age. Medicare covers persons under age 65 who are chronically disabled or have end stage renal disease (ESRD). The latter are not eligible for inclusion in TEFRA risk-contracting with HMOs and the former have been shown to consume a disproportionate amount of health care resources. Because the focus of this study was on the elderly, enrollees age 64 and younger were eliminated. Data Sources The time period for the analysis was one year from the date of initial enrollment in the senior health plan. Eligibility for services from the health plan always begins on the first of a month. For an elderly individual who joined in October 1991, claims activity was included through September 1992 while an elderly individual joining the health plan in February 1993 had a claim file retrieved through January 1994. For those who did not remain continuously enrolled during the study period, due either to death or disenrollment from the health plan, only data for the months during which they were enrolled in the health plan were included. All data for this investigation were provided by the HMO. The primary data source was the company's claims data files which were maintained on all health plan enrollees. Two separate data files were combined by member number to create an aggregate data set for analysis: a demographic data set and a hospital data set consisting of Medicare Part A reimbursable health services. Responses to the questionnaire were then added and linked by member number for the respondent population. The demographic data set on all 28,536 sample members (respondents and non-respondents) included date of birth, sex, county, zip code of residence, date of eligibility, and termination status, if 39 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. applicable, for each sample member. Information on race was not available because the health plan was precluded by federal law from asking for this information. Information on marital status was also not made available. For those who left the health plan, the date of termination was provided by the HMO. For those who remained in the health plan for the entire study period, a quasi-date of termination was computed by adding 365 days to the date of eligibility. The Medicare Part A claims data base contained financial and clinical information for each Part A service consumed (e.g., hospitalization, emergency room visit, and home health, and nursing home care). These data were complete and reliable. Claims for Part A services provided to enrollees were submitted by providers directly to the HMO for reimbursement. The extent to which services reimbursable under Medicare Part A were consumed by enrollees outside of the health plan and paid for out-of-pocket is unknown, but probably quite small. The Medicare Part A claims data files were converted to a member-base data file utilizing the member's unique health plan member number and the member's date of enrollment. Claims activity was sorted by admission date and summarized for each month of enrollment to produce the following variables— aggregate costs per month by type of service, utilization days per month by type of service, diagnostic mix, and procedures-directly from the claims files. The questionnaire data set was linked by member number to the demographic and Part A data. The questionnaire contained response data from 37 questions about health, social support, functional ability, prior health services utilization, and health habits (see Appendix B). No direct information about income was included in the questionnaire. Responses to the question on medication use were not provided as 40 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. these were determined to be unreliable and were eliminated by the data management personnel from the survey responses available for inclusion in this study. The health maintenance organization that sample members belonged to is a network model HMO that contracts with group practices to provide care to its enrollees. Some of these groups routinely submit outpatient encounter data while others submit data inconsistently or incompletely. Due to the unreliability and incomplete nature of these data, outpatient utilization was not included in the analysis. Therefore, no data on physician encounters or care outside of the hospital, nursing home, or from a home health agency were available. Variable Construction For this study a person-level summary file was created from the demographic, - claims, and questionnaire data sets for each member of the sample population. A file was created for each sample member that included demographic information on age, gender, residence, months of enrollment in the health plan, and if the individual died during the study period. The demographic variables defined for each sample member are displayed in Table 2.1. A few demographic variables require explanation. County of residence was reported as a variable in the original data set, but was found to be missing in many instances. Therefore, the zip code data provided for each sample member was used reconstruct the missing counties of residence data, which are referred to as Counties 1, 2, 3,4, and 5. Termination status data were used to create dummy variables which identified decedents and those sample members who voluntarily left the plan during the study period. For a subset of the sample, Respondents with Part A Costs, data were also 41 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.1 Demographic V ariables Variable (Code) Description Member Number (MEMNUM) Unique identifier for each sample member Reference Date (REFDATE) Date of eligibility Termination Status (TERM_ST) Enrollment as of 1 year after eligibility provided on the reason for termination. If a person died during the study year, hospital claims data were reviewed for discharge status to verify that the individual had expired. A discharge status of death was then used to determine the month of death. The date of termination from the health plan, supplied by the HMO’s data management division, was used to determine the month of death for those who died without a hospitalization or discharge status indicating death t The variables derived from the Part A claims data base for use in this study are displayed in Table 2.2. Part A claims data included the number of days of service, location of service, amount charged, the principal diagnosis, and up to four additional diagnoses, and the primary procedure and up to two additional procedures. Total days of care by provider were calculated by summing the service days using Age (AGE) Gender (GENDER) County (CNTYLOC) County of residence Age in years as of reference date Male or Female Termination Reason (TERM_R) Reason for leaving health plan Death (DEAD, DEADTR) Died during study period 42 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. location data to classify type of provider (hospital, skilled nursing home, or home health). Because individuals could receive more than one home health service on any day, total episodes of home health services were summed by month of service to provide a variable of the total number of episodes of home health care services received. Location service codes included inpatient hospital, outpatient hospital, emergency room, outpatient hospital setting, inpatient skilled nursing home, and the member's home. Costs were computed for each provider type by month of enrollment utilizing data for the amount charged and the location code. The amount charged was the amount submitted by the provider and corresponded to the billed charges for the service. The amount paid was not used for cost calculations because this figure varies considerably depending on co-payments and other adjustments based on contractual agreements between the HMO and the provider. Costs by provider type were then summed to produce total monthly cost of services for each month of the first year of enrollment. Months where no claims for services occurred were coded as zero. Part A claims data included information on diagnoses and procedures pertaining to each episode of care or admission using the International Classification of Diseases (I.C.D.-9) coding system. A principal diagnosis and up to four discharge diagnoses were provided for each claim episode. These were recoded into monthly dummy variables for 16 diagnostic organ system categories and 21 specific diagnoses of interest (Table 2.2). An individual with multiple diagnoses within the same diagnostic category (e.g., myocardial infarction and congestive heart failure both fall within circulatory diseases) was identified only as having a diagnosis within the category. No differentiation was made between those with single or multiple same-system diagnoses. A similar procedure was followed for the I.C.D.-9 procedure codes from 43 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.2. Part A Claims Data Variable (CODE) Description Cost Hospital Costs per Month (HCM1-12) Skilled Nursing Home Costs per Month (SNFCM1-12) Home Health Services per Month (HHCM1-12) Total Cost per Month (TCM1-12) Utilization Hospital Days per Month (HDY1-12) Skilled Nursing Home Days per Month (SNFD1-12) Home Health Services per Month (HHD1-12) Diagnostic Codes Categorical Codes (ICD-9-CM) 001.0- 139.9 140.0-239.9 240.0 - 279.9 280.0 - 289.9 290.0- 319.9 390.0 - 459.9 460.0-519.9 520.0 - 579.9 580.0 - 629.9 680.0 - 709.9 Total hospital charges for each month of enrollment Total skilled nursing home charges for each month of enrollment Total charges for home health services for each month of enrollment The sum of hospital, nursing home and home health charges for each month of enrollment Total days of inpatient hospital care for each month of enrollment Total days of skilled nursing home care for each month of enrollment Total episodes of home health services for each month of enrollment Infectious Diseases Neoplastic Diseases Endocrine System Disorder Hematologic System Disorder Nervous System Disorder Circulatory System Disorder Respiratory System Disorder Gastrointestinal System Disorder Genitourinary System Disorder Dermatologic System Disorder 44 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.2, cont. Variable (CODE) Description 710.0-739.9 Musculoskeletal System Disorder 740.0 - 759.9 Congenital Disorder 780.0 - 799.9 Nonspecific Signs 800.0 - 999.9 Injury or Trauma V01.0- V82.9 Otherwise Undefined Specific Diagnoses (ICD-9-CM) 174.0- 174.9 Breast Cancer 183.0- 183.9 Ovarian Cancer 185.0- 185.9 Prostate Cancer 250.0 - 250.9 Diabetes Mellitus 290.0 - 290.9 Dementia 293.0 - 293.9 Delirium 296.0 - 296.3 Depression 291.0-291.9, 303.0-303.9 Alcohol-related Disorder 331.0-331.9 Cerebral Degeneration 410.0-414.9 Myocardial Infarct (Heart Attack) 428.0 - 428.9 Congestive Heart Failure (CHF) 431.0-438.9 Cerebral Vascular Accident (Stroke) 480.0 - 487.9 Pneumonia 490.0 - 496.9 Chronic Obstructive Pulmonary Disease 733.0-733.9 Osteoporosis 780.0 Coma 820.0- 821.9 Hip Fracture V56.0 - V56.9 Dialysis V44.3 Colostomy Procedure Codes Categorical Codes (ICD-9-CM) 01.0-05.9 Nervous System 06.0 - 07.9 Endocrine System 08.0 - 16.99 Ophthalmologic (Eye) 18.00-20.99 Otologic (Ear) 21.00-29.99 Upper Respiratory System Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.2, cont. Variable (CODE) Description 30.00 - 34.99 Respiratory System 35.00 - 39.99 Circulatory System 40.00-41.99 Hematologic System 42.00 - 44.99 Gastrointestinal System 55.00 - 59.99 Genitourinary System 60.00 - 64.99 Male Genital System 65.00 - 72.99 Female Genital System 76.00 - 84.99 Orthopedic 85.00 - 86.99 Dermatologic (Skin) 87.00 - 99.00 Radiologic (X-ray) Specific Procedures (ICD-9-CM) 13.1 - 13.9 Cataract Repair/Extraction 31.10 - 31.29 Tracheostomy 35.20 - 35.29 Cardiac Valve Replacement 36.00 - 36.99 Coronary Bypass (CABG) 37.70 - 37.79 Pacemaker Placement 43.1 -43.2 Gastrostomy T ube Placement 60.00 - 60.69 Prostate Repair 79.35 Hip Fracture Repair each claim. A primary procedure and up to three additional procedures were recorded for each claim episode. These were recoded into monthly dummy variables for 15 categorical and 8 specific procedures. The presence or absence of diagnoses or procedures, both by system and specifically, were then calculated for the year. Again, no differentiation was made between having a diagnosis, categorically or specifically, one or more times. 46 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Questionnaire Variables Responses from 16 questions from the health assessment questionnaire were included as explanatory health status and health behavior variables for analysis in the study (Table 2.3). In some instances, variables were created directly from the questionnaire responses while other questionnaire responses required recoding for purposes of meaningful analysis. For example, respondents were asked to rate their health on a 4-point scale from excellent (I) to poor (4). These responses were used to create a series of four dummy variables: excellent, good, fair, and poor health. Conversely, responses to questions concerning 12 medical conditions for which medical care was being received were simply recoded into dummy variables for each condition. A summary variable was then created to indicate the number of pre-existing medical conditions for which a respondent was currently receiving treatment. In addition to self-reported health status and current medical problems, respondents were asked to self-rate their ability for 1 1 functional categories using a 3- point scale (1 = independent, 2 = need some assistance, 3 = need total assistance). These individual tasks correspond to the widely used Katz Activities of Daily Living (ADL) and Instrumental Activities of Daily Living (IADL) scales (Katz, Branch, Branson et al., 1983). Variables for ADL and IADL ability were created from the sum of responses to questions about functional ability relating to bathing, dressing, eating, toileting, and mobility (ADLs) and ability to perform housekeeping chores, prepare meals, take medications, manage money, shop, and use transportation (IADLs). Thus , two contiguous variables, one reflecting ADL and another reflecting IADL ability, were created as measures of functional ability. 47 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.3 Health Care Questionnaire Variables Variable (CODE) Description Health Status (HEALTHS!) Activities of Daily Living (ADLSCORE) Instrumental Activities of Daily Living (IADLSCOR) Medical Problems (MEDPROB) Physician Visits (MD VISIT) Hospital Use (HOSPUSE) Emergency Room Use (ERUSE) Home Health Nursing Use (HOMEUSE) Alcohol Use (ETOUSE) Years Smoked (SMOKE YR) Mental Health Symptoms (MENTPROB) Self-rated health status (1-4) Need for assistance in bathing, dressing, toileting, eating and mobility (5-15) Need for assistance in housekeeping, shopping, meal preparation, medication, money management, and transportation (6-18) Current treatment for 12 common chronic medical conditions Number of physician office visits during the prior year Number of times hospitalized in the prior year Number of times visited an emergency room in the prior year Number of times received home health nursing services during the prior year Use of alcohol on a regular basis (Y/N) Number of years smoked tobacco products Frequent presence of psychologic symptomatology Overwhelmed (OVERWHQ) Sadness (SADQ) Lonely (LONELYQ) Tense (TENSEQ) Depressed (DEPRESQ) Number of times a week feeling overwhelmed Number of times a week feeling sad Number of times a week feeling lonely Number of times a week feeling tense Number of times a week feeling depressed Exercise (XERCIS) Aerobic (AEROBQ) Bicycle Riding (BIKEQ) Running (RUNQ) Swimming (SWIMQ) Walking (WALKQ) Engages in regular exercise Number of times a week engages in aerobics Number of times a week bike rides Number of times a week runs Number of times a week swims Number of times a week walks Poverty (POOR) Advanced Directive (DPAHC) Receipt of MediCal (Medicaid) Has a living will or durable power of attorney for health care 48 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Indicators of health care utilization during the previous 12 months are often an indirect indicator of one's health status. In addition, prior utilization of health services has been shown to be a strong indicator of future utilization and costs (Levkoff et al., 1992). Variables were created to reflect how many times in the 12 months preceding the study period the respondents had seen a doctor, been hospitalized, been seen in an emergency room, and received home health nursing. Use of the emergency room, hospitalization for one or more days, or one or more home health nursing visits was then used to create a dummy variable indicating whether the respondent had prior use of any Part A service during the 12 months before the study. The number of physician visits during the previous year was kept as a continuous variable taken directly from each respondent's answer. Open-ended questions pertaining to personal health habits were included as additional indirect measures of health status. Members were asked about current and past alcohol consumption and smoking. A dummy variable was created to indicate regular alcohol consumption, defined as a response indicating consumption of more than 2 ounces of alcohol a week. A variable for the number of years of smoking (SMOKYR) was created directly from the responses to the question "How many years have you or did you smoke?" The respondents were also queried about the frequency of psychologic symptomatology and exercise. Dummy variables were created indicating the presence or absence of possible mental health problems based on the responses to questions regarding feelings of tenseness, sadness, anger, depression, loneliness, and being overwhelmed. These were used individually and summed to produce an indicator of frequent psychologic symptomatology defined as a total psychologic symptoms 10 or more times weekly. Similarly, dummy variables were created indicating regular 49 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. physical activity based on the question about weekly frequency of specific forms of exercise: jogging or running, walking, aerobics, swimming, walking, and bicycle riding. Responses totaling three or more times a week were used to indicate engaging in regular exercise. Data about income or economic status were limited to one question concerning MediCal eligibility. Information about the possession of an advanced directive (living will or Durable Power of Attorney for Health Care) was used to create a dummy variable indicating the presence of an advanced directive. Decedent and Survivor Populations Those individuals who died during the course of the study were identified using three separate data source variables: termination status at the end of the first year of enrollment, reason for termination, or for the members of the study who utilized Part A health care services, claims data for discharge status. Individuals who had no discharge status indicating that they had died and who did not terminate enrollment due to death were considered to have not died and comprised the survivor population. Data Analysis Techniques Multivariate statistics techniques were employed to model predictors of health care utilization and death. Because the sample is highly skewed with a majority of individuals incurring no Part A costs during the study period, a two-step approach was used (Duan, Manning, & Morris, 1983). First, logistic regression models were developed to identify predictors of having incurred costs. Next, regression models were estimated to identify predictors of cost among those who used services. The dependent variable was cost in dollars summed over specified time periods during the 50 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. first year of enrollment in the senior health plan (i.e., 12 months and last 3 months and month of life for decedents). Decedent costs were calculated from the month of death retroactively. The month of death was computed utilizing the discharge status and claim date for those decedents who used Part A services and died while receiving care. The date of termination from the health plan was employed to calculate the month of death for those decedents who used Part A services and died subsequent to receipt of such services. The month of death was unknown for decedents who did not utilize any Part A services. However, this latter group is known to have used no services during the study period and hence no services during the immediate period preceding death. For the survivor population, costs were summed for each month and averaged for the number of months of interest for the regression models. Sample members with no Part A services during the study had zero costs. For logistic regression models of health service utilization, the dependent variable was defined as having Part A costs versus not having costs during the time period of interest. For the regression models, the dependent variable was the amount of total costs for the specified type of service during the time period specified. Utilization and cost models were computed for the entire sample population and for the respondent population for the full 12 months from enrollment, for the last 3 months and the last month of life for those who died during their first 12 months in the health plan. This was done in order to compare and contrast effects on utilization and costs over shorter periods of time prior to death. Comparison models between the total sample and respondent populations were limited to the demographic variables of age, gender, and county of residence. Predictor variables for the claimant population in both samples included discharge diagnoses and procedures. For the respondent population, additional predictor variables included a wealth of measures of health 51 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. status drawn directly from the questionnaire responses: self-reported health status, chronic medical problems, functional status drawn from responses about ADLS and IADLs, recent health services utilization, and lifestyle factors (smoking, alcohol, exercise use) known to influence health and thereby indirectly affect utilization. Death was employed as a dichotomous dependent variable in logistic regression models to predict death over the first 12 months in the health plan. Simple models using demographic variables as predictors were constructed for the total and respondent populations. Complex models were then developed to include health status indicators from discharge diagnoses for both populations and questionnaire responses for the respondents. This was repeated to model high cost death, defined as total costs in excess of $10,000. 52 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER 3 DESCRIPTIVE RESULTS This chapter presents the descriptive results. First, the demographic characteristics and age composition of the total sample population and the respondent population are presented. Next, the decedent and survivor sub-populations are described. The distribution of costs during the full 12-month study period is then discussed. Cost results are presented for total Part A costs, as well as for the component parts of hospital and long-term care services for the year. The distribution of costs for decedents over different time periods preceding death is then contrasted to those of survivors for the 12-month study. The frequency of discharge diagnoses by diagnostic categories for the claimant populations is reviewed. Finally, respondents’ answers to the questions from the health assessment questionnaire are presented. Sample Characteristics The demographic characteristics of the sample population are presented in Table 3.1. Respondents comprised 20.8% of the total sample. Individuals who used some type of Part A service during the study year, the claimant sub-sample, accounted for 28% of the total sample. A slightly larger proportion of the non-respondents used Part A services (29.1%) than did the respondents (24.3%). The sample population ranged in age from 65 to 107 years with a mean age of 72.2 years and a median of 70.1 years. The respondent population was slightly 53 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.1. Age, Gender and Survival Characteristics of the Sample Population Total N = 28,535 Respondents n = 5,926 Non-respondents n = 22,610 No Claims Claims No Claims Claims No Claims Claims N 20,525 8,010 4,485 1,441 16,041 6,569 (%) (71.9) (28.1) (75.7) (24.3) (70.9) (29.1) Mean Age 71.64 73.72* 71.33 73.82* 71.37 73.39* % Female 57.7 54.4* 51.9 48.7* 59.3 55.7* Mean Age Female 71.72 74.24* 71.69 74.11* 71.64 73.96* Mean Age Male 70.87 73.09* 70.94 73.55* 71.01 72.73* % Died 4.3 8.7* 3.7 8.2* 4.5 8.7* Mean Age Decedents 75.67 76.61 75.44 77.20 76.18 76.49 Mean Age Survivors 71.14 73.44* 71.17 73.52* 71.12 72.56* *t > .001 or chi sq > .001 younger with a mean age of 71.9 years. Claimants were significantly older than non claimants by about 2 years in all populations. Females comprised the majority of the total sample population. Females also predominated in the other sub-samples, except among respondents with claims. The female population was on average a year older than the male population. Claimants of both genders were significantly older than those of the same gender without claims. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Five and one-half percent of the total sample population died during the study period. This is similar to the 5% annual death rate reported among Medicare beneficiaries annually. A greater proportion of the non-respondent population died (5.7%) than of the respondents (4.8%). A significantly greater proportion of claimants died compared to non-claimants among the total, non-respondent and respondent populations. Decedents were significantly older than survivors, but there was no significant difference in the age of decedents who used Part A services (claimants) compared to those who did not (non-claimants). The age distribution by 5-year cohort was similar in the total and respondent population (Table 3.2). The majority of both populations were between 65 and 74 years of age and one-fourth were between 75 and 84 years of age. Only a minority (5 Table 3.2. Five Year Age Cohort Distribution of the Total and Respondent Population3 U.S. Total Respondents Population3 Cohort Age Number Percent Number Percent Percent 65-69 years 12623 44.2% 2669 45.0% 30.98% 70-74 years 6950 24.4% 1502 25.3% 26.25% 75-79 years 4556 16.0% 942 15.8% 19.81% 80-84 years 2668 9.4% 510 8.6% 12.81% 85+ years 1719 6.0% 303 5.1% 10.15% Total 28536 5926 “ U.S. Dept of Commerce, Bureau of the Census, 1994 55 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. to 6%) were 85 years or older. The age distribution by gender within the total and respondent populations was also very similar. The age distribution of the sample population was somewhat younger than that reported for the U.S. population during 1990 (U.S. Bureau of the Census, 1993). In both the total sample and respondent populations, a greater proportion were 65 to 69 years and a smaller proportion were in each 5-year age cohort over 75 years of age, than in the U.S. population. Age-specific rates of death from both the total sample and respondent population contrasted to that reported for the U.S. population are shown in Table 3.3. With increasing age an increasing proportion of each age cohort died. More than twice the proportion of the old-old cohort (75 to 84 years) died as of the young-old cohort (65 to 74 years). Almost 17% of those in the oldest-old cohort, over age 85, died during the study. Thus, based on overall mortality, this sample appears to be as healthy as the elderly population of the United States. The death rate by 5-year age cohort is shown in Table 3.4. With increasing age, a greater proportion or each age cohort died. When just the claimant populations (total or respondents) were examined, a slightly larger percentage of each age cohort died than in the total population. Among the claimant populations a slightly greater proportion of the older age respondents (above age 75) died than in the total population. This is consistent with findings that those who use health services at any age are sicker and therefore more likely to die. Total Costs Total costs for Medicare Part A health services for the sample population ranged from $0 to $250,000 during the 12-month study period. Over 70% did not use any Part A service and therefore incurred no costs (Table 3.5). A slightly greater 56 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.3. Death Rate in the Total Sample and Respondent Populations Compared to National Census Data* Total Sample Age Cohort Population Respondents 1990 Population 65-74 yrs 3.68% 3.21% 2.65% 75-84 yrs 7.84 6.96 6.14 85+ yrs 16.93 16.83 15.04 Males 65-69 yrs 4.06 3.52 2.98% 70-74 yrs 6.16 5.42 4.53 75-79 yrs 7.92 7.61 6.96 80-84 yrs 11.82 10.13 10.42 85-89 yrs 19.75 10.98 14.99 90-94 yrs 23.36 21.05 21.43 Females 65-69 yrs 2.00 2.17 1.69 70-74 yrs 4.14 2.28 2.55 75-79 yrs 5.66 4.98 3.88 80-84 yrs 8.74 6.59 6.32 85-89 yrs 17.65 16.67 11.03 90-94 yrs 20.90 15.38 17.67 “ U.S. Dept, of Commerce, Bureau of the Census, 1990 57 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.4. Proportion of the Population that Died by Five Year Age Cohort Respondent Total All Claimants Claimants (N = 28,536) (n = 8,011) (n= 1,441) 65-69 yrs 370 2.9% 161 5.7% 26 5.4% 70-74 yrs 350 5.0 148 7.6 21 5.3 75-79 yrs 301 6.6 142 9.5 30 9.9 80-84 yrs 267 9.9 116 11.7 19 12.4 85+ yrs 291 16.9 131 17.1 23 18.8 Total 1,579 698 119 proportion of the total sample than respondents used some Part A service. Further, a greater proportion of the total sample (5.3%) had total costs in excess of $10,000 than among the respondents (4.0%). In both the total population and among respondents a small number of individuals incurred very costly care. About 2% of both the total sample population and the respondent population incurred total expenses in excess of $20,000 and less than 1% had costs totaling over $40,000. Average costs are slightly higher for the total population compared to the respondents. The average total cost for 12 months among the total sample population was $1,973 and in the respondent sample was $1,484. Average total costs per Part A user (claimant) were higher, but comparable between the total sample population and the respondents: the average total cost per Part A user was $7,028 in the total population and $6,103 in the respondent sample. 58 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.5. Cost Distribution by Type of Service for Total Sample and Respondent Populations Total Respondents Dollars Spent Number % Number % Total Costs Mean All $1973 $1484 Mean Claimants $7028 $6103 $0 20525 71.9 4485 75.7 $1-999 2389 8.4 445 7.5 $1,000-4,999 2940 10.3 542 9.2 $5,000-9,999 1143 4.0 217 3.7 $10,000-19,999 817 2.9 139 2.3 $20,000-39,999 493 1.7 67 1 .1 $40,000-99,999 204 0.7 28 0.5 >$100,000 22 .08 3 0.1 Hospital Costs Mean All $1682 $1054 Mean Claimants $5994 $4334 $0 21693 76.0 5132 86.6 $1-999 1817 6.4 124 2.1 $1,000-4,999 2540 8.9 303 5.1 $5,000-9,999 1126 3.9 189 3.2 $10,000-19,999 775 2.7 106 1 .8 $20,000-39,999 426 1.5 54 0.9 $40,000-99,999 1 4 1 0.5 1 6 0.3 >$100,000 1 5 .05 2 0 m Skilled Nursing Home + Home Health LTC) Costs V Mean All $183 $121 Mean Claimants $652 $499 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.5, cont. Total Respondents Dollars Spent Number % Number % $0 26657 93.4 5617 94.8 $1-999 1023 3.6 179 3.0 $1,000-4,999 530 1 .9 90 1.5 $5,000-9,999 214 0.8 23 0.4 $10,000-19,999 83 0.3 14 0.2 $20,000-39,999 26 0.1 3 0.1 Hospital use was also greater and more costly on average within the total sample ($1,682) than in the respondent population ($1,054) due in part to differences in use rates. One-quarter of the total sample had hospital costs, but only 14% of the respondents required hospital care during the study. Average hospital costs among decedents in both the total sample and the respondents were double those for survivors. Long-term care costs were considerably less than costs for hospital services. The average long-term care cost among the total sample was $183 and among the respondents was $121. A small proportion (5 to 6%) of both populations used long term care services, either nursing home or home health care, that met eligibility for coverage through the health plan. Less than 1% of each population required long-term care services in excess of $10,000, with the majority who used such long-term care services having costs of less than $1,000. Among those who used long-term care services, those who died had slightly lower average total costs than those who survived (not shown). 60 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Total costs by 5-year age cohort for both the total sample population and the respondents are shown in Table 3.6. Similar data for hospital and long-term care costs by age cohort is found in Appendix D. Both populations demonstrated a similar pattern of increasing cost by age due to a greater proportion of each advancing age cohort incurring costs. In the younger elderly cohorts, 75 to 80% of the population had no Part A costs. Among those over 85 years, 55% of the total population and 60% of respondents had no Part A costs during the year. Within each age cohort, a decreasing proportion of the population fell into each higher cost category beyond $1,000. However, the percentage of each age cohort that required higher cost care increased with age. Despite this latter effect, the majority of elderly who incurred high cost care, whether defined as greater than $ 10,000 or $20,000, were among the younger elderly due to the size of these cohort groups. Seventy percent of those with high costs in both the total and respondent population were less than 80 years of age. Only 12% of those with high cost care were in the oldest old ages over age 85). Total Costs and Death Decedent costs were higher than those of survivors for all categories of care and across all age groups. Table 3.7 presents a comparison of the average costs for decedents and survivors in the claimant populations. The average 12-month cost of care for decedents who used Part A services was $15,062 in the total population and $14,048 in the respondent sample. Average 12-month costs for claimant survivors were considerably less-$6,262 and $5,387 in the total and respondent populations, respectively. Twelve-month total and hospital costs for decedents in the total population declined with increasing age. In contrast, those for survivors rose with 61 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.6. Total Costs by Five Year Age Cohort Age cohort in years Amount Spent 65-69 70-74 75-79 80-84 85+ Total a. Total Sample $0 9,793 5,014 3,068 1,699 951 20,525 -47.71“ -24.43 -14.95 -8.28 -4.63 -71.9 [77.59]b [72.14] [67.35] [63.23] [55.32] $1-999 917 586 459 255 172 2,389 -38.38 -24.53 -19.21 -10.67 -7.2 -8.37 [7.27] [8.43] [10.08] [9.49] [10.01] $1,000-4,999 1055 732 541 350 262 2,940 -35.88 -24.9 -18.4 -11.9 -8.91 -10.3 [8.36) [10.53] [11.88] [13.03] [15.24] $5,000-9,999 385 284 202 135 137 1,143 -33.68 -24.85 -17.67 -11.81 -11.99 -4.01 [2.88] [3.77] [4.07] [4.64] [7.23] $10,000-19,999 254 176 163 121 103 817 -31.09 -21.54 -19.95 -14.81 -12.61 -2.86 [2.01] [2.53] [3.58] [4.50] [5.99] $20,000-39,999 146 110 83 84 70 493 -29.61 -22.31 -16.84 -17.04 -14.2 -1.73 [1.16] [1.58] [1.82] [3.13] [4.07] $40,000-99,999 63 42 36 42 21 204 -30.88 -20.59 -17.65 -20.59 -10.29 0.71) [0.50] [0.60] [0.79] [1.56] [1.22] >$100,000 9 6 3 1 3 22 -40.91 -27.27 -13.64 -4.55 -13.64 -0.08 [0.07] [0.09] [0.07] [0.04] [0.17] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.6, cont Age cohort in years Amount Spent 65-69 70-74 75-79 80-84 85+ Total TOTAL 12,622 6,950 4,555 2,687 1,719 28,53! -44.24 -24.36 -15.96 -9.42 -6.02 >$10,000 472 334 285 248 197 1536 -30.7 -21.7 -18.6 -16.1 -12.8 [3.7] [4.8] [6.3] [9.2] [11.5] > $20,000 218 158 122 127 94 719 -30.3 -22 -17 -17.7 -13.1 [1.7] [2.3] [2.7] [4.7] [5.5] b. Respondent Sample $0 2,195 1,112 640 357 181 4485 -48.94 -24.79 -14.27 -7.96 -4.04 -75.7 [82.84] [74.03] [67.94] [70.00] [59.74] $1-999 151 126 97 40 31 445 -33.93 -28.31 -21.8 -8.99 -6.97 -7.51 [5.66] [8.39] [10.30] [7.84] [10.23] $1,000-4,999 187 150 110 57 38 542 -34.5 -27.68 -20.3 -10.52 -7.01 -9.15 [7.01) [9.99] [11.68] [11.18] [12.54] $5,000-9,999 72 54 41 24 26 217 -33.18 -24.88 -18.89 -11.06 -11.98 -3.66 [2.70] [3.60] [4.35] [4.71] [8.58] $10,000-19,999 41 33 33 18 14 139 -29.5 -23.74 -23.74 -12.95 -10.07 -2.36 [1.54] [2.20] [3.50] [3.53] [4.62] $20,000-39,999 18 14 14 1 1 10 67 63 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.6, cont Age cohort in years Amount Spent 65-69 70-74 75-79 80-84 85+ Total -26.87 -20.9 -20.9 -16.42 -14.93 -1.13 [0.67] [0.93] [1.49] [2.16] [3.30] $40,000-99,999 4 1 1 7 3 3 28 -14.29 -39.29 -25 -10.71 -10.71 -0.47 [0.15] [0.73] [0.74] [0.59] [0.99] >$100,000 1 2 0 0 0 3 -33.33 -66.67 0 0 0 -0.05 [0.04] [0.13] [0.] m [0.] TOTAL 2,669 1,502 942 510 303 5926 -45.04 -25.35 -15.9 -8.61 -5.11 >$10,000 64 60 54 32 27 237 -27 -25.3 -22.8 -13.5 -11.4 [2.4] [4.0] [5.7] [6.3] [8.9] > $20,000 23 27 21 14 13 98 -23.5 -27.6 -21.4 -14.3 -13.3 [.08] [.02] [.02] [.03] [.04] = % of dollar category, b [ ] = % of age cohort increasing age. Long-term care costs increased with increasing age for both decedents and survivors. Among the respondent population with Part A use, the trend in the pattern of costs among survivors and decedents is not as clear as it is with all claimants. In the decedent respondent population, total costs increased from the 65 to 69-year old cohort 64 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.7. Average Cost by Five Year Age Cohort of Decedents and Survivors with Part A Use Total Population Respondents N = 8,010 n = 1441 Age Cohort Decedents Survivors Decedents Survivors Total Mean $15,061 $,6262 $14,048 $5387 70-74 years 16,242 5,786 18,835 5788 75-79 years 13,091 6,205 12,539 5412 80-84 years 14,089 7,662 12,466 5922 85+ years 13,980 7,835 11,327 6388 65-69 years $17,291 $5,775 $15,554 $4663 Hospital Mean $13,256 $5,301 $11,596 $3,681 65-69 years $15,916 $5,151 $13,481 $3,336 70-74 years 14,580 4,915 14,966 3,955 75-79 years 11,223 5,240 10,717 3,712 80-84 years 12,661 6,217 11,070 3,954 85+ years 11,220 5,889 7,970 3,768 LTC Mean $3,585 $582 $1,160 $493 65-69 years $921 $309 $577 $195 70-74 years 1,252 451 2,530 509 75-79 years 1,407 561 596 399 80-84 years 1,151 1,069 469 536 85+ years 2,289 1,466 1,879 1,258 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. to the 70 to 74-year old cohort and then declined, resulting in an overall decline in 12- month total costs between the youngest-old and oldest-old decedents. Average hospital costs among decedent respondents with Part A use increased and then dropped with each successive 5-year cohort, but overall showed a decline with increasing age. Total and hospital costs of respondent survivors showed a varied pattern between 5-year cohorts, but one of an overall increase in average costs with increasing age. Long-term care costs increased overall among respondent decedents and survivors with increasing age, but rose in the 70 to 74-year old cohort among both decedents and survivors for some unexplained reason. The most likely explanation for these between-cohort variations among the respondents with Part A use is an undefined bias within the respondent population. Decedents were more likely to have used some Part A service and therefore had a higher proportion with costs than survivors (Table 3.8). Among decedents, the proportion with costs declined the shorter the time period preceding death used for analysis. Slightly less than one-half the decedent population had costs for services during the 12-month study. During the last month of life, only one-quarter of the decedents used Part A services. A greater proportion of decedents than survivors incurred high cost care. Over the 12-month study period, 10% of decedents had costs totaling in excess of $10,000 and 4% used Part A services costing in excess of $40,000. In contrast, less than 1% of survivors required such high cost care. Diagnoses The frequency of diagnoses by diagnostic category for the claimant populations is shown in Table 3.9. The relative frequency of diagnostic categories is consistent 66 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.8. Cost Distribution of Decedents and Survivors among the Total and Respondent Populations Dollars 1 2 months Last 3 months Last month 12 months Total Population $0 55.8% 62.3% 74.7% 72.9% $1-499 4.1 5.4 4.8 86 $500-999 12.4 11.2 8.6 10.2 $1000-9,999 8.1 7.4 5.8 3.8 $10,000-19,000 8.9 7.2 3.7 2.5 $20,000-39,000 6.8 4.6 2.1 1.4 $40,000-99,999 3.6 1.8 0.4 0.6 >$100,000 0.4 0.1 0 0.1 > $20,000 10.8 6.5 2.5 2.1 > $40,000 4 1.9 0.4 0.7 Respondents $0 58.4% 63.3% 75.9% 76.6% $1-499 3.9 4.9 3.5 7.7 $500-999 9.8 10.8 8 9.1 $1000-9,999 10.5 9.8 7 3.3 $10,000-19,000 8.4 4.9 2.8 2 $20,000-39,000 4.9 4.5 2.1 0.9 $40,000-99,999 4.2 1.7 0.7 0.3 >$100,000 0 0 0 0 > $20,000 9.1 6.2 2.8 1.2 > $40,000 4.2 1.7 0.7 0.3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.9. Diagnoses by Frequency of Occurrence in the Total Sample and Respondent Populations with Part A Claims Category Total Respondents Total Decedents Survivors Total Decedents Survivors Circulatory 45.3% 76.5 42.3 44.1 76.0 41.0 Congenital 1.2 1.7 1 .2 0.8 2.3 0.7 Dermato logic 4.2 7.6 3.9 4.1 8.5 3.7 Endocrine 22.5 46.1 20.2 21.0 46.5 18.6 Gastroentrologic 22.9 26.4 22.6 23.7 26.4 23.4 Genitourinary 17.9 24.4 17.2 16.2 24.8 15.4 Hematologic 10.4 18.6 9.7 10.0 16.3 9.5 Infectious 6.1 15.0 5.2 4.7 11.6 4.1 Injury 24.2 25.2 24.1 22.7 21.7 22.8 Musculoskeletal 18.9 16.2 19.2 19.3 20.9 19.1 Neurologoic 17.8 20.9 17.5 15.0 16.3 14.9 Neoplasm 17.8 29.7 16.7 18.3 33.3 16.8 Psychiatric 8.8 15.2 8.2 6.9 14.7 6.2 Respiratory 25.2 56.2 22.3 25.3 59.7 22.0 between the total claimant sample and the respondents, suggesting a similar pattern of illness between the two populations. Circulatory conditions were the most frequent reason for use of Part A services in both populations. In descending order of frequency the next most common categories of illness were respiratory problems, injuries, gastrointestinal disorders, and endocrine problems. Congenital problems were the least frequent category of discharge diagnosis among all claimants. The relative frequency of diagnoses was similar in both the decedent and survivor claimant populations as in the entire sample. However, endocrinologic, psychiatric, respiratory Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and neoplastic problems were much more common among decedents than in survivor or total sample population. Questionnaire Responses Among the respondent population, the frequency of responses to the health assessment questionnaire items are presented in Table 3.10. Responses are shown for the total respondent population, decedents, and survivors. The responses reflected the self-report of individuals just prior to entering the health plan and thus the study period. The majority of respondents rated their health as "good. Almost two-thirds of all respondents and survivors rated their health "good," but only 50% of decedents described their health as good upon entering the health plan. Only 1% of respondents rated their health as "poor," but almost 8% of respondents who died had described their health as "poor." The vast majority of respondents reported no functional limitations based on self-declared impairments in ADLs or IADLs. A greater proportion of all respondents and the survivors reported no functional limitations than did respondents who subsequently died. Overall, many more decedent respondents (14%) indicated impairment in one or more ADLs. Forty percent of decedents reported a limitation in an IADL compared to only just over 10% of all respondents or respondent survivors. The frequency of chronic medical problems requiring ongoing care was higher among decedent respondents than among all respondents or survivors. One-third of all respondents reported no chronic medical problems while only 20% of decedents had none. Over 20% of decedents had reported three or more chronic medical problems, 69 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.10. Questionaire Responses as a Percentage of the Population Respondent Respondent Variable All Respondents Survivors Decedents Health Status Indicators Health Status Excellent 20.9% 21.6% 8.0% Good 60.8 61.3 50.4 Fair 17.1 16.2 33.9 Poor 1.2 0.9 7.7 ADL Impairment None 97.9% 98.5% 86.0% 1-2 0.7 0.5 3.8 3+ 1.4 1 10.2 IADL Impairment None 86.5% 87.9 59.1 1-2 8.1 7.7 16.8 3-4 3.6 3.1 14.3 5+ 1.7 1.3 9.8 Chronic Medical Problems Requiring Treatment None 35.7% 36.5 20.6 1 34.2 34.2 33.2 2 17.8 17.6 22.4 3 7.4 7.2 12.2 4 2.9 2.7 5.9 5+ 2 1.8 5.6 Part A Use During the Prior 12 Months Yes 19.3 18.4 35.7 No 80.7 81.6 64.3 Number of Physician Visits During the Prior Year Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.10, cont. Respondent Respondent Variable All Respondents Survivors Decedents None 16.4 16.6 12.8 1 18.6 19 11.7 2 19.1 19.5 12.5 3 15.6 15.6 15.1 4-7 16.1 15.7 23.8 8-10 7.4 7.2 10.9 11-19 5.3 5 10.9 20+ 1.5 1.5 2.3 Frequent Mental Health Symptoms No 71.8 71.8 71.3 Yes 28.2 28.2 28.7 Life Style Indicators Regular Alcohol Consumption No 74.3 73.8 82.9 Yes 25.7 26.1 17.1 Number of Years Smoking 0 27.1 27.2 26.2 I-10 12.4 12.5 10.1 II-20 16.9 17 14.7 21-30 14.9 15 11.9 31-40 12.7 12.6 14.3 41+ 16 15.7 22.7 Engages in Regular Exercise No 69.2 69.9 55.2 Yes 30.8 30.1 44.8 Demographics Poverty Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.10, cont. Respondent Respondent Variable All Respondents Survivor: Decedents No 98.2 98.2 98.2 Yes 1.8 1.8 1.8 Miscellaneous Completed an Advanced Directive No 63.4 63.4 62.9 Yes 36.6 36.6 37.1 but only 12% of survivor respondents reported a similar large number of chronic medical problems Self-reported use of Part A services during the 12 months prior to entering the health plan was much more frequent among decedents than all respondents or the survivors. Just over one-third (35%) of respondents who died reported use of Part A services during the prior year compared to less than 20% of survivors. There was suggestion of more frequent physician visits during the prior year reported by decedents than by survivors or in the total respondent population. It is of interest to note that about 16% of respondents reported no physician office visits during the year prior to enrolling in the health plan with a slightly lower percentage (12.8%) of decedents reporting no physician visits during the prior year. The report of frequent mental health symptoms such as anxiety and depression was similar among all respondents and the decedents and the survivors. Just under 30% of respondents in all categories reported frequent mental health symptoms. 72 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Decedents were different from all respondents and survivors in their self-report of regular alcohol consumption and exercise, but not in years of smoking. A greater proportion of decedents reported less alcohol consumption and more frequent exercise than did all respondents or the survivors. A small number (1.8%) of respondents reported they received MediCal Medicaid) and thus can be classified as poor. The proportion indicating they were poor did not vary among those who died or survived. One-third of respondents reported having completed some type of advanced directive prior to joining the health plan (living will or Durable Power of Attorney for Health Care). The proportion was constant among survivors and decedents. 73 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER 4 PREDICTORS OF UTILIZATION OF HEALTH CARE SERVICES Logistic regression models were developed to answer the question "Is age a predictor of use of health care services?" Simple sequential models were developed utilizing demographic variables common to all sample members to predict use of any type of Part A service, for hospital use and for long-term care (home health and skilled nursing home) use. These models were calculated separately for the entire sample population and for the respondent population. Use of services over three different time periods was examined: the entire 12-month study period, the last 3 months of life for decedents, and the last month of life for decedents. Results of simple sequential logistic regression models from the 12-month time period are presented. (See Appendix C for details of 3-month and last month of life models.) The results of these models are summarized using normalized odds ratios calculated for males age 65 and 85 as representative of the young-old versus oldest-old, and for male survivors and decedents of both ages. This allows examination of both the age and death effect on the probability of use of health care services. Odds ratios represent the probability of having utilized the type of health care service in question during the specified time period. Finally, results from more complex logistic regression models in the respondent population that include health status and other variables derived from the health care assessment are reviewed. 74 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Probability of Part A Service Use Total Population The results of logistic regression models on predictors of use of any Part A service (hospital, home health, or skilled nursing home) are presented in Table 4.1. Age was a significant predictor of use in Model I in which AGE was the only regressor, and remained so in more complex models (Models II-VI). For every year above age 65, there was a 5% increase in the probability of using some type of Part A health service. The estimated effect of age on having used some type of Part A service was impacted very minimally when other variables were added to the model. Death was also a strong predictor of use of some type of Part A service. In Model IV, those who died were 74% more likely than those who did not die to have used some type of Part A service even when age and gender were considered. The effect of death on using Part A services increased significantly when the interactive term "age x death" was added to the model (OR 1.74 increased to OR 2.75). Those who died were more than two and one-half times more likely than those who survived to have used a Part A service, but this effect decreased about 4% for every year of age above 65. Stated differently, the "age x death" interaction term indicated that the probability of Part A use increased by 5.5% per year of age for survivors, but only 1.5% per year of age for decedents. Respondent Population The models discussed above were re-estimated using only respondents to the health care questionnaire. Among the respondent population the age effect was slightly greater, and the death effect similar, to that of the total population in 75 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 4.1 Logistic Regression Odds Ratio of Having Part A Costs During the 12 Month Study Period Model Age Gender Age2 Died Age x Dead County-2 County-3 County-4 County-5 a. Total Sample (N = 28,536) I 1.049** II 1.050** 1.198** HI 1.058** 1.196** 1 IV 1.057** 1.179** 1 1.739** V 1.055** 1.177** 1 2.754** 0.960** VI 1.058** 1.186** 1 2.756** 0.960** 1.329** 1.142* 0.904* 0.840* b. Respondent Population (N = 5,923) 1 1.058** 2 1.059** 1.193* 3 1.094** 1.183* 0.998* 4 1.096** 1.164* 0.998* 5 1.094** 1.163* 0.998* 2.392** 0.979 6 1.096** 1.171* 0.998* 2.370** 0.98 1.161 1.101 0.0914 0.839 + = p >.05, * = p >.01, ** = p>.001 a \ predicting use of Part A services (Table 4.1b). Respondents were about 6% more likely with each additional year of age above age 65 to use some type of Part A service. When the interactive term age2 was added to the model, the age effect increased to 9.5% and remained at this level when additional demographic variables were added to the model. Respondents who died were almost twice as likely (OR 1.895) as those who did not die to use Part A services. When age at death was entered into the model ("age x death" interaction term), death became an even stronger predictor of having used Part A services among respondents. For every year above age 65 at death, respondents were 2% less likely to use Part A services. While this difference was similar to the results incurred in the larger sample, it was not statistically significant in the smaller respondent population. Normalized Odds Ratios of Total Part A Costs Total Population Table 4.2 presents the odds ratios calculated using Model V above for males ages 65 and 85 who died and did not die, normalized to survivors age 65. For the total sample population, the probability of having Part A service costs during the first year of enrollment increased with both age and death. With 20 additional years of age, the probability of using Part A services during 1 year rose over two and one-half times. This relative increase in Part A service use with age increased slightly more when comparing use over an average 3 months or 1-month period in the total population. The probability of Part A service use was also relatively higher for decedents compared to survivors. Those who died at age 65 were 2.75 times (275%) more 77 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.2. Normalized Odds Ratios for Total Costs of Males 12 Months 3 Months Last Month Alive Dead Alive Dead Alive Dead Total Population Age 65 1.000 2.754 1.000 2.202 1.000 1.279 Age 85 2.892 3.245 2.88 2.515 2.967 3.129 Respondents Age 65 1 2.392 1 1.81 1.000 1.202 Age 85 3.228 5.085 3.287 4.312 3.223 2.054 likely than those who did not die to have used some form of Part A service in the period of up to 12 months preceding their death. Decedents age 85 were also more likely to have used Part A services during the year than were survivors age 85. Age had a somewhat different effect on survivors and decedents. The effect of 20 years of age on Part A service use during the year was similar to that of death among the young-old (OR 2.7). Survivors 85 years of age were 2.7 times (269%) more likely to have used Part A services than survivors age 65. However, in the decedent population, an additional 20 years of age increased the likelihood of having used Part A services by only 18%. Similarly, those who died at age 85 were 20% more likely than those who died at age 65 to have used Part A services. Therefore, the effect of age on the probability of Part A use was reduced in the decedent population and the effects of dying were significantly reduced in the old-old population. The age effect increased among survivors and declined among decedents in the total population for use of Part A services as the time period of analysis declined. 78 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Survivors age 85 were almost three times more likely to use Part A services in an average month compared to just over two and one-half times during the 12-month period. Decedents age 65 were a little more than twice as likely (OR= 2.202) to have used Part A services during the last 3 months, and only 28% more likely during the last month of life as survivors of the same age to have used such services during the year in the health plan. Decedents at age 85 were less likely than survivors of the same age to use such services during their last 3 months or last month of life than survivors age 85. The death effect was diminished when comparing the probability of use for the total population over shorter times periods. In the last 3 months of life older decedents remained somewhat more likely than younger decedents to have used Part A services. This difference declined when examined for the last month of life. The probability of use of Part A services among decedents in the last month of life was very similar at age 65 and age 85 (OR 1.28 and 1.30, respectively). Respondents Among respondents both age and death affected the probability of use of Part A services and both had a slightly stronger effect than in the total population. Decedent respondents at age 65 were almost two and one-half (OR 2.39) more likely to use Part A services during the first year of enrollment than were respondents age 65 who did not die. Decedent respondents at age 85 were five times more likely than survivors at age 65, and almost two times more likely than respondent survivors at age 85, to use services. Respondent survivors at age 85 were a little more than three times more likely than those 65 years of age to use Part A services during the first 12 months in the health plan. 79 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The age effect among survivor respondents remained constant but the age effect declined among decedent respondents when the probability of use was examined for shorter time periods. Survivor respondents were a little more than three times more likely to use Part A services with an additional 20 years of age. The difference in the probability of Part A use declined between decedents' age 65 compared to survivors of the same age the shorter the time period. Decedent respondents at age 65 were 80% more likely to have used Part A services in the last 3 months of life and only 20% more likely to use such services than were survivors of the same age during the entire 12 months. The death effect among respondents declined in a pattern similar to that of the total population. Respondents who died at age 85 were significantly more likely than respondents who died at age 65 to use Part A services. Respondents who died at age 85 had a fourfold increase (OR 4.312) in the probability of Part A use during the last 3 months of life compared to respondents age 65 who did not die. Respondents who died at age 65 were only 20% more likely to use Part A services during their last month of life than survivor respondents age 65 were to use such services during the entire year. Decedent respondents at age 85 were 80% more likely than those age 65 to use Part A services. Respondent survivors at age 85 were more likely to use Part A services during their first year in the health plan than were respondents age 85. Hospital Use Total Population The probability of hospital use showed a pattern similar to that of all Part A services for both the total population and the respondent population (Table 4.3). 80 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 4.3 Logistic Regression Odds Ratio o f Having Hospital Costs for 12 Month Study Period Model Age Gender Age2 Died Age X Dead a. Total Population (N = 28,536) 1 1.043** 1 1 1.044** 1.198** III 1.050** 1.197** 1.000 IV 1.049** 1.175** 1.000 1.997** V 1.047** 1.173** 1.000 3.150** 0.960** VI 1.049** 1.181** 1.000 3.154** 0.963** b. Respondent Population (N = 5,923) I 1.050** 1 1 1.051** 1.233* III 1.081** 1.223* 0.999 IV 1.083** 1.185* 0.998* 3.029** V 1.081** 1.183* 0.999* 3.891** 0.978 VI 1.083** 1.181* 0.999* 3.926** 0.977 + = p >.05, * = = p >.01, ** = p > .001 County-3 County-4 1.259** 1.099* 0.913* 0.859 0.997 1.028 County-5 0.828** 0.906 o o Among the total sample population, AGE alone was a significant predictor of having received hospital services (Model I). For every year over age 65, there was a 4% increase in the probability of requiring hospital care. This increased minimally when gender, death, or county of residence was included (Models II-VI). Death was a strong and slightly more powerful predictor of hospital use than it was of any Part A use. Those who died were twice as likely as those who did not die to have used hospital services. When age at death was considered, the age effect increased over 100%. However, with every additional year of age over 65 at the time of death, there was a statistically significant 4% decline in the probability of using hospital care. Respondents The probability of hospital use among respondents during the year was similar to that for Part A use among respondents. It also followed the same pattern as seen in the total population for both hospital and Part A use. The age effect was mild and significant while the death effect was stronger in the respondent than in the total population. For each year of age above age 65, respondents were 5% more likely to use hospital services (Table 4.3b). This effect was not altered when gender was considered but increased slightly when the interactive term age2 was added to the model. There remained an 8% increased probability of using hospital services among respondents with each additional year of age above 65 even when death, age at death, and county of residence were considered. Respondents who died were three times more likely than those who survived to have used hospital services (Model IV). When age at death was added into the equation (Model V), the death effect increased. Respondents who died were almost 82 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. four times as likely to require hospital care as those who did not die, but with each additional year of age at death, respondents were slightly less likely (2%) to use hospital services. Normalized Odds Ratios for Hospital Use Total Population The probability of use of hospital care increased with both age and death, but the effects of each differed depending on the time frame of reference (Table 4.4). Increased age increased the probability of hospital use over the year and the last 3 months of life for both survivors and decedents, but had minimal effect on use of hospital services by decedents during their last month of life. During the first year of enrollment, the probability of hospital use increased five and one-half times (550%) with 20 additional years of age among survivors. The age effect increased among survivors when examined over a 3-month period, but declined when the probability of hospital use over any one month was considered. The age effect on the probability of hospital use among decedents followed a similar pattern; however, during the last month of life, younger elderly decedents were actually more likely to have used hospital services than were those 20 years older. The effect of dying on the probability of hospital use differed depending on the time period examined. Over 1 year both young-old and oldest-old decedents were more likely than survivors of similar age to have used hospital services. Oldest old decedents had the highest probability of hospital use during the year and the last 3 months of life. During the last 3 months of life, the probability of hospital use by decedents was seven and one-half times higher for those age 85 and almost four times higher for those age 65 compared to survivors age 65. The probability of decedents 83 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.4. Normalized Odds Ratios for Hospital Costs of Males 12 Months 3 Months Last Month Alive Dead Alive Dead Alive Dead Total Population Age 65 1.000 3.150 1.000 2.524 1.000 1.533 Age 85 2.392 3.325 2.382 2.474 2.397 1.421 Respondents Age 65 1.000 3.892 1.000 2.885 1.000 2.396 Age 85 2.686 6.651 2.713 5.740 2.792 3.028 using hospital services in the 3 months prior to death was slightly less than during the year for both the young-old (age 65) and oldest-old (age 85). In the 3 months preceding death, decedents age 65 were just under four times (OR=3.82) more likely to have used hospital care than survivors were during the entire year. The relative difference in the probability of hospital use was much less when decedents and survivors age 85 were compared. Oldest-old survivors were six times more likely to use hospital services during the year than young-old survivors and oldest-old decedents seven and one-half times more likely. When the probability of use in the last month of life of decedents was compared to that of survivors for the year, the differences were reduced. Survivors age 85 were actually more likely to use hospital care during the year than were decedents of the same age or younger to do so during their last month of life. During the last month of life, decedents age 65 were only 53% more likely to use hospital services than survivors and age 65. Decedents age 85 were slightly less likely than decedents age 65 (OR 1.42) to use hospital care in the month prior to death. 84 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Respondents Among the respondent population, the impact of age on the use of hospital services was comparable to age in the total population. During the first year of enrollment, 20 years of age among respondents had an effect similar to that of dying on the probability of using hospital services. Respondents who died at age 65 were six times (OR 5.998) as likely as those who survived to have used hospital services. Those who died at age 85 were just over six times (OR 6.291) more likely than those who survived at age 65 to require hospital care during the first year in the health plan. Decedents at age 85 had over twice the probability of decedents age 65 to have used hospital services. During the last 3 months of life, the effect of age and death appeared similar to that observed during the year. Decedent respondents age 65 were almost five times (OR 4.862) more likely to have required hospital care during their last 3 months of life than survivors of the same age were during their first year in the health plan. Decedent respondents age 85 were eleven times (OR 10.95) more likely to use hospital services in their last 3 months of life, and had twice the probability of having hospital care than did decedents age 65 (OR 4.862). The difference in the probability of hospital use during the last month of life for decedent respondents compared to survivors was much less than during the other time periods. Respondents who died at age 65 had more than twice the increased probability of hospital care in their last month of life than did survivors age 65 during the year of study. Decedent respondents age 85 had a slightly higher probability of hospital use during their last month of life than decedents age 65. Unlike in the total population, decedents age 85 had an increased probability of hospital use in the last month of life compared to the probability of survivors using hospital care during the 85 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. year. Respondent decedents age 85 were three times more likely to use hospital care during their last month of life than survivors age 65 during the entire year, but only a little more likely to have used such services than did survivors age 85 during the entire year of study. Long-term Care Use Total Population The probability of using long-term care services showed a different pattern than that for Part A service use or hospital care (Table 4.5). For every year of age above 65 there was a 9% increase in the probability of using long-term care services (Model I). This was not altered when other demographic variables were added to the model. Decedents were twice as likely as survivors to use long-term care services (Model IV). When age at death was added to the model, the effect of death on use of long-term care services increased. Those who died were four and one-half times more likely than those who did not die to use long-term care services, but with each additional year of age at death, there was a 6% decline in the probability of use of such services. Respondents The age and death effects on long-term care service use had a similar pattern among respondents in the total population, but were slightly stronger among the respondents. Among respondents each additional year of age above 65 increased the likelihood about 10% of having used long-term care services (Table 4.5b, Model I). This increase was not affected by gender (Model II), but declined slightly (by 2%) when the interactive term age2 was added to the model (Model III). 86 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4 .5 Logistic Regression Odds Ratio o f Having L T C Costs for th e 1 2 Month Study. 87 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Having died was a statistically significant and powerful predictor of long-term care services use among respondents. Respondents who died were almost three times as likely (OR 2.977) as those who did not die to have used long-term care services (Model IV). The probability of using long-term care services increased even more, to six times, for respondent decedents compared to respondent survivors, when age at death was considered. Additionally, with each additional year of age at death above 65, there was a statistically significant 5% decline in the probability of long-term care services use among respondents (Model V). Normalized Odds Ratios of Long-term Care Use Total Population The probability of use of long-term care services (skilled nursing home or home health care) increased with both age and death during all time periods, but declined for decedents compared to survivors as death approached (Table 4.6). Among the total population during the year of study decedents were more likely than survivors of the same age to use long-term care services. Survivors age 85 were five times more likely to use long-term care services during their first year in the health plan than were survivors age 65. Decedents age 85 had a slightly greater probability of long-term care use during the year than did decedents 20 years younger. During the last 3 months of life the probability of long-term care use increased with age for survivors, but was not much different for decedents by age. Survivors age 85 were almost two and one-half (OR 2.38) times more likely than survivors age 65 to use long-term care services. Decedents age 85 were slightly less likely to use long term care services during their last 3 months of life than were decedents age 65 to use such services during their last 3 months of life. 88 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.6. Normalized Odds Ratios for Long-term Care Costs of Males 12 Months 3 Months Last Month Alive Dead Alive Dead Alive Dead Total Population Age 65 1.000 4.608 1.000 3.817 1.000 1.667 Age 85 5.900 8.047 6.080 7.527 6.104 3.804 Respondents Age 65 1.000 5.998 1.000 4.862 1.000 1.867 Age 85 6.291 13.366 6.328 10.949 6.383 4.665 The probability of long-term care use in the last month of life increased significantly with age. Among survivors 20 years of age the probability of long-term care use increased sevenfold. Decedents age 85 were three times more likely than those age 65 to use long-term care in the last month of life. Decedents at age 65 had a slight increase in the probability of long-term care use compared to survivors. However, among the oldest-old, decedents actually had a lower probability of long term care use in their last month of life compared to the probability of use of such services by the oldest-old survivors during the year. Respondents The age and death effect on respondents' probability of long-term care use was similar to that of the total population with the death effect being somewhat stronger in the respondents than in the total population. Survivors age 85 were 2.7 times more likely than survivors age 65 to use long-term care services during the year of study or 89 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. during a 3-month period. The age effect increased among survivors for any month with survivors age 85 six times more likely than those age 65 to have used long-term care services during a month. The death effect among respondents' use of long-term care services was more pronounced among the oldest-old than in the total population. During the first year of enrollment in the health plan, decedent respondents age 65 were almost four times (OR 3.892) more likely than survivors age 65 to use long-term care services. The probability of long-term care use during the study year and during the last 3 months of life among respondents who died at age 85 was almost twice that of those who died at age 65. In the last month of life, respondents who died had 4.66 times the probability of long-term care use compared to only 86% increase among those age 65 compared to survivors age 65. However, the probability of long-term care use among oldest-old survivors during any month was greater than that among oldest-old decedents in their last month of life. Other Significant Effects Gender In addition to age and death, both gender and county of residence were seen in the sequential models (Tables 4.1,4.3, and 4.5) to have an effect on the probability of service utilization. In the models predicting Part A service use (Table 4.1), males at any age above age 65 were 20% more likely than females to use Part A health services. This effect decreased slightly, to 18%, when death was added (Model IV) and remained at this level when age at death was considered. Gender remained a statistically significant predictor of Part A service use among respondents (Table 4.1b). as it was in the total population. Male respondents were 19% more likely than 90 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. female respondents at any age above 65 to use Part A services. The gender difference declined to 16% when death was taken into consideration. Men were significantly more likely than women, even when controlling for age, death, and residence, to require hospital care (Table 4.3). Males were 20% more likely than females at any age above 65 to require hospital care. The gender difference in the probability of hospital use declined slightly (OR 1.175) when death was considered (Model IV). There was no statistically significant difference between the probability that males or females used long-term care services. Males and females were equally as likely at any age over age 65 to use long-term care services. However, when death was added to the models, males became slightly less likely than females to use such services. County of Residence The probability of having used some type of Part A service or hospital service differed significantly for residences of the five counties in Southern California (Table 4.1a and 4.2a, Model VI). Residents of Counties 2 and 3 were more likely, and those of Counties 4 and 5 less likely, than those of County 1 to have used either Part A or hospital services. Among the respondent population, there was no statistically significant difference in the probability of Part A or hospital use by county of residence. Residents of all counties except County 5 were significantly more likely than those in County 1 to use home health or skilled nursing home services. Those residing in County 3 were 74% more likely, while those in Counties 2 and 4 were 33% and 57%, respectively, more likely than residents of County I to use long-term care services. There was no statistically significant difference in use of longer term care services between residents of Counties 5 and 1, but those in the former were slightly less likely to use such services. Respondents who resided in County 3 were 91 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. statistically more likely (65%) than respondents of County 1 to use skilled nursing home or home health care services, even when controlling for age, gender, and death. Respondents who resided in County 2 and County 4 were also more likely, but by a smaller and statistically insignificant amount, than those in County 1 , to use such services. On the other hand, respondents in County 5 were less likely to use long-term care services. Models were also calculated that included interaction terms for gender with age and death, but neither was statistically significant nor did their inclusion alter the results, so these are not presented here. Complex Models The results of more elaborate logistic regression models to predict use of health care services among the respondent population during the year are presented in Tables 4.7, 4.8, and 4.9. These models utilized direct and indirect measures of health status derived from the health assessment questionnaire to predict the probability of Part A, hospital, and long-term care use during the first 12 months of enrollment for the respondent population. If the age effect seen in the simple logistic regression models we the result of age serving as a statistical proxy for health status, then adding health status measures to the models should reduce the estimated effects of age in both absolute value and statistical significance. Age and death remained statistically significant predictors of service utilization among respondents even while controlling for health status over the 12 months. Both age and death showed a similar pattern for the probability of different service utilization. For every year over age 65 respondents were about 8% more likely to use Part A or long-term care services and about 7% more likely to receive hospital care. 92 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.7. Logistic Regression Odds Ratio on Having Part A Costs for 12 or 3 Month Periods for Respondents (N = 5,923) 12 Months 3 Months Last Month Variable Odds Ratio Odds Ratio Odds Ratio Age 1.088** 1.088** 1.088** Gender 1.264** 1.233* 1.208* Age2 0.998** 0.998** 0.998** Died 1.894* 1.348 0.876 Age x Dead 0.975 0.983 0.966 County-2 1.136 1.159 1.159 County-3 1.070 1.065 1.089 County-4 0.908 0.913 0.909 County-5 0.826+ 0.824+ 0.794+ Good Health 1.518** 1.506** 1.526** Fair Health 2.101** 1.988** 2.008** Poor Health 1.815+ 1.692 1.522 ADL Score 1.116+ 1.131+ 1.117+ IADL Score 1.025 1.015 1.014 Medical Problems 1.109** 1.116** 1.129** Part A Use Last Year 1.447** 1.451** 1.469** Doctor Visits 1.009 1.009 1.007 Mental Health Symptoms 1.079 1.093 1.080 Regular Alcohol Use 0.961 0.947 0.951 Years Smoking 1.004+ 1.004+ 1.004+ Regular Exercise 0.895 0.875 0.864+ Poverty 0.996 1.001 0.920 Adv. Directive 1.220* 1.228* 1.218* +p > .05, * p > .01, **p . 01 93 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.8. Logistic Regression Odds Ratio on Having Hospital Costs for 12 or 3 Month Periods for Respondents (N = 5,923) 12 Months 3 Months Last Month Variable Odds Ratio Odds Ratio Odds Ratio Age 1.072** 1.073** 1.079** Gender 1.225* 1.218+ 1.214+ Age2 0.998+ 0.998+ 0.998* Died 3.016** 2.175* 1.827+ Age x Dead 0.978 0.985 0.962 County-2 0.837 0.853 0.865 County-3 0.967 0.959 0.981 County-4 1.041 1.039 1.046 County-5 0.871 0.856 0.832 Good Health 1.499** 1.498** 1.500** Fair Health 2.030** 2.010** 1.969** Poor Health 2.188* 2.158* 1.741 ADL Score 1.051 1.060 1.050 IADL Score 1.013 1.012 1.009 Medical Problems 1.122* 1.126** 1.141** Part A Use Last Year 1.235+ 1.236+ 1.259+ Doctor Visits 1.003 1.003 1.001 Mental Health Symptoms 1.060 1.063 1.061 Regular Alcohol Use 0.928 0.916 0.907 Years Smoking 1.006* 1.006* 1.006* Regular Exercise 0.880 0.857 0.846 Poverty 0.883 0.897 0.772 Adv. Directive 1.208+ 1.216+ 1.202+ +p> .05, *p > .01, **p>.001 94 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 4.9. Logistic Regression Odds Ratio on Having Long Term Care Costs for 12 or 3 Month Periods for Respondents (N = 5,923) Variable 12 Months Odds Ratio 3 Months Odds Ratio Last Month Odds Ratio Age 1.088** 1.093** 1.093** Gender 0.773 0.740+ 0.714* Age2 1.000 0.999 0.999 Died 3.405** 2.528* 0.828 Age x Dead 0.954+ 0.955+ 0.963 County-2 1.214 1.205 1.246 County-3 1.617 1.582+ 1.803* County-4 1.143 1.107 1.191 County-5 0.860 0.802 0.778 Good Health 1.409 1.507 1.422 Fair Health 1.706+ 1.817+ 1.775+ Poor Health 2.458+ 2.829* 2.448+ ADL Score 1.058 1.061 1.105 IADL Score 1.090* 1.095* 1.087* Medical Problems 1.117* 1.131* 1.156+ Prior Part A Use 1.308 1.300 1.363+ Doctor Visits 1.011 1.010 1.010 Mental Symptoms 1.383+ 1.317+ 1.321+ Alcohol Use 0.715+ 0.727 0.723 Smoking 1.008+ 1.009* 1.009* Exercise 0.710* 0.695* 0.684* Poverty 1.224 1.207 0.748 Adv. Directive 1.139 1.133 1.133 +p > .05, *p > .01, **p > .001 95 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Respondents who died were three times more likely to have used either hospital or long-term care services but only 90% more likely to have used any Part A service during their first 12 months in the plan than were respondents who did not die. Gender was also a significant predictor of all types of service utilization over the 12 months even when controlling for health status. Male respondents were a little more than 20% more likely to use some type of Part A or hospital care while female respondents were 23% more likely to use long-term care services. Variability in service utilization persisted by county of residence in the complex models. Regardless of health status, respondents in County 5 were less likely than those in County 1 to use any types of service, but the difference was statistically significant only for any Part A service use. Regardless of health status, respondents in Counties 2 and 3 remained slightly more likely to use some type of Part A service or long-term care, but less likely to use hospital care than those in County 1. Self-reported health status was a significant predictor of health service utilization. Respondents who characterized their health as good were 50% more likely to use Part A and hospital services, and 40% more likely to use long-term care during the first 12 months in the plan than were those who said their health was excellent. Those with self-reported fair health were twice as likely to use Part A or hospital services and 70% more likely to use long-term care than respondents who reported excellent health. Respondents with self-reported poor health were twice as likely to use hospital care and two and one-half times more likely to use long-term care services than those with excellent health. Respondents with self-reported poor health were 80% more likely to use some Part A service than those with excellent health, but the difference was not statistically significant. 96 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Among the indirect indicators of health status, the number of medical problems was a significant predictor for Part A, hospital, and long term-care use (Tables 4.7, 4.8, and 4.9). Use of Part A services during the year prior to joining the health plan was a significant predictor of Part A and hospital service use, but not long-term care use. Having an advanced directive was also a significant predictor of both Part A and hospital use. Regular alcohol use and exercise and the number of frequent mental health symptoms significantly decreased the probability of using long-term care services during the year's study. The same model incorporating health status indicators taken from the health assessment questionnaire was re-estimated for the last 3 months and the last month of life for decedents for each of the types of service utilization (Tables 4.7,4.8, and 4.9). Results differed only for the death effect on the probability of service utilization over shorter time periods. Death decreased in both power and significance as a predictor of service utilization for all types of care. For both Part A use and long-term care use those who died were actually less likely (OR <1) to have used services in the last month of life than were survivors. Minimal differences in the power or significance were found for the other demographic and health status predictor variables included in the model. Odds ratios on the probability of service utilization for each of these time periods for the different types of service utilization are in Appendix C. 97 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER 5 PREDICTORS OF COST OF CARE Ordinary least squares (OLS) regression models were developed to identify predictors of cost for those elderly HMO members who used services (the claimant population). Models were computed for total costs, hospital costs, and long-term care (skilled nursing home and home health) costs for both the total claimant population and the respondent claimant population. Models were calculated for three different time periods: the full 12-month study period, the last 3 months of life of decedents/average 3 months of survivors, and the last month of life for decedents/average month for survivors. Detailed results are presented here for the 12 months and exclude the last 3 months and last month of life which can be found in Appendix D. Results from simple models, including only those demographic variables common to both populations, are presented. These regression equations were used to calculate average costs for survivors and decedents at ages 65 and 85 in order to better contrast the relative impact of death and age on costs in each of the three time periods. Next, in an attempt to factor in the differential impact of varying health conditions on total costs, complex models that include diagnostic variables were computed. Results of these models for both the total sample and respondent population for the 12-month study are shown. Finally, models were developed for claimants from the respondent population that include additional health status and life-style variables to test the hypothesis that age may serve as a statistical proxy for health status in the simple 98 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. models. Specifically, if the observed age effect in the simpler models represents age serving as a statistical proxy, then adding health status measures derived from discharge diagnoses and other health status indicators from the health assessment questionnaire responses should reduce the estimated impact of age on costs. Total Costs All Claimants Results of simple OLS regression equations for total costs of all claimants are presented in Table 5.1a. A total of 8,010 elderly from a total sample of 28,536 (28.1%) used Part A services during their first year of enrollment. Age and gender alone explained 0.6% of the variance in total costs among those with Part A costs during their first year of enrollment in the health plan. However, age loses its significance as a predictor of total costs when the interactive term "age2 " is added to the model (Model III). The age effect changed significantly when having died is entered as a regressor. Total Part A costs were significantly higher for claimants who died compared to those who survived (Model IV). Costs were, on average, $8,500 higher for claimants of the same age and gender who died compared to those who did not die. When age at death was considered, the cost differential between claimants who died and survived increased to $ 11,000 at age 65 (Model V). However, for every additional year of age at death, total Part A costs actually declined by $240. Respondent Claimants Simple regression models for predicting total Part A costs among the respondents who used services are presented in Table 5.1b. Similar results were found using the 99 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 5.1. OLS Regression Simple Models on Total Costs for Claimants During the 12 Month Study Model Variable I. 1 1 . 1 1 1 . IV. V. VI. a. Total Population (N = 8,010) R2 .0039** .0058** .0059** 0.0394** 0.0410** .0454** Intercept 6035.17** 5466.60** 5674.20** 5321.28** 5210.92** 5253.41* Age 114.34** 120.70** 50.53 43.12 34.86 60.44 G ender 1126.16** 1139.78** 878.81* 866.71* 903.06* A ge2 Died Age x Dead County-2 County-3 County-4 County-5 3.14 1.52 8532.05** 3.06 11218.** -239.51* 1.91 11352.** -246.11* 145.89 -727.44 847.51 + -1877.47** b. R espondent Population (N = 1,440) R2 .0031* .0031 .0032 .047** .051** .054* Intercept 5320.82** 5406.59** 5539.45** 5220.33* 5082.13** 5155.09* Age 88.67+ 89.20+ 49.09 51.12 42.05 34.32 G ender -159.24 -146.27 -405.18 -421.77 -363.25 A ge2 1.63 -0.53 1.37 1.56 Died 8559.19** 12043.** 12285.** A ge x Dead -298.41+ -310.87+ County-2 1254.87 County-3 -703.26 County-4 -20.11 County-5 -874.48 + p > .05, * p > .01, **p> .001 o o smaller sample of respondents with claims. Age, and age when considered with gender, was a significant predictor of total costs, but a less powerful predictor than among total claimants. Total Part A costs for respondents with claims increased on the average $88 with every year of age over age 65. However, as with the total sample of claimants, age lost its statistical significance as a predictor of total Part A costs among respondents with claims when the interactive term age2 was added to the model. The effect on total costs of having died was very similar in the respondent and total claimant samples. Total Part A costs increased significantly for decedent respondents with claims compared to survivor respondents. Respondents with claims who died had total costs about $8,500 higher than those who survived of the same age and gender. When age at death was considered, the cost differential between decedent respondents with claims and survivor respondents with claims increased to $12,000. However, similar to the total claimant population, total Part A costs actually declined with increasing age at death. For every additional year of age at death, total costs declined about $300 for respondents with claims compared to respondents with claims who survived of the same age and gender. Computed Models for Part A Costs Table 5.2 compares the average total costs for males ages 65 and 85 who died and did not die during the three time periods of study using Model V in Table 5.1. Costs were affected by both age and death, but costs clearly rose much more with death than with additional years of age in all time periods considered. Overall, young-old decedents with use of Part A services during the first year in the health plan were more costly than oldest-old decedents with Part A service use during any time period examined. 101 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5.2. Computed OLS Regression Results in Dollars for Total Costs of Males With Claims for Part A Service Use 12 Months Alive Dead 3 Months Alive Dead Last Month Alive Dead Total Population Age 65 $6077.63 $17295.63 $1517.52 $12939.52 $459.46 $9251.16 A ge 85 $7999.82 $14427.62 $1959.45 $10651.64 $635.43 $6970.20 Respondents Age 65 $4660.24 $16703.24 $1150.62 $13315.62 $377.91 $12032.91 A ge 85 $6047.30 $12122.16 $1307.49 $8947.52 $523.41 $6436.79 +p > .05, *p > .01, **p> .0 0 1 Costs for survivors were lower than costs for decedents and younger-old survivors spent less than oldest-old survivors. Total Part A costs of those age 65 who did not die were lowest across all three time periods. Survivors age 85 with Part A costs had higher total costs by more than $1,900 than survivors age 65, reflecting the relatively small age2 effects in Model V. The average Part A costs of claimants who died at age 65 was almost three times higher than that of survivors of the same age and about $3,000 higher than those who died at age 85 during the first 12 months in the health plan. While decedents age 85 had on average double the Part A costs of survivors age 85, the cost of dying decreased with age. This pattern of decreasing costs of dying with increasing age was constant across all time periods and within the respondent population. The pattern of total Part A costs in the last 3 months of life of decedents compared to the average 3-month costs of survivors who used services in the total population was similar to that observed 102 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. over the entire 12 months. The highest average Part A costs were among decedents, with those age 85 having slightly lower total Part A costs than decedents age 65. Among survivors, those age 85 had somewhat higher total costs than those age 65. The age effect on costs for decedents was most evident in the last month of life. Decedents age 65 had on average almost double the total Part A costs of decedents age 85 during their last month of life. The average total Part A costs during the last month of life for decedents age 85 was ten times the average monthly cost for survivors of the same age. Among the young-old decedents, costs during the last month of life for those age 65 were twenty times the average monthly Part A costs for survivors age 65. The pattern of average Part A costs among respondents who used Part A services by age and survivorship was very similar to that found in the total population. Respondents who died had higher average Part A costs at both age 65 and 85 compared to survivors of the same age in all three time periods. Respondent survivors with Part A service use at 85 years of age had slightly higher average Part A costs than those age 65. In the last month of life, the average total Part A costs of decedent respondents age 65 was double that of decedent respondents age 85. Hospital Costs Total Claimants Predictors of total hospital costs for those in the total sample population who used hospital services during the first 12 months in the HMO are shown in Table 5.3a. The power and significance of predictors for hospital use were very similar to those seen for total Part A use because costs derived from hospital care comprised the majority of total Part A costs. Age alone accounted for about 0.2% of the variance in hospital costs. The predictive power of the model increased slightly to 0.5% when age 103 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 5.3. OLS Regression Simple Models on Hospital Cost for Claimants During the 12 Month Study Models Variable I II III IV V VI a. Total P opulation (N = 6,841) R 2 .0019** Intercept 6388.23** A ge 73.40** G ender Age2 Died A ge x Dead County-2 County-3 County-4 County-5 b. R espondent Population (N = 972) .0047** 5737.67** 81.20** 1272.57** R2 Intercept Age Gender A ge2 Died A ge X Dead County-2 County-3 County-4 County-5 .0012 7364.66** 56.34 .0019 7692.38** 54.62 -594.80 .0047** 5816.52** 53.50 1278.90** 1.26 .0019 7711.33** 48.77 -591.65 0.24 .040** 5442.44** 48.04 1040.89** -0.40 7715.36** .0465** 7175.96** 58.63 -793.35 -2.47 768.16** .0419** 5327.20** 38.51 1029.20** 1.29 10272.** -232.06** .051** 6978.58** 47.24 -812.45 0.12 10730* -291.46 .0466** 5392.07* 67.96 1081.54** -0.04 10413.** -237.85** 290.97 -803.73 738.94 -1803.64** 6999 47 -584 -0 11183 -313 4271 -697 -885 068** 86* * 06 .83 .03 ,00** .21 + .94* .37 .19 -1159.04 + = p > .05, * = p > .01, ** = p > .001 o and gender are considered. Similar to the models for total Part A costs, the complete demographic model explained about 5% of the variance in hospital costs for those who used hospital services. Hospital costs increased a small amount with each additional year of age. When gender was considered among claimants of the same age, hospital costs increase slightly more, about $80 with each additional year of age. The significance of age disappeared and the increase in total hospital costs declined when the interactive term age2 and death were added to the model (Models IV and V). Death was a significant and large predictor of total hospital costs for those who required hospital care during their first year in the health plan. Claimants who used hospital care and died had costs about $7,700 higher than those who required hospital care and survived. When age at death was considered, claimants who used hospital care and died had even higher hospital costs, $10,000, than those who survived. However, for every year of age at death, there was a significant decline in hospital costs of slightly more than $200 for claimants of the same age and gender. Respondent Claimants The simple OLS regression models for hospital costs among respondents who used hospital services explained only a small amount of the variance in hospital costs. Models I, II, and III, in which age, gender and age2 were considered explained less than 1% of the variance. Only when death was considered did the models of total hospital costs among respondents who used hospital care become significant and, similar to the models (IV and V), explain about 5% of the variability in hospital costs. Unlike in the total claimant population, among respondents who used hospital care neither age alone nor age and gender were significant predictors of total hospital 105 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. costs. For every year of age over age 65, total hospital costs increased a small amount ($50) for individuals of the same gender who used hospital services. Even when death was considered respondents with hospital claims had only slightly higher costs with every additional year of age (Model IV). Death added significantly to hospital costs among claimant respondents at an amount similar to that in the total hospital claimant population. Respondents with hospital claims who died had total hospital costs that were almost $7,500 greater than those who required hospital care and survived. The difference in total hospital costs between respondents with hospital claims who died and those who survived rose to $ 10,000 when age at death was considered. However, total hospital costs decreased by about $300 for every year of age at death for respondents of the same age and gender who used hospital care and died compared to those who survived. Computed Models for Hospital Costs The pattern of hospital costs among decedents and survivors by age was very similar to that of total costs for both the total population and respondents (Table 5.4). Decedents who used hospital care in the months prior to death cost more than those of the same age who required hospitalization and survived. Further, total hospital costs increased with age among survivors, but were inversely related to age among decedents. In both the total population and among respondents, survivors age 85 who used hospital services had slightly higher average monthly hospital costs than did those who died at age 85. In all three time periods hospital costs are highest for younger-old decedents, those age 65, who used hospital services (Table 5.4). During the first 12 months in the health plan, decedents age 65 who required hospital care had costs $10,000 more than 106 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5.4. Computed OLS Regression Results in Dollars for Hospital Costs o f Males with Claims for Hospital Services 12 M onths 3 M onths Last M onth Alive Dead Alive Dead Alive Dead Total Population Age 65 $6356.91 $16623.91 $1580.12 $12616.12 $472.00 $8721.68 Age 85 $7368.24 $12891.19 $1874.39 $9976.13 $577.74 $6389.56 R espondents Age 65 $6166.13 $16896.13 $1506.05 $15010.05 $416.11 $11417.11 Age 85 $7157.97 $12059.90 $1595.04 $9628.20 $575.55 $6261.97 those of the same age who used hospital care and survived. Decedents age 85 who used hospital services during their first year of enrollment had average costs about $5,000 higher than survivors age 85 who used hospital services. The differential in decedent versus survivor hospital costs remained greatest for the young-old in all three time periods. Long-term Care Costs Total Claimants OLS regression models of the cost of long-term care services for those who used such services during their first year of enrollment are presented in Table 5.5. The amount of variance in long-term care costs explained by the models was considerably less than in the models for all Part A costs or hospital costs. Age alone and age with gender (Models I and II) predicted less than 1% of the total variance in long-term care costs among those in the total sample who used skilled nursing home or home health 107 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. T able 5.5. O L S Regression Sim ple Models o n Total Long-term C are Costs fo r Claimants D uring th e 1 2 M onth Study 4 . C -C « * b $ * c n m * * in * • — « m cn O n o o • — r-* o o 0 0 — ON in o in O n —• O n < * ■ © c n — CN c n — rr NO — in © in N O— O s — r-» — p* O ^ 0 0 c n O 00* O n c n o c n c n © c n in © r* in NO ON i 0 0 1 00 0 0 CN NO NO ^r i o o in r»* N" < * ■ * — CN o O n CN i p - r* CN i i I i i * * * p » O n * * © c n c n OO ON in i n o NO 0 0 0 0 0 0 o O NO in CN — o CN CN — * — o o — o o NO rr © o ^5* in CN o 0 0 O n NO NO m i SO n NO n - O s — i * * T T * O s * o — c n p- o o c n O m s o NO CN 0 0 * s o rr o o s o v o in * r - * ON ON * * ^ t OO * 0 ^ — 0*^ O ~ ON CN CN ‘ ^ N O § S O 0 0 • CN * * T T * * 00 * * o 0 0 OO 0 0 O n * m in o — o o n s o o o o CN 0 0 * O O * * ■ * — ' < n I I °° » © s _ o _ g "3 s . © 0- £ « ttJ * * * * m O n — - os c n © C * ^ so o CN < n NO * o < n p - c n NO ON o Os — — c n CN 0 0 NO o c n c n 0 0 t t 1 1 O i n i n i ^ — o ~ o ^ • c n r-* o o c n ON i n 0 0 0 0 CN O O n o m 1 * * o C - o o NO o CN NO O O c n o c n NO ON O n O c n g ^ * o * * • 5 0 0 0 5 5 © m in 3 ~ c n f t , c n c n O 2 ft* a. o C J T J 03 " O o C 0 0 U — — < O < Q 4) B V T J B ~ ^ . O O i « i i & K e c c s 0 3 3 3 3® e o o o o o . <UUOU js 8 M £ 1 ^ ri U O C ' 0 0 < U - a C O Q u n m m ^ I I I I * C S G C w u O 3 3 3 r 00.2 : c o o O O Oi rs , — « u >*# w w w w a : £ < o < Q < u u u c o o A Q. * * A Q. I I * in' o A a . I I + 108 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. care services. The predictive power of the models increased minimally when death and the interactive terms age2 and "age x death" were added (Models III, IV, and V). The full demographic model (VI) predicts only 1.5% of the variance in long-term care services among all claimants who used such services. Age was a significant predictor of long-term care costs alone and with gender (Models I and II). The age effect lost its statistical significance as a predictor of such costs when the interactive term age2 was factored into the model (Model III). However, the strength of the age effect remained fairly constant with each additional year of age over 65, adding about $67 to long-term care costs. Death was not a statistically significant predictor of long-term care costs among claimants who used long-term care. However, long-term care costs were about $1,000 higher for those who used such services and died compared to those who survived. Age at death caused a 50% decrease in the death effect on costs, but itself had no significant impact on the cost of long-term care services (Model V). Respondent Claimants The models of long-term care costs among respondents who utilized long-term care services explained less than 1% of the variability in long-term care costs among this population. Neither age nor death were significant regressors. Long-term care costs increased slightly, $33, among respondents with each year of age over 65 (Model I). The age effect was not impacted by gender but did increase slightly, by $10, when death was considered (Model III). Long-term care claimants in the respondent population who died had higher costs than survivors, but the difference was not statistically significant. The costs difference between respondent decedents and survivors who used long-term care doubled to 109 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. $1,000 but remained statistically insignificant when age at death was considered (Model V). Among respondents who used long-term care services, age at death was also not a statistically significant predictor, but there was a small decrease in such costs with each additional year of age at death. Computed Models for Long-term Costs The average calculated combined skilled nursing home and home health costs, among those who used such services, at age 65 and 85 are shown in Table 5.6 for 12, 3, and 1 -month periods. Long-term care costs were considerably lower than hospital costs and comprised a much smaller proportion of total Part A costs. The pattern of costs among those who used long-term care services differed from that observed in total Part A and hospital costs. In the total population that used long term care, long-term care costs were highest for decedents age 85 during the first 12 months of enrollment and in the last 3 months of life compared to the average 3-month costs for those who survived. However, long-term care costs in the last month of life were highest among decedents age 65 compared to those of decedents age 85 or average month costs of survivors at both ages. Among the respondents who used such services, decedents age 65 had the highest average long- term care costs in all three time periods examined. Long-term care costs among those who used these services and did not die were consistently higher for those age 85 compared to those age 65. Decedents age 65 who used long-term care services also had higher costs than survivors age 65 who used such services among both populations over all three time periods. 110 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5.6. Computed OLS Regression Results in Dollars for Long-term Care Costs o f Males with Claims for Long-term Care Services 12 Months Alive Dead 3 M onths Alive Dead Last Month Alive Dead Total Population Age 65 $2023.00 $2484.49 $511.06 $2056.82 $187.42 $2508.14 Age 85 $3239.76 $3715.97 $766.47 $2896.49 $279.00 $1959.12 Respondents Age 65 $1756.43 $2925.23 $291.50 $2404.98 $199.41 $2902.69 Age 85 $2589.49 $2776.05 $509.61 $1824.83 $203.59 $1759.36 Other Effects Gender Several other results are of interest. For example, gender and county of residence affected not only the probability of health service utilization as discussed in Chapter 4, but also influenced costs for those who used health care services. There was a significant difference in total costs by gender. Males at any age over age 65 had higher total costs by a little more than $1,000 than did females of the same age (Table 5.1, Model II). When death was considered along with age (Model IV), the gender differential declined by a little more than $100, but remained statistically significant. The gender differential in total Part A costs was not affected further when age at death or county of residence was considered. There was no statistically significant difference in total costs by gender among respondents with claims (Table 5.1b). Total Part A costs for male respondents with claims tended to be lower than for females. Interactive 111 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. terms of gender with age and death were applied to the model but were not significant and did not alter the findings so are not presented here. Among claimants who used hospital care, there was also a significant difference in the total cost of such care by gender. Males who required hospital care had significantly higher total hospital costs than females, by $1,200. The gender effect among respondents for hospital costs was just the opposite. Male respondents of the same age who used hospital services had lower costs by $600, but when compared to female respondents of the same age who used these services the difference was not significant. Female respondents with hospital claims who died had higher costs than males of similar characteristics by almost $800, but the difference was not statistically significant. Gender was not a statistically significant predictor of long-term care costs in either population. Male long-term care claimants in both the total and respondent population had slightly higher costs than female long-term care claimants. However, in the respondent population when death was considered, the gender difference in long term care costs almost disappeared. County of Residence The cost of care was significantly affected by the county of residence which presumably was the same county where the care was received. Total Part A costs differed significantly for claimants who resided in Counties 4 and 5 compared to claimants in County 1 . Claimants in County 4 had higher overall total costs and those in County 5 had lower total costs than claimants who resided in County 1 . Claimants 112 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. who lived in Counties 2 and 3 also tended to have higher total Part A costs than those in County 1 , but the difference was not statistically significant. There was no statistically significant difference in total Part A costs for respondents with claims who resided in different counties. Respondents with claims who resided in County 2 tended to have higher total costs and those from the three other counties tended to have lower total costs than respondents with claims in County I. Respondents with hospital claims who lived in County 2 had significantly higher costs than those in County 1 . Respondents who used hospital services in the three other counties had lower costs than those in County 1, but the differences were not statistically significant. County of residence was a significant factor only in the cost of long-term care for residents of County 5 in the total population who used this type of service. Residents of County 5 who required long-term care had costs that were about $1,000 less than residents of the same age and gender in County 1 who used long-term care. Residents in Counties 2 and 3 also had lower long-term care costs, and those in County 4 had minimally higher costs than those in County 1, but the differences were not statistically significant. Among respondents who received long-term care, costs did not differ significantly by county of residence. Long-term care claimant respondents who resided in Counties 2 and 4 had higher long-term care costs, and those in County 3 had lower costs than those in County 1 . Complex Models Tables 5.7, 5.8, and 5.9 present the results of more complex OLS regression models for total Part A costs, hospital costs, and long-term care costs over the entire 113 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. first year of enrollment in the health plan. Model VII shows the result of adding diagnostic categories for claimants in both the total population and respondent population. In Model VIII, additional predictor variables of health status and life-style are added for the respondent population. Total Part A Costs With the addition of diagnostic categorical variables, the predictive power of the models for total costs among those who used Part A services increased dramatically (Table 5.7). From explaining only 5% of the variability in the simple models, the model now explains 34% of the variability in total costs among the total population and 36% of the variability among the respondents. When diagnostic variables were added to the model (Model VII), age and age2 declined in value as predictors of total Part A costs among those who used services. In both the total and respondent population, total Part A costs actually declined with each additional year of age over 65. Among the respondents when health status and life style variables were included (Model VIII), total Part A costs increased a small, but insignificant, amount with each additional year of age over 65. The interactive term age2 was not significant in any of the complex models for either population. In the complex model that included diagnostic categories, age2 was negative indicating that total Part A costs actually declined at older and older ages. Among the respondents the decline increased slightly when health status along with diagnoses were considered, indicating that even when these factors were considered costs declined by a small amount with each year compared to those a year younger. 114 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5.7 Complex OLS Regression Models for Total Costs o f Those with Claims for Part A Services Total Population Respondents Variable Model VII Model VII Model VIII R2 0.34** 0.36** 0.37** Intercept -2425.80** -2312.34* -2151.46 Demographic Age -55.97 -8.46 48.17 Gender 772.27* -22.04 407.71 Age2 -0.10 -1.67 -4.08 Died 4107.53** 6124.81** 5925.42** Age x Dead -133.03+ -246.85+ -256.15+ County-2 261.34 1708.68+ 1686.72+ County 3 -613.98 -36.40 69.37 County-4 780.87+ 594.04 645.80 County-5 -1421.15** -354.07 -337.86 Diagnostic Hematologic 5436.85** 6954.94** 6871.31** Circulatory 4466.98** 3786.49** 3747.82** Congenital 3528.97** 1040.25 1619.64 Endocrine 3958.66** 3006.15** 3004.60** G-I 1925.75** 1382.46+ 1352.51 + G-U 3544.88** 3899.37** 4063.52** Infectious 5814.45** 4852.63** 4620.12** Injury 4995.79** 4444.89** 4449.50** Psychiatric 772.47 1109.93 1193.53 Orthopedic 2578.40** 2217.17** 2080.67** Neurologic 3760.60** 3721.37** 3584.22** 115 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5.7, cont. Total Population Respondents Variable Model VII Model VII Model VIII Neoplasm 2072.59** 2735.57** 2638.03** Respiratory 5296.45** 4877.70** 4668.48** Dermatologic 6246.87** 2668.26+ 2654.82+ Health Status Good Health -44.47 Fair Health -194.99 Poor Health 5451.49* ADLs -433.49 IADLs 260.04 Medical Problems 143.60 Prior Part A Use -1038.21 Doctor Visits -22.60 Mental Symptoms 1408.31* Alcohol Use -503.92 Tobacco Use -3.30 Regular Exercise 18.98 Poverty -1573.43 Advanced Directive -248.42 +p > .05, *p > .01, **p > .001 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Death remained a powerful and significant predictor of total Part A costs for those who used Part A services during their first year of enrollment even when diagnostic categories were considered. Total Part A costs among decedents in the total population who used Part A services were $4,000 higher than survivors, and among respondents $6,000 higher, when other demographic factors and the diagnostic category of illness were the same. Age at death continued to have a negative effect on total Part A costs over the year. With every additional year of age at death, total Part A costs of the total claimant population decreased by $130 even when diagnostic category of illness is considered. Among respondents, the decline in total Part A costs with increasing age at death was even greater. Gender remained a significant predictor of total Part A costs in the total population but not among claimant respondents. Males in the total population had total Part A costs that were $770 greater than those of females even when the diagnostic category of illness was taken into consideration. Among male respondents who used Part A services, total costs were only about $20 lower than female respondents who used such services when discharge diagnoses were considered, but were $400 greater for men when health status and life-style variables from the questionnaire were added to the model. Among the respondents the gender differences were not statistically significant. Among the diagnostic categories all except psychiatric diagnoses were statistically significant predictors of increased total Part A costs among those who used such services during their first year of enrollment in the health plan. Respiratory, hematologic, infectious diseases, and injuries had the greatest increase on total Part A costs among those who used Part A services in both sample populations. Among all 117 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. those with Part A costs, dermatologic conditions were the most costly category of illness. The addition of health status and life-style variables as reported by the respondents upon their entrance into the health plan did not add much to the models' ability to predict total health care costs (Model VIII). Total Part A costs increased slightly with additional years of age over 65 and decreased in the same amount among those who died when these additional predictors were added compared to the previous model that included only demographic and diagnostic category variables. Respondents with self-reported poor health, as compared to those with excellent health, had significantly higher total Part A costs by $5,000 during the first year. Among the other health status and life-style variables, only frequent mental health symptoms added significantly to total Part A costs. Hospital Costs Similar to the complex models of total Part A costs, the addition of diagnostic category variables increased the predictive power of the models for total hospital costs to about 30% (Model VII) (Table 5.8). Age became insignificant and caused a slight decrease in hospital costs over the year for those who used hospital services when diagnoses were considered. With each additional year of age, hospital costs actually declined compared to those of persons one year younger, as indicated by the negative coefficient on the interactive term age2 in all the complex models. As in total Part A costs, death had a significantly positive effect on total hospital costs. Those who died and used hospital services during their first year in the health plan had significantly higher hospital costs in both the total population and among the respondents. The death effect was slightly greater among the decedent respondents 118 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5.8. Complex OLS Regression Models for Hospital Costs of Those with Part A Claims Total Population Respondents Variable Model VII Model VII Model VLI R2 0.34** 0.34** 0.34** Intercept -1775.96** -2092.26** -1007.04 Demographic Age -63.72 -13.94 32.62 Gender 737.36** -52.56 179.31 Age2 -1.2 -2.32 -3.66 Died 4343.71** 6421.62** 6289.02** Age X Dead -146.88* -235.52+ -238.54* County-2 215.92 1438.48+ 1375.96+ County-3 -658.97 -1.40 96.71 County-4 585.71 301.06 317.96 County-5 -1266.44** -248.63 -202.34 Diagnostic Hematologic 5074.25** 5722.81** 5673.45** Circulatory 4029.85* 3371.20** 3384.84** Congenital 3310.88** 2386.09 2732.64 Endocrine 3451.82** 2355.51** 2480.97** G-I 1778.32** 1082.52*** 1022.18* G-U 3017.13** 3230.13** 3348.07** Infectious 4832.57** 3288.43** 3075.13* Injury 4094.32** 3356.42** 3380.70** Psychiatric 394.78 1375.26 1503.57 Orthopedic 1755.85** 1249.92*** 1134.54*** 119 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5.8, cont. Total Population Respondents Variable Model VII Model VII Model VIII Neurologic 2785.08** 2163.82** 2133.63** Neoplasm 1599.13** 1531.25* 1416.05* Respiratory 4753.20** 4050.51** 3852.37** Dermatologic 4451.61** 2098.28*** 2190.67*** Health Status Good Health -79.88 Fair Health -352.98 Poor Health 5071.63* ADLs -410.56 IADLs 137.34 Medical Problems -53.53 Prior Part A Use -687.72 Doctor Visits -9.97* Mental Symptoms 1256.81* Alcohol Use 7.68 Tobacco Use -5.75 Regular Exercise 0.44 Poverty -1439.86 Advanced Directive -263.89 +p > .01, *p > .01, **p > .001 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. than in all decedents ($6,422 compared to $4,344). With each additional year of age at death, hospital costs declined about $150 in the total populations and $230 in the respondent populations, statistically significant for both populations. The gender effect was different in the two populations. Males in the total population who used hospital services had significantly higher hospital costs by $700 than females even when diagnostic category of illness was considered. Among the respondent population, the gender differential in hospital costs was not significant nor as large. Total hospital costs increased significantly for those who used hospital services among the total population with all diagnoses except psychiatric. In the respondent population, congenital problems lost its statistical significance as a predictor of hospital costs and the level of significance declined for gastroenterologic, orthopedic, and neoplastic problems compared to the total population. Respiratory and hematologic conditions increased hospital costs the greatest among both populations. In the total population infectious diseases, circulatory problems, and injury also were among the most costly contributors to hospital costs while among the respondents infectious diseases were not quite as costly. Self-reported poor health compared to excellent health was a significant and powerful predictor of total hospital costs among respondents, adding $5,000 to hospital costs during the year. Endorsement of mental health symptoms also added significantly to hospital costs among respondents who used hospital services. No other self-reported health status or life-style variables were significant predictors of hospital costs. 121 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Long-term Care Costs The addition of diagnostic categories of illness increased the predictive ability of the models of total long-term care costs to 10% (Table 5.9). The complex models were significant in predicting these costs among those who used skilled nursing home or home health services among both the total sample population and the respondents. The demographic variables remained insignificant predictors of total long-term care costs over the year for those who used nursing home or home health care in both populations (Model VII). Total long-term care costs were not affected by increasing age over 65 and age was not a significant predictor of such costs in either population. Long-term care costs declined among decedents in the total population by $143 but increased about $300 among respondents when diagnostic categories of illness were considered. They increased slightly less, only $193, among respondents when health status and life-style variables were added to the model. Age at death had a minimal positive impact on total long-term care costs in the total sample population who used such services and a minimal negative impact in the respondent population. Gender continued to have the opposite effect on long-term care costs in the complex models as it had in the earlier simpler ones. Males had lower long-term care costs than females in both the total sample and respondent population. The gender differential was largest (-$180) in the respondents with long-term care use when health status was also taken into account. In the total population, all conditions contributed significantly to long-term care costs except congenital and gastroenterologic problems. Dermatologic conditions significantly increased total long-term care costs during the 12 months for the total population. Among the respondents with Part A use, fewer categories of illnesses were significant contributors to long-term care costs. Hematologic conditions were the most 122 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5.9 Complex OLS Regression Models of Long-Term Care Costs for Those With Part A Claims Total Population Respondents Variable Model VII Model VII Model VII R2 Intercept Demographic Age Gender Age2 Died Age x Dead County-2 County-3 County-4 County-5 Diagnostic 0.14** -731.46 1.78 -37.54 1.15 -143.58 14.59 -6.94 36.96 132.34 -146.28 0 . 10* * -418.25+ -4.49 -177.77 1.23 288.55 -9.77 105.13 126.75 166.26 32.75 0. 11** -1066.30* 9.291 -110.48 0.20 193.45 -14.60 134.55 142.89 170.29 36.79 Hematologic 261.99* 663.44** 659.39** Circulatory 400.84** 157.53 152.83 Congenital 300.19 -727.03 -675.38 Endocrine 469.95** 436.11* 386.10 G-I 104.44 148.66 193.13 G-U 441.86** 265.17 283.81 Infectious 773.52** -116.03 -145.71 Injury 780.82** 431.55** 424.56* Psychiatric 457.30** 103.35 -5.61 Orthopedic 708.47** 522.29** 510.09** Neurologic 783.19** 710.50** 666.65* Neoplasm 254.45** 402.11* 423.05* Respiratory 436.01** 365.27* 383.11* Dermatologic 1731.20** 275.71 460.47 123 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5.9, cont. Total Population Respondents Variable_________________ Model VII_______ Model VII Model VII Health Status Good Health -60.18 Fair Health 73.61 Poor Health 498.50 ADLs -1.89 IADLs 117.30** Medical Problems -10.70 Prior Part A Use -98.35 Doctor Visits -10.49 Mental Symptoms -124.72 Alcohol Use -105.66 Tobacco Use -0.14 Regular Exercise 13.44 Poverty 363.81 Advanced Directive -29.74 +p > .05 *p > .01, **p > .001 124 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. costly contributors to long-term care costs among the respondents. In both population samples, neurologic conditions, injury, and orthopedic problems also significantly increased long-term care costs. Having a congenital or infectious disease diagnosis actually had a negative impact on long-term care costs in the respondent population. The only health status and life-style variable that contributed to the predictive ability of the model on total long-term care costs was IADL impairment. For each reported impairment in an IADL, long-term care costs increased about $100. A number of the other health status or life-style variables were actually associated with lower long-term care costs, but none was statistically significant in predicting such costs (Model VIII). 125 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER 6 PREDICTING DEATH AND HIGH COST DYING Up to now the focus of this research has been on predictors of the use and cost of Part A services. Results of models presented thus far demonstrate that among the elderly, both age and death play a role in health care service utilization and the resulting costs. In order to better understand what leads to death, and in particular high cost dying, models of predictors of death and high cost death were developed. Simple models using age and the limited number of other demographic variables available were first presented. Then, in order to evaluate the effect of health status on age as well as the overall predictive value of the models, indicators of health status were added as available for the total claimant population and the respondents. Predictors of Death Logistic regression models predicting death in the total and respondent populations are presented in Table 6.1. Age alone was a small, but significant, factor in predicting who will die during their first year in the health plan in both populations. With each additional year of age the probability of dying increased 9%. In the total population, the age effect on death was not affected by the addition of other demographic variables. However, in the respondent population, the power and significance of age diminished when other demographic variables 126 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 6.1 Odds Ratio on the Probability of Dying During the 12-month Study Variable Model I Model II Model III Model IV a. Total sample population (N = 28,536) A ge 1.090* 1.079** 1.88** A ge2 1.001 1.001 G en d er 1.617** 1.580** C ounty-2 0.859 0.825 C ounty-3 1.222+ 1.202+ C ounty-4 0.972 0.984 C ounty-5 0.962 1.975 C laim /N o Claim 1.742** P seudo R2 0.05 0.06 0.07 b. Respondent population (N = 5,926) A ge 1.090** 1.055+ 1.048+ 1.054+ A ge2 1.001 1.001 1 G ender 1.635** 1.588** 1.756** C ounty-2 1.049 1.016 1.053 C ounty-3 1.104 1.096 1.079 C ounty-4 0.922 0.927 0.968 C ounty-5 0.789 0.806 0.751 C laim /N o Claim 1.833** 1.413* Responses G ood H ealth 1.837* Fair H ealth 3.052** Poor H ealth 9.220** A D Ls 1.008 IA D Ls 1.134** Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 6.1, cont. Variable Model I Model II Model III Model IV Number of Medical Problems 1.025 Prior Part A Use 1.403 * Physician Visits 1.002 Mental Health Symptoms 0.749 Alcohol Use 0.703+ Smoking 1.008+ Regular Exercise 0.837 Poverty 0.831 Advanced Directive 0.849 Pseudo R2 0.05 0.06 0.07 0.13 +p > .05, *p >.01, **p > .001 were added to the model. When health status indicators drawn from the questionnaire data were added, the age effect was not altered. Table 6.2 presents the odds ratio on the probability of dying when only those who used Part A services in both the total and respondent population are considered. In both claimant populations the age effect was slightly weaker than in the total population. Among the claimant populations the age effect declined about 1% when additional demographic variables were entered into the model. When health status indicators, discharge diagnoses, and/or self-reported response variables were added to the models the power and significance of age decreased further both populations. Having a discharge diagnosis in one of several categories was highly predictive of death among both the claimants and the respondent populations. 128 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 6.2. Odds Ratio on the Probability of Dying During the 12-month Study for Those Who Used Part A Services V ariable M odel 1 M odel 2 M odel 3 M odel 4 M odel 5 a . T o ta l c la im a n ts (N = 8,011 A ge 1.054** 1.042* 1.026 A ge2 1.001 1.001 G en d er 1.481** 1.293* C ounty-2 0.838 0.918 C ounty-3 1.284 1.405+ C ounty-4 0.932 0.91 C ounty-5 1.068 1.092 D iag n o stic C a te g o ry C irculatory 2.811** C ongenital 1.025 D erm atology 1.318 E ndocrine 1.958** G -I 0.778* G -U 0.938 H em atology (B lood) 1.164 Infections 1.723** Injury 0.837 M usculoskeletal 0.599** Psychiatric 1.099 N eurologic 1.108 N eoplasm 2.268** R espiratory 2.684** Pseudo R2 0.02 0.03 0.16 b. R e sp o n d e n ts w ith claim s (N = 1,441) A ge 1.068** 1.045 1.059 1.037 1.05 A ge2 1.001 1 1.001 1 G ender 1.477+ 1.549+ 1.475 1.553 C ounty-2 0.611 0.629 0.725 0.721 129 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 6.2. cont. Variable Model 1 Model 2 Model 3 Model 4 Model 5 County-3 0.888 0.954 0.989 1.102 County-4 0.913 0.931 0.831 0.836 County-5 1.201 1.173 1.201 1.238 Responses Good Health 1.893 0.749 Fair Health 1.781 1.431 Poor Health 4.758* 2.381 ADLs 0.967 0.976 IADLs 1.119+ 1.157* Number Medical Problems 0.911 0.86 Prior Part A Use 1.293 1.24 Physician Visits 0.992 0.995 Mental Health Symptoms 0.792 0.787 Alcohol Use 0.716 0.788 Smoking 1.01 1.003 Regular Exercise 1.023 1.253 Poverty 0.887 1.228 Advanced Directive 0.896 1.083 Diagnostic Category Circulatory 2.725** 2.909** Congenital 3.763 3.867 Dermatology 1.563 1.455 Endocrine 2.137** 2.099+ G-I 0.702 0.799 G-U 1.172 1.184 Hematology (Blood) 0.757 0.768 Infections 1.737 1.63 Injury 0.738 0.749 Musculoskeletal 0.691 0.667 Psychiatric 1.361 1.098 Neurologic 0.82 0.788 Neoplasm 2.941** 3.120** Respiratory 2.989** 2.977** Pseudo R2 0.05 0.06 0.13 0.12 0.17 **p > .001, *p > .01, +p > .05 130 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Circulatory, endocrine, respiratory, and neoplastic conditions were consistently found to be significant predictors of death. These same four diagnostic categories of illness remained significant predictors of death even when self-reported health indicators were added to the model among the respondent population. Having received care for an injury, or musculoskeletal or gastroenterologic condition decreased the likelihood of dying in both populations . In the respondent population, only two indicators of health status were significant predictors of dying within the first 12 months in the health plan. Those who reported their health status as poor were significantly more likely to die than those who reported excellent health. Those with poor health were much more likely (4.7 times) and those with fair health (78%) more likely to die within twelve months than were those who said their health was excellent upon entering the plan. Those who identified their health as good were actually slightly less likely than those with excellent health to die, but the difference was not statistically significant. Increasing difficulties with IADLs, but not ADLs, also significantly increased the probability of dying among the respondents. The probability of dying decreased for those who reported frequent psychologic symptoms, regular alcohol use, regular exercise, were poor, or had an advanced directive. The pseudo R2 was computed to compare the overall predictive power of the models. With increased model complexity, the amount of variability increased in both populations. The simple models that included only demographic variables explained 5% or less of the variability in the probability of dying. When additional health status variables were added, a substantial increase in the variability of the probability of dying is produced. Among the total claimant population, demographic variables and discharge diagnostic categories explained 16% of the probability of dying, while in the 131 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. respondent population these same variables explained slightly less, only 12%, of the probability of dying. In the respondent population the probability of dying was explained equally well by the addition of either set of health status indicators: diagnostic categories or self-reported response variables. When both sets of health status indicators were employed there was a small increase in the predictive power of the model. Predictors of High Cost Death The same equations were re-estimated to predict high cost death, defined as expenditures in excess of $10,000 and $20,000 during the study period. The number of individuals with costs in excess of $20,000 were so few among the respondent population that these equations could not be calculated. Therefore, only the results of equations that predict the probability of dying with total costs in excess of $ 10,000 are shown here (Table 6.3). Age alone is a small, but significant, predictor of dying with high costs for those who used Part A services. Among the total claimant population, the probability of dying with high costs increased about 9% with each additional year of age and remained significant, even when gender and county of residence were considered. Among the respondents the probability of dying increased just 3% with each additional year of age and, when the additional variables were entered into the equation, age lost both its power and significance. The addition of health status indicators, either discharge diagnostic categories or self-reported indicators, caused a further decline in the effect and significance of age as predictor of high cost death in both claimant populations. 132 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 6.3. Odds Ratio on the Probability of Dying with Costs in Excess of $10,000 During the 12-month Study V ariable Model I M odel 2 Model 3 Model 4 M odel 5 a. Total claimants (N = 8,011) Age 1.085** 1.091** 1.021 Age2 1.000 1.00 G ender 1.423* 1.081 County-2 1.407 1.489 County-3 1.461+ 1.575+ County-4 0.892 0.975 County-5 1.021 1.213 Diagnostic Category Circulatory 6.322** Congenital 1.496 D erm atology 2.101** Endocrine 2.560** G-I 1.068 G-U 1.261 H em atology (Blood) 1.571* Infections 2.069** Injury 1.413+ M usculoskeletal 0.804 Psychiatric 1.147 Neurologic 1.650** N eoplasm 3.007** Respiratory 5.491** Pseudo R2 .03 .04 .41 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 6.3, cont. Variable M odel I Model 2 M odel 3 M odel 4 M odel b. Respondents with Claims (N = 1,441) Age 1.026** 1.011 1.026 0.997 1.011 Age2 1.001 1 1 0.998 Gender 1.144 1.182 1.12 1.21 County-2 0.427 0.427 0.588 0.647 County-3 1.173 1.221 1.456 1.533 County-4 0.933 0.972 1.095 0.995 County-5 0.917 0.883 0.905 0.801 Responses Good Health 2.324 1.907 Fair Health 4.294 2.966 Poor Health 17.137* 9.172+ ADLs 0.811 0.714 IADLs 1.168+ 1.285* Num ber Medical Problems 0.842 0.772 Prior Part A Use 1.574 1.492 Physician Visits 0.992 1.01 Mental Health Symptoms 0.8771 0.634 Alcohol Use 0.823 1.042 Smoking 1.016+ 1.006 Regular Exercise 0.998 1.188 Poverty 2.058 1.419 Advanced Directive 0.745 0.951 Diagnostic Category Circulatory 2.414+ 2.782+ Congenital 4.538 4.334 Dermatology 3.588* 3.548+ Endocrine 2.006+ 2.137+ G-l 0.813 0.928 G-U 2.722+ 2.895* Hematology (Blood) 0.825 0.996 Infections 1.541 1.574 Injury 1.313 1.548 Musculoskeletal 1.286 1.154 Psychiatric 1.837 1.736 Neurologic 0.808 0.823 Neoplasm 1.996 2.151 + Respiratory 9.141** 8.038** Pseudo R2 0.01 0.02 0.1 0.3 0.36 **p > .001, *p > .01, +p > .05 134 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Certain diagnostic categories of illness are consistent strong predictors of costly deaths. In both populations, respiratory illnesses were strong predictors of high cost death. Respondents with respiratory problems were nine times more likely to die with high costs, and those in the total population with respiratory problems were five and one-half timesmore likely to die with costs in excess of $10,000. Circulatory diseases increased the probability of dying with high costs sixfold in the total claimant population and about two and one-half times in the respondents. Dermatologic conditions were also found to increase the probability of high cost death two to threefold. This probably represents the cost of care for those who are debilitated with decubitus (bed sores) and therefore meet Medicare criteria for skilled nursing services either in the home or the nursing home. Several health status and life-style indicators taken from respondents' self- reports upon entering the health plan were found to be predictors of high cost death. Self-reported poor health is the most potent predictor of high cost death during the ensuing 12 months. Respondents who identified their health as poor were 17 times as likely to die with high costs as those who reported excellent health. Those with fair or good health were more likely than those with excellent health to have a high cost death, but the differences were not statistically significant. Impairment of IADLs and smoking also significantly increased the likelihood of dying with high costs. When diagnostic categories were considered along with questionnaire responses, self- reported health status declined, but IADL impairments increased, as predictors of high cost death. The predictive power of the models for explaining the probability of high cost death increased with increasing model complexity. When only age and demographic variables were considered, less than 5% of the probability of high cost death was 135 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. explained. The inclusion of demographic variables and health status, as reflected by discharge diagnoses categories, explained about 36% of the probability of dying with institutional costs in excess of $10,000. Death, and in particular costly death, were not totally unpredictable event. The presence of certain categories of illness and self-identification of one’s health as poor were strong predictors of dying within 12 months. However, age within an elderly population was not a very good predictor of death or high cost death. 136 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER 7 DISCUSSION OF RESULTS AND THEIR IMPLICATIONS This study lends support to the hypothesis that age-based rationing of health care appears to be occurring already, albeit implicitly. The study expands our understanding about the role age and death play in the use and cost of health care services by the elderly and contributes to knowledge on the effect that managed health care has on the utilization of health care services by the elderly. By using multivariate analysis, this study goes beyond most of the previous research on the subject and examines simultaneously the role of age and death within an older population on both utilization and the cost of care, and the role of age in predicting death. Policy Recommendations The results of this study strongly suggest that the use of advanced age as a rationing criterion would result in the continued expenditure of resources on younger- old individuals who are sick and dying at the expense of limiting resources to the healthy oldest-old. Support for a policy recommendation to achieve cost containment using age as a criterion for limiting access to care is not found. Such a policy would be difficult to justify ethically or economically based on the findings reported here. Rather, the results of this study suggest that policies designed to contain health care costs among the elderly should focus on limiting care to those who are dying or have health conditions known to lead to a costly death. The results of this study and their policy implications are perhaps best understood in the context of a series of questions. 137 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Is there evidence that age-based rationing or limitations in the care of the elderly is already occurring? If elderly individuals at the oldest ages who died were receiving aggressive, cost-intensive treatment prior to death, one would expect to see a strong relationship between both age and death on hospital costs, the location of the most technologically advanced and costly treatments. The fact that the age effect on hospital costs is rather small (5-9% per year) and that the death effect is considerably larger (200-500%) suggests that this may not be the case. Indeed, it appears that the expensive, high- technology interventions, while not frequently used, are more likely to be used in the course of dying and are done so more often for the younger elderly. Temkin-Greener and colleagues (1992) suggested "that much of the cost for older decedents may be more a function of aging and related debility than of medical interventions associated with death" (p. 699). The research reported here supports the first half of the statement, but would suggest that it is death, or the dying process, not age, that fuels rising health care costs. The cost for medical care of older decedents is more a function of dying than of age, but dying with less aggressive medical interventions. And the cost of older survivors is a function of debility and illness, and not age itself. This is consistent with prior descriptive studies of utilization in the fee- for-service sector that showed usage associated with dying rather than aging (Lubitz & Prihoda, 1984; Lubitz & Riley, 1987; McCall 1984; Scitovsky 1984, 1988). The fact that the same trends in utilization and cost are found in a managed care system suggest these are not an aberration of economic incentives, but represent more fundamental patterns of need and care. For those elderly members of the health plan who used Part A services, the effect of age and death, separately and together, on costs was even more pronounced than it was for utilization. This is most easily seen in the computed models that 138 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. compared costs between young-old and oldest-old survivors and decedents. Costs increased slightly with age, but they rose dramatically for those who died, regardless of age. For decedents, costs declined with increasing age at death, while they rose slightly with age for survivors. These multivariate models confirm the findings from the descriptive literature (Gomick et al., 1993; Riley & Lubitz, 1987; Scitovsky, 1984, 1988). Again, this suggests that patterns reported in the fee-for-service sector are not due solely to the economic incentives of that system. If age-based rationing were the norm, one would expect to see costs decline with increasing age for survivors as well as decedents. The cost increase observed with increasing age for survivors suggests age-based limitations on care are not a systemic occurrence, based either on practice patterns or policy (implicit or otherwise). Rather, the findings reported here suggest there is limitation on the extent of care for the dying elderly, which might represent a limited application of age-based rationing. Age appeared to have little effect on the resources expended, as measured by cost, until there is some indication of impending death which, it can be theorized, might be easier to anticipate with increasing age. Death in our society is a costly event and there exists what has been called the "high cost of dying" (Ginzberg, 1980). The results reported here provide convincing evidence that it is a higher cost of dying for those elderly who die at younger ages. There is also what might be termed a "high cost of living" or "high cost of aging"; elderly survivors at older ages have higher health care costs than younger elderly survivors. The data here suggest that some age-based rationing of care to the dying occurs, but cannot answer whether or not this is appropriate. Once an older plan member is receiving Part A services, there appear to be limitations on the extent of care as 139 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. reflected by lower costs with increasing age. It can be argued that less costly, and therefore less aggressive, care for older-old decedents represents the prudent evaluation by providers to allow those who have lived a full life to avoid needless pain and suffering of a prolonged death. On the other hand, it might also be that providers are using age as an implicit indicator of health status and offering less aggressive care to individuals at the oldest-old ages simply because of their age. The complex models that include health status measures indicate that, for those with better than poor health by self-report, this may not be an accurate assumption. If this is the case then age- based rationing may already be occurring as a tacitly sanctioned policy among providers. Would the adoption of age-based rationing policies control health care costs of the elderly? The results of this study strongly argue against policies to limit care based solely on age as a means for controlling the cost of health care for the elderly. Age was shown to be a significant, but weak, predictor of utilization, cost, and death. Death appears to be the primary force with a much stronger effect than age on both utilization and cost of health care among the elderly, something that was not evident from previous research. Further, the influence of age appears to have been overstated in previous research leading to erroneous interpretations about the influence of increasing age on health care costs. Age and death also interact with a negative effect on costs, something that has not been documented in prior research. With each additional year of age there is a small increase in the probability of use of all Part A services that is somewhat more pronounced for long-term care than for hospital services. This is consistent with previous research on the impact that increasing age has on utilization in the fee-for-services sector (Fisher, 1980; Roos et 140 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. al., 1987; Waldo & Lazenby, 1984; Wolinsky et al., 1987). Increasing age among elderly enrollees in a managed care health plan exhibited similar patterns on utilization and did not appear to be a limiting factor in access to care. Again this suggests that age is not being used any differently to affect the care of the elderly in two very different delivery systems. The small effect each additional year of age has on the probability of service utilization was not altered by the introduction of variables controlling for health status upon entering the plan, even as the period prior to death diminished. With increasing age there was a persistent small increase in the likelihood of using health services, again suggesting that rationing of access to care based on age did not appear to be taking place. However, when the cost of care for those who use services was considered, controlling for health conditions and health status reduced the importance of age as a contributor to cost. This pattern suggests that it is an older person’s health status, as reflected by self-report and discharge diagnoses, that affects the cost of care. Thus, it is the elderly with poorer health (by self-report or by types of illnesses) who have higher costs, not those of older ages. It should be kept in mind that there is an increased likelihood of poorer health with increasing age. Thus, it becomes apparent that it is the linkage between increasing age and impaired health, not age directly, that leads to higher utilization rates and costs with increasing age. In this context the high cost of care at the end of life for some of the elderly might be better understood as ordinary medical care of the very sick, not extraordinary or excessive care of the very old or dying. One can envision the desire of insurers and health care delivery systems, especially the soon-to-be insolvent Medicare system, to identify populations of high utilization and cost who are, with a high degree of certainty, likely to die. If a formula 141 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. could be developed (based on salient factors from the research literature) to identify such a population, it is not too far-fetched to imagine it being used to limit or exclude such individuals from all, or certain types, of care. The proliferation of severity of illness measures (Iezzoni et al., 1995; Knaus et al. 1991; Steen et al., 1993) and their use by providers (hospitals), payers (insurance companies, managed care organizations), and government is a move in just this direction. While age would most likely be a consideration in such a formula, it should be evident from this research that when the elderly are examined as a heterogeneous population rather than compared en masse to younger populations, age alone is a small contributing force to health care utilization and costs. Mechanic (1985) suggested that within a managed health care system, access to health care services could be limited for the elderly. Indeed, one of the incentives within a managed health care system is to prevent excessive use of costly services which usually translates into institutional acute care services (Rosenfeld, 1996). The results of this research do not point to any limitation in access to the more costly services, such as hospital care, by age within the elderly population. Longitudinal data and data on costly services such as ICU care would be required to explore more completely the issue of access to costly services by age. In an effort to control costs, HMOs caring for younger populations were the first to show that costly hospital services could be substituted successfully with less expensive non-hospital care (home health, skilled nursing home, and outpatient services). Medicare and private insurance carriers have followed suit, but Schwartz and Mendelsohn (1992) suggested that the majority of the cost savings that can be achieved by elimination of unnecessary hospital stays and days has already occurred in both the fee-for-service and managed health care systems. However, within the 142 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. managed care system studied here, some additional cost-shifting away from costly institutional services seems evident when compared to care in the Medicare fee-for- service sector. A smaller proportion of study participants had both no Part A costs and no costs in excess of $10,000, than were reported for Medicare beneficiaries in 1990 (Helbing, 1993). To what extent such shifting of services can occur with an older population, and at what cost or benefit, is not known. Further, whether this represents a limitation in access to care for all elderly enrollees cannot be addressed by this study. Can costs be saved by rationing care to those who die? The decline of costs among decedents of increasing age suggests that, for those older individuals who require hospital or long-term care services, the care that is provided is less aggressive (i.e., less cost-intensive) the older the elderly person. How these limitations in care to the dying occur is beyond the scope of this study. Whether a decision (if there is an actual decision) is made to withhold certain types of costly care and by whom remains unanswerable. Providers, appropriately or not, acting overtly or implicitly, may be limiting the type and range of care provided to those of older-old ages (Barondess et al., 1988; Scitovsky, 1984). Or it may be that patients are exerting their wishes, directly or through advanced directives. Families may also be exerting their desires to limit care when they perceive a very old member to have lived out their years or to be on a downward course of failing health and function. The latter scenario seems less likely given the recent findings of the SUPPORT study (1995) in which the wishes of patient and their families to limit care were generally ignored. However, that study involved individuals of younger ages within the fee-for-service sector where incentives to be more aggressive remain. 143 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The results presented here suggest that death, when examined retrospectively, is a powerful predictor of the use of certain types of health care services (Part A Medicare benefits) and the costs associated with such care. Further, death and high cost death are predictable outcomes associated with a limited number of health status indicators. This supports development of policies that would restrict the type of care for those individuals who, regardless of the intervention and its costs, will die. Such policies, directed at limiting resources that would be expended on those who will die anyway, would be ethically acceptable and could be anticipated to be publicly acceptable if applied fairly to all, not just the aged (Annas, 1985). Among the elderly, death is a much more potent predictor of costs than is age. However, it is death identified retrospectively. If our goal is too develop policies that will contain costs by limiting the high expenditures associated with dying, research must move away from the easy targets of age and costs, and focus on the factors associated with death and dying. Research that focuses only on costs associated with death among the elderly is itself age-biased. While increasing age is strongly associated with an increased probability of dying, this association misdirects attention from the desired goal: to identify prospectively those who will die and expend large amounts of resources in the process, so that these resources can be "saved" and used for a better benefit elsewhere. To effectively control health care costs among the elderly, results from this study suggest that we need to be able to better manage death (Sulmasy, 1995). But to do so we must increase our knowledge and understanding about death among the elderly and non-elderly. To target only the elderly is to approach the problem with an inherent age-bias that does not take into account the heterogeneity of the older population, the high costs of dying for younger populations, and the generally low cost 144 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of care for the vast majority of the elderly during most of their later years of life. Outcome research on the common underlying diseases associated with dying are needed so that clinical decision-making about interventions is based on a true understanding of the costs and benefits involved. Such research should begin by examining the specific illnesses and interventions for those categories of disease (e.g., respiratory, circulatory, and neoplastic) that are linked with high cost death. Research is also required to better elucidate the wishes and desires of different cohorts, so that this information can be incorporated into standards of care. Such standards of care, as they are established, need to written with age as one of many factors, not the overriding or only consideration. The death effect is tempered slightly when health status and life-style indicators are considered and is more pronounced the longer the time period prior to death used for analysis. This supports prior findings that older people at any age who die are more likely to be involved in receiving hospital or long-term care in the year preceding their death than are survivors. Intuitively this is not surprising, since death for many in our society is increasingly the final event in a cumulative process of deteriorating health. Some will die from their acute illness in the hospital and thus have hospital use during the last month of life. Others will survive the hospitalization and die subsequently some weeks or months later, but not in the immediate post acute-care period. What additional policy changes are suggested by this study? The minority of the population who consume a disproportionate share of the resources represent those with chronic medical conditions who have on-going need for intervention. The challenge is for health care systems to develop better ways to deal 145 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. with people who have chronic diseases and physical or mental disabilities. In order to provide appropriate care and not "waste" resources, the goal must be to differentiate prospectively those for whom costly interventions are beneficial from those in whom they are futile or of limited or no quantifiable benefit (Esserman et al., 1995). Cost-effectiveness research provides a framework in which managed health care systems and society can begin to make such evaluations and comparisons between outcomes (Kaplan, 1995; Weinstein, 1995). The focus in such equations is not just on years of life (mortality), but also on the quality of life in terms of years of wellness or disability that are produced as the result of any specific intervention. The move to managed care, especially for the elderly population, increases the importance and urgency of making decisions regarding what services will be funded and for whom, based not just on cost, but on sound information about the outcomes (Duggan, 1989). Within such a model, medical benefit criteria measured by the quality and length of life gained, not age, becomes the standard by which intervention choices and decisions are made (Kilner, 1989). It is important to note the differences that emerge when hospital and long-term care services are disentangled from total Part A costs. Not surprisingly, hospital services, which constitute the bulk of Part A use and cost, follow the same pattern as total Part A use and costs. However, some important differences emerge when long term care services are examined separately. Both age and death were stronger predictors of long-term care use and cost than they were in the models of hospital or total Part A costs. Although nursing home and home health comprise a small fraction of the costs of hospital or the total health care budget, based on the limited Medicare reimbursement criteria for these services, they are increasingly being used to substitute for more costly hospital care. Policies, similar to those for hospice, that encourage 146 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. prospective identification of those who will die and restrict services provided to this population would probably foster further substitution of less cost-intensive non hospital care. This might result in expanded definitions for coverage of these services within the Medicare guidelines or encourage managed care organizations to develop these programs as part of their expanded benefits package. The results reported here demonstrate a pattern of utilization and cost among elderly enrollees in an HMO during a limited period after initial enrollment in a managed care plan that is consistent with what has been reported from the fee-for- service sector: a small number of enrollees are responsible for the majority of health care costs, costs for decedents are higher that those for survivors, and costs for younger decedents are higher than for older decedents. Further, this research has shown that a few categories of illness are associated with dying and dying with increased costs. This strongly suggests that if the goal is to control costs over a short period of time, managed health care organizations and other health care systems would be wise to develop policies and programs that substitute less costly services for this identified population of the elderly. However, if the goal is to control costs for conditions that affect large numbers of the elderly and lead to high costs over a longer period of time (e.g., Alzheimer's disease, arthritis), then very different types of programs and policies will need to be considered. The data reported here on the distribution of costs within a population of new elderly enrollees in a managed health care system are strikingly similar to that reported previously in the literature from the Medicare fee-for-service sector (Lubitz & Prihoda, 1984; Lubitz & Riley, 1984; McCall, 1984) and in Canada (Roos & Shapiro, 1981; Roos et al., 1989, 1984). The pattern of "small group, high usage" persists even in this changed health care environment. Only 30% of enrollees used any institutional 147 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. services during their first year in the health plan and a majority of these costs were attributable to use by a small percentage of enrollees. The forces behind these consistent patterns remain unclear, but probably represent the interaction or combined effect of diseases, interventions, and provider behavior. The changing health care system to one that favors a managed care orientation will have little, if any, impact on the phenomenon of a concentration of resources of a small proportion of the population. The hope is that, in a health care system where there are no incentives to provide more care than minimally necessary, the high cost segment can be better "managed" and costs contained. However, if these individuals represent the very ill, then even within a managed care environment the very nature of their conditions will warrant interventions. The gender and geographic differences in utilization and costs were also similar to what has been reported in the fee-for-service environment. The fact that males have consistently been found to have higher utilization rates for hospital care has been attributed to gender biases in provider behavior and poorer health status as measured by more co-morbidities in men compared to women (Bickell et al., 1992; Mutran & Ferraro, 1988). Higher utilization of long-term care by women is probably due to their increased longevity and the lack of a spouse to provide supportive care. The small geographic area variations found here between the two counties is a phenomena that has been well-documented (Wennenberg, 1973). These findings suggest that the clinical and social factors that affect the practice patterns in the fee-for-service environment persist within a managed care program despite very different economic incentives. To change these types of practice patterns requires changing not just the payment or delivery system, but a change of provider behavior. Managed care organizations attempt to do just this through their quality review and oversight. 148 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. However, as Lo (1995) suggested, what may be needed is a drastic change in physician culture that must begin in the medical schools. Direct and indirect measures of health status contribute to improved understanding of the variability in use and cost of services, but still leave much unaccounted for. The indicators from the questionnaire data that were significant predictors of utilization among the respondent population are consistent with what has been reported in previous research involving the elderly. It is well established that utilization during the prior year is a good predictor of services (Anderson & Knickman, 1984, Densen et al., 1959). Therefore, it is not surprising that use of hospital, emergency room, or home health services in the prior year is once again demonstrated in this study to be a significant predictor for use of services. Prior use is, however, not a predictor of cost among those who use services. The presence of chronic health problems has been shown to play a significant role in utilization among the elderly in HMOs (Haug, 1980; Hibbard & Pope, 1986), as it was in this study. Given the increased out-of-pocket costs associated with frequent medical care in the fee-for-service sector, it is very likely that among the elderly with chronic medical problems that require ongoing care, the monetary incentives to join a Medicare HMO are very real. As the HMO senior market continues to expand, the health plans will have to increasingly deal with how to manage care for an increasingly frailer population with multiple chronic medical problems. The federal policy of continuous open enrollment for Medicare HMOs allows Medicare recipients much greater flexibility to change coverage than exists among other insured populations. What effect, if any, the disenrollment of significant numbers of enrollees from the plan, either to other plans or the fee-for-service sector, has on the provision of care by Medicare HMOs is unknown. Further, for the 149 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Medicare beneficiary, the economic incentive to change plans once a medication benefit limit has been reached is very strong. Policy changes with enrollment windows and commitment periods along with annual limitations on certain benefits that are portable between plans are likely areas for change in the next few years. The findings reported here on the effect of perceived health status on utilization are consistent with what has been reported previously, especially among the elderly (McFarland, 1988; Wolinsky et al., 1987). The probability of using hospital or long term care services is significantly increased for those with a self-report of fair or poor, as compared to excellent, health. Report of poor health also contributes significantly to costs for those who use services. Freeborn et al. (1990), in a study of elderly enrollees in an HMO, also found that consistently high users of care were more likely to report fair or poor health. Although this study was based on elderly enrollees in a managed health care system, the data still reflect the inherent acute care bias and limitations of the Medicare system. The use of long-term care services is under-reported, since most of the payment for nursing home, and all the cost of supportive living (i.e., board and care, retirement, homemaker assistance) remains outside of Medicare reimbursement, even within a Medicare HMO system. A comprehensive managed care system in which the full range of care in all settings is covered might produce different results (Roosetal., 1987). Federal policies to promote advanced directives were adopted with the expressed hope that such documents would be utilized to promote autonomy and limit undesired and unnecessary care. Conflicting evidence exists in the literature to date that advanced directives have achieved these goals. The self-reported possession of such a document among the respondents increased the likelihood of utilization, decreased 150 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. costs slightly, and did not impact on the effect of death as a predictor of care. However, having an advanced directive did decrease the likelihood of having a costly death. While these results do not reflect on whether this effect was due to clinicians' awareness of and compliance with an advanced directive, they do strongly argue for promotion of advanced directives as an indirect means of controlling costs. Data on use and cost of health care services by the elderly has led to the notion that "if we want to stem the rise in medical care costs, medical care expenditures at the end of life provide an excellent target for cost-containment efforts" (Scitovsky, 1984, p. 592). This study lends credence to this statement while pointing out the fallacy of equating end-of-life with increasing age. To control medical care costs at the end of life, we should examine commonalities between the cost of dying of people at all ages based on similarities in health conditions, not age. Sampling Issues This study was based on a convenience sample of new elderly enrollees in an HMO in Southern California. The findings are representative of new enrollees in similar health care plans in this region of the country during the early 1990s. Elderly enrollees in health plans in other regions of the country may differ considerably from this population, as may enrollees from other time periods. The genralizability of these results to populations from other regions and in other time periods therefore is questionable. The exclusion of Medicare beneficiaries under age 65 limits the comparability of these results to studies and other populations of Medicare beneficiaries that include these younger, disabled individuals. The issue of selection bias, both favorable and non-favorable, among elderly enrollees in HMOs has received considerable attention by the media and in the 151 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. scientific literature (Eggers & Prihoda, 1982; Luft, 1982; Retchin et al., 1992). Medicare enrollees in the early HMO demonstration projects tended to be younger and healthier (on a number of measures) than the general population. Others suggest that the economic incentives to join an HMO attracts those with multiple chronic health problems who incur considerable regular out-of-pocket expenditures which are eliminated by joining such a plan. The fact that the age distribution and mortality rate of the sample population in this study did not differ significantly from that reported for the general population over age 65 suggests any selection bias here, if present, is too subtle to be detected. Voluntary disenrollment from the health plan during the study was significant (12%), but lower than the 20% figure reported in Medicare risk contracting nationally (Brown et al., 1993). During the time period of this study, approximately 20% of Medicare beneficiaries in California were enrolled in an HMO (McMillan, 1993) and competition already existed in the Southern California market for membership from Medicare enrollees between several health plans. In subsequent years competition has increased further in this region while in other regions of the country the entry of HMOs into the health care marketplace serving the elderly is just beginning. Familiarity with Medicare HMOs may partially explain the lower disenrollment rate. The racial composition of the sample population remains unknown. The racial composition of the population over age 65 in the counties from which the sample was drawn is known to be predominantly (75-90%) Caucasian, with some counties having significant elderly populations of other racial groups (Hispanic, Black, and Asian) (Dept, of Commerce, Bureau of the Census, 1994). Whether any racial bias exists in the population who select to join the HMO and what effect, if any, race has on the research questions examined here, remains unanswerable. 152 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. It can be assumed that the respondent population contains some selection bias, but this remains undefined and difficult to quantify. Questionnaire respondents generally were not as sick or frail as non-respondents, since members of the former group with ill health would be more likely to be hospitalized or institutionalized at the time of the survey and therefore not available to respond, although proxy responders were allowed. Since the respondents were drawn from new voluntary enrollees in the health plan, it is likely that the very sick or frail are under-represented in this population. The similarity of the death rate between this sample population and that of the general U.S. population suggests that a bias to a healthier population, based on mortality, was not evident. On the other hand, a smaller proportion of the respondents without claims died than among the total or non-respondents without claims, suggesting that respondents who did not use Part A services may have been healthier than other non-claimants. Data Limitations The use of claims-based data has inherent limitations which impact the type of questions that can be answered and methodologies that can be employed. Claims- based data are retrospective in nature and provide only a rough sketch, at best, as to what occurred during episodes of illness. Identifying information about the individual was minimal with data on parameters that have been shown to affect health service utilization (race, marital status, education, income) not routinely included. Detailed information about specific health service use such as intensive care days, hours of surgical time, laboratory tests, and medication use are often not included or are unreliable. 153 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Claims-based data do not include any direct measurement of health status, an especially important factor when trying to understand patterns of use and cost prior to death. An attempt to alleviate this limitation was provided by linking the claims-based data to the questionnaire responses. The questionnaire was completed upon entering the plan and reflects self-report at that time, not at the time of utilization or illness. Ideally what is needed is a comprehensive medical information system with detailed utilization and personal health information that can be followed over time. The advent of computerized charting and user-friendly large computer systems should potentially greatly increase the quantity and the quality of the data available for population-based studies to examine the questions posited here. Managed care organizations are fertile ground for the development and implementation of such systems which offer great potential for applicability to activities such as utilization review and quality assurance. Conclusions As suggested by Lawlor (1995) in his review of Medicare managed care, if such a system change is to be truly effective in controlling costs, change will need to occur within the Health Care Financing Administration that oversees Medicare, the health plans and their providers, and among beneficiaries. I would suggest that an even greater change is necessary, one to a comprehensive system where all health care services are capitated and the health and illnesses of the elderly are really "managed." However, knowledge on the risks and benefits of different interventions must be acquired for the most common (and costly) conditions so that truly informed clinical decisions can be made by both providers and patients. Cost is only one factor, along with age, quality of life, and others that must be considered in such an equation. 154 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In an attempt to control health care costs, especially within the Medicare program, politicians seem always too anxious to find a "quick fix." The targeting of "high utilizers" and the very old provides such an attractive target. However, the numbers of high users remains small and the patterns of use and cost under different economic incentives are remarkably similar. It is difficult from current research to foresee vulnerable areas where costs can be significantly reduced without resulting either in otherwise preventable deaths or the unintended consequence of increased morbidity and additional costs. Targeting of the very old would also not have the intended large amounts of cost savings. For, as demonstrated in this study, death at increasing old age is actually less costly than death at younger older ages. The one remaining possibility for significant cost savings (benefit) that would be politically, economically, and ethically justifiable is to target the dying, if they could be accurately identified prospectively. 155 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. REFERENCES Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Aaron, H. J., & Schwartz, W. B. (1984). The painful prescription: Rationing hospital care. Washington, DC: Brookings Institution. Aaron, H. J., & Schwartz, W. B. (1990). Rationing health care: The choice before us. Science, 247,418-422. Anderson, G., & Knickman, J. R. (1984). Patterns of expenditures among high utilizers of medical care services: The experience of Medicare beneficiaries from 1974-1977. Medical Care, 22, 143-1499. Annas, G. J. (1984). The prostitute, the playboy and the poet: Rationing schemes for organ transplantation. American Journal o f Public Health, 74(8), 843-845. Awoke, S., Marton, C., & Parrott, M. (1992). Outcomes of skilled cardiopulmonary resuscitation in a long term care facility: Futile therapy? Journal o f the American Gerontology Society, 40(6), 593-595. Ayanian, J. Z., & Epstein, A. M. (1991). Differences in the use of procedures between men and women hospitalized for coronary artery heart disease. New England Journal o f Medicine, 325, 221-225. Barer, M. L., Evans, R. G., Hertzman, C., & Lomas, J. (1987). Aging and health care utilization: New evidence on old fallacies. Social Science Medicine, 24(10), 851-872. Barondess, J. A., Kalb, P., Weil, W. B., Cassel, C., & Ginzberg, P. (1988). Clinical decision-making in catastrophic situations: The relevance of age. Report of a conference convened by the American Geriatrics Society, April 17-18, 1987. Journal o f the American Gerontology Society, 36, 919-937. Barry, R. L. (1991). Mandatory, universal age-based rationing of scarce medical resources. In R. L. Barry & G. V. Bradley (eds.), Set no limits (pp. 3-13). University of Illinois Press. Bayer, R., Callahan, D., Fletcher, J., Hodgson, T., Jennings, B., Moses, D., Sieverts, S., & Veatch, R. (1983). The care of the terminally ill: Morality and economics. New England Journal o f Medicine, 309(21), 1490-1494. Bellamy, P., & Oye, R. K. (1987). Admitting elderly patients to the ICU: Morality and Economics. New England Journal o f Medicine, 309(21), 1490-1494. Berk, M. L., & Monheit, A. C. (1993). The concentration of health expenditures: An update. Health Affairs,'Winter, 145-149. Berk, M. L., Monheit, A. L., & Hagan, M. N. (1988). Analysis of the distribution of health expenditures for the United States 1928-1980. Health Affairs, Fall, 46- 60. 157 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Berki, S. E., Lepowski, J. N., Wyszewianski, L. et al. (1985). High-volume and low- volume users of health services, United States, 1980. National medical care utilization and expenditure survey. Series C, Analytical Report No. 2. DHHS Pub. No. 86-20402. National Center for Health Statistics, Public Health Service. Washington, DC: U.S. Government Printing Office. Binstock, R. J., & Post, S. G. (1991). Old age and the rationing of health care. InR. J. Binstock & S. G. Post (eds.), Too oldfor health care (pp. 1-12). Baltimore, MD: The Johns Hopkins University Press. Blank, R. H. (1988). Rationing medicine. New York: Columbia University Press. Bloom, B. S., & Kissick, P. D. (1980). Home and hospital cost of terminal illness. Medical Care, 18(5), 560-564. Brock, D. B., Foley, D. J., & Losonczy, K. G. (1987). A survey of the last ninety days of life. Procedures o f the American Statistics Association, Social Statistics Section (pp. 306-311). Brown, R. S., Clement, D. G., Hill, J. W., Retchin, S. M., and Bergesten, J. W. (1993). Do health maintenance organizations work for Medicare? Health Care Financing Review, 75(1), 7-23. Bums, R., Graney, M., & Nichols, L. (1989). Predictors of in-hospital cardiopulmonary outcomes. Archives o f Internal Medicine, 149, 1318-1321. Byrne, C., Vernon, P., Cohen, J. J. (1994). Effect of age and diagnosis on survival of older patients beginning chronic dialysis. Journal o f American Medical Association, 271(1), 34-36. Callahan, D. (1984). Rationing health care: Will it be necessary? Can it be done without age or disability discrimination? Issues in Law and Medicine, 5(3), 354- 366. Callahan, D. (1987). Setting limits: Medical goals in an aging society. New York: Simon and Schuster. Callahan, D. (1988). Allocating health resources. Hastings Center Report, April- May, 14-20. Carlisle, D. M., Siu, A. L., Keeler, E. B., McGlynn, E. A., Kahn, K. L., Rubenstein L. V., & Brook, R. H. (1992). HMO vs fee-for-service care of older persons with acute myocardial infarction. American Journal o f Public Health, 52(12), 1626- 1630. Chassin, M. R. (1982). Costs of outcomes of medical intensive care. Medical Care, 20(2), 165-179. 158 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Childress, J. F. (1979). Rationing of medical treatment. In W. T. Reich (ed.). Encyclopedia o f Biomedical Ethics (pp. 1114-1119). New York: Oxford University Press. Christensen, S. L., Long, H., & Rodgers, J. (1987). Acute health care costs for the aged Medicare population: Overview and policy options. Milbank Memorial Fund Quarterly, 55(3), 397-425. Churchill, L. R. (1987). Rationing health care in America: Perceptions and principles o f justice. Notre Dame, ID: University of Notre Dame Press. Civetta, M. M. (1973). The inverse relationship between cost and survival. Journal o f Surgical Research, 14, 265-269. Clark. (1985). The social allocation of health care resources: Ethical dilemmas in age group competition. The Gerontologist, 25, 119-120. Culler, S. D., Callahan, C. M., & Wolinsky, F. D. (1995). Predicting hospital costs among older decedents over time. Medical Care, 53(11), 1089-1105. Cunningham, F. C., & Williamson, J. W. (1980). How does the quality of health in HMOs compare to that in other settings? An analytic literature review. 1958- 1979. Group Health Journal, 1,4-25. Daniels, N. (1986). Why saying no to patients in the United States is so hard: Cost containment, justice and provider autonomy. New England Journal o f Medicine. 314(21), 1380-1383. Daniels, N. (1988). Am I my parents keeper? An essay on justice between the young and the old. New York: Oxford University Press. Damay, A. J. (ed.). (1994). Statistical record o f older Americans. Detroit: Gale Research Inc. Davis, F. A. (1988). Medicare hospice benefit early program experiences. HCFR, 9(4), 99-111. Davis, K., & Rowland, D. (1986). Medicare policy; New directions for health and long-term care. Baltimore, MD: The Johns Hopkins University Press Densen, P. M. (1992). Tracing the elderly through the health care system. ACPHR, No. 282-87-0049. Washington, DC; USDHHS. Densen, P. M., Shapiro, S., & Einhom, M. (1959). Concerning high and low utilizers of service in a medical plan and the persistence of utilization levels over a three- year period. Milbank Memorial Fund Quarterly, 37, 100-109. Detsky, A. S., Strieker, S. C., Mulley , A. G., & Thibault, G. E. (1981). Prognosis, survival and the expenditure of hospital resources for patients in an intensive care unit. New England Journal o f Medicine, 305, 667-672. 159 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Duan, N., Manning, W. G., & Morris, C. N. (1983). A comparison of alternative models for the demand of medical care. Journal o f Business and Economic Statistics, 1, 115-126. Duggan, J. M. (1989). Resource allocation and bioethics. Lancet, April 78,772-773. Eddy, D. M. (1991). What care is essential, what services are basic. Journal o f the American Medical Association, 265(6), 782-788. Eddy, D. M. (1991). Rationing by patient choice. Journal o f the American Medical Association, 265(1), 105-108. Edmunds, L. H., Stephenson, L., & Ratcliffe, M. (1988). Open heart surgery in octogenarians. New England Journal o f Medicine, 319, 131-136. Eggers, P., & Prihoda, R. (1982). Pre-enrollment reimbursement patterns of Medicare beneficiaries enrolled 'at risk’ HMOs. Health Care Financing Review, ¥(1), 55- 71. Ellencweig, A. Y., & Pagliccia, N. (1994). Utilization patterns of cohorts of elderly clients: A structural equation model. HSR, 29(2), 225-245. Emmanuel, E. J., & Emanual, L. L. (1994). The economics of dying: The illusion of cost savings at the end of life. New England Journal o f Medicine, 550(8), 540- 544. Esserman, L., Belkora, J., & Lenert, L. (1995). Potentially ineffective care: A new outcome to assess the limits of critical care. Journal o f the American Medical Association, 274(19), 1544-1551. Evans, R. W. (1991). Advanced medical technology and elderly people. In Binstock & Post (eds.), Too oldfor health care? (pp. 44-74). Baltimore MD: The Johns Hopkins University Press. Evans, R. W. (1983). Health care technology and the inevitability of resource allocation and rationing decisions. Part I, Journal o f the American Medical Association, 249(15), 2047-2053; Part II, Journal o f the American Medical Association, 249(16), 2208-2219. Evashwick, C., Rowe, G., Diehr, P. & Branch, L. (1984). Factors explaining the use of health care services by the elderly. HSR, 19(3), 357-374. Fink, A., Siu, A., Brook, R., Park, R., & Solomon, D. (1987. Assuring the quality of health care for older persons: An expert panel’ s priorities. Journal o f the American Medical Association, 258, 1905-1904. Fisher, C. R. (1980). Differences by age groups in health care spending. Healthcare Financing Review, I, 65-90. 160 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Forster, L. E., & Lynn, J. (1980). Predicting lifespan for applicants to inpatient hospice. Archives o f Internal Medicine, 148, 2540-2543. Freeborn, D. K., Pope, C. R., Mullooly, J. P., & McFarland, B. H. (1990). Consistently high users of medical care among the elderly. Medical Care, 28(6), 527-540. Freeborn, D. K., Pope, C. R., McFarland, B. H. & Siegenthaler, L. (1988). Consistently high and low elderly users of medical care. National Center for Health Services Research, Program Note, Executive Summary. Fuchs, V. R. (1984). Though much is taken: Reflections on aging, health and medical care. Milbank Memorial Fund Quarterly, 62, 143-166. Furrow, L., Wartman, S. A., & Brock, D. (1988). Science, ethics and the making of clinical decisions. Journal o f the American Medical Association, 259(21), 3161- 3167. Gagnon, R. P. (1978). Office visits by persons aged 65 and over: National ambulatory medical care survey. United States, 1975. Advance Data from Vital and Health Statistics. Garfmkle, S. A., Riley, G. F., & Iannachione, V. G. (1988). High-cost users of medical care. Health Care Financing Review, 9(4), 41-52. Gaumer, G. C., & Stavins, J. (1992). Medicare use in the last ninety days of life. Health Service Research, 26(6), 725-752. Gibbs, J., & Newman, J. (1982). Study of health services used and costs incurred during the last six months of a terminal illness. Contract No. HEW-100-79- 0110. Chicago IL: Blue Cross and Blue Shield Association. Ginzberg, E. (1980). The high cost of dying. Inquiry, 77,293-295. Gomick, M., McMillan, A., & Lubitz, J. (1993). A longitudinal perspective on patterns of Medicare payments. Health Affairs, Summer, 141-150. Gomick, M., Greenberg, J. N., Eggers, P. W., Dobson, A. (1985). Twenty years of Medicare and Medicaid: Covered populations use of benefits, and program expenditures. HCFR ***** Grimes, D. S. (1987). Rationing health care. Lancet, 1(8533), 615-616. Guralnik, J. M., LaCroix, A. Z., Branch, L. G., KasI, S. V., & Wallace, R. B. (1991). Morbidity and disability in older persons in the years prior to death. American Journal o f Public Health, 81(4), 443-447. Haber, 1986. Rationing is a reality. Journal of the American Gerontology Society, 34, 761-763. 161 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Hahn, B., & Lefkowitz, D. (1992). Annual expenses and sources ofpayment for health care services (AHCPR Publ. No. 93-0007). National Medical Expenditure Survey Research Findings 14, Agency for Health Care Policy and Research. Rockville, MD: Public Health Service. Helbing, C. (1992). Medicare program expenditures. Health Care Financing Review Annual Supplement, 23-27. Hertzman, C., & Hayes, M. (1985). Will the elderly really bankrupt us with increased health care costs? Canadian Journal o f Public Health, 76,373-377. Hibbard, J. H., & Pope, C. R. (1986). Age differences in the use of medical care in an HMO. Medical Care, 24(f), 52-66. Hing, E. (1987). Use of nursing homes by the elderly: Preliminary data from the 1985 National Nursing Home Survey. Advance Data No. 135. Hyattsville MD: National Center for Health Statistics. Hosking, N., Warner, M., Lobell, C., Offord, K., & Melton, J. (1989). Outcomes of surgery in patients ninety years of age and older. Journal o f the American Medical Association, 267:1909-1915. Iezzoni, L. I., Ash, A. S., Schwartz, M., Daley, J., Hughes, J. S., & Mackieman, Y. D. (1995). Predicting who dies depends on how severity is measured: Implications for evaluating patient outcomes. Annals o f Internal Medicine, 123,763-770. Jahnigen, D. W., & Binstock, R. H. (1991). Economic and clinical realities: Health care for elderly people. In Binstock & Post (eds.), Too oldfor health care? Controversies in medicine, law, economics and ethics (pp. 13-43). Baltimore MD: The Johns Hopkins University Press. Johnson, R. E., Mullooly, J. P., & Greenlick, M. R. (1990). Morbidity and medical care utilization of old and very old persons. Health Service Research, 25(4), 64- 65. Kaplan, R. M. (1995). Utilization assessment for estimating quality-adjusted life years. In F. Sloan (ed.), Valuing health care: Costs, benefits and effectiveness o f pharmaceutical and other medical technologies (pp. 31-60). London: Cambridge University Press. Kapp, M. (1989). Health care tradeoffs based on age: Ethically confronting the "R" word. The Pharos o f Alpha Omega Alpha, 52(3), 2-7. Kapp, M. (1991). Rationing health care: Legal issues and alternatives to age-based rationing. In Barry & Bradley (eds.), Set no limits: A rebuttal to Daniel Callahan's proposal to limit health care for the elderly (pp. 71-91). Katz, S., Branch, L. G., Branson, M. H., Papsidero, J. A., Beck, J. C., & Greer, D. S. (1983). Active life expectancy. New England Journal o f Medicine, 309,1,218- 224. 162 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Kemper, P., & Murtaugh, C. M. (1991). Lifetime use of nursing home care. Mew England Journal o f Medicine, 324(9), 595-600. King, K. M. (1990). Health care costs at the end of life. Congressional Research Service, Report for Congress. Kilner, J. F. (1989). Age criteria in medicine. Archives o f Internal Medicine, 149, 243-246. Kilner, J. F. (1990). Who lives, Who dies? Ethical criteria in patient selection. New Haven CN: Yale University Press. Knaus, W. A., & Thibault, G. E. (1981). Intensive care units today. InB. J. McNeal & E. G. Cravalho (eds,), Critical issues in medical technology (pp. 192-215). Boston: Auburn House Publishing. Knaus, W. A., Wagner, D. P., & Lynn, J. (1991). Short term mortality predictions for critically ill hospitalized adults: Science and ethics. Science, 254, 389-394. Knaus, W. A., Wagner, D. P., Draper, E. A., Zimmerman, J. E., Bergner, M., & Bastos, P. G. (1991). The APACHE III prognostic system. Risk prediction of hospital mortality for critically ill hospitalized patients. Chest, 100, 1619-1636. Kovar, M. G. (1983). Expenditures for the medical care of elderly people living in the community throughout 1980. National Medical Care Utilization and Expenditure Survey, Data Report No.4. Hyattsville MD: National Center for Health Statistics. DHHS Pub. No. (PHS) 84-20000. Lamm, R. D. (1987). Ethical health care for the elderly: Are we cheating our children? In Smeeding (ed.), Should medical care be rationed by age? Totowa, NJ: Rowan and Littlefield. Lawlor, E. F. (1995). Public policy, the costs of death, and Medicare. The public policy and aging report, 6(A), 1-3, 10-11. Lawlor, E. F. (1995). Is Medicare managed care the answer? Public Policy and Aging Report, 6(5/6), 1-4, 15-17. Lester, T. Medicine versus economics. New England Journal o f Medicine, 313(\0i), 611-614. Levkoff, S., Wetle, T., Coakley, E., & Kasper, J. (1992). Determinants of expenditures and use of services among elderly HMO enrollees. HMO Practice, 6(3), 6-12. Levkoff, S., Welte, T., Coakley, E., Jorda, H., Hartline, D., & Szmuszkovicz, D. (1988). Predictors of service utilization by Medicare HMO enrollees. HMO Practice, 2(2), 58-65. 163 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Lo, B. (1995). Improving care near the end of life: Why is it so hard? Journal o f the American Medical Association, 274(20), 1634-1636. Lubitz, J. (1990). Use and costs of Medicare services in the last year of life, 1976 and 1985. Health Care Financing Administration, Office of Research and Demonstration (unpublished). Lubitz, J., & Prihoda, R. (1984). Use and cost of medical services in the last two years of life. Health Care Financing Review, 5(1), 117-131. Lubitz, J., & Riley, G. F. (1993). Trends in Medicare payments in the last year of life. New England Journal o f Medicine, 328(15), 1092-1096. Lubitz, J., Beebe, J., & Baker, C. (1995). Longevity and Medicare expenditures. New England Journal o f Medicine, 332(15), 999-1003. Luft, H. S. (1982). Health maintenance organizations and rationing of medical care. Milbank Memorial Fund Quarterly, 60( 1), 268-306. McCall, N. (1984). Utilization and costs of Medicare services by beneficiaries in their last year of life. Medical Care, 22, 329-342. McCombs, J. S., Kasper, J. D., & Riley, G. F. (1990). Do HMO's reduce health care costs? A multivariate analysis of two Medicare HMO demonstration projects. Health Service Research, 25(4), 593-613. McFarland, B. H., Freeborn, D. K., Mullooly, J. P., & Pope, C. R. (1985). Utilization patterns among long-term enrollees in a prepaid group practice health maintenance organization. Medical Care, 25(11), 1221-1233. McMillan, R. M., Lubitz, J., McBean, A. M., & Russell, D. (1990). Trends and patterns in place of death for Medicare enrollees. Health Care Financing Review, 12( 1), 1-7. Mechanic, D. (1985). Cost containment and the quality of medical care: Rationing strategies in an era of constrained resources. Milbank Memorial Fund Quarterly/Health and Society, 63,453-475. Mechanic, D. (1979). Future issues in health care, social policy and the rationing o f medical services. Free Press Menzel, P. T. (1990). Strong medicine: The ethical rationing o f health care. Oxford University Press. Merrill, J. C., & Cohen, A. B. (1986). The emperor's new clothes: Unraveling the myths about rationing. New England Journal o f Medicine, 314(21), 1383-1385. Mushkin, S. J. (1974). Terminal illness and incentives for health care. In S. J. Mushkin (ed.), Consumer incentives for health care (pp. 183-216). New York: N. Watson. 164 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Mutran, E., & Ferraro, K. (1988). Medical need and the use of services among older men and women. Journal o f Gerontology, 43(5), SI 62-171. Office of Health Economics. (1979). Scarce resources in health care. Milbank Memorial Fund Quarterly, 57(2), 265-87. Olson, (1989). Medical implications of setting limits: Using age as a criterion. St. Louis University Law Journal, 33(3), 603-606. Pawlson, L. G., Glover, J. J., & Murphy, D. J. (1992). An overview of allocation and rationing: Implications for geriatrics. Journal o f the American Gerontology Society, 40(6), 628-634. Peris, T. T., & Wood, E. R. (1996). Acute care costs of the oldest old. Archives o f Internal Medicine, 156, 754-760. Pira, P. A., & Lutins, T. (1973). Utilization and reimbursement under Medicare for persons who died in 1967 and 1968. Health Insurance Statistics, USDHEW, Social Security Administration, DHEW Publ. No. (SSA) 74-11702. Polliack, M. F., & Shavitt, N. (1977). Utilization of hospital in-patient services by the elderly. Journal o f the American Gerontology Society, 25, 364-368. Reiman, A. S. (1990). The trouble with rationing. New England Journal o f Medicine, 323(13), 911-913. Retchin, S., Clement, D. G., Rositer, L. F., Brown, G., Brown, R., & Nelson, L. (1992). How the elderly fare in HMOs: Outcomes from the Medicare competition demonstrations. Health Service Research, 27(5), 6561-6569. Riley, G. F., & Lubitz, J. (1989). Longitudinal pattern of Medicare use by cause of death. Health Care Financing Review, 11(2), 1-12. Riley, G. F., Lubitz, J., Prihoda, R., & Stevenson, M. A. (1986). Changes in the distribution of Medicare expenditures among aged enrollees, 1969-1982. Health Care Financing Review, 7(3), 53-63. Roehrig, C. B. (1990). Rationing: Not "if," but "how" (Editorial). The Internist, 5,1. Roos, N. P., Shapiro, E., & Tate, R. (1989). Does a small minority of the elderly account for a majority of health care expenditures? A sixteen year perspective. Milbank Memorial Fund Quarterly, 67(3), 347-369. Roos, N. P., Montgomery, P., & Roos, L. L. (1987). Health care utilization in the years prior to death. Milbank Memorial Fund Quarterly, 65(2), 231-253. Roos, N. P., Shapiro, E., & Roos, L. L. (1986). Aging and the demand for health services: Which aged and whose demand? The Gerontologist, 24, 31-36. 165 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Roos, N. R., & Shapiro, E. The Manitoba Longitudinal Study on Aging: Preliminary findings on health care utilization by the elderly. Medical Care, 19(6), 644-657. Rosenfeld, A. S. (1996). Managed care for the elderly: What's the problem? Where's the problem? The Public Policy and Aging Report, 7(2), 1-3). Rosenthal, G. E., & Landefeld, E. S. (1993). Do older Medicare patients cost hospitals more? Archives o f Internal Medicine, 153, 89-96. Schick, F. C., & Schick, R. (eds.). (1994). Statistical handbook on aging Americans (p. 133). Oryx Press. Schneider, E. L., & Guralnik, J. M. (1990). The aging of America: Impact on health care costs. Journal of the American Medical Association, 263( 17), 1335-1340. Schneider, E. L. (1989. Options to control the rising health care costs of older Americans. Journal of the American Medical Association, 62, 591-608. Schneider, E. L., & Brody, J. (1983). Aging, natural death and the compression of morbidity: Another view. New England Journal of Medicine, 263(\1), 1335- 1340. Schroeder, S. A., Showstack, J. A., & Schwartz, J. (1981). Survival of adult high cost patients. Journal of the American Medical Association, 245, 1446-1449. Schroeder, S. A., Showstack, J. A., & Roberts, H. E. (1979). Frequency and clinical description of high cost patients in 17 acute care hospitals. New England Journal of Medicine, 300, 1306-1309. Schwartz, W. B.. & Aaron, H. J. (1985). Health care costs: The social tradeoffs. Issues in Science and Technology, 1(2), 39-46. Schwartz, W. B., & Joskow, P. L. (1978). Medical efficacy vs. economic efficiency: A conflict in values. New England Journal o f Medicine, 299, 1462-1464. Schwartz, W. B., & Mendelson, D. N. (1992). Why managed care cannot contain hospital costs— without rationing. Health Affairs, Summer, 100-107. Scitovsky, A. A. (1984). The high cost of dying: What do the data show? Milbank Memorial Fund Quarterly/Health and Society, 62(A), 591-608. Scitovsky, A. A. (1988). Medical care in the last twelve months of life: The relation between age, functional status and medical care expenditures. The Milbank Memorial Fund Quarterly, 66(A), 640-666. Scitovsky, A. A. (1994). The high cost of dying, revisited. The Milbank Memorial Fund Quarterly, 72(A), 561-591. Scitovsky, A. A., & Capron, A. M. (1986). Medical care at the end of life: The interaction of economics and ethics. Annual Review o f Public Health, 7, 59-75. 166 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Scotto, J., & Chiazze, L. (1974). Third National Cancer Survey: Hospitalization and Payment to Hospitals. Part A.: Summary. DHEW Publ. No. NIH-76-1094. Bethesda, MD: National Institutes of Health. Siegler, M. (1984). Should age be a criterion in health care? Hastings Center Report, 14,24-25. Siu, A., Brook, R., & Rubenstein, L. (1986). Medicare capitation and quality of care for the frail elderly. Health Care Financing Review Annual Supplement, 57-63. Smeeding, T. M. (ed.). (1987). Should medical care be rationed by age? Totowa, NJ: Rowan and Littlefield. Starfield, B., van den Berg, B. J., Steinwachs, P. M., Katz, H. P., & Horn, S. D. (1979). Variations in utilization of health services by children. Pediatrics, 63(4), 633-641. Strauss, M. J., LoGerfor, J. P., Yaltatzie, J. A., Temkin, N., & Hudson, L. D. (1986). Rationing of intensive care unit services: An everyday occurrence. Journal of the American Medical Association, 255, 1143. Sulmasy, D. P. (1995). Managed care and managed death. Archives o f Internal Medicine, 155, 133-136. SUPPORT Principal Investigators. (1995). A controlled trial to improve care for seriously ill hospitalized patients: The study to understand prognoses and preferences for outcomes and risks of treatments (SUPPORT). Journal o f the American Medical Association, 274(20), 1591-1598. Sutton, J. P., & Aliberti, E. M. (1994). Characteristics of high-cost older adults who are newly enrolled in an HMO. Medical Interface, July, 65-71. Sutton, G. F. (1965). Hospitalization in the last year of life, United States. Vital and Health Statistics, 22( 1). Dept. HEW Pub. No. NIH-76-1094. Bethesda, MD: National Institutes of Health. Temkin-Greener, H., Meiners, M. R., Peltz, E. A., & Szydolowski, J. S. (1992). The use and cost of health services prior to death: A comparison of the Medicare only and the Medicare-Medicaid elderly population. The Milbank Memorial Fund Quarterly, 70(4), 679-701. Thibault, G. E., Mulley, A. G., Barnett, G. O., Goldstein, R. L., Reder, B. A., Sherman, E. L., & Skinner, E. R. (1980). Medical intensive care: Indications, interventions and outcomes. New England Journal o f Medicine, 302, 938-942. Thomas, C., & Kelman, H. R. (1990). Health service use among the elderly under alternative health service delivery systems. Journal o f Community Health, 15(2), 77-92. 167 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Timmer, E. J., & Kovar, M. G. (1971). Expenses for hospital and institutional care during the last year of life for adults who died in 1964 or 1965. Vital and Health Statistics, Series 22, No. 11. Hyattsville, MD: USDHEW, Public Health Service. Turnbull, A. D., Carlong, G., Baron, R., Sichel, W., & Young, C. (1979). The inverse relationship between cost and survival in the critically ill cancer patient. Critical Care Medicine, 7, 20-23. United States Congress. Office of Technology Assessment. (1987). Life sustaining technologies and the elderly. Washington, DC: U.S. Government Printing Office. United States Congress. (1984). Technology and the aging: Rationalizing or rationing? Washington, DC: U.S. Government Printing Office. United States Department of Commerce, Bureau of the Census. (1994). 1990 census o f population and housing. CP90-AOA-US. United States House of Representatives. Select Committee on Aging. Health Care Costs for America's Elderly, 1977-1988. Washington, DC: U.S. Government Printing Office. Veatch, R. M. (1988). Justice and the economics of terminal illness. Hastings Center Report, 18, 34-40. Waldo, D„ & 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, < 5(1), 1-30. Waldo, D. R., Sonnefelt, S. T., McKusick, D. R., & Arnett, R. H. (1987). Health expenditures by age group 1977 and 1987. Health Care Financing Review, 70(4), 111-120. Wenneberg, J., & Gittlesohn, A. (1973). Small area variations in health care delivery. Science, 182, 1102-1107. Weiner, J. M. (1990). Rationing America: Overt and covert. Presented at "Rationing America's Health Care," The Brookings Institute. Weinstein, M. D. (1995). From cost-effectiveness ratios to resource allocation: Where to draw the line? In F. A. Sloan (ed.), Valuing health care. Costs, benefits and effectiveness o f pharmaceuticals and other medical technologies (pp. 77-97). London: Cambridge University Press. Wetle, C., & Levkoff, S. (1988. Geriatric medical decisions: Factors influencing allocation of scarce resources and the decision to withhold treatment. The Gerontologist, 28, 336-343. 168 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Wolinsky, F. D., & Johnson, R. J. (1991). The use of health services by older adults. Journal o f Gerontology, 46, S304-S312. Wolinsky, F. D., Coe, R. N., & Mosely, R. (1986). A cohort analysis of the use of health services by elderly Americans. Journal o f Health and Social Behavior, 27, 209-219. Wolinsky, F. D., Coe, R. N., & Mosely, R. (1987). The use of health services by elderly Americans: Implications from a regression-based cohort analysis. In Ward & Tobin (eds.), Health in aging; Sociological issues and policy predictions. Springer Publications. Wolinsky, F. D., Culler, S. D., Callahan, C. M., & Johnson, R. J. (1994). Hospital resource consumption among older adults: A prospective analysis of episodes, length of stay, and charges over a seven year period. Journal o f Gerontology, 49, S240-S252. Wunderlich, G. S., & Sutton, G. F. (1966). Episodes and duration of hospitalization in the last year of life, United States, 1961. Vital and Health Statistics, Series 22, No. 2. Public Health Service Publ. No. 1000. Public Health Service, Washington, DC: U.S. Government Printing Office. Wyszewianski, L. (1986). Financially catastrophic and high cost cases: Definitions, distinctions and their implications for policy formulation. Inquiry, 23, 382-394. Zook, D., & Moore, F. (1980). High cost users of medical care. New England Journal o f Medicine, 502(18), 996-1002. Zweibel, N. R., Cassell, C. K., & Karrison, T. (1993). Public attitudes about the use of chronological age as a criterion for allocating health care resources. The Gerontologist, 539(1), 74-80. 169 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. APPENDIX A Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. When the original sampling dates (October 1, 1991 to May 1 1992) were utilized to reconstruct the survey population, only 58% of those who completed the survey could be correctly identified as having been eligible for the study. In order to construct an accurate accounting of non-respondents, additional attempts to recreate the original sample population by identifying the known respondent population were made by support personnel at the HMO utilizing a variety of eligibility or enrollment data parameters. By employing eligibility dates of October 1, 1991 through May 1, 1992 (8 months), 91.3% of the respondent population was correctly captured. This created a sample population of 33,264 enrollees by eligibility dates. However, 1,079 respondents were not captured by eligibility dates. When these are added, a total sample population of 34,343 is created. Eliminating the sample population members under age 65 (1,873) produces a total sample population of 32,470 new enrollees age 65 and older (see Figure 2.1). Sample members who did not remain enrolled in the health plan for one year and did not die were eliminated from the sample population. Thus the total sample population for this study includes 28,536 new enrollees over age 65 who remained in the health plan for one year from the date of enrollment or who died during their first year of enrollment in the plan. 171 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure A. 1. Construction of Sample Population < 65 > 65 M ailed qu estionnaire 1,108 20,050 21,158 o u tsid e dates 1,079 w ith in dates 11,290 R 12,369 R 499 N R 609 N R D 1,373 <65 1,873 N R 8,789 N R 8,180 >65 ? 13,185 N R D by 21,974 dates R+N RE 34,343 R 11,870 N R D 20,600 >65 32,470 R = R espondents (returned questionnaire E R = E lderly Respondents (> 65) N R = N on-respondents (did not return questionnaire) N R D = N on-respondent enrollees selected by enrollm ent dates 172 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. APPENDIX B Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Health Inventory Name:_________________________________ Medicare # :____ Phone:_________________________________ Doctor's Name:__ Address:_______________________________ Medical Group:__ City:_________________State_____ Zip_______ Age:________ Sex: This section asks for information about your medical history and the medications that you take. This information will help us evaluate your need for any of the special programs that we offer. 1. In general, how would you rate your health? (Circle one number.) Excellent ........................................................ 1 G o o d .............................................................. 2 Fair ................................................................ 3 Poo ................................................................ 4 2. Please ✓ the □ of all those conditions for which you are currently receiving medical treatment. □ Arthritis □ Urinary Problems □ Ankle/Leg Swelling □ Heart Problems □ Bowel Problems □ Breathing Problems □ Digestive Problems □ Diabetes/High Blood Sugar □ High Blood Pressure □ Cancer-Where? □ Memory Problems □ Other___________ □ Mental Problems ____________________ Would you like to receive educational material on those conditions you checked? □ Yes □ No 3. During the past 9 weeks, how much did pain interfere with your normal activities (including both outside the home and housework)? (Circle one number) Not at a ll.................................................... 1 A little b it.................................................... 2 Moderately................................................ 3 Quite a b it.................................................. 4 Extremely.................................................. 5 4. Where is the pain located? (Circle all numbers that apply.) Head.............................. 1 Knees................................... 6 Heart ............................ 2 Back..................................... 7 Neck.............................. 3 F e e t................................... 8 Arm s.............................. 4 Stomach............................... 9 Le g s.............................. 5 O th e r............................... 10 174 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This section asks about your medication history. Please list both prescribed and over- the-counter medications you routinely take. If you need more room, please use a separate page. Medication name Reason for medication How long on this medication? Any problems taking this medication? This section asks about your social support system and your home environment. This will help us determine what kind of additional support you may require, if the need arises. (Circle one number.) 6. I live: Alone........................................................................ 1 With Spouse.......................................................... 2 With family, who?................................. 3 Other, explain....................................... 4 7. I live in: House...................................................................... 1 Apartment.............................................................. 2 Mobile hom e.......................................................... 3 Board & care.......................................................... 4 Nursing facility........................................................ 5 Other, explain......................................... 6 8 . My residence is on the: Ground flo o r.......................................................... 1 Upper floor ............................................................ 2 9. In your home, please indicate with ✓ all □ that are available to you. □ Stairs □ Wheelchair ramps □ Elevators □ one of the above 10. What, if any, of the following equipment are you currently using on a daily or weekly basis? (Circle all number that apply.) Oxygen ........................................................................ 1 Breathing machine........................................................ 2 Hospital Bed ................................................................ 3 Wheelchair.................................................................... 4 Other, explain................................................ 5 11. Do you currently use Meals on Wheels? QYes QNo 175 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The section will inform us whether or not your health limits you in performing activities of daily living and to what degree you may or may not require added assistance. Circle 1,2, or 3 on each line. 12. Need Some Need Total Independent Assistance Assistance ........................ 2 3 ........................ 2 3 ........................ 2 3 ........................ 2 3 ........................ 2 3 ........................ 2 3 ........................ 2 3 ........................ 2 3 ........................ 2 3 ........................ 2 3 1 1 Bathing .................... Dressing .................. Eating ...................... Toileting.................... Mobility .................... Taking Medications .. Meal Preparation___ Housekeeping Chores Shopping & Errands . Transportation.......... Money Management . 13. If you require support with the above mentioned activities, who primarily assists you? (Circle one number.) If you do not require support, please skip to the next questions. Spouse 1 Hired Attendant.......................... 4 Other relative — 2 Other, explain_____________ 5 Neighbor/friend .. 3 14. In the past 12 months how many times have you seen a doctor? times 15. In the past 12 months how many times haveyou been hospitalized? times 16. In the past 12 months how many times wereyou in the emergency room? times 17. In the past 12 months how many times haveyou required home health nursing? times This section will explore your current individual health habits. 18. How many meals do you each day? /day 19. How many ounces of alcohol do you consume each day?________/day Each w e e k ? _________/week 20. Are you a current smoker? □ Yes □ No 21. Are you a past smoker? □ Yes □ No 22. How many years have you or did you smoke? years 23. Do you now or have you in the past lived with a smoker? □ Yes □ No 24. Is your cholesterol □ Normal □ High □ Don't know 25. Is your blood pressure □ Normal □ High □ Don't know 176 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Please complete each section fully by selecting a number on a scale of 0 to 10 (0 = never, 5 = occasionally, 10 = often) in each line. a. Stress b. Exercise How many times each week do you feel: How many times each week do you: Tense Depressed Jog or Run Swim Angry Overwhelmed Walk Bicycle Sad Lonely Aerobics None □ Yes □ No □ Yes □ No □ Yes □ No □ No □ Don't 26. Do you have a VCR? (For health education materials) 27. Do you have a (living will) Durable Power of Attorney? 28. Do you have long-term care insurance? 29. Are you currently receiving MediCal Stickers? □ Yes (State funded medical coverage) This section wili inform us of the health screenings that you might require. 30. Have you had any of the following immunizations in the past year? Tetanus shot □ Yes □ No □ Don't know Flu vaccine □ Yes □ No □ Don't know Pneumococcal pneumonia vaccine QYes QNo □ Don't know 31. Have you ever had a rectal exam? □ Yes □ No □ Don't know (Circle the number to the right that applies for questions 32 through 34. 32. How long has it been since your last physical exam? Less than 1 year 1 Within 1-3 years 2 Greater than 3 years 3 33. How long has it been since you received an eye examination? Less than 1 year 1 Within 1-3 years 2 Greater than 3 years 3 34. How long has it been since you received a hearing examination? Less than 1 year 1 Within 1-3 years 2 Greater than 3 years 3 177 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. FOR WOMEN ONLY (Circle the number that best describes your answer.) Years 35. When was your last breast exam by a medical provider? 1 2 3+ 36. When was your last mammogram? 1 2 3+ 37. When was your last Pap smear? 1 2 3+ THANK YOU Did you complete this questionnaire?__________ Did someone complete it for you?_____________ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. APPENDIX C 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. Appendix Table C. 1. Logistic Regression Odds Ratio of Having Part A Costs During the Last 3 Months of Life for Decedents Compared to During the 12-month Study for Survivors Model Age Gender Age2 Died Total Sam ple (N - 28,536) I. 1.047** II. 1.048** 1.184** 1 1 1 . 1.056** 1.183** 1.000 IV. 1.056** 1.174** 1.000 1.321** V . 1.053** 1.173** 1.000 2.202** VI. 1.056** 1.181** 1.000 2.201** Age x Dead County-2 County-3 County-4 County-5 0.956** 0.956** 1.336** 1.143" 0.909+ 0.848** b. R espondent Population (N = 5,923) 1. 1.057** 2. 1.058** 1.176* 3. 1.095** 1.165* 0.998* 4. 1.096** 1.154+ 0.998* 1.524** 5. 1.094** 1.153+ 0.998* 1.810** 0.985 6. 1.096** 1.161+ 0.998* 1.791* 0.985 1.176 1.099 0.912 0.848 +p, *p > .01, **p > .001 o o o Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission. Appendix Table C.2. Logistic Regression Odds Ratio of Having Hospital Costs for the Last 3 Months of Life for Decedents Compared to the 12-month Study Period for Survivors Age x Dead County-2 County-3 County-4 County-5 Mode 1 Age Gender Age2 Died Age x Dead a. Total Population (N = 28,536) I. 1.042** II. 1.043** 1.184** III. 1.048** 1.183** 1.000 IV. 1.048** 1.169** 1.000 1.531** V. 1.045** 1.167** 1.000 2.525** 0.957** VI. 1.047** 1.175** 1.000 2,526** 0.957** b. Respondent Population (N = 5,923) I. 1.050** 1 1 . 1.051** 1.215** III. 1.081** 1.205* 0.999 IV. 1.083** 1.176+ 0.998+ 2.417** V. 1.082** 1.175+ 0.999+ 2.884** 0.985 VI. 1.083** 1.173+ 0.999+ 2.909** 0.984 +p > .05, *P>.01, **p>.001 1.265** 0.919+ 0.831" 0.875 0.989 1.025 0.892 o o Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Appendix Table C.3. Logistic Regression Odds Ratio of Having Ltc Costs for the Last 3 Months of Life for Decedents Compared to the 12-month Study for Survivors Mode 1 Age Gender Age3 Died Total Population (N = 28,536) I. 1.093** II. 1.093** 0.955 1 1 1 . 1.090** 0.955 1.000 IV. 1.090** 0.935 1.000 1.781** V. 1.087** 0.932 1.000 3.817** VI. 1.090** 0.942 1.000 3.809** Respondent Population (N = 5,923) I. 1.101** II. 1.099** 0.674* III. 1.091** 0.675* 1.000 IV. 1.092** 0.654** 1.000 2.341** V . 1.089** 0.651** 1.000 4.862** VI. 1.093** 0.656** 1.000 4.744** Age x Dead County-2 County-3 County-4 County-5 0.946** 0.945** 1.384* 1.534* 1.129 0.830+ 0.949* 0.949+ 1.206 1.615+ 1.079 0.832 +p > .05, * p > .0 1 ,* * p > .0 0 1 o o to Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Appendix Table C.4. Logistic Regression Odds Ratio of Having Part a Costs During the Last Month o f Life for Decedents Compared to During the 12-month Study for Survivors Mode 1 Age Gender Age! Died Age x Dead County-2 County-3 County-4 County-5 i. Total Sample (N = 28,536) 1 . 1.044** II. 1.045** 1.163** III. 1.058** 1.161** 0.999 IV . 1.058** 1.170** 0.999+ 0.727** V . 1.056** 1.168** 1.000 1.279 0.953** VI. 1.058** 1.177** 1.000 1.275+ 0.953** 1.334** 1.155+ 0.908* 0.839** ). Respondent Population (N = 5,923) 1. 1.053** 2. 1.053** 1.141 + 3. 1.098** 1.128 0.998** 4. 1.098** 1.133+ 0.998* 0.828 5. 1.095** 1.131 + 0.998* 1.202 0.969 6. 1.097** 1.140+ 0.998* 1.187 0.969 1.174 1.122 0.907 0.819 +p > .05, *p> .01, **p > .001 00 Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Appendix Table C.5. Logistic Regression Odds Ratio of Having Hospital Costs for the Last Month of Life for Decedents Compared to the 12-month Study Period for Survivors Mode 1 Age Gender Age2 Died Age x Dead County-2 County-3 County-4 a. Total Population (N = 28,536) I. 1.039** II. 1.040** 1.168** 1 1 1 . 1.051** 1.166** 0.999 IV. 1.051** 1.169** 1.000 0.888 V. 1.048** 1.168** 1.000 1.533** 0.954** VI. 1.051** 1.176** 1.000 1.532** 0.954** 1.267** 0.920* 0.824** b. Respondent Population (N = 5,923) I. 1.045** II. 1.046** 1.197+ II. 1.090** 1.183+ 0.998* I. 1.091** 1.170+ 0.998* 1.170+ V. 1.088** 1.167+ 0.998* 2.396** 0.962+ VI. 1.090** 1.166+ 0.998* 2.415** 0.962+ 0.882 1.009 1.031 +p > .05, * p > .0 l,* * p > .0 0 1 0.869 00 4 ^ Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Appendix Table C.6. Logistic Regression Odds Ratio of Having Ltc Costs for the Last Month of Life for Decedents Compared to the 12-month Study for Survivors Mode 1 Age Gender Age3 Died a. Total Population (N = 28,536) I. 1.095** 1 1 . 1.095** 0.971 III. 1.104** 0.970 1.000 IV. 1.103** 0.989 1.000 0.588** V. 1.099** 0.988 1.000 1.329 VI. 1.103** 0.998 1.000 1.315 b. Respondents Population (N = 5,923) I. 1.098** 1 1 . 1.095** 0.633** 1 1 1 . 1.092** 0.633** 1.000 IV. 1.091** 0.635** 1.000 0.909 V. 1.087** 0.633** 1.000 1.867 VI. 1.092** 0.637** 1.000 1.807 +p > .05, *p > .01, **p >.001 Age X Dead County-2 County-3 County-4 County-5 0.946** 0.946** 1.416** 1.569** 1.079 0.782* 0.954 0.955 1.244 1.810* 1.157 0.804 o o L /l APPENDIX D 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. Appendix Table D. 1. OLS Regression Simple Models on Total Costs for Claimants During the Last 3 Months for Decedents and Average 3 Months for Survivors M odel Variable 1 II III IV V VI a. T o tal P o p u la tio n (N = 7,907) R2 .0040** .0048** .0057** 0.2047** 0.2071** .2092** Intercept 1864.24** 1685.84** 1851.95** 1454.24 1398.63** 1398.54** A ge 52.74** 54.73** -1.67 -0.06 -4.71 2.26 G ender 354.88* 365.96* 122.72 118.89 132.17 A ge2 2.53 1.04 1.34 1.02 Died 9901.42** 11422.** 11466** A ge x Dead -136.52** -138.15** C ounty-2 211.72 C ounty-3 -250.00 C ounty-4 220.00 C ounty-5 -566.55* 00 '-J > . R espondent Population ( N = 1,426) * * in r-* CN * * m VO * CN CM rr in 0 0 CN in 0 v vO CN O s O m 0 0 VO in © CN CM vO in Ov i i cm C M CM t m CM CM CN i * * CN in N * CN * * c n r- m V O * * * in in CN 0 0 in C M r n © 0 0 in v d n * CN CN V O CM rr i CN i CM i m * * vO * * c n vO © vO m CN — * © 0 0 O s Ov © © t m m i n r r » © © * in * r ^ 0 0 © 0 0 0 0 © © CM r r o o i n CM © © i r f CM + 0 0 * * + © O s CM w mm © © i n © © c n 0 0 t T © O s CM t * w -m * m * + © 0 0 m m rn © © 0 0 c n © m © o o CL 0 ) o 0 3 H cc a u C O < " O e a C O < T 3 Q J T 3 O Q x o 0 0 < CN m & c 3 O C J c 3 O U i * G 3 O a * e 3 O U a. * * trf © A a. + 188 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. Appendix Table D.2. OLS Regression Simple Models on Hospital Cost for Claimants During the Last 3 Months for Decedents and Average 3 Months for Survivors Model Variable 11 1 1 1 IV V I a. Total Population (N = 6,744) R2 Intercept Age Gender Age2 Died Age x Dead County-2 County-3 County-4 County-5 .0025** 2001.35** 39.70** .0037** 1797.48** 42.17** 400.37* .0040** 1906.53** 3.53 409.45* 1.76 .2047** 1467.51** 7.39 187.05 -0.36 9321.77** .2082** 1398.34** 1.34 181.46 0.67 11036.** -156.71** .2 10 1 ** 1415.54** 9.08 198.32 0.31 11078.** -158.09** 249.86 -304.13 165.60 -519.18* o o so CN O O ' CN * * © ^ r r - 0 0 * * * m 0 0 0 0 a s © m r*^ i n * VO m © © £ v CN ©* vO © c n Ov 0 0 CN CN m r ^ 0 0 CN r - • — 2 i © SO CN * — N* CN m i c n 1 i * * VO * * 0 0 * — • r>- vO m * m CN CN •— i n m VO OO r n o s CN o o © i VO m CN CN i m 1 * * m * m * r-* m c n * — m 0 0 * CN 00 r - OO © * © S O * CN ^ r i O s 1 m s o m CN 1 © Ov i n © * * © i n r - Ov © CN CN v q OO CN © m CN i VO m i c o U C N ' V 9 s Jj .2 jo Js e 2 > O OO r * - I I Z s o G m O C u a *9 tn tS CN * i n * © m 0 0 © r - SO oo* c n rf- © N* © 0 0 © CN i I s * * CN * O — — © r- © * r^ * ■ ^ r CN rr o' J5 CL V f c > V w c < u 0 0 < a > -a c CD o o c o < -a a > •o e o O Q X 4 > 00 < CN i * £ £ O U 3 O U * c 3 O CJ m i £> c 3 o U o o A Q. * * irT o A a. + 190 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. Appendix Table D.3. OLS Regression Simple Models on Long-term Care Costs for Claimants During the Last 3 Months for Decedents and Average 3 Months for Survivors M odel Variable 1 1 1 1 1 1 IV V VI a. T o tal P o p u la tio n (N = 1,808) R2 .0117** .0120** .0126** .1249** .1266** .1314** Intercept 610.88** 578.96** 647.32** 517.12** 544.22** 581.97** Age 25.86** 26.30** 8.50 7.09 8.03 9.27 G ender 67.09 70.25 -28.81 -33.16 -22.32 A ge2 0.70 0.43 0.24 0.17 D ied 1928.93 1545.75 1563.14** A ge x Dead 29.21 28.31 County-2 7.70 C ounty-3 -97.25 C ounty-4 53.74 County-5 -365.17* v O r-* CN c n * * r-* Os i n c n r-* s o c n 0 0 — o Os CN CN © CN v d © <> Os c n v d o 0 0 1 N - o o ON 0 0 i a 1 CN i CN CN 1 t * * + * c n CN CN 0 0 CN CN * - • ^*1 O s i n c n c n © c n a s r-* 0 0 i m t T i CN i i n + * © r-* © c n CN © o o i n c n c n © ON CN © O n ' CN — © i SO i n CN i n c n o p CN VA CN vO CN c n O n 00 T f O O r - * c n * © O ' © CN © s o © CN ^ r — r - s o i c o O m • ju D J3 o > .2 •S -5 H > O n * O n * n o n o i i q - v o * s = w so ■ # * e o s 2 & a. 4 ) a & 5 i> .o * 2 < a > 7 3 C O o < u 00 7 3 1 ) < Q ■ a cd 4) a CN c n * ?s c c o 3 3 0 0 O O < C J U * c 3 O U * a 3 o V o o Q . * * * n p A a . + 192 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. Appendix Table D.4. OLS Regression Simple Models on Total Costs for Claimants During the Last Month for Decedents and Average Month for Survivors M odel Variable 1 1 1 1 1 1 IV V VI a. Total Population (N = 7,711) R2 .0024** .0030** .0030** .2899** .2946** .2961** Intercept 723.40** 650.94** 656.66** 453.15** 415.36** 401.56** Age 21.24** 22.05** 20.09 15.49 13.85 17.89 G ender 144.98+ 145.37+ 44.76 44.10 51.04 A ge2 0.09 -0.69 -0.25 -0.43 Died 6425.30** 8791.71** 8807.39** Age x Dead -122.85** -123.85** C ounty-2 80.93 County-3 1.51 C ounty-4 116.45 C ounty-5 -253.13* v O u > i . R espondent Population ( N = 300) * * © c n m c n m OO O ' * «- * O s i n r - CN ON c n c n * t — i n Os vO O O - J i n CN c n vO * — • —— i CN o o 0 0 m 0 0 CN VO CN i c n “ * ■ * c n * + * m 0 0 vO CN O ' * oo c n VO © CN CN * p i n CN © i n VO — 1 m oo c n VO CN i * * SO m * CN i n CN VO c n O m i n — - CN CN - J c n SO —• • — < m i c n oo O ' O * * © i n c n © © oo c n © v d © CN i © © * * © © © c n v q vO © c n c n r - © * © * © c n c n © v q © c n c n CL u c* 3 l. < D T 3 N C < D 00 O oo < a < • T 3 O -o a a > Q x C O < C M i & c 3 O CJ * s 3 O CJ * e 3 O CJ in i c 3 o O o © A o. * * < / " T p A o. + 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. Appendix Table D.5. OLS Regression Simple Models on Hospital Cost for Claimants During the Last Month for Decedents and Average Months for Survivors M odel Variable I II (II IV V VI a. T o tal P o p u la tio n (N = 6,569) R2 .0016* .0022* .0022* .2772* .2822* .2837 * * * * * * Intercept 755.95** 698.71** 679.13** 467.55** 423.99** 407.33** Age 17.51** 18.44** 25.47 17.38 16.63 21.28 G ender 152.63+ 151.00+ 49.53 48.01 57.91 A ge2 -0.72 -0.95 -0.53 -0.74 Died A ge x Dead County-2 County-3 County-4 County-5 6920.72** 8249.68** -121.87** 8260.01** -122.50** 136.99 10.63 96.88 -249.74+ TT r-* CN CN m * * * O r^ r - n c n in 0 0 o o cn * "'T CN CN o © * * in o O Ov v d cn Ov c n i CN vO N " o r * * 0 0 CN 1 Ov CN r * —• c n wn c n m OV CN in * c n c^* Ov vO CN * CN in S in 0 0 c n VO i © vO i o CN i in * CN * * c n in Ov o o c n Ov Ov r r OO in * n Cv o v c n in —• in 1 • — vO i o o * o * o c n Ov in OV o 0 0 Ov o o © © in VO cn VO t o o o VO in r*- * * O v O O O v c n o — o — T^- c o CJ Q 3 g j .2 3 }a t= > e •2 § 3 § 3 f i . O ft- c -3 fi O f i * 0 5 V O S * * o CN 00 Q. C J o 0 0 < 0 > T 3 c CJ O o 0 0 < “ O CJ - o C Q CJ Q CJ 00 < CN i & c 3 O CJ c 3 O CJ * 3 3 O CJ 3 O CJ o © A f i . * * in © A f i . + 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. Appendix Table D.6. OLS Simple Models on Term Costs for Claimants During the Last Month for Decedents and Average Month for Survivors Model Variable 1 1 1 1 1 1 IV V V I R2 .0040* .0051* .0051+ .2234** .2275** .2337 Intercept 259.50** 227.20** 239.15** 182.94** 168.59** 225.50** Age 8.12* 8.59* 5.43 3.94 2.67 3.68 Gender 68.49 69.06 14.78 18.83 23.53 Age2 0.12 -0.04 0.10 0.05 Died 1878.09** 2320.73** 23567.00** Age x Dead -32.03** -34.12* County-2 -112.37 County-3 -120.88 County-4 4.18 County-5 -29.47 Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission. Table D-6, Cont. Variable I II III b. R e sp o n d e n t P o p u la tio n (N = 278) R 2 .0014 .0017 .0024 Intercept 251.93+ 233.72 279.45 Age 5.23 5.35 -5.22 j 1 G ender 46.30 53.43 Age2 0.37 Died Age x Dead County-2 County-3 C ounty-4 County-5 +p > .05, **p > .001 IV V VI .1642** .1765** .1840** 230.74 194.86 283.47 -6.37 -7.39 -9.09 -16.90 4.55 -8.40 -0.17 0.38 0.40 1862.01** 2736.26** 2736.26** -57.37 -59.46 -148.93 -220.64 38.34 -111.96

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

An analysis of postacute treatment and outcome differences between Medicare fee-for-service and managed care

PDF

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

PDF

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

PDF

Statistical analysis of health care expenditure and health status in the well-elderly study

PDF

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

PDF

The effect of rate regulation and price competition on labor efficiency in ancillary and nursing service areas of hospitals in California, 1983-1991

PDF

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

PDF

Patient life in a nursing home: an experiential study

PDF

State legislation concerning individuals with dementia: An evaluation of three theoretical models of policy formation

PDF

Futile medical treatment: an analysis of physician decision-making

PDF

Causal estimation of the effect of medication compliance on health outcomes

PDF

Evaluating differences in patient satisfaction with inpatient hospital care categorized by diagnostic related groups

PDF

The use of transaction cost economics, fuzzy set and systems theories in the development of a three-dimensional pricing model of asset specificity and contracts, and applied to health care

PDF

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

PDF

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

PDF

Religious-based managed care: A Wesleyan paradigm for reforming health care

PDF

Mental models and attachment systems: Predictions of health and relational outcomes of the elderly

PDF

Family adaptation to a child with attention deficit hyperactivity disorder

PDF

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

PDF

Evaluation of the Care Advocate Program: Bridging managed care and home community -based services

Asset Metadata

Creator Segal-Gidan, Freddi I. (author)

Core Title 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

Degree Doctor of Philosophy

Degree Program Gerontology and 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-c17-283546

Unique identifier UC11352871

Identifier 9733132.pdf (filename),usctheses-c17-283546 (legacy record id)

Legacy Identifier 9733132.pdf

Dmrecord 283546

Document Type Dissertation

Rights Segal-Gidan, Freddi I.

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