University of Southern California Dissertations and Theses

Uncovering hidden figures: disparities in dementia burden in an aging America

(USC Thesis Other)

Uncovering hidden figures: disparities in dementia burden in an aging America

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Uncovering Hidden Figures:
Disparities in Dementia Burden in an Aging America

by

Yi Chen

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
PUBLIC POLICY AND MANAGEMENT

May 2022

Copyright 2022 Yi Chen
ii
ACKNOWLEDGEMENTS
I am extremely grateful to my Chair, Prof. Julie Zissimopoulos, and committee members
Prof. Alice Chen and Prof. Eileen Crimmins. They provided enormous intellectual and emotional
support along this journey. I will always remember what they have taught and inspired me, as a
researcher, as a human being.
My appreciation also goes to Jillian Wallis and Patricia Ferido, who have kindly assisted
with my access to the restricted data used in this dissertation, as well as Prof. Geoffrey Joyce and
Bryan Tysinger for their insights into my research.
I would like to thank my colleagues and friends, Sushant Joshi, Zeewan Lee, Johanna
Thunell, Jianhui Xu, Bo Zhou, and Yingying Zhu, for making my life at USC a wonderful time.
I would not be able to complete my study without the love and support from my family.
Big thanks to my parents and parents-in-law for their patience, encouragement, and many things
beyond that. I am indebted to my husband, Pu, who has been by my side through good and bad
times and sacrificed a lot to let me pursue my dream. Special thanks to Ethan, for being our son
and for making us stronger. I hope he will be proud of me.
Finally, gratitude to all that enabled me to embark on this journey. It is one of the best
decisions I ever made.

iii
Table of Contents

ACKNOWLEDGEMENTS ............................................................................................................ ii
LIST OF TABLES ......................................................................................................................... iv
LIST OF FIGURES ....................................................................................................................... vi
Chapter 1 INTRODUCTION ......................................................................................................... 1
Chapter 2 Analysis of Dementia in the U.S. Population using Medicare Claims: Insights from
Linked Survey and Administrative Claims Data ............................................................................ 7
INTRODUCTION ..................................................................................................................................... 7
METHODS ................................................................................................................................................ 9
RESULTS ................................................................................................................................................ 13
DISCUSSION ......................................................................................................................................... 17
TABLES & FIGURES ............................................................................................................................ 22
APPENDIX ............................................................................................................................................. 28
Chapter 3 Socio-Economics Status and Survival after Dementia Diagnosis: Education, Wealth,
and Health Differences at Diagnosis............................................................................................. 34
INTRODUCTION ................................................................................................................................... 34
METHODS .............................................................................................................................................. 37
RESULTS ................................................................................................................................................ 40
DISCUSSION ......................................................................................................................................... 43
TABLES & FIGURES ............................................................................................................................ 47
APPENDIX ............................................................................................................................................. 51
Chapter 4 Do Individuals Skimp on Their Own Health Care after Spouse’s Dementia
Diagnosis? ..................................................................................................................................... 61
INTRODUCTION ................................................................................................................................... 61
BACKGROUND ..................................................................................................................................... 64
DATA, SAMPLE, AND MEASURES ................................................................................................... 70
EMPIRICAL STRATEGY...................................................................................................................... 73
RESULTS ................................................................................................................................................ 75
DISCUSSION ......................................................................................................................................... 78
TABLES & FIGURES ............................................................................................................................ 84
APPENDIX ............................................................................................................................................. 90
Chapter 5 CONCLUSION ............................................................................................................ 92
REFERENCES ............................................................................................................................. 95

iv
LIST OF TABLES

Chapter 2 Analysis of Dementia in the U.S. Population using Medicare Claims:
Insights from Linked Survey and Administrative Claims Data

Table Page
1. Sample Characteristics in Years 2000 and 2008………………………… 22
2. Concordance in Dementia Prevalence and by Race 2000-2008
(N=31,186)………………………………………………………………. 22
3. Concordance in Dementia Incidence and by Race 2000-2008
(N=9,623)………………………………………………………………... 24
Supp 1. Odds Ratios for Concordance in Prevalence
Relative to ‘Dementia, Both Measures (N=31,186)……………………... 28
Supp 2. Predicted Probability of Concordance in Dementia Prevalence
by Race, Education, Sex, and Year (N=31,186)………………………… 29
Supp 3. Characteristics of Sample at the Time of Incident Dementia Ascertained
by HRS (N=1,161)………………………………………………………. 30
Supp 4. Changes in Cognitive Score around Incident Dementia in HRS………... 31
Supp 5. Odds Ratios for Concordance in Incidence
Relative to ‘Less than 2-year Time Difference between Diagnosis and
Incident Dementia in HRS’ (N = 1,161)………………………………… 32
Supp 6. Predicted Probability of Concordance in Dementia Incidence
by Race, Education, Sex, and Year (N=1,161)………………………….. 33

Chapter 3 Socio-Economics Status and Survival after Dementia Diagnosis: Education,
Wealth, and Health Differences at Diagnosis

Table Page
1. Sample Characteristics………………………...………………………… 47
2A. Health Differences at Diagnosis by Education………………………….. 48
2B. Health Differences at Diagnosis by Wealth Quartile.…………………… 48
3. Length of Survival Time (Days)………………………………………… 49
4. Hazard Ratios for Mortality……………………………………………... 49
Supp 1. ICD-9 diagnostic codes for ascertaining dementia.……………………... 51
Supp 2. Time difference between date of incident diagnosis and the closet HRS
interview date……………………………………………………………. 51
Supp 3. Sample Selection………………………………………………………… 51
Supp 4. Sample Characteristics under Different Time Windows………………… 52
Supp 5. Hazard ratios for mortality, using [-12m, 6m] matching window…….… 53
v
Supp 6. Hazard ratios for mortality, restricting to incident diagnosis during wave
5-9 (2000-2008)………………………………………………………….. 55
Supp 7. Hazard ratios for mortality, no interaction of ADL and IADL, and
removing IADL………………………………………………………….. 57
Supp 8. Hazard ratios for mortality, using claim-ascertained comorbidities (only
showing results with health controls)……………………………………. 59

Chapter 4 Do Individuals Skimp on Their Own Health Care after Spouse’s Dementia
Diagnosis?

Table Page
1. Summary Statistics………………………………………………………. 84
2. Event Study Regression Results…………………………………………. 86
3. Event Study Regression Results in Stratified Samples………………….. 89
Supp 1. Identifying Spouses of PLWD…………………………………………... 90
Supp 2. CPT Codes for Identifying Health Care Procedure……………………… 90
Supp 3. Unadjusted Health Care Utilization over Time.…………………………. 91

vi
LIST OF FIGURES

Chapter 2 Analysis of Dementia in the U.S. Population using Medicare Claims:
Insights from Linked Survey and Administrative Claims Data

Figure Page
1. Predicted Probability of Concordance in Prevalent Dementia, by
Race/Ethnicity, Education, and Sex (N=31,186)………………………... 23
2. Timing of Diagnosis in Claims Data Relative to Incident Dementia
Based on Cognitive Scores in HRS Data (N=1,161)……………………. 25
3. Predicted Probability of Relative Timing of Incident Dementia, by Race,
Education, and Sex (N = 1,161)…………………………………………. 26

Chapter 4 Do Individuals Skimp on Their Own Health Care after Spouse’s Dementia
Diagnosis?

Figure Page
1. Unadjusted Health Care Utilization over Time…………………………. 85
2. Changes in Health Care Use, from Event Study Models………………... 86
1
Chapter 1
INTRODUCTION
Dementia is a family of neurodegenerative diseases that progressively impairs cognition,
function, and thus independence. In 2016, 6.49 million of Americans aged 65 years and older are
estimated to have dementia (Zissimopoulos et al 2018). Due to the aging of the Baby Boomers and
longevity gains at older ages, dementia, for which advanced age is an identified risk factor
(Alzheimer’s Association 2021; Castellani et al. 2010), is expected to affect 11.66 million older
Americans by 2040 (Zissimopoulos et al 2018). This poses unprecedented challenges to
individuals and the U.S. health care system. Ranking the 7
th
leading cause of mortality in the U.S.
(National Center for Health Statistics 2021), dementia also substantially contributes to disability,
as the 19
th
leading cause to Years Lived with Disability (US Burden of Disease Collaborators
2018). From an economic perspective, the annual health care cost associated with dementia
amounts to $40,000 to $56,000 per patient or roughly $200 billion in total (Zissimopoulos et al
2015; Hurd et al 2013), which is largely driven by use of costly health care services such as
inpatient and post-acute care (Lin et al 2016; Leibson et al 2015; Langa et al 2004). Another major
cost component is the value of informal caregiving, as most persons living with dementia (PLWD)
in the U.S. are community-dwelling and rely heavily on unpaid help from their family members
(Chi et al 2019; Kasper et al 2016). The annual informal cost is estimated to be $30,000 per case
or $126 billion for the whole society (Zissimopoulos et al 2015). To mitigate the public health
impacts of dementia, the National Alzheimer's Project Act (NAPA) effective in 2011 has led to
strategic efforts along the continuum of dementia care, including prevention, diagnosis, treatment,
care, and support for PLWD and their caregivers (U.S. Department of Health and Human Services
2021; National Institute on Aging 2020).
2
Amid the substantial burden on the aggregate level, certain groups bear a disproportionate
burden of dementia. For example, both national and regional studies show an elevated risk of
dementia among racial/ethnic minorities and females (Aranda et al 2021; Matthews et al 2019;
Mehta et al 2017; Tang et al 2001). After dementia diagnosis, non-Whites outlive Whites by 2-3
years on average (Chen et al 2022; Mayeda et al 2017; Helzner et al 2008), and hence require
greater resources to support the additional years of life. The nature of these differences can be
multifaceted, involving biological, behavioral, environmental, and health-care processes (National
Institute of Health, 2018). Growing evidence suggests disparities in dementia care, which may
drive and amplify disparities in dementia burden. Racial/ethnic minorities are reported to be more
susceptible to mis- and under-diagnosis (Gianattasio et al 2019; Livney et al 2011; Cooper et al
2010). In terms of post-diagnosis care, minorities have a lower rate of using specialty care (Drabo
et al 2019) and anti-dementia pharmacological treatment (Drabo et al 2019; Thorpe et al 2016),
and a higher likelihood of being admitted to low-quality nursing homes (Aranda et al 2021; Rivera-
Hernandez et al 2019). However, important gaps remain in the knowledge of disparities in
dementia care and burden, such as incomplete understanding about the existence, magnitude, and
mechanism of heterogeneity throughout the whole dementia care continuum. Addressing these
gaps advances understanding of dementia-related disparities and informs priorities to directing
limited resources to those who are disproportionately affected by dementia.
Specifically, findings on under-diagnosis of racial/ethnic minorities are often drawn from
studies 1) comparing cognition measured in representative surveys with dementia diagnosis in
administrative claims (Gianattasio et al 2019; Amjad et al 2018; Ostbye et al 2008; Pressley et al
2003), and 2) comparing clinically-assessed dementia diagnosis, i.e. the gold standard, with
claims-based diagnosis in non-representative samples (Taylor et al 2002). The best available study
3
balances sample representativeness and diagnostic accuracy by contrasting the gold-standard
diagnosis in the nationally representative Aging, Demographics, and Memory Study (ADAMS),
with claims-based diagnosis (Taylor et al 2009). The study reports similar level of diagnosis
completeness by race/ethnicity, which may be a result of limited sample sizes of minorities in the
ADAMS sample (n=240). More importantly, previous studies lack a longitudinal perspective, as
they only examine presence of diagnosis rather than timing of it. Considering benefits that a timely
diagnosis renders to dementia patients and families, to name some, early mitigation of safety risks,
and participation in clinical trials (White et al 2021; Dubois et al 2016), identifying disparities in
the timing of diagnosis reveals missed opportunities for reducing burden in some groups.
Survival after dementia diagnosis, indicative of the duration of dementia and time with
care, is a parameter of dementia burden. Health economics theories and the cognitive reserve
hypothesis lead to opposing predictions about the independent effect of education on post-
diagnosis survival (Grossman 1972; Stern et al 1995), while hypothesis regarding financial
resources is more consistent with the well-known SES gradients in the general population.
Understanding nuances between education and financial resources provides insights into the nature
of disparities in post-diagnosis survival and informs interventions for improving well-being of all
PLWD. Nevertheless, prior studies fail to explicitly disentangle the intertwined relationships
between education, financial resources, and health differences at dementia diagnosis (e.g. Bebe et
al 2019; Russ et al 2013) when explaining disparities in survival after diagnosis. They suffer from
additional methodological limitations, including non-representative samples (e.g. Wolfsen et al
2001; Qiu et al 2001), inaccurate measure of survival from incident diagnosis (e.g. Bebe et al 2019;
Contador et al 2017), and insufficient length of follow-up (e.g. Andersen et al 2010; Pavlik et al
2006).
4
Families are profoundly affected by dementia. Substantial value of caregiving time (Coe
et al. 2018; Zissimopoulos et al. 2015; Hurd et al. 2013; Lin & Neumann 2013) and negative health
impacts (Chen et al 2020b; Roth et al 2019; Allen et al 2017; Schulz et al 2004) imposed on family
members of PLWD are well recognized. It is less clear how health care utilization changes around
the dementia diagnosis of loved ones, which is an important input for one’s own health production.
Considering the time and financial strains dementia places, both caregiving and non-caregiving
family members, the former in particular, face the trade-offs between satisfying the care needs of
PLWD and taking care of their own. As a result, families regardless of caregiving status may alter
the pattern of health care use, which has implications for their own health and downstream health
care costs. The scant literature fails to tease out within-individual variations in health care use that
are not attributable to the dementia diagnosis of their loved ones (e.g. Rahman et al 2019; Gilden
et al 2017; Goren et al 2016). Furthermore, none of them tests for differential responses across
individuals with different constraints, return on health investment, and preference.
In this three-essay dissertation, I address many of these knowledge gaps. In collaboration
with Dr. Bryan Tysinger, Dr. Eileen Crimmins, and Dr. Julie Zissimopoulos, the first essay titled
“Analysis of Dementia in the U.S. Population using Medicare Claims: Insights from Linked
Survey and Administrative Claims Data” analyzes disparities in dementia diagnosis cross-
sectionally and longitudinally, by comparing claims-based diagnosis to survey-based cognition for
diverse subgroups. Conditional on low cognition at a point in time, racial/ethnic minorities are
twice as likely as Whites to be undiagnosed. Following substantial cognitive decline, racial/ethnic
minorities are more vulnerable to no-diagnosis and delayed diagnosis; this longitudinal aspect of
diagnosis disparities has not been investigated previously. Over 2000-2008, we observe reduced
5
likelihood of under- and delayed- diagnosis, suggesting improvement in diagnostic practice.
However, gaps across race/ethnicity remain the same.
In collaboration with Dr. Eileen Crimmins and Dr. Julie Zissimopoulos, the second essay
“Socio-Economics Status and Survival after Dementia Diagnosis: Education, Wealth, and Health
Differences at Diagnosis” disentangles the independent effects of education and wealth on
survival after dementia diagnosis. We follow a broadly representative cohort with incident
dementia diagnosis for up to 18 years and account for comorbidity, function, and cognition at
diagnosis, in testing hypotheses of education and wealth effects. Persons with higher education
and wealth preserve more intact function and cognition and have less comorbidity burden when
first diagnosed. While education is not associated with survival after dementia diagnosis, the most
affluent group have a 17% reduction in mortality risk relative to the least, which is sensitive but
robust to adjustment for impairment and comorbidity. “In the absence of one instrument that acts
on them (author: SES aspects) all” (Deaton 2002), the wealth-related but not education-related
disparities revealed in this essay suggest opportunities to improve quantity and quality of life for
the deprived.
The final essay titled “Do Individuals Skimp on Their Own Health Care after Spouse’s
Dementia Diagnosis?” investigates broader impacts of dementia on families, in terms of outcomes
and affected subjects. Specifically, I examine changes in health care utilization, an input for health
production, among elderly spouses of PLWD irrespective of caregiving status. I improve the
estimates using an event-study framework with individual and time fixed effects to reduce
unobservable confounders. For the first time, I explore differential trajectories among
subpopulations defined by sex, race/ethnicity, education, wealth, health, and caregiving status. In
the first few years after dementia diagnosis, spouses of PLWD stay on their usual health care
6
schedule and spending level. The pattern holds for most subgroups, except for males, non-Hispanic
Whites, and those with more wealth who reduce the number of doctor visits by 1 to 2 in the year
of spouse’s incident diagnosis. The forgone ambulatory care does not pose immediate detriment,
for instance in the form of elevated risk of hospitalization.
Taken together, this dissertation advances the literature of disparities in dementia burden
in the U.S., by investigating the magnitude and mechanism of heterogeneity throughout the
dementia care continuum, ranging from diagnosis and care of PLWD to family support. With the
guidance of health economics theories, this series of study identify hypotheses of groups
disproportionately affected by dementia, use novel linkage of longitudinal representative survey
and administrative claims, and apply rigorous methods to test for these hypotheses. In an aging
and diversifying America, understanding dementia-related disparities illuminates areas for
achieving health equities and informs optimal allocation of limited resources.

7
Chapter 2
Analysis of Dementia in the U.S. Population using Medicare Claims:
Insights from Linked Survey and Administrative Claims Data

INTRODUCTION
Accurate estimates of the prevalence and incidence of dementia, how they are changing
over time and for whom, are essential for quantifying disease burden and for preparing health and
long-term care systems for the inevitable increase in cases. Yet, there is no single data source for
doing so. In the absence of dementia tracking through a national screening program, the main
sources for estimating dementia in the US are nationally representative surveys and health care
claims.
Medicare claims are an important data source for identifying and tracking rates of
diagnosed disease over time in the older US population because the program provides health
insurance for about 97% of older Americans from the age of 65 years until death. The number of
diagnosed cases in the Medicare records however, may underestimate the actual burden of disease
if individuals do not seek treatment for symptoms or request cognitive assessments, providers do
not recognize symptoms and/or undertake assessment, or choose not to report it because of a lack
of treatments that can change the course of the disease (Bradford et al 2009; Chodosh et al 2004;
Valcour et al 2000). Nationally representative surveys are another key source for estimating
population dementia prevalence. The Health and Retirement Study (HRS) (Zissimopoulos et al
2018; Langa et al 2017; Zissimopoulos et al 2014) and the National Health and Aging Trends
Study (NHATS) have repeatedly used cognitive tests to measure dementia prevalence as well as
onset in nationally representative cohorts. Cognitive tests for dementia ascertainment from surveys
8
have been criticized for focusing heavily on language and memory (Kotwal et al 2015), being
sensitive to education level (Spering et al 2012) and for their limited ability to differentiate mild
cognitive impairment from dementia (Knopman et al 2010).
Prior validation studies were limited by small or non-representative samples. Taylor et al
(2009) compared dementia diagnoses in Medicare claims to clinical examinations in the 2001-
2003 Aging Demographic and Memory Study (ADAMS), a small sub-sample of the HRS with
few minority respondents, and reported sensitivity of 85 percent and a specificity of 89 percent.
Other validation studies comparing claims and clinical assessment were based on non-
representative samples and similarly reported co-existence of false positive and negative diagnoses
in claims data (Wilkins et al 2007; Taylor et al 2002)
Studies that compared claims-based diagnoses with survey-based cognitive assessments
for dementia ascertainment in samples broadly representative of the older US population had
opposite findings. Two studies reported higher dementia ascertainment in Medicare claims data
compared to survey-based ascertainment (Ostbye et al 2008; Pressley et al 2003). In contrast,
Amjad and colleagues (2018) reported that 60 percent of respondents with ‘probable’ dementia in
2011 NHATS data had formal diagnosis in three-year Medicare claims. None of these studies
addressed measurement error in dementia ascertainment based on survey data that two recent
studies showed lead to an upward bias in dementia ascertainment (Zissimopoulos et al 2018;
Freedman et al 2018). Nor did they address measurement error in claims data due to ‘rule-out’
diagnosis of reversible dementia symptoms (e.g. visual or auditory problems, vitamin B12
deficiency, thyroid disturbance). Although study results were inconsistent, they generally found
lower level of agreement in dementia prevalence across data sources among individuals with
9
advanced age (Taylor et al 2009; Wilkins et al 2007; Savva et al 2015), mild dementia (Taylor et
al 2002; Ostbye et al 2008), ad lower education (Amjad et al 2018).
In this study we analyzed dementia prevalence and incidence in a large sample of
individuals broadly representative of the older US population from the HRS with data linkages to
their Medicare claims records from 2000-2008. We improved upon the methods used in prior
studies by requiring verification of dementia in both survey and claims-based data sources to
reduce measurement error. We added to prior literature an analysis of how (dis)agreement in
dementia prevalence is changing over time and for which populations. This is the first study to
quantify concordance in incidence of dementia and the timing of diagnosis after substantial
cognitive decline, as well as racial/ethnic, socioeconomic and sex differences in this timing. The
findings illuminate the value and caveats of using Medicare claims and cognitive measures from
survey data for studying dementia in the U.S. population. This is particularly important given the
absence of clinical assessments in nationally representative, large, and longitudinal samples.
Meanwhile, diagnoses in Medicare claims reflect clinical practice. Improved understanding of who
is diagnosed and when may aid policies to reduce disparities in dementia diagnosis.

METHODS
Study Population
We use data from the HRS linked to respondents’ Medicare claims from the beginning of
2000 to the end of 2008. HRS is a nationally representative longitudinal study that has surveyed
Americans over 50 years of age and their spouses since 1992. Respondents are interviewed
biennially, on topics of health, health care usage, employment, economic, and family. A key
feature of the HRS study design is oversampling of African Americans and Hispanics and weights
10
may be used for providing a nationally representative sample. Minority response rates at baseline
and in longitudinal follow-ups have been equal to or better than that of majority Whites (O’fstedal
et al 2005). Eighty-eight percent of HRS respondents consented to the linkage of their survey
responses to their Medicare claims records (St.Clair et al 2017). Our sample is restricted to HRS
respondents age 67 and older, with linked claims data and at least two years of continuous fee-for-
service (FFS) enrollment yielding 10,450 unique persons and 31,186 person-waves. The mean
follow-up was 2.98 HRS interview waves.

Dementia Measures and Outcomes
Cognitive tests were administered at each wave to respondents using an adapted version of
the Telephone Interview for Cognitive Status (TICS). When missing for self-respondents, the
measures were imputed by HRS as described by Fisher et al (2015). Around 6.2 percent of self-
respondents in our study sample had at least one imputed scores for cognitive tests. When a
respondent does not, or cannot perform the cognitive assessment, dementia was determined using
information provided by a proxy respondent, typically a spouse or other family member and the
interviewer (O’fstedal et al 2005). We followed prior studies in the classification of dementia
which is based on the concordance of HRS cognitive functioning scores and consensus diagnosis
of dementia in a subset of HRS respondents who had extensive neuropsychological assessment in
ADAMS (Langa et al 2017; Crimmins et al 2011). An individual was classified as having dementia
based on a low score (0-6 out of 27) on test items that evaluate memory and concentration and
executive function: immediate and delayed word recall, counting back from 100 by 7’s, and
counting back from 20 (Zissimoupoulos et al 2018; Langa et al 2017; Crimmins et al 2011).
Among respondents with a proxy, dementia is based on number of limitation with instrumental
11
activities of daily living, interviewer impairment rating from 0 (none) to 2 (cognitive limitations)
and proxy informants’ impairment rating from 0 (none) to 4 (poor).
To reduce measurement error in dementia ascertainment based on cognitive scores, we
required one wave with dementia and evidence of continued cognitive impairment in the next
consecutive wave (Zissimopoulos et al 2018; Freedman et al 2018). If the respondent with one
wave of dementia died before the next wave, he or she was assumed to have dementia before dying.
Once we identified a respondent as having ‘verified’ dementia, we assumed dementia in all
subsequent waves.
Providers that bill Medicare use codes for patient diagnoses. The first code listed is the
primary diagnosis and more than one diagnosis code is allowed. In Medicare claims, we
ascertained dementia based on the Chronic Conditions Data Warehouse (CCW) algorithm for
Alzheimer's disease or related disorders or senile dementia using the following ICD-9 diagnosis
codes: 331.0, 331.11, 331.19, 331.2, 331.7, 290.0, 290.10, 290.11, 290.12, 290.13, 290.20, 290.21,
290.3, 290.40, 290.41, 290.42, 290.43, 294.0, 294.10, 294.11, 294.20, 294.21, 294.8, and 797.
Additional diagnostic codes were also included, to account for dementia with Lewy bodies,
cerebral degeneration, senile psychosis, and dementia classified elsewhere: 331.82, 331.89, 331.9,
290.8, 290.9, 294.9. CCW algorithm requires at least one inpatient, facility, home health or
outpatient claim with one of the above diagnosis codes during a three-year lookback period.
Similar to the verified measure in HRS, we additionally required a second diagnosis claim over
the study period to rule out reversible dementia symptoms.
The main outcome of interest is the (dis)agreement between the two measures of dementia
for an individual. Agreement at a point in time (prevalent dementia) was defined as having the
same dementia status across the data sources during the years between two consecutive HRS waves,
12
approximately two years. We assessed agreement in incident dementia similarly by comparing
dates of incidence dementia based on the two measures. Incident dementia using claims data is
the earliest dementia diagnosis date on a claim conditional on no prior diagnosis (verified by
subsequent diagnosis as described above). Incident dementia based using HRS is the earliest
survey date of dementia based on cognitive assessment conditional on not having dementia as
measured by scores in the prior waves. Again, verification of ‘new’ dementia with subsequent
low cognitive status in the next wave is required.

Explanatory Variables
Also included in the analysis are: age, sex, and race (black, Hispanic, non-Hispanic White),
highest level of education (less than high school, high school, college and above), marital status
(married or not), the presence of chronic conditions and diseases (stroke, heart disease, diabetes,
and hypertension), health care utilization (binary indicator for a physician visit during the past two
years), and survival (indicator for whether died between survey waves).

Statistical Analysis
We applied HRS sampling weights to quantify concordance in dementia prevalence from
2000-2008. We used multinomial logistic regression to quantify demographic and socioeconomic
factors associated with concordance in dementia prevalence, adjusting for survival into the next
wave, physician visits, and a linear time trend. An interaction term between race and time was
tested separately to see whether there were differential time trends by race.
We quantified dementia incidence in a sample without prior dementia based on either
measure. We also selected a subsample with incident dementia between 2000 and 2004 based on
13
cognitive tests from the HRS (N=1,161), and analyzed whether and when a dementia diagnosis
occurred based on claims data. We quantified timing and applied sampling weights. We used
multinomial logistic regression to quantify the socioeconomic and demographic factors associated
with the timing of diagnosis relative to incident dementia based on cognitive decline from ‘no
dementia’ to ‘new’ dementia based on scores from cognitive tests.
As a sensitivity check, we modified the definition of dementia in the following ways: 1)
required any subsequent verification of dementia in HRS (rather than that at the next consecutive
wave); 2) required no verification for diagnosis in claims; and 3) used an augmented list of
diagnostic codes including dementia symptoms (ICD-9 codes: 780.93, 784.3, 784.69, and 331.83).
When defining agreement, we allowed for a longer period for diagnosis or HRS dementia
(extending by approximately 2 years). We also added control for household wealth in multivariate
analyses.

RESULTS
Sample characteristics
Table 1 reports the cross-sectional characteristics of the respondents in years 2000 and
2008. Characteristics in this linked sample were compared to that in the full HRS sample aged 67
and above. In 2000, the linked sample was comparable to the full HRS sample in terms of sex,
education, marital status, and cardiovascular profiles. The linked sample is more likely to non-
Hispanic White than the full HRS sample.

14
Concordance in prevalent dementia
We reported concordance in prevalent dementia for persons according to four categories:
(a) person does not have dementia, both measures, (b) has dementia, both measures, (c) has
dementia based on cognitive tests only, and (d) has dementia based on diagnosis only, during years
between two consecutive HRS survey waves. The first two categories were considered as
agreement. There was concordance in prevalent dementia for 86.1 percent of the respondents based
on the two measures (Table 2). Dementia prevalence ascertained by both measures was 7.2 percent
while survey-based cognitive tests only was 6.9 percent and by diagnosis only was 7.0 percent.
Thus, only half of dementia cases identified by one source had dementia ascertained by the other
measure. Whites had higher concordance (no dementia both or has dementia both) than blacks
and Hispanics (W= 88.1 percent; B= 74.9 percent; H= 70.8 percent). ‘Has dementia both measures’
was more prevalent in racial/ethnic minorities than Whites (W= 6.7 percent; B= 12.2 percent; H=
9.4 percent). The dominant disagreement type among Whites was ‘dementia by diagnosis only’,
while that among blacks and Hispanics was ‘dementia by cognitive tests only’. Whites and
Hispanics had a similar proportion of dementia by diagnosis only (W= 7.2 percent; H= 8.1 percent),
nearly twice as high as that for blacks (B= 4.3 percent).
Figure 1 shows the results from multinomial logits models of concordance across dementia
measures, illustrated by predicted probabilities of each outcome separately by race/ethnicity,
education and sex. Odd ratios are provided in Supplementary Table 1. Blacks were 3.8 times as
likely and Hispanics 2.9 times as likely as Whites to have dementia identified by cognitive test
only (B= 0.138; W= 0.036; H= 0.105). The likelihood of having ‘dementia by diagnosis only’
was 0.062 for Whites, 0.055 for blacks and not statistically different (P value= .268). The
probability of ‘dementia by diagnosis only’ for Hispanics was 0.103 and statistically different than
15
Whites (P value= .004). We found no differential time trends by race.
Individuals with less than high school education (0.135) were more likely to have dementia
ascertained by cognitive tests only than individuals with a high school (0.039) or college (0.018)
education. There were no differences by education in the predicted probability of ‘dementia,
diagnosis only.’ Males were slightly more likely to have ‘dementia by cognitive test only’ (P
value= .014), and there was no sex difference in the likelihood of ‘dementia by diagnosis only’ (P
value= .52). Over time, respondents were less likely to have ‘dementia by cognitive test only’
(Supplementary Table 2). All results were robust to varying definitions of dementia and of
agreement, and to adding wealth controls to the models.

Concordance in incident dementia
Table 3 reported concordance in dementia incidence, during a 4-year time window (i.e. 2
years backward and 2 years forward). Roughly 83 percent of individuals had agreement in
dementia incidence across the measures. Concordance among racial/ethnic minorities was lower
than that among Whites (W = 84.5 percent; B= 73.8 percent; H = 72.9 percent). Among Whites,
dementia incidence based on cognitive tests only (3.2 percent) was lower than that based on
diagnosis only (12.3 percent). In contrast, among blacks, dementia incidence based on cognitive
tests only (15.2 percent) was higher than that based on diagnosis only (11.1 percent). Among
Hispanics, the rates were similar for cognitive test only (13.2 percent) and diagnosis only (13.9
percent).
We analyzed diagnosis among a subsample of respondents with incident dementia
ascertained with cognitive tests between years HRS survey years 2000 and 2004. We quantified
groups of individuals who were: (1) diagnosed more than 2 years prior to incident dementia, (2)
16
diagnosed 2 years or less prior to incident dementia, (3) diagnosed 2 years or sooner after incident
dementia, (4) diagnosed more than 2 years after incident dementia and before the end of the study
period (December 31, 2008), (5) died 2 years or sooner after incident dementia without a diagnosis,
(6) died more than 2 years after incident dementia and before end of the study period (December
31, 2008), without a diagnosis, and (7) survived to end of the study period (December 31, 2008),
without a diagnosis. A significant proportion (22.3%) were diagnosed prior to incident dementia
as measured by cognitive tests. About 85 percent were either diagnosed with dementia or died
during the study period (Figure 2). The remaining 15.3 percent of the sample were on average
followed for 5.9 years without receiving a diagnosis.
Descriptive characteristics of individuals in each of the 7 groups described above are
provided in Supplementary Table 3 and Table 4. Figure 3 illustrates with predicted probabilities
the results of a multinomial logit model of the factors associated with timing of diagnosis relative
incident dementia as ascertained through cognitive test. The estimated model combined the groups
(2) and (3) where the time difference between incident dementia and diagnosis is two years or less,
combined the groups (5) and (6), persons who died without a diagnosis, and adjusted for sex, age
group, race, education, marital status, doctor visit during the past two years, and a linear time trend.
Blacks (0.196) and Hispanics (0.236) had a significantly higher odds of surviving without
diagnosis than Whites (0.094) and the differences were statistically significant (P valueB< .000; P
valueH= .003). Blacks and Hispanics were less likely than Whites to be diagnosed over 2 years
prior to incident dementia based on cognitive test (P valueB= .027; P valueH= .022). Blacks (0.301)
were less likely be diagnosed within 2 years of incident dementia compared to Whites (0.395) and
there was no statistical differences between Hispanics and Whites (P value= .055). Blacks were
more likely (0.141) than Whites (0.093) to receive diagnosis 2 years after cognitive-test-based
17
dementia incidence (P value= .048); no significant difference was detected between Hispanics and
Whites (P value= .205).
Disparities were prominent between those with less than high school education and those
with any college education (Figure 3). Individuals with a college education were more likely to
have a diagnosis before dementia was first detected by cognitive tests (P value= .036), and less
likely to have received a diagnosis more than two years after (P value= .005), and of surviving
without diagnosis (P value< .000). Results were not statistically different for those with high
school diploma compared to no high school degree with the exception of a lower likelihood of
surviving without diagnosis.
Sex was not associated with existence and timing of diagnosis (Figure 3. There was a
decrease in the likelihood of diagnosis more than 2 years after incident dementia, as well as an
increase in that of surviving without diagnosis (Supplementary Table 6). These results were not
qualitatively different when definition of dementia was modified and with wealth adjustments.

DISCUSSION
Utilizing a nationally representative sample of older Americans from the longitudinal
Health and Retirement Study with linkages to their health care claims, we found that at a point in
time, ascertained dementia from survey-based cognitive tests and dementia diagnosis from
Medicare claims produced similar prevalence estimates at the population level (14%). However,
only half of these individuals were identified as having dementia by both measures. This level of
agreement in dementia prevalence at the individual level was consistent with previous literature
(Amjad et al 2018; Ostbye et al 2008; Pressley et al 2003). Racial/ethnic minorities, individuals
with less than a high school education, and males were more likely than Whites, college educated
18
individuals and females, respectively, to have been identified as with dementia based on cognitive
tests only. In contrast, dementia ascertained by diagnosis only was no different across education
groups, for males relative to females and blacks relative to Whites. Hispanics were more likely
than Whites to have diagnosed dementia only (but not ascertained by cognitive tests).
To our knowledge, this is the first study to examine concordance in the timing of incident
dementia. We found almost one-quarter had a diagnosis of dementia two years or more before
dementia was indicated by cognitive tests. This was more prevalent among college-educated
person than those with lower levels of education. Among respondents with incident dementia
between the years 2000-2004 (ascertained with cognitive assessment), 85 percent of these
respondents were diagnosed or had died by 2008. Although only 15 percent had not yet been
diagnosed, it was more common among blacks and Hispanics than Whites. These findings suggest
significant disparities in dementia diagnosis by race/ethnicity, both in terms of presence and timing
of diagnosis. Unmeasured factors such as physician behavior and patient preferences may be
related to the racial/ethnic differences and are key areas for future research (Hill et al 2015;
Mukadam et al 2013; Livney et al 2011).
The study also reflects racial/ethnic disparities in dementia risks. Studies using in-depth
clinical examinations for dementia ascertainment reported mixed evidence on elevated risk of
dementia for blacks, in geographically restricted samples (Katz et al 2012; Hebert et al 2010;
Fitzpatrick et al 2004; Tang et al 2001). Using data from ADAMS, Plassman and colleagues
(Plassman et al 2011) found no black-White difference in dementia risk, yet the estimation was
based on a small sample of blacks. In this study, we observed higher rates of dementia prevalence
for blacks and Hispanics compared to Whites based on either diagnosis in claims data or cognitive
test in survey data.
19
Following cognitive decline, college-educated individuals had a lower odds of surviving
without diagnosis compared to individuals with less than high school education. The college-
educated were more likely assessed as having ‘normal’ cognition after receiving a dementia
diagnosis, and to have a shorter lag between low cognition and diagnosis. These results are
consistent with a cognitive reserve hypothesis (Stern et al 2006 & 1999), contending that education
would mitigate the symptoms of dementia, such as impaired cognition, until dementia is at a more
advanced stage. Furthermore, highly educated individuals may be more likely to be diagnosed than
low educated individuals as a result of better access to and utilization of health care services.
Several studies have called for an adjustment for education in cognitive tests (Spering et al 2012;
Crum et al 1993). However, trade-offs between standardization of test and precision of estimation
require further investigations.
Females were more likely than men to have agreement in prevalent dementia across
measures consistent with empirical findings of a higher risk of dementia, which may be driven by
genetic differences, social-cultural factors, or mortality selection (Mielke et al 2014; Azad et al
2007; Ruitenberg et al 2001). We did not find sex difference in disagreement across measures (i.e.
cognitive tests only or diagnosis only).
Over time, we observed potential improvement in diagnostic practice between 2000 and
2008, as shown by the shrinking likelihood of having prevalent ‘dementia by cognitive tests only’,
coupled with that of having diagnosis more than 2 years after incident dementia based on cognitive
tests. Continued efforts are needed to alleviate barriers to diagnosis, including increased access to
care, or improvement in physicians’ knowledge about dementia and willingness to diagnose
(Bradford et al 2009), especially for groups vulnerable to missed diagnosis. A timely diagnosis not
only confers benefits to patients and families afflicted with dementia (Barnett et al 2014; Borson
20
et al 2013; Carpenter et al 2008), but also reduces long-term care spending to the health care system
(Long et al 2014; Weimer et al 2009).
There exist several limitations in this study. Although broadly representative, this sample
does not include individuals in Medicare HMOs, who are more likely to be racial/ethnic minorities
and younger (McGuire et al 2011), and only includes respondents consenting for linkage to
Medicare claims who tend to be younger, non-White, and wealthier (Sakshaug et al 2014; Sala et
al 2014). Measurement error in ascertaining dementia is reduced by requiring a second dementia
ascertainment, and by examining change in cognition, rather than cross-sectional variations in
cognition. However, some subtypes of dementia may manifest in symptoms that are not well-
detected by the set of cognitive tests in the HRS but may be diagnosed by a clinician. Measurement
error may vary by race and education. For example, if cognitive batteries in the HRS are less
sensitive to cognitive decline among high-educated individuals relative to low-educated, these
individuals would have a lower likelihood of being in our subsample analysis of incident diagnosis
after cognitive decline and thus less likely to be at risk of ‘no diagnosis.’ Similarly, non-Whites
may be more likely than Whites to be categorized incorrectly with cognitive decline. Thus,
disparities by education and race/ethnicity in onset of dementia without diagnosis may be over-
stated.
In conclusion, Medicare claims data yield equal prevalence estimates as nationally,
representative survey data. These data are important data resources for researchers quantifying
dementia in the U.S. population and how it is changing over time. However, disparities in
concordance of measures by race and education level shed light on data limitations in both survey
and claims data. Blacks, Hispanics, and persons with low education are at risk of having no or
delayed diagnosis. Using survey data containing cognitive tests to measure dementia may under-
21
identify incidence among Whites and college educated persons. Methodological advances for
identifying dementia by cognitive assessment in surveys are needed. Policy change such as
inclusion of cognitive assessment in the new Medicare Annual Wellness Visit and reimbursement
for this visit, may be improving recognition of dementia in clinical practice and across diverse
population.

22
TABLES & FIGURES

Table 1. Sample Characteristics in Years 2000 and 2008

HRS-Claims Linked Sample HRS 67+ Sample P Values
2000 2008 2000 2008 2000 2008
N
6,142 5,706
9,404 10,285

Age

0.169 0.000
67 to 74 42.6% 39.3% 46.7% 45.0%

75 to 84
42.8% 41.1% 40.4% 38.4%

85 and above 14.6% 19.6% 12.9% 16.6%

Mean (SD), years
76.8 (6.80) 77.6 (7.17) 76.2 (6.78) 76.6 (7.16) 0.049 0.000
Female 59.8% 60.0% 59.2% 58.1% 0.369 0.149
Race

0.010 0.000
White 86.8% 87.5% 86.4% 84.6%

Black 9.0% 8.0% 8.5% 8.4%

Hispanic 4.1% 4.5% 5.1% 7.0%

Education

0.893 0.761
Less than high school 35.5% 25.9% 35.3% 28.3%

High school & equivalent 32.1% 34.1% 31.5% 33.2%

College and above 32.4% 40.0% 33.2% 38.4%

Not Married/Partnered 47.5% 47.2% 46.9% 45.1% 0.474 0.041
Cardiovascular risk factors

Stroke 12.8% 13.4% 12.1% 12.8% 0.653 0.041
Heart disease 32.0% 34.7% 30.3% 32.7% 0.222 0.001
Diabetes 15.0% 21.5% 15.4% 22.1% 0.493 0.919
Hypertension 52.8% 64.8% 52.0% 64.6% 0.672 0.279
Died between this and next wave 12.2% 11.3% 11.9% 12.4% 0.021 0.028
Notes: HRS 67+ sample requires 1) age>=67 years, and 2) responded to HRS interview. HRS-claims linked sample additionally
requires continuous FFS enrollment for at least 2 years. The reported percentages are weighted, using wave-specific HRS
sampling weights to adjust for survey design. P values indicate level of statistical difference in characteristics between HRS 67+
sample and HRS-claims linked sample.

Table 2. Concordance in Dementia Prevalence and by Race 2000-2008 (N=31,186)

All Whites Blacks Hispanics
No dementia, both measures 78.9% 81.4% 62.7% 61.4%
Dementia, both measures 7.2% 6.7% 12.2% 9.4%
Dementia, cognitive test only 6.9% 4.8% 20.8% 21.1%
Dementia, diagnosis only 7.0% 7.2% 4.3% 8.1%

Concordance in prevalent dementia 86.1% 88.1% 74.9% 70.8%
N 31,186 25,504 3,953 1,728
Notes: Agreement is based on the same dementia status during the years between two consecutive HRS waves.

23
Figure 1. Predicted Probability of Concordance in Prevalent Dementia,
by Race/Ethnicity, Education, and Sex (N=31,186)

A. Race/Ethnicity

B. Education

24
C. Sex

Notes: Predicted probabilities of each concordance category are based on estimates from multinomial logistic regression,
adjusting for sex, age group, race, education, marital status, survival in two years, doctor visit during the past two years, and a
linear time trend; error bars show 95% confidence intervals of predictions. Black dots indicate statistical difference between a
probability and that for whites, less than high school education, or male, at a significance level of 0.05. Number of observations
in regression reduced from 31,186 to 31,117 due to missing values in covariates.

Table 3. Concordance in Dementia Incidence and by Race 2000-2008 (N=9,623)

All Whites Blacks Hispanics
No incident dementia, both measure 79.1% 81.4% 68.2% 69.3%
Incident dementia, both measure 3.4% 3.1% 5.6% 3.5%
Incident dementia, cognitive test only 5.2% 3.2% 15.2% 13.2%
Incident dementia, diagnosis only 12.2% 12.3% 11.1% 13.9%

Concordance in incident dementia 82.5% 84.5% 73.8% 72.9%
N 9,623 7,883 1,201 538
Notes: Agreement is defined as having incident dementia by both measures, during a 4-year time window (i.e. 2 years backward
and 2 years forward). Dementia based on cognitive test requires evidence of continued cognitive impairment in the next
consecutive HRS wave, or death before the next wave. Dementia diagnosis ascertained by Medicare claims requires observing a
second diagnosis or death as of 12/31/2008, as verification. This sample excludes respondents with incident dementia based on
HRS prior to wave 5 (i.e. year 2000) given time horizon of this study, and excludes those with incident dementia based on HRS
at wave 9 (i.e. year 2008) due to availability of linked claims data up to 12/31/2008.

25
Figure 2. Timing of Diagnosis in Claims Data Relative to Incident Dementia Based on
Cognitive Scores in HRS Data (N=1,161)

Notes: DX= diagnosis coded in Medicare claims. This subsample is limited to respondents who were ascertained as dementia by
HRS cognitive tests for the first time during HRS 2000, 2002, or 2004 waves. From the left to the right, outcomes are: (1)
diagnosed more than 2 years prior to incident dementia in HRS, (2) diagnosed 2 years or less prior to incident dementia in HRS,
(3) diagnosed 2 years or sooner after incident dementia in HRS, (4) diagnosed more than 2 years after incident dementia in HRS
and before 12/31/2008, (5) died 2 years or sooner after incident dementia in HRS without a diagnosis in claims, (6) died more
than 2 years after incident dementia in HRS and before 12/31/2008 without a diagnosis in claims, and (7) survived to 12/31/2008
without a diagnosis in claims. The reported percentages are weighted, using wave-specific HRS sampling weights to adjust for
survey design.

26
Figure 3. Predicted Probability of Relative Timing of Incident Dementia,
by Race, Education, and Sex (N = 1,161)

A. Race/Ethnicity

B. Education

27
C. Sex

Notes: DX= diagnosis coded in Medicare claims. Five outcome groups in the multinomial logistic regression include: 1)
diagnosed more than 2 years prior to incident dementia in HRS, 2) less than 2 years time difference between incident dementia in
claims data and in HRS, 3) diagnosed more than 2 years after incident dementia in HRS and before 12/31/2008, 4) died before
12/31/2008 without a diagnosis in claims, and 5) survived to 12/31/2008 without a diagnosis in claims. This figure omits
estimates for group (4). Predicted probabilities of each concordance category are based on estimates from multinomial logistic
regression, adjusting for sex, age group, race, education, marital status, doctor visit during the past two years, and linear time
trend; error bars show 95% confidence intervals of predictions. Black dots indicate statistical difference between a probability
and that for whites, less than high school education, or male, at a significance level of 0.05. Number of observations in regression
reduced from 1,161 to 1,152 due to missing values in covariates.

28
APPENDIX

Supplementary Table 1. Odds Ratios for Concordance in Prevalence
Relative to ‘Dementia, Both Measures (N=31,186)

OR [95%CI]
No Dementia, both
measures
Dementia only in
HRS
Dementia only in Medicare
claims
Female 0.709*** 0.585*** 0.683***
[0.597, 0.842] [0.480, 0.714] [0.558, 0.837]
Age
(ref. Aged 67 to 74)

Aged 75 to 84 0.221*** 0.377*** 0.672**

[0.182, 0.269] [0.301, 0.471] [0.529, 0.854]
Aged 85 + 0.063*** 0.311*** 0.411***
[0.051, 0.079] [0.243, 0.397] [0.315, 0.537]
Race
(ref. Whites)

Black 0.386*** 1.735*** 0.405***

[0.317, 0.469] [1.413, 2.131] [0.309, 0.531]
Hispanic 0.574*** 1.949*** 1.118
[0.424, 0.778] [1.420, 2.676] [0.774, 1.616]
Education
(ref. Less than high school)

High school 2.657*** 0.656*** 2.389***

[2.211, 3.192] [0.525, 0.820] [1.915, 2.979]
College and above 3.238*** 0.362*** 2.957***
[2.684, 3.905] [0.280, 0.469] [2.370, 3.689]
Single 0.749*** 0.844 0.945
[0.634, 0.885] [0.692, 1.028] [0.771, 1.157]
Died before the next wave 0.137*** 0.400*** 0.397***
[0.122, 0.152] [0.350, 0.458] [0.346, 0.455]
Visited doctors during past two
years
0.551*** 0.459*** 1.424
[0.409, 0.744] [0.331, 0.636] [0.922, 2.201]
Linear time trend 0.976 0.933** 0.995
[0.940, 1.014] [0.890, 0.977] [0.949, 1.043]
Constant 135.5*** 18.13*** 1.384
[87.48, 209.8] [11.04, 29.76] [0.767, 2.497]
Observations 31,117
Notes: Variables are measured at the specific HRS wave. *** denotes P value <.001, ** P value <.01, and * P value <.05.
Number of observations in regression reduced from 31,186 to 31,117 due to missing values in covariates.

29
Supplementary Table 2. Predicted Probability of Concordance in Dementia Prevalence
by Race, Education, Sex, and Year (N=31,186)

Predicted Probability of Concordance in Dementia Prevalence

No dementia, both
measures
Dementia, both
measures
Dementia, cognitive test
only
Dementia, diagnosis
only
Whites 0.869 [0.862, 0.876] 0.033 [0.030, 0.037] 0.036 [0.032, 0.040] 0.062 [0.057, 0.067]
Blacks 0.735 [0.710, 0.759]***
0.073 [0.061,
0.084]*** 0.138 [0.120, 0.156]*** 0.055 [0.044, 0.066]
Hispanics 0.743 [0.706, 0.780]*** 0.049 [0.036, 0.063]* 0.105 [0.084, 0.126]***
0.103 [0.075,
0.130]**

Less than high
school 0.739 [0.723, 0.755] 0.068 [0.060, 0.076] 0.135 [0.122, 0.147] 0.059 [0.051, 0.066]
High school 0.869 [0.858, 0.880]***
0.030 [0.025,
0.035]*** 0.039 [0.033, 0.045]*** 0.062 [0.055, 0.069]
College and
above 0.892 [0.883, 0.901]***
0.025 [0.021,
0.029]*** 0.018 [0.015, 0.022]*** 0.065 [0.057, 0.072]

Male 0.853 [0.842, 0.864] 0.031 [0.026, 0.035] 0.052 [0.045, 0.058] 0.065 [0.057, 0.072]
Female 0.852 [0.843, 0.861]
0.043 [0.039,
0.048]*** 0.042 [0.038, 0.047]* 0.062 [0.056, 0.068]

Year 2000 0.853 [0.844, 0.862] 0.036 [0.032, 0.040] 0.050 [0.045, 0.056] 0.061 [0.055, 0.067]
Year 2002 0.853 [0.845, 0.860] 0.037 [0.033, 0.041] 0.048 [0.044, 0.053]* 0.062 [0.057, 0.067]
Year 2004 0.853 [0.846, 0.860] 0.038 [0.034, 0.042] 0.046 [0.042, 0.050]** 0.063 [0.059, 0.068]
Year 2006 0.853 [0.845, 0.860] 0.039 [0.035, 0.043] 0.043 [0.039, 0.048]** 0.065 [0.059, 0.070]
Year 2008 0.853 [0.843, 0.862] 0.040 [0.035, 0.045] 0.042 [0.037, 0.047]** 0.066 [0.059, 0.072]
Notes: Predicted probabilities of each concordance category are based on estimates from multinomial logistic regression,
adjusting for sex, age group, race, education, marital status, survival in two years, doctor visit during the past two years, and
linear time trend; 95% confidence intervals of predictions are displayed in brackets. Asterisks indicate level of statistical
difference across race, educational attainment, and Sex, with the top group of each variable being reference; for years from 2000
to 2008, test of statistical difference is performed between this year and the previous year. *** P value <.001, ** P value <.01,
and * P value <.05.

30
Supplementary Table 3. Characteristics of Sample at the Time of Incident Dementia
Ascertained by HRS (N=1,161)

Diagnosed within study period Never diagnosed within study period

1. DX
more than
2 years
prior to
incident
dementia
in HRS
2. DX 2
years or
less prior
to
incident
dementia
in HRS
3. DX 2
years or
sooner
after
incident
dementia
in HRS
4. DX
more than
2 years
after
incident
dementia
in HRS
and before
12/31/2008
5. Died 2
years or
sooner
after
incident
dementia
in HRS
without a
DX in
claims
6. Died more
than 2 years
after incident
dementia in
HRS and
before
12/31/2008
without a DX
in claims
7.
Survived
to
12/31/2008
without a
DX in
claims
n (% of N) 22.3 22.4 13.8 10.5 7.7 8.0 15.3
Age (%)

67-74 12.0 12.1 15.9 23.3 28.9 34.7 48.3
75-84 44.0 49.6 46.3 50.9 32.7 30.9 39.5
85+ 44.0 38.3 37.8 25.8 38.3 34.4 12.3
Mean Years (SD) 82.9(7.11) 82.6(7.00) 82.5(7.54) 80.1(7.37) 80.7(8.22) 79.3(8.76) 75.8(7.12)
Male (%) 33.1 28.3 37.8 38.6 48.9 54.7 39.0
Race (%)

White 83.8 87.2 78.5 63.6 79.5 72.7 51.2
Black 12.5 10.2 16.1 27.8 15.6 20.7 34.3
Hispanic 3.7 2.6 5.4 8.6 4.9 6.6 14.5
Education (%)

Less than high
school
47.5 46.1 53.8 70.9 42.3 72.0 79.1
High school 23.3 28.8 29.1 20.5 37.7 20.4 15.3
College and above 29.2 25.1 17.1 8.6 20.0 7.6 5.6
Doctor Visit during
the past 2 years
96.7 96.8 96.7 93.6 97.3 84.9 93.7
Disease Prevalence

Stroke (%) 37.8 33.3 26.5 21.1 35.5 26.1 12.9
Heart disease (%) 41.8 49.0 48.7 42.3 64.9 43.2 32.1
Diabetes (%) 15.6 20.6 31.0 17.5 32.0 35.1 24.9
Hypertension (%) 58.7 59.2 64.4 62.5 66.2 54.8 60.8
Nursing Home
Residency
35.1 43.6 18.2 3.6 27.4 15.8 1.9
Mean Time Difference
(yr)
-4.64 -1.05 0.85 4.00 0.81 3.70 5.88
Notes: DX= diagnosis coded in Medicare claims. This subsample is limited to respondents who were ascertained as dementia by
HRS measure for the first time during HRS 2000, 2002, or 2004 waves. From the left to the right, outcome groups are: (1)
diagnosed more than 2 years prior to incident dementia in HRS, (2) diagnosed 2 years or less prior to incident dementia in HRS,
(3) diagnosed 2 years or less after incident dementia in HRS, (4) diagnosed more than 2 years after incident dementia in HRS and
before 12/31/2008, (5) died 2 years or less after incident dementia in HRS without a diagnosis in claims, (6) died more than 2
years after incident dementia in HRS and before 12/31/2008 without a diagnosis in claims, and (7) survived to 12/31/2008
without a diagnosis in claims. Mean time difference in years is calculated from the date of diagnosis minus the date of HRS-
ascertained dementia onset. The former is based on the date provided in a claim with a diagnosis of incident dementia. The latter
is based on the corresponding HRS survey date when dementia was ascertained by HRS for the first time.

31
Supplementary Table 4. Changes in Cognitive Score around Incident Dementia in HRS

Received no DX and survived (n=177) Received DX (n=801)
Transition in
respondent
type
Wave prior to
incident
dementia
Wave of incident
dementia
Change
in score
Wave prior to
incident
dementia
Wave of
incident
dementia
Change in
score
Self to self
(n=508)
9.92 5.03* -5.01 10.20 4.69 -5.48
[9.38, 10.45] [4.76, 5.30]
[-5.60, -
4.42] [9.88, 10.51] [4.52, 4.85] [-5.82, -5.14]
Self to proxy
(n=226)
10.55 8.45 N/A 10.53 9.02 N/A
[6.18, 14.9] [5.51, 11.40] N/A [10.10, 10.97] [8.78, 9.25] N/A
Proxy to
proxy
(n=183)
2.62 6.45*** 3.69 3.51 8.47 4.63
[1.58, 3.66] [5.91, 6.98]
[2.33,
5.05] [3.20, 3.81] [8.18, 8.75] [4.17, 5.08]
Proxy to self
(n=61)
1.79* 5.39* N/A 3.10 3.98 N/A
[0.85, 2.74] [4.78, 6.00] N/A [2.39, 3.82] [3.40, 4.56] N/A
Notes: DX= diagnosis coded in Medicare claims. Cognitive score and dementia criteria for self-respondents: 0-6 "Dementia", 7-
11 "CIND", 12-27 "Normal". Cognitive score and dementia criteria for proxy-respondents: 6-11 "Dementia", 3-5 "CIND", 0-2
"Normal". We divided the subsample based on cross-wave respondent type: self to self (n= 508), self to proxy (n=226), proxy to
proxy (n=183), and proxy to self (n=61). Changes in cognitive score were only calculated for individuals with the same
respondent type across wave. P values were calculated for the significant difference in mean score between ‘received no DX and
survived’ and ‘received DX’. *** denotes P value <.001, ** P value <.01, and * P value <.05.

32
Supplementary Table 5. Odds Ratios for Concordance in Incidence
Relative to ‘Less than 2-year Time Difference between Diagnosis and Incident Dementia in
HRS’ (N = 1,161)

DX more than
2 years prior
to incident
dementia in
HRS
DX more than 2 years
after incident dementia
in HRS and before
12/31/2008
Died before
12/31/2008
without DX
Survived to
12/31/2008
without DX

Female 0.972 0.838 0.520*** 0.934
[0.694, 1.362] [0.549, 1.279] [0.359, 0.753] [0.629, 1.386]
Race (ref. Whites)

Black 0.952 1.985** 1.421 2.737***

[0.615, 1.475] [1.225, 3.217] [0.905, 2.229] [1.762, 4.253]
Hispanic 0.821 2.092 1.541 3.458***
[0.390, 1.727] [0.998, 4.387] [0.738, 3.215] [1.786, 6.695]
Education (ref. Less than
high school)

High school 0.789 0.678 1.097 0.561*

[0.526, 1.183] [0.400, 1.150] [0.706, 1.703] [0.338, 0.930]
College and above 1.225 0.378** 0.664 0.242***
[0.819, 1.833] [0.183, 0.782] [0.388, 1.136] [0.113, 0.516]
Age (ref. Aged 67 to 74)

Aged 75-84 1.003 0.712 0.372*** 0.266***

[0.606, 1.660] [0.412, 1.231] [0.228, 0.608] [0.168, 0.421]
Aged 85+ 1.171 0.454** 0.488** 0.120***
[0.708, 1.936] [0.250, 0.823] [0.299, 0.795] [0.0689, 0.209]
Visited doctor during the past
2 years
1.007 0.448 0.272** 0.482

[0.352, 2.882] [0.159, 1.264] [0.116, 0.639] [0.182, 1.277]
Linear time trend
1.069 0.550*** 0.780* 1.511***

[0.880, 1.297] [0.420, 0.719] [0.624, 0.976] [1.190, 1.920]
Constant
0.423 39.57*** 18.85*** 0.224

[0.0837, 2.139] [5.856, 267.3] [3.765, 94.41] [0.0368, 1.359]
Pseudo R
2

0.0931
Observations
1,152
Notes: DX= diagnosis coded in Medicare claims. *** denotes P value <.001, ** P value <.01, and * P value <.05. Number of
observations in regression reduced from 1,161 to 1,152 due to missing values in covariates.

33
Supplementary Table 6. Predicted Probability of Concordance in Dementia Incidence
by Race, Education, Sex, and Year (N=1,161)

Predicted Probability of Concordance in Dementia Incidence

DX >2 years
prior to incident
dementia in HRS
Less than 2 years
time difference
between DX and
incident dementia
in HRS
DX more than 2
years after incident
dementia in HRS
and before
12/31/2008
Died before
12/31/2008
without DX
Survived to
12/31/2008
without DX
Whites
0.256 [0.222,
0.290]
0.395 [0.357, 0.433] 0.093 [0.070, 0.115]
0.163 [0.134,
0.191]
0.094 [0.071,
0.117]
Blacks
0.186 [0.134,
0.237]*
0.301 [0.241,
0.361]*
0.141 [0.096, 0.184]*
0.176 [0.128,
0.225]
0.196 [0.145,
0.248]***
Hispanics
0.153 [0.071,
0.234]*
0.287 [0.184, 0.390] 0.141 [0.068, 0.214]
0.183 [0.094,
0.271]
0.236 [0.144,
0.328]**

Less than
high school
0.214 [0.180,
0.247]
0.333 [0.294, 0.372] 0.127 [0.099, 0.155]
0.163 [0.132,
0.193]
0.163 [0.131,
0.196]
High school
0.197 [0.147,
0.247]
0.388 [0.326, 0.451] 0.100 [0.061, 0.139]
0.208 [0.154,
0.262]
0.107 [0.066,
0.147]*
College and
above
0.331 [0.261,
0.401]**
0.421 [0.348,
0.495]*
0.061 [0.024,
0.098]**
0.137 [0.084,
0.189]
0.050 [0.017,
0.083]***

Male
0.213 [0.172,
0.253]
0.332 [0.285, 0.379] 0.109 [0.078, 0.140]
0.229 [0.187,
0.271]
0.117 [0.085,
0.149]
Female
0.241 [0.205,
0.276]
0.387 [0.346, 0.427] 0.107 [0.081, 0.132]
0.139 [0.111,
0.167]***
0.127 [0.099,
0.156]

Year 2000
0.200 [0.163,
0.238]
0.340 [0.295, 0.385] 0.181 [0.142, 0.219]
0.202 [0.163,
0.241]
0.077 [0.053,
0.101]
Year 2002
0.231 [0.203,
0.259]*
0.367 [0.335, 0.398]
0.107 [0.086,
0.128]***
0.170 [0.146,
0.194]
0.125 [0.101,
0.149]***
Year 2004
0.249 [0.205,
0.292]
0.369 [0.320, 0.412]
0.059 [0.037,
0.081]***
0.133 [0.100,
0.167]**
0.190 [0.148,
0.232]***
Notes: Predicted probabilities of each concordance category are based on estimates from multinomial logistic regression,
adjusting for sex, age group, race, education, marital status, doctor visit during the past two years, and linear time trend; 95%
confidence intervals of predictions are displayed in brackets. Asterisks indicate level of statistical difference across race,
educational attainment, and Sex, with the top group of each variable being reference; for years from 2000 to 2004, test of
statistical difference is performed between this year and the previous year. *** P value <.001, ** P value <.01, and * P value
<.05.

34
Chapter 3
Socio-Economics Status and Survival after Dementia Diagnosis:
Education, Wealth, and Health Differences at Diagnosis
INTRODUCTION
There exist well-documented socio-economic status (SES) gradients in mortality linked to
a variety of behavioral, psychosocial, financial, and environmental factors (Hamad et al 2020;
Petrovic et al 2018; Zajacova & Lawrence 2018; Stringhini et al 2010; Adler & Stewart 2010). As
a multidimensional construct, however, distinct aspects of SES may affect mortality risks through
different pathways. Education, one of the most used measures of SES, is proposed to directly
increase productive and allocative efficiency of health investment thus leading to longer life
(Grossman 1972; Cutler & Lleras-Muney 2010). Financial resources, mainly income and wealth,
likely extend length of life via purchasing goods that promote health, such as nutrition, health care,
etc. (Deaton 2003).

Since education and financial resources are also highly correlated (Card 1999),
they may affect health via the direct effect of each other, reinforcing the need to isolate their
independent effects when examining SES-health relationships.
When it comes to post-diagnosis mortality of persons living with dementia (PLWD),
nuances across SES aspects may differ from those in the general population. In contrast to the
protective effects of education argued by health economists (Grossman 1972), cognitive reserve
hypothesis contends that the highly educated possess cognitive reserve to compensate for
underlying brain atrophy (Stern et al 1995); as a result, education protects against dementia
pathology and when dementia symptoms appear, individuals with higher education or reserve have
more advanced pathology than counterparts, which likely shorten their survival after dementia
diagnosis. Financial resources per se do not entail properties of cognitive reserve, whose impact
35
on post-diagnosis mortality is more consistent with the corresponding impact in the general
population. Considering post-diagnosis survival being an indicator of 1) health and well-being of
PLWD, 2) duration of dementia, and 3) time with care, it is important to disentangle nuanced roles
of education and financial resources, to inform policies for reducing disparities in post-diagnosis
survival and in dementia burden.
The empirical literature suggests differences in the direction and magnitude of impacts on
post-diagnosis survival across SES measures. Studies on education and survival after dementia
diagnosis generally concluded with insignificant effects of education (Piovezan et al 2020;
Contador et al 2017; Meng & D’Arcy 2012; Paradise et al 2009; Llina`s-Regla et al 2008; Guehne
et al 2006; Pavlik et al 2006; Brookmeyer et al 2002), while a few reported reduced (Korhonen et
al 2020) or elevated mortality risk among the highly educated putatively owing to cognitive reserve
(Tom et al 2015; Freels et al 2002; Stern et al 1995). Most studies on the economic aspect of SES
(i.e. income and wealth) found inverse associations with mortality risk (Jitlal et al 2021; Korhonen
et al 2020; Bebe et al 2019; van de Vorst et al 2015); others with small samples reported
insignificant effects (Chen et al 2014; Meng et al 2012; Aneshensel et al 2000). Yet methodological
limitations make these studies insufficient for understanding nuances across SES dimensions and
for guiding policy levers. First and foremost, they frequently omitted factors related to both SES
and mortality. Controlling for comorbidity, one of such factors (Crimmins & Seeman 2004), is not
a standard practice in these studies (e.g. Bebe et al 2019; Russ et al 2013). Neglecting factors that
are differentially correlated with aspects of SES is more problematic. Dementia severity is one
example, which is associated with education according to the cognitive reserve hypothesis but less
so with financial resources. However, only a few assessed dementia severity (Rountree et al 2012;
Agüero-Torres et al 2002; Freels et al 2002; Stern et al 1995) and obtained severity measures at
36
the expense of sample size and representativeness. Other methodological shortcomings include
upward bias arising from mixing up prevalent dementia with incident cases (Chen et al 2020a;
Piovezan et al 2020; Bebe et al 2019; Contador et al 2017; Andersen et al 2010; Pavlik et al 2006;
Agüero-Torres et al 1999), and censoring bias from insufficient follow-up (Ono et al 2021; Guehne
et al 2006; Fritsch et al 2001).
In this study, we addressed these gaps to advance understanding of independent effects of
education and financial resources. We used a broadly representative sample of older Americans
newly diagnosed with dementia and followed them longitudinally for up to 18 years. We separately
analyzed the association of education and financial resources with survival, conditional on health
differences at dementia diagnosis that are correlated with both SES and mortality, including
comorbidity and function/cognition as proxy for dementia severity. We chose household wealth
over income for measuring financial resources for this elderly retired population, to better reflect
their consumption and purchasing power (Braveman et al 2005). We found both highly educated
and high wealth persons had less functional/cognitive impairment and comorbidity burden at
diagnosis, but only ranking in the top quartile of household wealth conferred longer survival after
diagnosis. These findings suggested unlike in the general population, education did not make one
live longer, while wealth-based advantage persisted. Improving quantity and quality of life after
dementia diagnosis for the poor may be particularly promising in achieving health equities among
PLWD.

37
METHODS
Data and sample
This study used data from the Health and Retirement Study (HRS) linked to Medicare
claims from 1991 to 2012. HRS is a nationally representative survey of respondents aged 51 and
older in the United States. Data collected by the HRS include demographics, measures of economic
status, cognition, and health. Medicare claims data include enrollment files, and claims from Part
A (inpatient, skilled nursing facility, hospice, and home health care) and Part B (outpatient care).
The sample is drawn from HRS respondents who agreed to have their Medicare claims
linked (88%). Using Medicare claims, we selected individuals with an incident dementia diagnosis.
To identify incident cases, respondents were continuously enrolled in fee-for-service (FFS)
Medicare during the year of first diagnosis, and the two preceding years to ensure a wash-out
period with no dementia diagnoses. A dementia diagnosis was determined by the International
Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes (see
Supplementary Table 1 for a complete list of dementia diagnostic codes). We also required the
first diagnosis being verified by a second diagnosis in two years or death in one year to rule out
reversible causes of dementia symptoms. This yielded a cohort of individuals with incident
dementia diagnosis during 1993-2012 (N=3,754).
We required respondents with an incident dementia diagnosis to have an HRS interview
within 12 months of the diagnosis and matched them to their closest HRS interview wave
(N=2,625). The mean time difference between incident diagnosis and HRS interview before and
with this restriction is described in Supplementary Table 2. Data on cognition was consistent across
HRS survey waves beginning in 2000 thus we excluded respondents whose matched HRS
interview was prior to 2000. The final sample consisted of 1,873 adults aged 67 and older with an
38
incident dementia diagnosis between 2000 and 2012 and who were followed until December 31th,
2018 or death. The sample selection process is described in Supplementary Table 3.

Measures
The outcome, time from incident dementia diagnosis to death, was measured using the
death date provided in the 2018 public HRS, which is a combination of the National Death Index
and the HRS exit interviews (RAND Center for the Study of Aging, 2016).
Education and wealth are the key predictors of interest. Education was measured in
categories for highest level achieved (less than high school, high school/some college, and
bachelor’s degree or above). Wealth was measured by the location in the distribution of household
wealth among persons with incident dementia diagnosis, with 4
th
quartile being the highest wealth
and 1
st
quartile the lowest. Household wealth was the value of real estate, vehicle, business, liquid
asset (including retirement account, stock, cash, and bond) less all debt for a respondent and her
spouse. We chose to use quartile based on the wealth distribution among the incident sample to
show the rank of wealth in a more homogeneous population.
Functional impairment at diagnosis was measured using HRS questions “because of a
health or memory problem do you have any difficulty with…” five activities of daily living (ADLs,
including bathing, eating, dressing, walking across a room, and getting in/out of bed) and five
instrumental activities of daily living (IADLs, including using a telephone, taking medication,
handling money, shopping, and preparing meals). We used two indices for ADL and IADL
limitations, each of which was the sum of number of ADLs or IADLs a respondent had difficulties
completing and thus ranged from 0 to 5. To increase statistical power in analysis, we collapsed the
two indices into 3-category variables: no difficulty, 1-2 difficulties, and 3-5 difficulties. Cognitive
39
impairment at diagnosis were based on performance of cognitive tests administered at HRS
interviews for those who were able to self-respond, and on proxy-based measures of cognition for
those who were unable to. We then classified respondents as with dementia, cognitively
impairment but no dementia (CIND), and normal based on methods described elsewhere (Langa
et al 2017).
Other covariates were also extracted from HRS, including age, sex, race/ethnicity, self-
reported stroke, diabetes, hypertension, and heart disease at the time of dementia diagnosis.

Statistical Analysis
We described comorbidity, function, and cognition at the time of dementia diagnosis for
different education and wealth levels, and performed Chi-square tests for statistical differences in
their distributions by education and wealth. We used Cox Proportional Hazard models to quantify
the relationship between SES, as measured by education and household wealth- and post-diagnosis
mortality. We sequentially introduced demographics, comorbid conditions, interaction of ADL
and IADL limitations, and cognitive impairment to the Cox models and compared coefficients on
education and wealth across models.
We assessed robustness of results to a shorter length of time between dementia diagnosis
and nearest HRS interview of no more than 12 months prior to incident dementia diagnosis and no
more than 6 months after. Imposing a more restrictive time window after diagnosis may reduce
measurement error of impairment. For instance, the initiation of anti-dementia medications after
diagnosis may relieve symptoms; thus, functional and cognitive impairment measured in the post-
diagnosis period can understate the level of deficits at dementia diagnosis. To mitigate concerns
for differential censoring by SES, we performed another sensitivity check in individuals with
40
incident dementia during 2000-2008 instead of 2000-2012 in the main analysis. This increased the
lower bound of follow-up length from 6 years to 10 years as we followed them to 2018 and reduced
bias due to differential censoring, if existed. Third, we tested alternative functional forms of
impairment, for example ADL and IADL without interaction. Considering IADL being part of
cognitive measure for proxy respondents in HRS, we also removed IADL from the model to
address concerns for correlation between IADL and cognition. Finally, we tested the robustness of
results to alternative measures of comorbidity, by replacing the HRS-based self-reported diagnoses
of chronic conditions with claims-based ones.

RESULTS
Sample Characteristics
Among the 1,873 respondents, mean age at incident diagnosis was 83.16 (Table 1).
Consistent with sex differences in risk of dementia, females were disproportionately represented
in this sample with incident dementia (64.23%). Forty-one percent did not have a high school
diploma, and 14% completed at least college education. Household wealth was highly
heterogeneous: an average individual in the bottom quartile of wealth distribution was in debt
(mean = -$1,208) while a counterpart in the top quartile possessed over $1 million (mean =
$1,040,945). Stroke, hypertension, diabetes, and heart disease were prevalent, with the self-
reported prevalence being 26.84%, 64.65%, 23.58%, and 45.50%, respectively.
When first diagnosed with dementia, 48.26% and 41.60% of the respondents were free of
ADL difficulty and IADL difficulty, respectively, according to self- or proxy- report. Only 39.59%
of respondents with dementia diagnosis were classified as “having dementia” based on HRS
41
cognitive measures, and a significant proportion (30.82%) scored normal in either cognitive tests
or in informant-based scales.
Sample characteristics were similar across choice of time window for matching the nearest
HRS interview (Supplementary Table 4).

Function, Cognition, and Comorbidity by Education and Wealth
There was an inverse relationship between SES and level of impairment within 12 months
of dementia diagnosis. While 40.57% of those who did not finish high school had no difficulty in
all 5 ADLs, the proportion was 62.45% among those with at least a bachelor’s degree (Table 2A).
SES gradients in ADL limitations were even more striking by wealth quartile: in the bottom
quartile, 28.98% of the respondents were free of ADL limitations, whereas 62.79% in the top
quartile had difficulty in none of the 5 ADL tasks (Table 2B). The same pattern was observed for
IADLs, but to a lesser extent (No IADL difficulty less than high school = 33.72% v. No IADL difficulty
BA and above = 53.08%; No IADL difficulty bottom quartile = 29.42% v. No IADL difficulty top quartile =
50.42%). About 45% of the highly educated and 23% of the lower educated scored normal using
HRS cognitive measures. In the top wealth quartile, 41% were classified as normal cognition; in
contrast, the portion in the bottom quartile was 21%. All these differences were statistically
significant. In summary, higher education and wealth group exhibited less impairment in function
and cognition than the lower group when first diagnosed with dementia.
In terms of comorbid conditions at first diagnosis, higher education was associated with
significantly lower prevalence of stroke, diabetes, hypertension, and heart disease (Table 2A). The
pattern in general held for wealth, except for the insignificant difference in the distribution of heart
disease by wealth quartile (p=0.077, Table 2B).
42
Length of Survival Time and Adjusted Mortality Risks
The mean of unadjusted survival after diagnosis was 4.1 years (1,504 days), and the median
was 3.1 years (1,145 days) (Table 3). On average, an individual with at least a bachelor’s degree
had similar length of post-diagnosis survival as someone with less than high school education
(1603.6 v. 1439.8 days, p=0.086). An average individual in the top wealth quartile outlived
someone in the bottom quartile by 451 days (p<0.000).
Modeling results for mortality are shown in Table 4. Consistent with the unadjusted pattern,
univariate models (Model 1 and 2 for education and wealth, respectively) showed insignificant
education difference (HR high school= 0.905, 95% CI: 0.816-1.003; HR BA+= 0.864, 95% CI: 0.744-
1.002) and significant wealth gradients in survival (HR 2nd quartile = 0.832, 95% CI: 0.728-0.951;
HR 3rd quartile = 0.809, 95% CI: 0.706-0.928; HR top quartile= 0.694, 95% CI: 0.605-0.795). In Model
3 where both (and only) education and wealth were included, the effect of wealth slightly
diminished, which was consistent with the weak correlation between years of education and log
household wealth (r=0.419). After controlling for sex, age at diagnosis, race/ethnicity and calendar
year of diagnosis (Model 4), education remained insignificant (HR high school= 0.947, 95% CI: 0.844-
1.063; HR BA+= 0.959, 95% CI: 0.811-1.113), and the wealth effects were all mitigated (HR 2nd
quartile = 0.881, 95% CI: 0.768-1.010; HR 3rd quartile = 0.862, 95% CI: 0.746-0.997; HR top quartile =
0.735, 95% CI: 0.629-0.858).
In Model 5, we introduced indicators of comorbid conditions at dementia diagnosis.
Differences between the 2nd and bottom quartiles was fully absorbed. The survival advantage of
ranking in the 3
rd
and top quartiles shrank (HR 3rd quartile = 0.862, 95% CI: 0.746-0.997; HR top quartile
= 0.735, 95% CI: 0.629-0.858). When cognitive status and the interaction of ADL and IADL
(Model 6) were accounted for, only ranking in the top quartile was associated with a lower
43
mortality risk (HR top quartile = 0.832, 95%CI: 0.709-0.976), and its risk reduction decreased from
26.5% in the absence of impairment to 16.8%. ADL appeared to play a more important role in
determining survival than IADL did: with intact ADL function, even those with 3-5 IADL
limitations had similar mortality as the least impaired (HR=1.212, 95%CI: 0.941-1.560);
irrespective of IADL impairment, those with 3-5 ADL limitations always had 1.5-1.7 times higher
mortality risk than the reference group. Cognition exhibited a protective effect (HR CIND = 0.818,
95%CI: 0.708-0.946; HR normal = 0.803, 95%CI: 0.684-0.941).
All results were robust to using a more restricted matching window [-12m, 6m]
(Supplementary Table 5), excluding those with dementia diagnosis during 2009-2012 and hence a
shorter follow-up (Supplementary Table 6), removing IADL or the interaction of ADL and IADL
(Supplementary Table 7), and using claim-based comorbid conditions (Supplementary Table 8).

DISCUSSION
Following a broadly representative cohort of older Americans with incident dementia for
up to 18 years and controlling for comorbidity, cognition, and function when first diagnosed, we
demonstrated that those ranking in the top quartile of household wealth had a 17% reduction in
mortality risk compared to the lowest quartile while there was no difference for the middle two
wealth quartiles. Education level was not associated with survival after dementia diagnosis with
or without adjustment for demographic and health differences. The null effects of education and
the beneficial effects of wealth were in accordance with the main body of literature on SES
differences in survival after dementia diagnosis (Piovezan et al 2020; Contador et al 2017; Llina`s-
Regla et al 2008; Guehne et al 2006; Pavlik et al 2006; Brookmeyer et al 2002; Jitlal et al 2021;
Korhonen et al 2020; Bebe et al 2019; van de Vorst et al 2015).
44
Persons with higher education and wealth had more intact function and cognition and fewer
comorbidities when first diagnosed, compared to their counterparts. This observation has potential
policy implications. First, consistent with earlier work using Danish registry (Petersen et al 2021)
and a small Canadian clinic sample (Qian et al 2014), it may suggest the both high education and
wealth groups receive earlier dementia diagnoses, due to better awareness of own symptoms and
of dementia (Lian et al 2017; Brooker et al 2014; Bradford et al 2009) and more resources for
seeking diagnostic care (Petersen et al 2021; Balogh et al 2015). Interventions for early detection
of dementia should be targeted at the low-SES group. Second, function and cognition indicate
degree of disability and amount of care needed. As such, the finding on SES differences in
impairment highlighted the importance of supporting care provision to the low-SES group, who
already required greater level of care when first diagnosed with dementia and meanwhile possessed
fewer resources.
The observed inverse relationship between SES and level of impairment also lined up with
life table studies reporting compression of life years with cognitive impairment among the highly
educated (Cha et al 2021; Crimmins et al 2018; Lièvre et al 2008).
We then examined impacts of education and wealth on post-diagnosis survival independent
of impairment and comorbidity at diagnosis. The head start in function and cognition of the highly
educated was not translated into longer post-diagnosis survival relative to the lower educated, for
which the cognitive reserve hypothesis can be a possible explanation. According to the hypothesis,
those with high education have more reserve to mitigate clinical manifestations of dementia and
to postpone diagnosis to later dementia stage. Thus, the highly educated may have greater
neuropathological severity at diagnosis and/or faster decline after diagnosis, despite of better
function and cognition when first diagnosed (Stern et al 1995). This hypothesis can only be
45
explicitly tested with data on brain pathology or clinical stage of dementia. Additionally, the fact
that the low-educated lived similar length of life after dementia diagnosis but in worse function
and cognition suggested they bear a disproportionate burden from dementia.
Estimates of wealth effect on survival were sensitive to the inclusion of comorbidity and
impairment at diagnosis in the regression models. While these health differences partly explained
wealth effects, the most affluent group had a robust risk reduction in mortality than the most
deprived, exhibiting a threshold fashion of relationship rather than gradients. After controlling for
the presence of comorbid conditions, there could be remaining differences in underlying health
that drove the observed survival advantage among the rich. For instance, resource-related
disparities in quality of disease management exist among the general population due to differential
access to medical care, food, transportation (Hamad et al 2020; Syed et al 2013), and can
exacerbate among PLWD with frailty and disabilities. Differential health and mortality risk may
also arise from disadvantages accumulated over the life cycle, such as allostatic load (Castagne et
al 2018; Duru et al 2012).
There are several limitations to the study. To capture complete records of diagnoses, the
sample was restricted to Medicare FFS beneficiaries and excluded the Medicare Advantage
population, which has recently grown to account for 42% of Medicare population (Freed et al
2021). However, this restriction had limited impact on the representativeness of the sample, as
FFS enrollees represented roughly 80% of older Americans during the incidence period 2000-2012
(McGuire et al 2011). In addition, regarding the interpretation of differences in diagnosis timing,
functional and cognitive symptoms are not perfect measure of clinical severity or stage of dementia.
In the absence of clinical-pathological data, we employed these two heavily weighted components
in widely used clinical instruments for staging dementia, e.g. Clinical Dementia Rating (CDR).
46
Third, all covariates including impairment were measured at the timing of incident diagnosis.
Future studies can incorporate time-varying measures for insights into origins and trajectories of
SES disparities. Forth, considering the low-SES may be under-diagnosed due to inferior access to
care, we might have selected a subgroup of them with better access to care, introducing downward
bias of the disparities. Finally regarding wealth measures, this work focused on the total net asset
value of an elderly couple, thereby excluding the contribution of their children to the financial
resources available to them. Future work may also examine differential effects by type of assets,
e.g. housing, liquid asset, transportation, to further unpack the nature of wealth-related advantage.
To sum up, in this population-based study, we found highly educated and high wealth
persons exhibited fewer functional and cognitive limitations as well as comorbidities at incident
diagnosis, but only wealth conferred longer survival. These findings suggested survival disparities
by financial resources and highlighted the importance of achieving equities via improving the well-
being of the deprived. To inform such interventions, more research is needed to understand the
mechanisms underlying survival differences by wealth, as well as the nuance of wealth and
education, especially placing them along the continuum of dementia care.
47
TABLES & FIGURES

Table 1. Sample Characteristics

N 1,873
Race

Whites 76.40%
Blacks 15.27%
Hispanics 6.83%
Other races 1.49%
Age at Diagnosis (yr)

67-69 4.11%
70-74 11.80%
75-79 17.83%
80-84 24.29%
85+ 41.96%
Mean(sd) 83.16(7.35)
Female 64.23%
Education

< High school 41.32%
High school/Some college 44.69%
>= Bachelor’s degree 13.99%
Mean (sd) Household Wealth ($)

All 319,771 (1,004,648)
Bottom (1
st
) quartile -1,208 (20,980)
2nd quartile 39,836 (24,593)
3rd quartile 176,838 (64,277)
Top (4
th
) quartile 1,040,945 (1,790,359)
Comorbid Conditions

Stroke 26.84%
Hypertension 65.65%
Diabetes 23.58%
Heart disease 45.50%
ADL

No difficulty
48.26%
1-2 difficulties
25.68%
3-5 difficulties
26.06%
IADL

No difficulty
41.60%
1-2 difficulties
25.70%
3-5 difficulties
32.71%
Cognition

Dementia
39.59%
CIND
29.59%
Normal
30.82%
Proxy-Reported
28.24%
Notes: Wealth quartile was based on the relative position in this sample with incident dementia diagnosis.

48
Table 2A. Health Differences at Diagnosis by Education

< High school High school/Some college >= BA P-Value
N 774 837 262

ADL

p<0.000
No difficulty
40.57% 50.96% 62.45%

1-2 difficulties
26.87% 27.58% 16.09%

3-5 difficulties
32.56% 21.46% 21.46%

IADL

p<0.000
No difficulty
33.72% 45.32% 53.08%

1-2 difficulties
25.32% 26.86% 23.08%

3-5 difficulties
40.96% 27.82% 23.85%

Cognition State (self+proxy)

p<0.000
Dementia
52.58% 32.34% 24.23%

CIND
24.81% 33.77% 30.38%

Normal
22.61% 33.89% 45.38%

Stroke 28.68% 27.48% 19.08% p=0.008
Hypertension
69.90% 63.80% 58.02% p=0.001
Diabetes
29.33% 20.79% 15.27%
p<0.000
Heart disease
49.48% 43.85% 38.17% p=0.003
Notes: P-values indicate level of significant differences in distributions of all health characteristics by education categories.

Table 2B. Health Differences at Diagnosis by Wealth Quartile

Bottom quartile 2nd quartile 3rd quartile Top quartile P-Value
N 454 496 441 482

ADL

p<0.000
No difficulty
28.98% 47.38% 53.18% 62.79%

1-2 difficulties
31.64% 25.60% 25.91% 19.96%

3-5 difficulties
39.38% 27.02% 20.91% 17.26%

IADL

p<0.000
No difficulty
29.42% 38.71% 47.73% 50.42%

1-2 difficulties
26.33% 27.22% 23.64% 25.42%

3-5 difficulties
44.25% 34.07% 28.64% 24.17%

Cognition State (self+proxy)

p<0.000
Dementia
52.32% 42.54% 35.00% 28.75%

CIND
26.49% 30.24% 31.36% 30.21%

Normal
21.19% 27.22% 33.64% 41.04%

Stroke 31.94% 27.22% 23.36% 24.69% p=0.017
Hypertension
69.16% 68.55% 68.48% 56.22%
p<0.000
Diabetes
30.84% 24.80% 20.41% 18.26%
p<0.000
Heart disease
48.46% 47.78% 43.99% 41.29% p=0.077
Notes: P-values indicate level of significant differences in distributions of all health characteristics by wealth quartiles.

49
Table 3. Length of Survival Time (Days)

All Education Household Wealth

< High
school
High school/Some
college
>=
Bachelor’s
degree
1
st

quartile
2nd
quartile
3rd
quartile
4th
quartile
N 1873 774 837 262 454 496 441 482
Mean 1504.1 1439.8 1532.4 1603.6 1266.0 1480.0* 1543.4** 1717.3***
25th
Percentile
440 402 455 481 340 417 497 571
Median 1145 997 1222 1283 794 1086 1306 1431
75th
Percentile
2296 2204 2332 2418 1910 2268.5 2316 2483
Notes: *** p<0.001, ** p<0.01, * p<0.05 indicate difference in means across education and wealth categories (relative to less
than high school and bottom quartile, respectively).

Table 4. Hazard Ratios for Mortality
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Education
Only
Wealth
Only
Edu +
Wealth
Demo-
graphics
Comor-
bidity
Function+
Cog
High school / Some college (r.t. less
than HS) 0.905

0.977 0.906 0.947 1.021

[0.816,
1.003]

[0.877,
1.089]
[0.809,
1.015]
[0.844,
1.063]
[0.907,
1.149]
BA and above 0.864

0.992 0.882 0.959 1.051

[0.744,
1.002]

[0.844,
1.166]
[0.747,
1.040]
[0.811,
1.133]
[0.888,
1.246]
Household wealth 2nd quartile (r.t.
bottom quartile)

0.832** 0.835** 0.859* 0.881 0.947

[0.728,
0.951]
[0.730,
0.955]
[0.750,
0.984]
[0.768,
1.010]
[0.825,
1.088]
Household wealth 3rd quartile

0.809** 0.814** 0.827** 0.862* 0.936

[0.706,
0.928]
[0.707,
0.938]
[0.716,
0.956]
[0.746,
0.997]
[0.808,
1.084]
Household wealth top quartile

0.694*** 0.699*** 0.676*** 0.735*** 0.832*

[0.605,
0.795]
[0.603,
0.810]
[0.580,
0.788]
[0.629,
0.858]
[0.709,
0.976]
Female

0.715*** 0.730*** 0.704***

[0.644,
0.793]
[0.656,
0.812]
[0.632,
0.784]
Age at Diagnosis 70-74 (r.t. 67-69)

1.271 1.254 1.299

[0.942,
1.716]
[0.926,
1.699]
[0.957,
1.763]
Age at Diagnosis 75-79

1.383* 1.333 1.363*

[1.036,
1.845]
[0.997,
1.784]
[1.018,
1.826]
Age at Diagnosis 80-84

1.788*** 1.769*** 1.783***

[1.350,
2.368]
[1.331,
2.353]
[1.338,
2.375]
Age at Diagnosis 85+

2.793*** 2.816*** 2.633***

[2.119,
3.681]
[2.128,
3.725]
[1.984,
3.495]
Blacks (r.t. whites)

0.928 0.921 0.908
50

[0.804,
1.070]
[0.797,
1.066]
[0.785,
1.052]
Hispanics

0.769* 0.803* 0.783*

[0.625,
0.948]
[0.651,
0.991]
[0.634,
0.967]
Stroke

0.874 1.077 0.970

[0.581,
1.313]
[0.963,
1.204]
[0.864,
1.090]
Hypertension when first diagnosed

1.171** 1.157**

[1.051,
1.305]
[1.038,
1.291]
Diabetes when first diagnosed

1.148* 1.138*

[1.018,
1.295]
[1.009,
1.284]
Heart disease when first diagnosed

1.259*** 1.232***

[1.138,
1.393]
[1.113,
1.363]
No ADL Diff # 1-2 IADL Diff (r.t. no #
no)

1.125

[0.943,
1.342]
No ADL Diff # 3-5 IADL Diff

1.212

[0.941,
1.560]
1-2 ADL Diff # no IADL Diff

1.136

[0.923,
1.399]
1-2 ADL Diff # 1-2 IADL Diff

1.479***

[1.240,
1.764]
1-2 ADL Diff # 3-5 IADL Diff

1.192

[0.968,
1.468]
3-5 ADL Diff # no IADL Diff

1.508*

[1.009,
2.255]
3-5 ADL Diff # 1-2 IADL Diff

1.672***

[1.333,
2.097]
3-5 ADL Diff # 3-5 IADL Diff

1.622***

[1.356,
1.939]
CIND (r.t. dementia)

0.818**

[0.708,
0.946]
Normal Cognition

0.803**

[0.684,
0.941]

Observations 1,873 1,873 1,873 1,873 1,873 1,873
Notes: *** p<0.001, ** p<0.01, * p<0.05. 95% CI in brackets.

51
APPENDIX

Supplementary Table 1. ICD-9 diagnostic codes for ascertaining dementia

Disease ICD-9 Codes
Alzheimer's disease 331.0
Pick's disease 331.11
Frontotemporal dementia 331.19
Senile dementia
290.0, 290.20, 290.21, 290.3,
331.2, 797
Presenile dementia 290.10, 290.11, 290.12, 290.13
Vascular dementia 290.40, 290.41, 290.42, 290.43
Dementia classified elsewhere 294.0, 294.10, 294.11, 294.8, 331.7
Unspecified dementia 294.20, 294.21

Supplementary Table 2. Time difference between date of incident diagnosis and the closet
HRS interview date

No Restrictions [-12m, 12m] [-12m, 6m]
N 3745 2625 2139
mean (days) 491.6 174.6 155.9
sd (days) 916.2 101.5 101.2
min (days) 0 0 0
p10 (days) 51 39 31
p25 (days) 124 88 72
p50 (days) 247 171 141
p75 (days) 404 261 236
p90 (days) 844 319 314
max (days) 7039 360 360
Notes: time difference between date of incident dementia diagnosis and the nearest HRS interview date is described in 3
scenarios: 1) no restrictions, 2) restricting the HRS interview to be no more than 12 months before or after the diagnosis date, and
3) restricting the HRS interview to be no more than 12 months before incident dementia diagnosis or no more than 6 months
after.

Supplementary Table 3. Sample Selection

Restriction N
In HRS-Medicare linked data 1991-2012 24,668
With dementia dx verified by another diagnosis or death 4,922
2-year washout period prior to the earliest verified diagnosis (continuous FFS
enrollment + no dementia diagnosis), and FFS enrollment in the year of incident
diagnosis
3,878
Aged at least 67 years when first diagnosed 3,745
Had an HRS interview within 12 months of the first diagnosis 2,625
The matched HRS interview was no earlier than 2000 (HRS Wave 5) 1,873

52
Supplementary Table 4. Sample Characteristics under Different Time Windows

No Restriction [-12m, 12m] [-12m, 6m]
N 2,517 1,873 1,528
Race

Whites 76.88% 76.40% 76.44%
Blacks 14.54% 15.27% 15.71%
Hispanics 6.79% 6.83% 6.35%
Other races 1.79% 1.49% 1.51%
Age at Diagnosis (yr)

67-69 4.29% 4.11% 3.93%
70-74 11.20% 11.80% 11.45%
75-79 17.48% 17.83% 17.87%
80-84 24.47% 24.29% 24.24%
85+ 42.55% 41.96% 42.54%
Mean(sd) 83.34(7.45) 83.16(7.35) 83.30(7.40)
Female 62.77% 64.23% 64.14%
Education

< High school 41.68% 41.32% 41.36%
High school/Some college 44.50% 44.69% 44.83%
>= BA 13.83% 13.99% 13.81%
Mean (sd) Household Wealth ($)

All 337,563 (1,069,642) 319,771 (1,004,648) 326,713 (1,074,672)
Bottom quartile -866 (18,454) -1,208 (20,980) -1,841 (23,913)
2nd quartile 41,326 (24,516) 39,836 (24,593) 39,264 (24,481)
3rd quartile 177,100 (64,682) 176,838 (64,277) 175,441 (65,030)
Top quartile 1,090,248 (1,900,935) 1,040,945 (1,790,359) 1,076,252 (1,934,908)
Comorbid Conditions

Stroke 26.39% 26.84% 25.57%
Hypertension 65.44% 65.65% 65.27%
Diabetes 24.04% 23.58% 23.46%
Heart disease 44.18% 45.50% 44.89%
ADL

No difficulty
49.66% 48.26% 49.44%
1-2 difficulties
25.53% 25.68% 25.97%
3-5 difficulties
24.81% 26.06% 24.59%
IADL

No difficulty
43.72% 41.60% 44.55%
1-2 difficulties
25.51% 25.70% 25.66%
3-5 difficulties
30.77% 32.71% 29.79%
Cognition

Dementia
38.60% 39.59% 37.31%
CIND
28.49% 29.59% 30.43%
Normal 34.22% 30.82% 32.26%
Proxy-Reported
27.37% 28.24% 25.52%
Notes: All three groups were restricted to matched HRS interview from Wave 5 and on.

53
Supplementary Table 5. Hazard ratios for mortality, using [-12m, 6m] matching window

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7

Educatio
n Only
Wealth
Only
Edu +
Wealth Base All
ADL #
IADL
Cogniti
on
ADL #
IADL + Cog
High school / Some college (r.t.
less than HS) 0.935

1.024 1.046 1.102 1.122 1.139

[0.834,
1.047]

[0.908,
1.155]
[0.920,
1.190]
[0.968,
1.255]
[0.983,
1.280]
[0.998,
1.300]
BA and above 0.845*

0.993 1.026 1.098 1.126 1.140

[0.716,
0.997]

[0.829,
1.190]
[0.851,
1.236]
[0.910,
1.324]
[0.932,
1.361]
[0.943,
1.378]
Household wealth 2nd quartile
(r.t. bottom quartile)

0.815** 0.812** 0.842* 0.900 0.869 0.905

[0.703,
0.945]
[0.700,
0.943]
[0.723,
0.981]
[0.771,
1.050]
[0.746,
1.013]
[0.775,
1.056]
Household wealth 3rd quartile

0.821* 0.818* 0.852 0.922 0.870 0.922

[0.706,
0.956]
[0.700,
0.957]
[0.726,
1.000]
[0.784,
1.085]
[0.741,
1.022]
[0.784,
1.085]
Household wealth top quartile

0.667**
* 0.664***
0.699**
* 0.782**
0.729**
* 0.785**

[0.573,
0.776]
[0.563,
0.784]
[0.587,
0.832]
[0.655,
0.935]
[0.612,
0.869]
[0.657,
0.939]
Female

0.727**
*
0.707**
*
0.715**
* 0.705***

[0.646,
0.819]
[0.627,
0.797]
[0.635,
0.805]
[0.625,
0.795]
Age at Diagnosis 70-74 (r.t. 67-
69)

1.077 1.035 1.104 1.053

[0.769,
1.507]
[0.739,
1.450]
[0.788,
1.547]
[0.751,
1.476]
Age at Diagnosis 75-79

1.138 1.106 1.139 1.106

[0.825,
1.569]
[0.802,
1.526]
[0.826,
1.571]
[0.802,
1.526]
Age at Diagnosis 80-84

1.536** 1.489* 1.546** 1.506*

[1.121,
2.105]
[1.086,
2.042]
[1.127,
2.121]
[1.098,
2.067]
Age at Diagnosis 85+

2.518**
*
2.250**
*
2.463**
* 2.274***

[1.849,
3.430]
[1.648,
3.073]
[1.807,
3.358]
[1.665,
3.106]
Blacks (r.t. whites)

0.987 0.972 0.948 0.952

[0.841,
1.158]
[0.828,
1.142]
[0.807,
1.114]
[0.809,
1.119]
Hispanics

0.906 0.849 0.906 0.854

[0.714,
1.148]
[0.669,
1.078]
[0.715,
1.149]
[0.673,
1.084]
Stroke when first diagnosed

1.074 0.965 1.024 0.965

[0.947,
1.217]
[0.847,
1.100]
[0.901,
1.163]
[0.846,
1.100]
Hypertension when first
diagnosed

1.202** 1.156* 1.205** 1.166*

[1.066,
1.355]
[1.024,
1.304]
[1.068,
1.359]
[1.032,
1.316]
Diabetes when first diagnosed

1.170* 1.170* 1.176* 1.170*

[1.024,
1.338]
[1.023,
1.338]
[1.029,
1.345]
[1.023,
1.338]
Heart disease when first
diagnosed

1.282**
*
1.238**
*
1.282**
* 1.245***

[1.146,
1.434]
[1.106,
1.386]
[1.146,
1.434]
[1.111,
1.394]
54
No ADL Diff # 1-2 IADL Diff
(r.t. no # no)

1.237*

1.234*

[1.014,
1.508]

[1.012,
1.504]
No ADL Diff # 3-5 IADL Diff

1.327*

1.188

[1.005,
1.753]

[0.889,
1.588]
1-2 ADL Diff # no IADL Diff

1.150

1.167

[0.922,
1.434]

[0.936,
1.456]
1-2 ADL Diff # 1-2 IADL Diff

1.498**
*

1.504***

[1.235,
1.816]

[1.238,
1.827]
1-2 ADL Diff # 3-5 IADL Diff

1.310*

1.155

[1.064,
1.614]

[0.920,
1.450]
3-5 ADL Diff # no IADL Diff

1.782*

1.796*

[1.116,
2.846]

[1.124,
2.869]
3-5 ADL Diff # 1-2 IADL Diff

1.547**
*

1.521***

[1.219,
1.963]

[1.196,
1.935]
3-5 ADL Diff # 3-5 IADL Diff

1.971**
*

1.720***

[1.654,
2.349]

[1.409,
2.101]
CIND (r.t. dementia)

0.751**
* 0.805**

[0.655,
0.861]
[0.687,
0.943]
Normal Cognition

0.704**
* 0.808*

[0.609,
0.814]
[0.679,
0.963]

Observations 1,528 1,528 1,528 1,528 1,528 1,528 1,528
Notes: 95% CI in brackets. *** p<0.001, ** p<0.01, * p<0.05.

55
Supplementary Table 6. Hazard ratios for mortality,
restricting to incident diagnosis during wave 5-9 (2000-2008)

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7

Educatio
n Only
Wealth
Only
Edu +
Wealth Base All
ADL #
IADL
Cogniti
on
ADL #
IADL + Cog
High school / Some college (r.t.
less than HS) 0.931

1.007 0.999 1.051 1.057 1.065

[0.832,
1.042]

[0.895,
1.133]
[0.881,
1.133]
[0.925,
1.195]
[0.929,
1.202]
[0.936,
1.212]
BA and above 0.888

1.029 1.008 1.066 1.103 1.095

[0.755,
1.045]

[0.861,
1.230]
[0.839,
1.213]
[0.885,
1.283]
[0.915,
1.330]
[0.908,
1.320]
Household wealth 2nd quartile
(r.t. bottom quartile)

0.793** 0.792** 0.836* 0.900 0.865 0.903

[0.686,
0.917]
[0.684,
0.916]
[0.721,
0.971]
[0.775,
1.047]
[0.745,
1.005]
[0.777,
1.050]
Household wealth 3rd quartile

0.804** 0.800** 0.834* 0.911 0.859 0.911

[0.693,
0.932]
[0.687,
0.932]
[0.713,
0.975]
[0.777,
1.067]
[0.734,
1.005]
[0.777,
1.067]
Household wealth top quartile

0.689**
* 0.682***
0.706**
* 0.795**
0.742**
* 0.802*

[0.594,
0.798]
[0.580,
0.801]
[0.596,
0.836]
[0.668,
0.946]
[0.625,
0.879]
[0.674,
0.955]
Female

0.712**
*
0.685**
*
0.701**
* 0.683***

[0.634,
0.801]
[0.609,
0.771]
[0.623,
0.788]
[0.606,
0.769]
Age at Diagnosis 70-74 (r.t. 67-
69)

1.269 1.287 1.332 1.308

[0.920,
1.750]
[0.932,
1.778]
[0.965,
1.839]
[0.947,
1.808]
Age at Diagnosis 75-79

1.360 1.404* 1.391* 1.406*

[1.000,
1.850]
[1.031,
1.912]
[1.022,
1.893]
[1.032,
1.915]
Age at Diagnosis 80-84

1.712**
*
1.739**
*
1.732**
* 1.736***

[1.267,
2.313]
[1.285,
2.354]
[1.281,
2.342]
[1.282,
2.351]
Age at Diagnosis 85+

2.837**
*
2.688**
*
2.774**
* 2.683***

[2.111,
3.811]
[1.996,
3.621]
[2.064,
3.728]
[1.992,
3.614]
Blacks (r.t. whites)

0.916 0.927 0.886 0.910

[0.780,
1.074]
[0.789,
1.089]
[0.755,
1.041]
[0.774,
1.070]
Hispanics

0.830 0.811 0.835 0.809

[0.661,
1.042]
[0.645,
1.019]
[0.665,
1.048]
[0.643,
1.017]
Stroke when first diagnosed

1.012 0.911 0.960 0.908

[0.895,
1.144]
[0.802,
1.035]
[0.848,
1.087]
[0.799,
1.032]
Hypertension when first
diagnosed

1.172** 1.144* 1.185** 1.154*

[1.043,
1.317]
[1.017,
1.287]
[1.054,
1.331]
[1.025,
1.299]
Diabetes when first diagnosed

1.184* 1.174* 1.197** 1.180*

[1.037,
1.352]
[1.028,
1.341]
[1.048,
1.366]
[1.033,
1.348]
Heart disease when first
diagnosed

1.277**
*
1.251**
*
1.275**
* 1.253***
56

[1.144,
1.425]
[1.120,
1.397]
[1.142,
1.423]
[1.122,
1.400]
No ADL Diff # 1-2 IADL Diff
(r.t. no # no)

1.133

1.121

[0.934,
1.375]

[0.924,
1.360]
No ADL Diff # 3-5 IADL Diff

1.348*

1.205

[1.030,
1.762]

[0.905,
1.605]
1-2 ADL Diff # no IADL Diff

1.102

1.116

[0.880,
1.380]

[0.891,
1.399]
1-2 ADL Diff # 1-2 IADL Diff

1.407**
*

1.388***

[1.164,
1.701]

[1.147,
1.681]
1-2 ADL Diff # 3-5 IADL Diff

1.403**

1.250

[1.146,
1.718]

[0.996,
1.569]
3-5 ADL Diff # no IADL Diff

1.621*

1.571*

[1.038,
2.532]

[1.005,
2.456]
3-5 ADL Diff # 1-2 IADL Diff

1.632**
*

1.586***

[1.265,
2.107]

[1.227,
2.050]
3-5 ADL Diff # 3-5 IADL Diff

1.853**
*

1.638***

[1.566,
2.192]

[1.338,
2.005]
CIND (r.t. dementia)

0.792**
* 0.894

[0.692,
0.906]
[0.758,
1.055]
Normal Cognition

0.679**
* 0.810*

[0.589,
0.783]
[0.677,
0.969]

Observations 1,502 1,502 1,502 1,502 1,502 1,502 1,502
Notes: 95% CI in brackets. *** p<0.001, ** p<0.01, * p<0.05.

57
Supplementary Table 7. Hazard ratios for mortality,
no interaction of ADL and IADL, and removing IADL

No Interaction of ADL
and IADL
No IADL

High school / Some college (r.t. less than HS) 1.018 1.008

[0.905, 1.145] [0.896, 1.134]
BA and above 1.049 1.041

[0.886, 1.243] [0.879, 1.232]
Household wealth 2nd quartile (r.t. bottom quartile) 0.946 0.953

[0.824, 1.087] [0.830, 1.094]
Household wealth 3rd quartile 0.938 0.937

[0.810, 1.086] [0.809, 1.085]
Household wealth top quartile 0.832* 0.842*

[0.709, 0.976] [0.718, 0.988]
Female 0.703*** 0.706***

[0.632, 0.783] [0.634, 0.786]
Age at Diagnosis 70-74 (r.t. 67-69) 1.283 1.285

[0.946, 1.741] [0.948, 1.743]
Age at Diagnosis 75-79 1.355* 1.345*

[1.012, 1.814] [1.005, 1.800]
Age at Diagnosis 80-84 1.765*** 1.754***

[1.325, 2.351] [1.318, 2.335]
Age at Diagnosis 85+ 2.594*** 2.592***

[1.956, 3.439] [1.956, 3.435]
Blacks (r.t. whites) 0.907 0.907

[0.784, 1.050] [0.784, 1.050]
Hispanics 0.783* 0.789*

[0.634, 0.967] [0.639, 0.974]
Stroke when first diagnosed 0.968 0.967

[0.862, 1.086] [0.862, 1.085]
Hypertension when first diagnosed 1.159** 1.161**

[1.040, 1.292] [1.042, 1.295]
Diabetes when first diagnosed 1.140* 1.141*

[1.011, 1.286] [1.011, 1.287]
Heart disease when first diagnosed 1.232*** 1.247***

[1.113, 1.363] [1.127, 1.379]
1-2 ADL Diff (r.t. no) 1.160* 1.219**

[1.018, 1.322] [1.078, 1.378]
3-5 ADL Diff 1.467*** 1.551***

[1.259, 1.708] [1.353, 1.778]
1-2 IADL Diff (r.t. no) 1.187*

58

[1.040, 1.354]

3-5 IADL Diff 1.105

[0.931, 1.311]

CIND (r.t. dementia)
0.824** 0.823**

[0.713, 0.952] [0.726, 0.935]
Normal Cognition 0.806** 0.796**

[0.687, 0.945] [0.694, 0.914]

Observations 1,873 1,873
Notes: 95% CI in brackets. *** p<0.001, ** p<0.01, * p<0.05.

59
Supplementary Table 8. Hazard ratios for mortality, using claim-ascertained comorbidities
(only showing results with health controls)

Model 4 Model 5 Model 6 Model 7 Model 8
Base All
ADL #
IADL
Cognitio
n
ADL #
IADL +
Cog No IADL
High school / Some college (r.t.
less than HS) 0.934 0.985 0.990 1.007 0.990

[0.833,
1.048]
[0.877,
1.106]
[0.881,
1.112]
[0.895,
1.132]
[0.881,
1.113]
BA and above 0.945 1.003 1.031 1.036 1.023

[0.800,
1.117]
[0.849,
1.186]
[0.871,
1.221]
[0.876,
1.227]
[0.864,
1.211]
Household wealth 2nd quartile
(r.t. bottom quartile) 0.877 0.934 0.909 0.940 0.948

[0.765,
1.005]
[0.814,
1.072]
[0.793,
1.042]
[0.820,
1.079]
[0.827,
1.088]
Household wealth 3rd quartile 0.854* 0.933 0.883 0.935 0.937

[0.739,
0.987]
[0.806,
1.081]
[0.764,
1.021]
[0.807,
1.083]
[0.810,
1.085]
Household wealth top quartile 0.723*** 0.812* 0.760*** 0.818* 0.831*

[0.619,
0.844]
[0.693,
0.953]
[0.650,
0.888]
[0.697,
0.960]
[0.709,
0.974]
Female 0.735*** 0.715*** 0.726*** 0.714*** 0.716***

[0.661,
0.817]
[0.643,
0.796]
[0.653,
0.807]
[0.642,
0.795]
[0.644,
0.796]
Age at Diagnosis 70-74 (r.t. 67-69) 1.221 1.253 1.279 1.273 1.258

[0.903,
1.651]
[0.925,
1.698]
[0.945,
1.732]
[0.939,
1.726]
[0.929,
1.704]
Age at Diagnosis 75-79 1.283 1.320 1.299 1.317 1.293

[0.959,
1.716]
[0.986,
1.768]
[0.971,
1.738]
[0.983,
1.764]
[0.966,
1.730]
Age at Diagnosis 80-84 1.653*** 1.673*** 1.672*** 1.683*** 1.650***

[1.245,
2.195]
[1.258,
2.225]
[1.258,
2.222]
[1.265,
2.240]
[1.241,
2.194]
Age at Diagnosis 85+ 2.644*** 2.490*** 2.586*** 2.500*** 2.446***

[2.000,
3.496]
[1.879,
3.300]
[1.955,
3.421]
[1.886,
3.314]
[1.847,
3.239]
Blacks (r.t. whites) 0.935 0.941 0.907 0.923 0.922

[0.809,
1.079]
[0.815,
1.088]
[0.785,
1.048]
[0.799,
1.067]
[0.798,
1.066]
Hispanics 0.767* 0.751** 0.773* 0.749** 0.753**

[0.621,
0.947]
[0.608,
0.927]
[0.626,
0.954]
[0.606,
0.926]
[0.609,
0.929]
Stroke when first diagnosed 1.148** 1.097 1.132* 1.100 1.097

[1.038,
1.270]
[0.991,
1.216]
[1.023,
1.252]
[0.993,
1.219]
[0.991,
1.215]
Hypertension when first
diagnosed 0.991 1.001 1.007 1.000 1.006

[0.835,
1.175]
[0.844,
1.189]
[0.848,
1.195]
[0.843,
1.188]
[0.847,
1.194]
Diabetes when first diagnosed 1.097 1.088 1.114* 1.097 1.098

[0.989,
1.217]
[0.981,
1.208]
[1.004,
1.236]
[0.988,
1.217]
[0.989,
1.218]
Heart disease when first
diagnosed 1.397*** 1.345*** 1.391*** 1.355*** 1.363***
60

[1.216,
1.606]
[1.170,
1.546]
[1.210,
1.600]
[1.178,
1.558]
[1.185,
1.567]
1-2 ADL Diff (r.t. no)

1.201**

[1.062,
1.357]
3-5 ADL Diff

1.538***

[1.343,
1.760]
No ADL Diff # 1-2 IADL Diff (r.t.
no # no)

1.157

1.149

[0.971,
1.379]

[0.964,
1.370]

No ADL Diff # 3-5 IADL Diff

1.357*

1.198

[1.068,
1.725]

[0.931,
1.544]

1-2 ADL Diff # no IADL Diff

1.098

1.107

[0.892,
1.350]

[0.900,
1.362]

1-2 ADL Diff # 1-2 IADL Diff

1.484***

1.489***

[1.246,
1.767]

[1.249,
1.775]

1-2 ADL Diff # 3-5 IADL Diff

1.337**

1.167

[1.110,
1.610]

[0.950,
1.434]

3-5 ADL Diff # no IADL Diff

1.530*

1.504*

[1.025,
2.284]

[1.007,
2.245]

3-5 ADL Diff # 1-2 IADL Diff

1.710***

1.668***

[1.366,
2.141]

[1.330,
2.093]

3-5 ADL Diff # 3-5 IADL Diff

1.845***

1.602***

[1.589,
2.143]

[1.345,
1.908]

CIND (r.t. dementia)

0.753*** 0.819** 0.830**

[0.667,
0.850]
[0.709,
0.946]
[0.732,
0.942]
Normal Cognition

0.694*** 0.803** 0.801**

[0.610,
0.789]
[0.685,
0.941]
[0.698,
0.918]

Observations 1,873 1,873 1,873 1,873 1,873
Notes: 95% CI in brackets. *** p<0.001, ** p<0.01, * p<0.05.

61
Chapter 4
Do Individuals Skimp on Their Own Health Care
after Spouse’s Dementia Diagnosis?
INTRODUCTION
Dementia places substantial burden on individuals and families. In the U.S., 85% of
persons living with dementia (PLWD) live in community as opposed to in residential facilities
(Chi et al 2019; Weir & Langa 2019). Over 90 percent of these non-institutionalized PLWD receive
help from family caregivers, with their spouse being a major source of caregiving (Kasper et al.
2016; Zissimopoulos et al. 2015). Impacts of dementia on family members are a significant
component of dementia burden.
Family members of PLWD incur enormous economic costs, generally recognized as the
value of caregiving time (Coe et al. 2018; Zissimopoulos et al. 2015; Hurd et al. 2013; Lin &
Neumann 2013), the annual costs of which amount to $126 billion in the U.S. (Zissimopoulos et
al 2015). Restricting scopes to caregiving time may understate the full impact of dementia on
families. First, it can underestimate the number of affected subjects, if dementia affects family
members who are not caregivers. Second, if impacts of dementia or caregiving accrue to aspects
other than time, such as health deterioration and long-term productivity loss, it would understate
the per capita cost. Due to shocks in financial resources and time ensuing spouse’s dementia
diagnosis, individuals may ration health care due to inward-shifting budget constraints.
Considering health care as an input for health production, forgone care may jeopardize health and
elevate downstream health care costs in the long run, a plausible pathway for explaining
documented health decline in caregivers/spouses of PLWD (Roth et al 2019; Allen et al 2017). On
the contrary, individuals may demand more health (care) due to increased marginal value of
62
healthy time, either to engage in caregiving tasks or to spend more time with their spouses with
dementia.
Consistent with this theoretical ambiguity, reported associations between being a spouse
of PLWD and health care utilization were mixed. The main body of literature drew cross-sectional
comparisons in outcomes such as doctor visits and health care costs for spouses/caregivers of
PLWD and controls, and found more use (Rahman et al 2019; Goren et al 2016; Suehs et al 2014;
Grä sel et al 2002), less use (Moore et al 2001; Shaw et al 1997), or similar use (Laks et al 2014;
Kolanowski et al 2004) by spouses/caregivers of PLWD. The only longitudinal study (Gilden et
al 2017) compared the Medicare spending of AD spouses with that of controls, who were selected
based on a limited array of characteristics (age, sex, race/ethnicity, and geography). This
longitudinal study and other cross-sectional ones failed to account for systematic differences
between dementia families and controls, partly contributing to conflicting findings. In addition,
none of these studies was representative, using convenience samples with less than 200
spouses/caregivers of PLWD (Rahman et al 2019; Martín-García et al 2017; Mausbach et al 2013;
Baumgmten et al 1997; Shaw et al 1997), or non-representative surveys (Goren et al 2016; Laks
et al 2014; Moore et al 2001; Grä sel et al 2002) or insurance databases (Gilden et al 2017; Suehs
et al 2014; Kolanowski et al 2004).
Beyond low-quality evidence of within-individual variations in health care use after
spouse’s dementia diagnosis, prior studies did not test for heterogeneous responses across
individuals subject to different constraints, return on health investment, and preference. For
instance, caregiving spouses may face greater strains in time schedule and meanwhile value
healthy time more than non-caregiving counterparts. More knowledge on heterogeneity by
63
caregiving would show if policy support such as respite care programs should be extended to non-
caregiving spouses; yet this was not studied previously due to data limitation (Gilden et al 2017).
To address these issues, the present study investigated broader family spillovers of
dementia, specifically the trajectory of health care utilization around spouse’s incident dementia
diagnosis, using the Health and Retirement Study (HRS) linked to Medicare claims, a nationally
representative panel with rich information on family, health(care), and caregiving. Building upon
the representative HRS, my findings had greater generalizability to American families afflicted
with dementia than extant studies. A broad range of health care outcomes were examined,
including preventive, ambulatory, and inpatient care with different cost and well-being
implications, as well as medical spending by various entities, to comprehensively assess dynamics
of health care use among spouses of PLWD. Employing an event-study approach, I improved
estimates of within-individual changes in these utilization outcomes, by including 1) individual
fixed effects (FE) to net out unobservable determinants/influencers of utilization that are time-
invariant, and 2) calendar time fixed effects to absorb time-specific unobservables. For the first
time, I examined whether temporal patterns varied by individual and household characteristics,
including socio-demographics, health, and caregiving roles. These are important contributions to
understanding how diverse individuals respond to their spouses’ incident dementia diagnosis and
to informing targeted interventions for supporting families of PLWD.
During a 5-year period before and after spouse’s dementia diagnosis, I observed stable
trends in the volume of flu shots and office-based evaluation and management (E/M) visits, a slight
increase in use of inpatient care, together with a rise in Medicare, out-of-pocket, and total medical
expenses in years following the event of interest. After adjusting for demographic, socioeconomic,
health characteristics at individual or spousal-dyadic levels in FE models, none of these utilization
64
outcomes deviated significantly from pre-diagnosis levels. This suggested an average person
whose spouse was newly diagnosed with dementia could maintain her care schedule without
substantial changes in out-of-pocket and federal health care spending, in the first few years after
diagnosis. While the pattern held for most subgroups, males, non-Hispanic Whites, and those in
the upper half of wealth distribution, reduced E/M visits by 1.61, 0.98, and 1.24, respectively, in
the year of spouse’s diagnosis. The changes represented 12%-20% decrease from years before
diagnosis and were transitory as they became insignificant in the next two years. For the three
subgroups, the reduction in ambulatory care did not elevate the use of costly inpatient care,
suggesting limited immediate harm of it. Racial/ethnic minorities, in contrast, increased use by 3.8
visits in years post diagnosis, equivalent to a 49% rise from baseline level. Over the study period,
there lacked evidence for differential responses by caregiving status.
This paper showed few changes in health care use among older Americans whose spouses
were newly diagnosed with dementia, indicating that short-term health deterioration, if exists, may
not be a result of forgone care. Meanwhile it signaled rationed care by subgroups such as males,
the burden and impacts of which may emerge, sustain, and exacerbate over a longer period.

BACKGROUND
Dementia and Family
Dementia poses financial challenges to patients and their families. Cross-sectionally,
PLWD have remarkably (ranging from 34% to 230%) higher total and out-of-pocket spending on
health care than dementia-free counterparts (Leibson et al 2015; Zhao et al 2008; Bynum et al
2004), driven by intensive use of inpatient and nursing home care (Oremus & Aguilar 2011; Langa
et al 2004). As revealed in longitudinal studies, health care utilization and costs of PLWD start to
65
surge shortly before dementia onset and continue to surpass that of controls (Lin et al 2016; Zhu
et al 2015; Chen et al 2014; Suehs et al 2013). Higher severity of dementia is associated with
greater formal care costs (Cantarero-Prieto et al 2019). Assuming financial assets as a shared
resource within a family, dementia places strains on financial resources available to all family
members.
Existing frameworks for evaluating economic impacts of dementia recognize the
importance to incorporate indirect costs incurred by family members (Coe et al. 2018;
Zissimopoulos et al. 2015; Hurd et al. 2013; Lin & Neumann 2013). They in general monetize the
indirect costs as the value of informal caregiving using replacement cost (i.e. the cost of formal
caregiving if purchased in market) and foregone wage (i.e. caregiver’s income had caregiving time
been used for earning wage). The latter approach may not reflect the value of time spent by older
adults who are not in the labor force. For older adults, opportunity costs of caregiving provision
can accrue to other aspects of life, such as leisure and health production. Failure to include broad
spillovers on family members can underestimate societal costs of a disease as well as benefits of
interventions that confront that disease (Basu & Meltzer 2005).
There is a large body of literature on the health impacts of dementia on family members,
mainly caregivers. Studies consistently report worsened mental health in caregivers (Sheehan et al
2020; Ma et al 2017; Schulz et al 2004) and in spouses irrespective of caregiving status (Chen et
al 2020b). Less consensus is on physical health, with mixed findings on how biomarkers change
following caregiving (Leggett et al 2020; Roth et al 2019; Allen et al 2017; Roth et al 2015;
Fonareva et al 2014). It is also unclear whether health deterioration is associated with changes in
health care use.

66
Theoretical Framework
This section explains why and how in theory individuals, regardless of caregiving roles,
may change health care use after their spouse’s dementia diagnosis.
In Grossman’s model of demand for health (1972), individuals desire health for two
reasons. First, as a consumption commodity, health yields utility directly. Second, as an investment
commodity, it generates healthy time as a return. An individual chooses level of health that
maximizes lifetime utility subject to wealth and time constraints. On the one hand, considering the
notable amount of medical expenses (i.e. direct costs) and caregiving associated with dementia
described earlier, having a spouse with dementia places pressure on budget and time constraints
of an individual. The inward-shifting constraints elevate the marginal cost of producing health,
and thus lower the demand for health. On the other hand, affected individual may value healthy
time more, which can be used to perform caregiving tasks, take over household responsibility, or
spend time with their loved ones given the irreversible nature of dementia. In this case, an
individual demands more health due to higher marginal value of healthy time. This theoretical
ambiguity further complicates the prediction of health care, a derived demand for health, the path
of which also depends on the elasticity of marginal return of health and the depreciation rate of
health stock/investment. Another possibility hinges on the empirical observation of health
deterioration among family members. To compensate for high depreciation, individuals may
demand more health care to maintain the desired level of health stock, even though her demand
for health remains the same.
Subgroups may respond differentially to spouse’s dementia. Below are illustrations of
potential heterogeneity by characteristics.
67
Caregiving. Both caregiving and non-caregiving face time constraints, and the type and
degree of time constraints likely differ. Regardless of caregiving status, spouses of PLWD may
need to take over more household responsibility than previously. Beyond that, caregiving takes
additional time off one’s schedule. Holding all else equal, caregivers face both higher marginal
value of healthy time and greater strains in time available for health investment than non-caregivers.
Moreover, if as shown in the empirical literature caregiving is associated with additional harm in
mental or physical health, then caregivers may have stronger incentive to restore health stock with
health care and other health inputs. I expect to see different trajectories of health care use by
caregiving status, albeit in unknown direction.
Sex. Compared to males, females shoulder greater caregiving responsibility both at
extensive and intensive margins (Kasper et al 2015), and in general use more health care due to
different health perceptions and attitudes (Manuel 2018; Koopmans and Lamers 2007; Bertakis et
al 2000). Differences in time schedule and health beliefs suggest heterogeneity by sex.
Health status. Ehrlich & Chuma (1990) argues that higher health endowment extends the
optimal length of life and thus increase the demand for health. Individuals with better health may
be less likely to skimp on health investment than those with worse health. In terms of health change,
those with accelerated decline may consume more to repair depreciated health stock.
Socio-economic status (SES) proxied by wealth and education. The high-SES individuals
demand more health, through mechanisms that wealth increases consumption capacity and
education increases the efficiency of health production.
1
The group also possess more resources to
alleviate shocks in time constraint following spouse’s dementia diagnosis, by substituting with

1
In Grossman’s model of demand for health, SES may intervene via wage rate too. Individuals with higher wage face greater
opportunity cost of sickness and therefore demand more health. Due to the focus of the current paper on an elderly and retired
population, income heterogeneity is not assessed.
68
other market goods, such as paid caregiving. Thus, relative to the high-SES, disadvantaged
individuals are more likely to ration health care.
Race/ethnicity. Racial/ethnic minorities may demand different levels of health compared
to White counterparts, for reasons beyond their lower SES on average. It is well known that African
Americans and Hispanics have a strong cultural tradition to provide caregiving for family members
(Rote et al 2019; Dilworth-Anderson et al 2002), despite of worse health when become caregivers
than Whites (Chen et al 2020b; Pinquart & Sorensen, 2005). Minorities face barriers to health care
arising from language barriers (Grady et al 2015), discrimination, and systemic racism (Shepherd
et al 2018; Benjamins et al 2014; Paradise et al 2013), and often hold stereotypes against health
care providers and system (Mead et al 2009; Boulware et al 2003). Differential SES, caregiving
responsibilities, health status, and cultural expectation about health care system may yield different
responses by race/ethnicity.
An extra consideration is that heterogeneity may vary by types of health care. Using
education as an example, in addition to Grossman’s argument on education improving the
productive efficiency of the same set of inputs (1972), education can enhance allocative efficiency
in the choice of inputs (Rosenzweig and Schultz 1982; Deaton 2002). Highly educated individuals,
compared to counterparts, may have a stronger tendency to substitute the costly and intense
inpatient care with preventive and ambulatory care. Put differently, one with high educational
attainment may increase the use of preventive and ambulatory services and reduce inpatient care.

Empirical Literature
Knowledge about health care use among family members of PLWD is emerging but still
limited. Studies have examined various health care use outcomes, including doctor visits (Rahman
69
et al 2019; Laks et al 2014; Suehs et al 2014; Moore et al 2001; Baumgmten et al 1997),
hospitalization (Rahman et al 2019; Laks et al 2014; Moore et al 2001; Shaw et al 1997),
emergency room visits (Goren et al 2016; Laks et al 2014; Suehs et al 2014; Kolanowski et al
2004), prescription drug use (Rahman et al 2019; Martín-García et al 2017; Mausbach et al 2013),
and health care expenses (Gilden et al 2017; Suehs et al 2014; Kolanowski et al 2004; Moore et al
2001; Baumgmten et al 1997). The overall quality of evidence is low, mainly because their cross-
sectional comparison of caregivers/spouses of PLWD versus controls failed to account for
underlying differences in outcomes across the two groups. Given widely documented within-
couple correlations of health and health behaviors due to shared environment and belief, etc.
(Manne et al 2012; Di Castelnuovo et al 2008), the two groups of individuals may inherently differ
in care use pattern, for reasons that are unrelated to their spouse’s dementia diagnosis. One early
study (Baumgmten et al 1997) contrasted dementia caregivers to non-dementia caregivers who
might be more homogeneous regarding health and health care use; however, in a cross-sectional
setting, this merely reduced between-group differences rather than fully adjusting for them. The
only longitudinal study (Gilden et al 2017) attempted to overcome the permanent differences in
health care use between AD spouses and controls, by comparing their trends in Medicare spending,
under the implicit assumption that two groups would have the same trajectory in the absence of
spouse’s AD diagnosis. However, the control group was selected based on a limited set of
characteristics, including age, sex, race/ethnicity, and geography; observable determinants or
influencers of health care utilization, such as SES and unobservables such as health belief, were
not considered.
The quality of evidence is further impaired by using convenience samples with less than
200 spouses/caregivers of PLWD (Rahman et al 2019; Martín-García et al 2017; Mausbach et al
70
2013; Baumgmten et al 1997; Shaw et al 1997), non-representative surveys (Goren et al 2016;
Laks et al 2014; Moore et al 2001; Grä sel et al 2002) or partial insurance databases (Gilden et al
2017; Suehs et al 2014; Kolanowski et al 2004). Other than the three studies that measured health
care utilization with insurance claims (Gilden et al 2017; Suehs et al 2014; Kolanowski et al 2004),
all other studies used self-reported health care consumption, which is subject to measurement error.
Prior work also provided inadequate understanding on heterogeneity. Subgroups defined
by socio-demographics, health, and caregiving status may have heterogeneous responses to their
spouse’s dementia diagnosis, for the reasons described in my theoretic framework. No studies have
examined changes in health care use across subpopulations. For instance, they used “caregivers of
PLWD” and “spouses of PLWD” interchangeably and did not test for differential responses among
caregiving- and non-caregiving spouses. Lack of information on caregiving status was a major
obstacle (Gilden et al 2017).

DATA, SAMPLE, AND MEASURES
Data
This study used Health and Retirement Study (HRS) linked to Medicare claims, from 1991
to 2012. HRS is a nationally representative survey that interviews Americans aged over 50 years
old and their spouses biennially since 1992. Respondents are asked about demographics, family
structure, socio-economic status (SES), and health (care). Roughly 88% of HRS respondents
consent for the linkage to their Medicare claims, who are more likely to be younger, richer, and
racial/ethnic minorities (St. Clair et al, 2017; Sakshaug et al 2014; Sala et al 2014). For billing
purpose, fee-for-service (FFS) Medicare claims files include the International Classification of
Disease, Ninth Revision, Clinical Modification (ICD-9-CM) codes for classifying diagnoses and
71
the Current Procedural Terminology (CPT) for health care procedures. I accessed FFS inpatient,
outpatient, skilled nursing facility, hospice, home health, and durable medical equipment claims
files for the current analysis. Beneficiary characteristics, enrollment, and expenditures were
retrieved from annual Master Beneficiary Summary File (MBSF). There were 24,668 unique
respondents in the dataset.

Sample
Using ICD-9-CM codes in Medicare claims, I identified 4,559 individuals with an incident
dementia diagnosis and a subsequent diagnosis, which rules out reversible cognitive decline. The
verification method has been described elsewhere (Thunell et al 2019). Definitions of incident
dementia diagnosis can be found in Supplementary Table 1. Of that sample, 1,471 PLWD reported
having a spouse at an HRS interview wave W, which is set to be within 1 year around the incident
diagnosis.
To be included in the final sample, a spouse of PLWD must further 1) be dementia-free up
to the end of 2012, the latest year of the dataset, 2) age 67+ at PLWD’s incident diagnoses, 3)
continuously enroll in FFS Medicare in a 5-year period (i.e. 2 years prior to, in the year of, and 2
years after PLWD’s incident diagnosis, and 4) respond to three consecutive HRS waves, including
W, W-1, and W+1. These restrictions yielded a sample of 418 spouses of PLWD (2,090 person-
years), whose characteristics and health care use were observable during a 5-year period around
PLWD’s incident dementia diagnosis. Supplementary Table 1 provides details in sample selection.

72
Measures
The outcomes of interest included preventive, ambulatory, and inpatient care, as well as
medical spending in a given year. Preventive and ambulatory care were measured by use of annual
flu shot (binary) and number of outpatient E/M visits (continuous), respectively. Both were
identified by CPT codes listed in Medicare claims (see a full list of CPT codes in Supplementary
Table 2). Use of inpatient care (binary) was indicated by positive inpatient expenses in MBSFs.
Total medical expenses were calculated by adding up payment for inpatient, outpatient, skilled
nursing facility, hospice, home health, and durable medical equipment, incurred by Medicare,
beneficiary, and other payers, recorded in MBSFs. Medicare and out-of-pocket spending were the
sum of Medicare reimburse amount and beneficiary responsibility amount in all the
aforementioned care settings, respectively. I examined Medicare and out-of-pocket spending
separately, as they reflected burden on public insurance programs and on individual beneficiaries,
respectively. All costs were inflation-adjusted to 2010 US dollars (U.S. Bureau of Labor Statistics).
Individual without any use or spending were assigned zero in the corresponding outcomes.
The primary predictors were time dummies that indicated relative years from spouse’s
incident dementia diagnosis, i.e. the index. Such time dummies could take three forms: a) 5 yearly
indicators including 2 years prior to the index, 1 year prior to the index, year of index, 1 year after
the index, and 2 years after the index; b) 3 indicators including years prior to the index, year of
index, and years after the index; c) 2 indicators of years prior to the index and years after the index
(pooling year of index and two subsequent years).
Other variables used in regressions and/or for stratification were extracted from HRS
survey, including time-invariant variables such as sex, race/ethnicity, and educational attainment.
Due to the inconsistency between annual outcomes and biennial HRS interviews, I approximated
73
values of time-varying variables using those collected in the closest interview wave, if no HRS
interview took place in a year of observation.
2
Such variables included quartiles of household
assets and income in all HRS respondents at a given wave
3
, self-rated health status, spouse’s
difficulties in 5 activities of daily living (ADLs)
4
and in 5 instrumental activities of daily living
(IADLs)
5
, and whether being a spousal caregiver for ADLs, IADLs, or either
6
. The only exception
to this approximation was the age group of an individual and her diagnosed spouse, which was
calculated precisely using year of birth and year of observation.

EMPIRICAL STRATEGY
The empirical strategy followed a flexible event-study model (Jacobson et al 1993),
allowing for the association between spouse’s dementia diagnosis and health care utilization to
vary over time. I exploited the panel nature of HRS-Medicare linked data and included individual
fixed effects (FE) to absorb time-invariant unobservables that determined or influenced health care
use. This paper measured an aggregate change in health care use ensuing spouse’s dementia
diagnosis, allowing for unobserved factors to vary over time, as long as their changes were solely
related to the event of interest. Examples of such factors were underlying health, health belief, etc.
Therefore, the assumption was essentially the time invariability of remaining parts of such
unobservables, conditional on other covariates. For instance, I assume an individual would hold

2
Due to the sample restriction of continuous response to HRS wave W-1, W, and W+1, the maximum time difference between a
year of observation without HRS interview and an approximated HRS wave was roughly 1 year.
3
Household asset refers to the value of real estate, vehicle, business, liquid asset (including retirement account, stock, cash, and
bond) less all debt for a respondent and her spouse. We chose to use quartile based on the wave-specific wealth distribution in all
HRS respondents to reflect the rank in the American population.
4
HRS asks questions “because of a health or memory problem do you have any difficulty with…” the following five ADLs,
including bathing, eating, dressing, walking across a room, and getting in/out of bed). Spouse’s ADL limitation is measured by
the sum of number of ADLs a spouse has difficulties completing and is categorized into 0 (none), 1-2 (light limitation), and 3-5
(severe limitation).
5
Similar methods of generating spouse’s IADL limitation measure. The surveyed IADLs are using telephone, taking medication,
managing money, preparing meals, and grocery shopping.
6
Caregiving status is determined by whether an individual’s spouse reported receiving help from her for ADLs, IADLs, or either,
respectively. Caregiving intensity is measured by her spouse’s report of help hours per week provided by her.
74
the same health belief had her spouse not been diagnosed with dementia. Calendar time fixed
effects were also introduced to control for unobserved shocks in outcomes that were common to
all individuals.
The empirical model is as follows:
𝐻𝑒𝑎𝑙𝑡 ℎ𝐶𝑎𝑟𝑒 𝑖𝑡
= ∑ 𝛽 𝑘 ∙ 1(𝑡 − 𝑡 𝑖 ∗
)
𝑖𝑡
𝑘 =−2
𝑘 =2
+ 𝛿 ∙ 𝑋 𝑖𝑡
+ 𝛼 𝑖 + 𝜏 𝑇 + 𝘀 𝑖𝑡

where i indexed an individual, and t a time period (t= -2, -1, 0, 1, 2 in the scenario of 5 yearly
indicators). 𝐻𝑒𝑎𝑙𝑡 ℎ𝐶𝑎𝑟𝑒 𝑖𝑡
was individual i’s health care use at time t. 𝑡 𝑖 ∗
denoted the timing of
spouse’s incident dementia diagnosis or the index. 1(𝑡 − 𝑡 𝑖 ∗
)
𝑖𝑡
equaled to 1 if t = k year(s) relative
to the index. Two years prior to the index (t= -2) was chosen as the reference point, so there were
four yearly dummies in the case of 5 time periods. 𝛽 𝑘 , parameters of interest, represented changes
in outcome over time. 𝑋 𝑖𝑡
was a vector of time-varying characteristics, including age group,
household wealth and income quartiles, spouse's age group, as well as spouse's ADL and IADL
limitations. Individual fixed effects 𝛼 𝑖 , calendar year fixed effects 𝜏 𝑇 , and idiosyncratic error term
𝘀 𝑖𝑡
were also included.
This equation was estimated by linear FE models with robust standard errors clustered at
individual level, in the full sample for receiving flu shot or not, number of E/M visits, using
inpatient care or not, and log-transformed medical expenses to account for the right-skewed
distribution of spending. I also stratified the analysis by individual’s sex, race/ethnicity (non-
Hispanic White v. non-White), education (less than v. above high school), household wealth at
index (bottom or 2
nd
quartiles v. 3
rd
or top quartiles), self-reported health status at index (good v.
poor or fair health), and caregiving status at index (caregiver v. non-caregiver), to see if the
temporal trajectory differed in subpopulations.
75
RESULTS
Sample Characteristics
Table 1 reports characteristics of the study population generated by the sample selection
process described earlier (418 unique persons and 2,090 person-years), in comparison to spouses
of PLWD with only age restriction (845 unique persons and 4,225 person-years). Mean age of the
study sample was 79.31 (SD=6.06), significantly older than the unrestricted group (mean=77.78,
SD=6.38, p<0.001). The distribution of age band, however, was not different between the two
groups (p=0.132). The selected spouses of PLWD were predominantly female (59.57%) and non-
Hispanic White (85.41%). Less than 30% of them did not complete high school. Over half ranked
the 3
rd
(29.11%) or top (34.33%) quartiles of household wealth in the full HRS population. In this
retired population, the majority were in the bottom (15.02%) or 2
nd
(39.28%) quartiles of the
household income distribution. According to self-reported diagnoses of chronic conditions, the
prevalence of stroke, diabetes, hypertension, and heart disease were 9.81%, 16.36%, 56.17%, and
28.80%, respectively. The distribution of these characteristics was similar with and without sample
restrictions. The final study sample had a significantly smaller portion of caregivers than the
unrestricted group, regardless of helping with ADL (18.09% v. 24.53%, p=0.011), IADL (28.52%
v. 39.57%, p<0.001), or either (33.59% v. 45.14%, p<0.001). Conditional on being a caregiver, an
average caregiver in the study sample provided 26.54 (SD=46.11) hours of help per week;
caregiving burden at the intensive margin was not significantly different between the two groups
(p=0.167).

76
Unadjusted Health Care Use
Figure 1 and Supplementary Table 3 show the unadjusted trends in health care utilization
over time. During the 5-year period, use of flu shot was consistently limited, with slightly over
half of older adults being vaccinated at any point in time. The average number of E/M visits was
also stable throughout. The dip in the index year (7.4, 95% CI: 6.8, 8.1) was not different from
pre-index years (p=0.445). The percentage of using inpatient care did not change meaningfully,
until in Year 2 when it reached 22%, a 6-percentage-point increase from Year -2 (15.8%, p=0.019).
Total medical expenses started to surge after the index, increasing by $2,483 (or 52.50%,
p =0.017) in Year 1 from that in Year -2. In Year 2, total medical expenditures almost doubled that
in Year -2 ($5,367 or a 94.31% increase, p<0.001). The expenses paid by Medicare shared the
same trajectory, with an increase of 52.10% in Year 1 (p=0.027) and of 97.78% in Year 2 (p<0.001),
relative to the spending level in Year -2. Graphically, out-of-pocket expenses went up steadily over
time; the increase was not statistically significant until in Years 1 and 2, equivalent to a 45.25%
(p=0.006) and 78.93% (p<0.001) increase from Year -2 level, respectively.

Event Study Models
I started with the four yearly dummies including 1 year prior to the index, year of index, 1
year after the index, and 2 years after the index, and treated 2 years prior to the index as the
reference. The specification with calendar year FE absorbed variations in outcomes in the post-
index periods, leading to the omission of the two post-index yearly indicators. To cope, I pooled
the two pre-years and the two post-years and used 3 time periods, i.e. a) years prior to the index,
b) year of index, and c) years after the index. Years prior to the index was chosen as the reference.
77
Estimates from linear FE models are shown in Table 2 and Figure 2. All models were
adjusted for age group, household wealth and income quartiles, spouse's age group, and spouse's
ADL/IADL limitations. Consistent with the stable trends in Figure 1, the likelihood of receiving a
flu shot did not change in the year of index (0.004, 95% CI: -0.074, 0.083) or in post-index years
(0.032, 95% CI: -0.099, 0.163). The reduction in number of E/M visits was trivial and insignificant
in the year of index (-0.615, 95% CI: -1.434, 0.205) and in the two years to follow (-0.286, 95%
CI: -1.623, 1.051). There was also limited change in the likelihood of using inpatient care (-0.012,
95% CI: -0.089, 0.065 in the index year; 0.003, 95% CI: -0.124, 0.130 in the post years).
With adjustment for covariates and log transformation, none of the spending outcomes
evolved significantly over time. Comparing point estimates might suggest a greater elevation from
baseline expenses in post years than in the index year. Total medical expense of spouses of PLWD
increased by 7.37 percent (95% CI: -21.1, 35.8) in the year of index; in contrast, the increase was
as high as 17.2 percent in the post-index years (95% CI: -29.6, 64). Similarly, increase in out-of-
pocket expense in the index year and in post-index years were 6.91 percent (95% CI: -15.9, 29.7)
and 18.5 percent (95% CI: -19.2, 56.3), respectively. Medicare payment had limited fluctuation,
with a 0.8-percent (95% CI: -30.5, 32.2) increase in the year of index, and a 3.9-percent (95% CI:
-47.7, 55.4) increase in the two years to follow.
All results were robust to alternative form of time dummies, i.e. pooling index year with
post-index years.

Subgroup Heterogeneity
To investigate heterogeneity in temporal patterns, I ran the fully adjusted event study model
in subpopulations, defined by sex, race/ethnicity, education, household wealth, self-reported health
78
status, and caregiving status (all measured at index if varying over time). Estimates of time
dummies are shown in Table 3. Null effect prevailed in most stratified analyses, except for E/M
visits. In the year of index, males had 1.61 fewer E/M visits (95% CI: -2.947, -0.263), a 20.5%
decrease relative to the pre-index level. Non-Hispanic Whites and those ranked in the 3
rd
or top
wealth quartile also saw doctor less often with a reduction of 0.98 (95% CI: -1.821, -0.416) and
1.24 (95% CI: -2.220, -0.261) visits, respectively, representing a 12.6% and 15.9% drop from pre-
index years. For the three subgroups, the magnitude of change in the following 2 years remained
similar but did not reach statistical significance. Spouses of PLWD with minority background, in
contrast, increased the number of E/M visits by 3.81 (95% CI: 0.683, 6.928) or by 49.1% compared
to years prior to the index. No differences were detected by education, wealth, subjective health
status, or caregiving status. For example, the magnitude of reduction in E/M visits was greater in
caregivers (index year= -1.404, 95%CI: -2.819, 0.011; post-index years= -2.335, 95%CI: -4.73,
0.06) than in non-caregivers (index year = -0.552, 95%CI: -1.627, 0.523; post-index years= -0.07,
95%CI: -1.77, 1.63); however, the effects were not statistically different.

DISCUSSION
Using a nationally representative panel linked to administrative claims, this study explored
whether spouses of PLWD change health care utilization ensuing PLWD’s dementia diagnosis, as
well as its implications for comprehensively understanding dementia costs incurred by family. In
a broadly representative sample of elderly spouses of PLWD, I found little evidence for notable
changes in health care utilization after spouse’s dementia diagnosis, during the year of index and
two years after index relative to two years prior to. This suggests an average person whose spouse
is newly diagnosed with dementia maintain her care schedule and medical spending level in the
79
first few years after diagnosis. Therefore, current frameworks for valuing dementia costs, which
do not incorporate a family health care aspect, do not substantially underestimate the indirect costs
of dementia. While estimates were not statistically significant in adjusted analyses, it is worth
noting that the temporal patterns may vary by type of care. For instance, flu shot was upward
trending from pre-index to post-index years suggesting more health investment, whereas E/M visit
experienced a dip during the year of index and did not recover to pre-index level in the second or
third year after diagnosis, which could be a sign of constraints and forgone care. Inpatient care was
stable. Total medical expense tended to increase, mainly incurred by older adults themselves,
rather than by federal spending. More research is needed to understand mechanisms underlying
the differences by service type, as well as how they affect the well-being of spouses of PLWD.
Beyond showing within-individual variations in health care use over time, this study
analyzed differential responses to spouse’s dementia diagnosis across individuals, for the first time.
Economic theories and empirical evidence from multiple disciplines suggest changes in health care
use likely differ across subgroup who face different schedules of marginal costs of health
investment, and of marginal benefits of staying healthy. Despite the theoretical prediction of higher
productive and allocative efficiency among the highly educated, individuals with at least high-
school education did not differ from those without. Due to lack of statistical power, the current
study was unable to assess alternative cut-offs for defining “high education”, e.g. college education,
which may better reflect the efficiency advantage of education. Contrary to the hypothesis of
relatively suppressed use in lower-SES groups due to low resources, those in the upper half of
wealth distribution had significantly (but transitorily) fewer E/M visits in the year of spouse’s
diagnosis, suggesting SES-based heterogeneity may be driven by mechanisms other than resource
constraints, such as differential emotional shocks. Non-Hispanic Whites reduced the number of
80
E/M visits in the year of index, while racial/ethnic minorities saw doctors more often in two years
after the index. The observed increase among minorities may be explained by their health being
disproportionately “penalized” by spouse’s dementia diagnosis and thus a stronger incentive to
restore health stock with health care. However, according to a representative and rigorous study
on racial/ethnic disparities in health changes around the onset of spousal dementia (Chen et al
2020b), African Americans and Hispanics exhibit similar rate of changes in depression and self-
reported health as Whites do. It remains possible if other aspects of health decline faster in
racial/ethnic minorities. When seeking spouse’s diagnosis, increased contact with the health care
system may mitigate minorities’ stereotypes about health care providers and the system,
encouraging health care use among minorities (Wharton et al 2017). These two mechanisms lead
to different policy recommendations and should be disentangled in future studies.
There was also sex difference. Males significantly reduced the number of E/M visits by
1.61 in the year of index. In contrast, females remained at the pre-index level, with a positive point
estimate. Wives’ dementia diagnosis was associated with more discontinuation of ambulatory care
for husbands, than husbands’ diagnosis for wives. Existing evidence shows reduced outpatient
visits led to hospitalization (Trivedi et al 2010). In the current paper, the forgone ambulatory care
by males in the year of index did not translate into higher likelihood of using inpatient care in the
same year (-0.038, 95% Ci: -0.168, 0.092) or in the two years to follow (-0.016, 95% CI: -0.220,
0.187).
7
In a longer term, however, failure to satisfying the care needs of male spouses of PLWD
may still increase the risk of hospitalization and downstream health care costs, which will
ultimately be borne by the whole society, via federal programs like Medicare.

7
This pattern held for the other two subgroups who also experienced forgone E/M visits, i.e. non-Hispanic Whites and those
ranked in the 3
rd
or top quartile of household wealth, suggesting limited immediate harm of care discontinuation.
81
Another contribution of this paper was the assessment of caregiving. Spouse and caregiver
were used interchangeably in literature, which obscures potential commonality and differences of
the two types of individuals. HRS linked to Medicare claims provided a unique opportunity to
study caregiving. The insignificant changes in utilization among all spouses, non-caregiving
spouses, or caregiving spouses, suggested lack of (differential) responses by caregiving status.
Since the intensity of caregiving increases over the course of dementia, the long-term effects may
accumulate and exacerbate. As shown in Coe & Van Houtven (2009), negative impacts of health
emerge 2 years after caregiving onset. Over time, on the contrary, caregivers may find coping
strategies to adapt to chronic exposures to stress and strains (Pearlin 1990). Caregiving can also
improve subjective well-being such as life purpose and fulfillment (Roth et al 2015; Brown &
Brown 2014). Thus, the long-term temporal pattern of health care use warrants further
investigation.
Several limitations should be noted. First, the validity of FE approach rests upon the time
invariability of unobservables. While this paper captured the aggregate changes in health care use
after spouse’s dementia diagnosis and therefore allowed for event-related changes in
unobservables, the models could not fully tease out systematic variations in unobservables that
were not attributable to spouse’s dementia diagnosis. To provide direct evidence for interventions
that support PLWD and families, causality and mechanisms should be established. Second, I
assumed “2 years pre-index” represented the usual level of health care utilization. Dementia
patients can undergo a prodromal stage with cognitive/functional decline and neuropsychiatric
symptoms, the length of which may last 4-6 years and vary by case (Vermunt et al 2019; Stella et
al 2014). Put differently, changes associated with dementia onset could have occurred before I
observed them in Year -2. Studies on medical expenses of PLWD (Lin et al 2016; Zhu et al 2015)
82
and spouses of PLWD (Gilden et al 2017) both conclude with stable pre-diagnosis trends until
several months prior to diagnosis. Together with the stability during Year -2 and Year -1 observed
in the present paper, this assumption was not overly restrictive. Third, to capture health care use
in Medicare claims and time-varying information in HRS, the sample required continuous FFS
enrollment and survey response. While this did not impair the representativeness of this sample of
PLWD spouses as shown in Table 1, it remarkably reduced the sample size and thus statistical
power. Future studies can include more recent HRS-Medicare linked data to create a larger panel
for detecting differences in full and stratified samples. Adding more years of data also yields a
longer panel, to examine 1) long-term trajectories of health care use for all and for diverse groups,
and 2) outcomes that do not recur annually, such as cancer screening. Use of these services may
require more inputs including time, out-of-pocket spending, and health literacy. As a result, the
corresponding consumption may be more sensitive to shocks, than flu shot and E/M visits with
low time and pecuniary costs. Suppression of these services may also lead to higher downstream
health care costs to both older adults and Medicare. Finally, heterogeneity was only analyzed by
characteristics at index, likely endogenous to health care use. For example, selection into
caregiving based on baseline health may yield downward bias if caregivers are healthier in the first
place (i.e., the healthy caregiver hypothesis, Fredman et al 2015), or upward bias if caregivers are
in worse shape initially (e.g. Chen et al 2020b showing black and Hispanic caregivers have worse
health at baseline than non-caregiving counterparts). Future investigations should incorporate
time-varying measures of these endogenous characteristics for further insights.
In summary, when their spouse was newly diagnosed with dementia, individuals were able
to stay on their usual pattern of health care utilization and level of spending. There were signals of
rationed care in certain subgroups, for instance males, suggesting the importance of supporting
83
spouses’ care needs when caring for PLWD. Investigation on long-term pattern and causal
mechanisms is needed for supporting diverse spouses of PLWD.

84
TABLES & FIGURES

Table 1. Summary Statistics

Spouse of PLWD
Included in Sample
Spouse of PLWD aged 67+ P-value
# of Persons 418 845

# of Person-years 2090 4225

Age

Mean(sd) yrs 79.31 (6.06) 77.78 (6.38) ***
67-74 36.99% 32.82%

75-84 50.53% 51.07%
85+ 12.49% 16.11%
Female 59.57% 58.53%

Race/Ethnicity

Whites 85.41% 80.21%

African Americans 8.61% 11.85%
Hispanics 5.26% 6.04%
Other 0.72% 1.90%
Education

Less than High School 29.67% 34.32%

High School/Some College 52.39% 50.95%
BA and above 17.94% 14.73%
Household Wealth

1st(lowest) Quartile 13.11% 15.91%

2nd Quartile 23.45% 23.94%
3rd Quartile 29.11% 29.35%
4th(top) Quartile 34.33% 30.80%
Household Income

1st(lowest) Quartile 15.02% 16.17%

2nd Quartile 39.28% 40.89%

3rd Quartile 29.33% 27.74%

4th(top) Quartile 16.36% 15.20%

Self-Reported Health

Stroke 9.81% 10.55%

Diabetes 16.36% 18.01%

Hypertension 56.17% 55.45%

Heart Disease 28.80% 31.28%

Caregiving

Help Spouses with ADL 18.09% 24.53% *
Help Spouses with IADL 28.52% 39.57% ***
Help Spouses with ADL or IADL 33.59% 45.14% ***
Help Hours per Week if Being a Caregiver, Mean(sd) 26.54 (46.11) 31.29 (50.73)
Notes: Compared spouses of PLWD who aged at least 67, those included in sample additionally had continuous FFS enrollment
in 2 years prior to, in the year of, and 2 years after PLWD’s first diagnosis (5 years in total), and responded to to HRS wave W,
W-1, and W+1 (3 HRS waves in total). P-values indicated level of significant difference between samples obtained from Chi-
square tests and t-tests (the latter for continous variable only, i.e. age years and help hours): *** p<0.001, ** p<0.01, * p<0.05.
Wealth and income quartiles were based on wave-specific position in the full HRS sample.

85
Figure 1. Unadjusted Health Care Utilization over Time

86
Table 2. Event Study Regression Results

(1) (2) (3) (4) (5) (6)
Outcome
Flu Shot
(binary)
E/M Visit
(continuous)
Inpatient
Care
(binary)
Medical
Expense-
Total
(log-
transformed
)
Medical
Expense-
Medicare
(log-
transformed
)
Medical
Expense-
OOP
(log-
transforme
d)
Index year (r.t. pre years) 0.004 -0.615 -0.012 0.074 0.008 0.069

[-0.074,
0.083]
[-1.434,
0.205]
[-0.089,
0.065]
[-0.211,
0.358]
[-0.305,
0.322]
[-0.159,
0.297]
Post years (r.t. pre years) 0.032 -0.286 0.003 0.172 0.039 0.185

[-0.099,
0.163]
[-1.623,
1.051]
[-0.124,
0.130]
[-0.296,
0.640]
[-0.477,
0.554]
[-0.192,
0.563]
Aged 75-84 (r.t. 67-74) 0.056 0.531 0.031 0.166 0.155 0.111

[-0.015,
0.127]
[-0.295,
1.357]
[-0.048,
0.110]
[-0.226,
0.558]
[-0.249,
0.559]
[-0.202,
0.424]
Aged 85+ (r.t. 67-74) 0.094 0.292 0.037 -0.049 -0.050 -0.027

[-0.049,
0.237]
[-1.260,
1.844]
[-0.102,
0.175]
[-0.670,
0.573]
[-0.704,
0.604]
[-0.500,
0.446]
2nd wealth quartile (r.t. bottom) -0.005 0.495 0.022 0.027 0.059 -0.0001

[-0.088,
0.077]
[-0.619,
1.608]
[-0.078,
0.122]
[-0.386,
0.440]
[-0.396,
0.514]
[-0.345,
0.345]
3rd wealth quartile (r.t. bottom) 0.076 0.802 0.042 -0.029 -0.0312 0.061

[-0.038,
0.190]
[-0.569,
2.172]
[-0.078,
0.163]
[-0.524,
0.467]
[-0.590,
0.528]
[-0.334,
0.455]
Top wealth quartile (r.t.
bottom) 0.056 0.807 -0.023 -0.402 -0.382 -0.254

[-0.070,
0.182]
[-0.708,
2.322]
[-0.147,
0.101]
[-0.961,
0.157]
[-1.003,
0.238]
[-0.709,
0.201]
2nd income quartile (r.t.
bottom) 0.019 0.052 -0.042 -0.058 -0.063 -0.045

[-0.020,
0.057]
[-0.409,
0.513]
[-0.124,
0.039]
[-0.202,
0.086]
[-0.215,
0.089]
[-0.165,
0.075]
3rd income quartile (r.t.
bottom) 0.022 0.861 -0.043 0.143 0.186 0.142

[-0.084,
0.128]
[-0.167,
1.888]
[-0.138,
0.052]
[-0.368,
0.654]
[-0.402,
0.773]
[-0.287,
0.571]
Top income quartile (r.t.
bottom) 0.074 -0.025 0.006 -0.191 -0.087 -0.115

[-0.123,
0.270]
[-1.308,
1.257]
[-0.102,
0.114]
[-1.083,
0.701]
[-1.028,
0.854]
[-0.850,
0.621]
Spouse aged 75-84 (r.t. 67-74) -0.0005 0.404 -0.022 -0.452 -0.931 0.014

[-0.131,
0.130]
[-3.935,
4.743]
[-0.415,
0.372]
[-1.515,
0.611]
[-2.167,
0.305]
[-0.816,
0.845]
Spouse aged 85+ (r.t. 67-74) -0.075 0.012 -0.027 -0.160 -0.722 0.233

[-0.223,
0.073]
[-4.444,
4.469]
[-0.428,
0.375]
[-1.278,
0.958]
[-2.019,
0.575]
[-0.649,
1.116]
Spouse had 1-2 ADL
limitations (r.t. none) -0.0356 -0.282 0.016 0.061 0.041 0.047

[-0.105,
0.034]
[-0.997,
0.434]
[-0.043,
0.076]
[-0.217,
0.339]
[-0.263,
0.345]
[-0.179,
0.273]
Spouse had 3-5 ADL
limitations (r.t. none) -0.039 -0.131 -0.064 -0.127 -0.074 -0.142

[-0.132,
0.053]
[-1.068,
0.805]
[-0.143,
0.015]
[-0.517,
0.263]
[-0.476,
0.328]
[-0.475,
0.190]
Spouse had 1-2 IADL
limitations (r.t. none) 0.052 -0.377 0.011 -0.069 -0.101 0.040
87

[-0.007,
0.110]
[-1.048,
0.294]
[-0.045,
0.067]
[-0.312,
0.174]
[-0.363,
0.160]
[-0.152,
0.232]
Spouse had 3-5 IADL
limitations (r.t. none) -0.008 -0.170 0.023 -0.057 -0.093 0.007

[-0.091,
0.075]
[-1.193,
0.852]
[-0.049,
0.094]
[-0.426,
0.312]
[-0.489,
0.303]
[-0.301,
0.315]

Constant 0.491** 5.312 -0.065 6.972*** 6.987*** 4.994***

[0.147,
0.835]
[-0.043,
10.67]
[-0.579,
0.448]
[5.399,
8.546]
[5.219,
8.755]
[3.745,
6.242]

Observations 2,090 2,090 2,090 2,090 2,090 2,090
R-squared 0.044 0.026 0.022 0.071 0.065 0.084
Number of hhidpn 418 418 418 418 418 418
Notes: Models were estimated with linear fixed effects regressions with individual and calendar time fixed effects. Confidence
intervals, shown in brackets, were calculated from robust standard errors clustered at individual level. *** p<0.001, ** p<0.01, *
p<0.05.

88
Figure 2. Changes in Health Care Use, from Event Study Models

Notes: Outcomes were likelihood of receiving flu shot, number of E/M visits, likelihood of using inpatient care, log-transformed
total medical expenses, log-transformed Medicare expenses, and log-transformed out-of-pocket expenses, in a year. Estimates
were obtained from linear fixed-effects models adjusting for age group, household wealth and income quartiles in the full HRS
population, spouse’s age group, spouses’ ADL and IADL function, and individual and calendar time fixed effects. Confidence
intervals were calculated from robust standard errors clustered at individual level.

89
Table 3. Event Study Regression Results in Stratified Samples

A. Flu Shot
(Binary)
B. E/M visits
(Continuous)
C. Inpatient Care
D. Total medical
expenses
(Log-transformed)
E. Medicare
expenses
(Log-transformed)
F. Out-of-Pocket
expenses
(Log-transformed)
# of
Uniqu
e
Indivi
duals
Index
year
Post
years
Index
year
Post
years
Index
year
Post
years
Index
year
Post
years
Index
year
Post
years
Index
year
Post
years
Full sample
0.004 0.032 -0.615 -0.286 -0.012 0.003 0.073 0.172 0.008 0.039 0.069 0.185
418

[-0.074,
0.083]
[-0.099,
0.163]
[-1.434,
0.205]
[-1.623,
1.051]
[-0.089,
0.065]
[-0.124,
0.130]
[-0.211,
0.358]
[-0.296,
0.640]
[-0.305,
0.322]
[-0.477,
0.554]
[-0.159,
0.297]
[-0.192,
0.563]

Stratified by sex

Male
-0.065 -0.094 -1.605* -1.464 -0.038 -0.016 -0.010 0.091 -0.013 -0.007 -0.020 0.084 169

[-0.195,
0.066]
[-0.303,
0.114]
[-2.947,
-0.263]
[-3.586,
0.657]
[-0.168,
0.092]
[-0.220,
0.187]
[-0.521,
0.501]
[-0.759,
0.940]
[-0.581,
0.555]
[-0.950,
0.937]
[-0.417,
0.377]
[-0.593,
0.760]

Female
0.070 0.146 0.046 0.593 0.008 0.032 0.096 0.248 -0.0119 0.107 0.116 0.297 269

[-0.029,
0.170]
[-0.025,
0.317]
[-1.030,
1.123]
[-1.159,
2.344]
[-0.089,
0.105]
[-0.134,
0.198]
[-0.229,
0.421]
[-0.299,
0.796]
[-0.368,
0.344]
[-0.487,
0.701]
[-0.156,
0.388]
[-0.150,
0.743]

Stratified by race/ethnicity

Non-Hispanic
Whites 0.0117 0.0339 -0.984* -0.983 -0.034 0.009 0.002 0.069 -0.071 -0.039 0.013 0.112
357

[-0.0765,
0.100]
[-0.114,
0.182]
[-1.821,
-0.146]
[-2.429,
0.464]
[-0.116,
0.049]
[-0.129,
0.146]
[-0.312,
0.315]
[-0.453,
0.591]
[-0.416,
0.274]
[-0.617,
0.540]
[-0.236,
0.263]
[-0.307,
0.532]

Other
race/ethnicity -0.044 0.003 1.322 3.806* 0.205 0.101 0.442 0.778 0.405 0.343 0.334 0.601 61

[-0.234,
0.146]
[-0.315,
0.321]
[-1.119,
3.763]
[0.683,
6.928]
[-0.044,
0.454]
[-0.267,
0.468]
[-0.310,
1.194]
[-0.414,
1.971]
[-0.467,
1.277]
[-1.022,
1.709]
[-0.264,
0.932]
[-0.398,
1.601]

Stratified by education

Less than high
school 0.033 0.049 -1.287 -0.827 0.112 0.121 -0.142 -0.210 -0.260 -0.425 -0.109 -0.057 124

[-0.124,
0.191]
[-0.217,
0.315]
[-2.926,
0.352]
[-3.760,
2.105]
[-0.043,
0.268]
[-0.168,
0.410]
[-0.731,
0.448]
[-1.291,
0.871]
[-0.937,
0.417]
[-1.634,
0.784]
[-0.572,
0.354]
[-0.906,
0.792]

High school
and above -0.001 0.044 -0.534 -0.382 -0.078 -0.065 0.066 0.194 0.017 0.092 0.066 0.171 294

[-0.092,
0.089]
[-0.110,
0.197]
[-1.514,
0.447]
[-1.917,
1.152]
[-0.166,
0.011]
[-0.208,
0.079]
[-0.253,
0.385]
[-0.332,
0.721]
[-0.323,
0.356]
[-0.475,
0.659]
[-0.193,
0.325]
[-0.258,
0.601]

Stratified by household wealth at index

1st(lowest) and
2nd quartile -0.067 0.006 0.021 1.463 0.136 0.207 0.037 0.064 0.008 0.011 0.074 0.169 153

[-0.209,
0.0754]
[-0.220,
0.233]
[-1.771,
1.813]
[-1.186,
4.111]
[-0.005,
0.276]
[-0.035,
0.449]
[-0.482,
0.555]
[-0.854,
0.982]
[-0.581,
0.597]
[-1.021,
1.042]
[-0.342,
0.490]
[-0.548,
0.885]

3rd or 4th(top)
quartile 0.046 0.051 -1.240* -1.527 -0.079 -0.078 0.064 0.266 -0.040 0.096 0.071 0.211 265

[-0.062,
0.153]
[-0.130,
0.232]
[-2.220,
-0.261]
[-3.213,
0.159]
[-0.175,
0.018]
[-0.232,
0.076]
[-0.308,
0.435]
[-0.288,
0.821]
[-0.429,
0.350]
[-0.493,
0.685]
[-0.230,
0.372]
[-0.246,
0.668]

Stratified by self-reported health status at
index

Poor/Fair 0.051 0.049 0.216 0.773 0.012 0.079 0.120 0.431 -0.034 0.367 0.214 0.343 119

[-0.126,
0.228]
[-0.223,
0.322]
[-1.311,
1.744]
[-1.886,
3.432]
[-0.184,
0.208]
[-0.262,
0.420]
[-0.379,
0.619]
[-0.456,
1.317]
[-0.569,
0.502]
[-0.608,
1.342]
[-0.215,
0.644]
[-0.356,
1.041]

Good -0.018 0.013 -0.740 -0.918 -0.019 -0.015 0.111 0.099 0.059 -0.101 0.055 0.106 299

[-0.113,
0.078]
[-0.144,
0.171]
[-1.610,
0.130]
[-2.375,
0.538]
[-0.100,
0.063]
[-0.144,
0.114]
[-0.235,
0.457]
[-0.461,
0.659]
[-0.317,
0.435]
[-0.710,
0.508]
[-0.225,
0.334]
[-0.357,
0.568]

Stratified by caregiving status at index

Not caregiver 0.023 0.035 -0.552 -0.067 -0.056 -0.068 -0.104 0.010 -0.150 -0.085 -0.091 0.028
278

[-0.082,
0.129]
[-0.146,
0.216]
[-1.627,
0.523]
[-1.765,
1.631]
[-0.153,
0.041]
[-0.228,
0.092]
[-0.474,
0.267]
[-0.591,
0.611]
[-0.562,
0.262]
[-0.745,
0.575]
[-0.397,
0.214]
[-0.470,
0.526]

Caregiver 0.020 0.064 -1.404 -2.335 0.020 0.029 0.304 0.355 0.146 0.161 0.358 0.470
140

[-0.123,
0.163]
[-0.179,
0.306]
[-2.819,
0.011]
[-4.732,
0.063]
[-0.128,
0.168]
[-0.231,
0.289]
[-0.348,
0.955]
[-0.816,
1.525]
[-0.568,
0.860]
[-1.119,
1.441]
[-0.151,
0.866]
[-0.448,
1.389]
Notes: Outcomes were likelihood of receiving flu shot, number of E/M visits, likelihood of using inpatient care, log-transformed
total medical expenses, log-transformed Medicare expenses, and log-transformed out-of-pocket expenses, in a year. Estimates
were obtained from linear fixed-effects models adjusting for age group, household wealth and income quartiles in the full HRS
population, spouse’s age group, spouses’ ADL and IADL function, and individual and calendar time fixed effects. Confidence
intervals were calculated from robust standard errors clustered at individual level.

90
APPENDIX

Supplementary Table 1. Identifying Spouses of PLWD

Restriction N
Individuals in HRS-Medicare linked data 24,668
Individuals w/ dementia diagnosis verified by another diagnosis or death 4,922
PLWD had 2-year washout period prior to the earliest verified diagnosis 4,559
PLWD responded to an HRS wave W within 1 year around incident dementia
diagnosis
3,682
PLWD had spouse at HRS wave W 1,471
Spouse never had verified dementia up to 12/31/2012 1,026
Spouse aged at least 67 when PLWD were first diagnosed 845
Spouse had continuous FFS enrollment in 2 years prior to, in the year of, and 2 years
after PLWD’s first diagnosis (5 years in total)
504
Spouse responded to to HRS wave W, W-1, and W+1 (3 HRS waves in total) 418

Supplementary Table 2. CPT Codes for Identifying Health Care Procedure

Procedure CPT Code
Flu Shot
90630, 90653, 90654, 90657,
90658, 90660, 90661, 90662,
90672, 90673, 90674, 90682,
90685, 90686, 90687, 90688,
90756, Q2034, Q2035, Q2036,
Q2038, G0008
Office or other outpatient E/M services, new patient 99201, 99202, 99203, 99204, 99205
Office or other outpatient E/M services, established patient 99211, 99212, 99213, 99214, 99215

91
Supplementary Table 3. Unadjusted Health Care Utilization over Time

Year -2 Year -1 Index Year Year 1 Year 2
% Receiving Flu Shot
Mean 55.3% 56.7% 56.0% 56.2% 53.8%
95% CI [50.5%, 60.0%] [51.9%, 61.5%] [51.2%, 60.8%] [51.4%, 61.0%] [49.0%, 58.6%]
P-value r.t. Year -2

0.677 0.835 0.781 0.677
# of Office Visits
Mean 7.8 7.7 7.4 7.8 8.2
95% CI [7.2, 8.4] [7.1, 8.3] [6.8, 8.1] [7.1, 8.4] [7.6, 8.8]
P-value r.t. Year -2

0.816 0.445 0.949 0.365
% Using Inpatient Care
Mean 15.8% 16.5% 17.2% 18.7% 22.0%
95% CI [12.3%, 19.3%] [13.0%, 20.1%] [13.6%, 20.8%] [14.9%, 22.4%] [18.0%, 26.0%]
P-value r.t. Year -2

0.787 0.589 0.280 0.019
Total Medical Expenses
Mean 5690.3 5897.5 6305.3 8678.0 11056.9
95% CI [4549.1, 6831.5] [4849.1, 6945.7] [5160.5, 7450.0] [6814.1, 10541.9] [8237.8, 13876.0]
P-value r.t. Year -2

0.869 0.623 0.017 0.000
Medical Expenses Paid by Medicare
Mean 4766.3 4892.6 5158.7 7249.4 9426.3
95% CI [3729.5, 5803.1] [3965.9, 5819.2] [4157.5, 6160.0] [5601.5, 8897.2] [6887.8, 11964.9]
P-value r.t. Year -2

0.910 0.726 0.027 0.000
Out-of-Pocket Medical Expenses
Mean 901.2 961.6 1118.9 1308.9 1612.5
95% CI [779.1, 1023.2] [827.0, 1096.1] [934.3, 1303.4] [1093.6, 1524.3] [1301.3, 1923.6]
P-value r.t. Year -2 0.682 0.140 0.006 0.000
Notes: Level of significant differences over time were the same when grouping time as pre years, index year, and post years.

92
Chapter 5
CONCLUSION
As a disease that leads to disability, dependency, and intensive use of (health) care
resources, dementia imposes remarkable burden on PLWD and their families. This dissertation
demonstrates persistent disparities in dementia burden and care among diverse PLWD and families
and informs tailored interventions for achieving health equities. Built upon broadly representative
samples, these findings are generalizable to older Americans afflicted with dementia.
Studying disparities is increasingly important in a diversifying America. For instance,
while non-Hispanic Whites remain the largest race/ethnicity group in today’s U.S., its proportion
has declined by 8.6% since 2010; the Hispanic population has grown 23% from 2010, reaching
62.1 million in 2020 (U.S. Census Bureau 2021). For planning purposes, explicit understanding of
heterogeneity across subpopulations is essential for precisely estimating the societal burden of
dementia, and for projecting future burden. Disparities in dementia burden and care also reveal
areas for improving population health, via directing resources to those who are disproportionately
affected by dementia.
The current studies delineate heterogeneities throughout the dementia care continuum,
ranging from diagnosis and care of PLWD to family support, and explore mechanistic pathways
underlying these disparities. The first chapter on diagnosis disparities by race/ethnicity
incorporates a previously neglected longitudinal perspective and shows racial/ethnic minorities are
more susceptible to under- and delayed- diagnosis, resulting in missed opportunities for improving
well-being with timely actions. The second chapter reveals greater comorbidity burden and
function/cognitive impairment at the time of incident dementia diagnosis among those with lower
education and wealth. Net of these health differences, the highly educated do not outlive the lower
93
educated, whereas the richest survive longer than the deprived after diagnosis. Finally, to
complement understanding of dementia impacts on families, the third chapters investigates
whether spouses of PLWD, regardless of caregiving status, experience disruption in health care.
Leveraging panel data, I show that spouses of PLWD do not skimp on or spend more on health
care ensuing incident diagnosis, yet certain groups such as males forgo ambulatory care. These
findings together highlight the importance of targeted interventions throughout dementia care
cycle, given limited public health resources. For example, promoting early detection among
racial/ethnic minorities and those with low education or wealth may be particularly fruitful in
achieving equities in dementia diagnostic care. There is potential to extend post-diagnosis life
expectancy for the deprived; while considering their resource disadvantage, additional support
may be warranted. When caring for PLWD, care providers should be aware of and sensitive to
potential trade-offs and strains faced by family members of PLWD. In the context of dementia, a
shift away from patient-centric care may benefit family members and ultimately PLWD as well.
The next phase of research should focus on causal and mechanistic evidence to inform
interventions. Disparities in dementia burden and care may arise from multifaceted domains
including biological, socio-cultural, contextual, and health-care related factors, and understanding
them likely require a life course approach. While the linkage of population survey and
administrative claims contains rich information for studying dementia-related disparities, as shown
in the current dissertation, it may be insufficient for disentangling these mechanisms. Future
studies should consider taking advantage of other data, such as clinic-pathological cohort studies
for dementia severity/stage, or qualitative data for lived experience of PLWD. Expanding the time
horizon of inquiry is another venue for future research. The stable trends in health care utilization
around spousal dementia diagnosis do not necessarily deny this pathway through which dementia
94
imposes burden on family members. Detrimental impacts may emerge and propagate over a longer
time, given the irreversible and progressive nature of dementia. Last but not least, other than
heterogeneity across subpopulation, there may exist differences within subpopulation, which is not
examined in this work due to small subsample size.
Despite limitations, this dissertation features unique strengths in advancing the knowledge
base of dementia-related disparities in the U.S., such as a longitudinal perspective, breadth of
examined outcomes, and the use of health economics theories as guidance for identifying
vulnerable groups. Meanwhile it opens promising avenues for future research. The findings shed
light on policy and practice to improve the well-being of diverse PLWD and families. With limited
public resources, learning what type of support to be allocated to whom is essential for us to
overcome the challenges posed by dementia.

95
REFERENCES

Adler, N. E., & Stewart, J. (2010). Health disparities across the lifespan: meaning, methods, and
mechanisms. Annals of the New York Academy of Sciences, 1186(1), 5-23.

Agüero-Torres, H., Fratiglioni, L., Guo, Z., Viitanen, M., & Winblad, B. (1999). Mortality from
dementia in advanced age: a 5-year follow-up study of incident dementia cases. Journal
of clinical epidemiology, 52(8), 737-743.

Agüero-Torres, H., Qiu, C., Winblad, B., & Fratiglioni, L. (2002). Dementing disorders in the
elderly: evolution of disease severity over 7 years. Alzheimer Disease & Associated
Disorders, 16(4), 221-227.

Al‐Janabi, H., Van Exel, J., Brouwer, W., Trotter, C., Glennie, L., Hannigan, L., & Coast, J.
(2016). Measuring health spillovers for economic evaluation: a case study in meningitis.
Health economics, 25(12), 1529-1544.

Allen, A. P., Curran, E. A., Duggan, Á., Cryan, J. F., Chorcoráin, A. N., Dinan, T. G., ... &
Clarke, G. (2017). A systematic review of the psychobiological burden of informal
caregiving for patients with dementia: Focus on cognitive and biological markers of
chronic stress. Neuroscience & Biobehavioral Reviews, 73, 123-164.

Alzheimer’s Associatiom. Alzheimer’s Association 2021 Facts and Figures Report.

Amjad, H., Roth, D. L., Sheehan, O. C., Lyketsos, C. G., Wolff, J. L., & Samus, Q. M. (2018).
Underdiagnosis of dementia: an observational study of patterns in diagnosis and
awareness in US older adults. Journal of general internal medicine, 33(7), 1131-1138.

Andersen, K., Launer, L. J., Dewey, M. E., Letenneur, L., Ott, A., Copeland, J. R. M., ... &
EURODEM Incidence Research Group. (1999). Gender differences in the incidence of
AD and vascular dementia: The EURODEM Studies. Neurology, 53(9), 1992-1992.

Andersen, K., Lolk, A., Martinussen, T., & Kragh-Sørensen, P. (2010). Very mild to severe
dementia and mortality: a 14-year follow-up–the Odense study. Dementia and geriatric
cognitive disorders, 29(1), 61-67.

Aneshensel, C. S., Pearlin, L. I., Levy-Storms, L., & Schuler, R. H. (2000). The transition from
home to nursing home mortality among people with dementia. The Journals of
Gerontology Series B: Psychological Sciences and Social Sciences, 55(3), S152-S162.

96
Aranda, M. P., Kremer, I. N., Hinton, L., Zissimopoulos, J., Whitmer, R. A., Hummel, C. H., ...
& Fabius, C. (2021). Impact of dementia: Health disparities, population trends, care
interventions, and economic costs. Journal of the American Geriatrics Society, 69(7),
1774-1783.

Arksey, H., & Hirst, M. (2005). Unpaid carers’ access to and use of primary care services.
Primary Health Care Research & Development, 6(2), 101-116.

Azad, N. A., Al Bugami, M., & Loy-English, I. (2007). Gender differences in dementia risk
factors. Gender medicine, 4(2), 120-129.

Balogh, E. P., Miller, B. T., & Ball, J. R. (2015). Improving diagnosis in health care.

Barnett, J. H., Lewis, L., Blackwell, A. D., & Taylor, M. (2014). Early intervention in
Alzheimer’s disease: a health economic study of the effects of diagnostic timing. BMC
neurology, 14(1), 1-9.

Basu, A., & Meltzer, D. (2005). Implications of spillover effects within the family for medical
cost-effectiveness analysis. Journal of health economics, 24(4), 751-773.

Baumgarten, M., Battista, R. N., Infante-Rivard, C., Hanley, J. A., Becker, R., Bilker, W. B., &
Gauthier, S. (1997). Use of physician services among family caregivers of elderly
persons with dementia. Journal of clinical epidemiology, 50(11), 1265-1272.

Bebe, A., Siersma, V., Kragstrup, J., Nielsen, A. B. S., Møller, A., Søndergaard, J., ... &
Waldorff, F. B. (2019). The effect of economic assets on mortality in patients with
dementia: A population-based cohort study. Journal of Alzheimer's Disease, 67(4), 1245-
1253.

Benjamins, M. R., & Whitman, S. (2014). Relationships between discrimination in health care
and health care outcomes among four race/ethnic groups. Journal of Behavioral
Medicine, 37(3), 402-413.

Bertakis, K. D., Azari, R., Helms, L. J., Callahan, E. J., & Robbins, J. A. (2000). Gender
differences in the utilization of health care services. Journal of family practice, 49(2),
147-147.

Borson, S., Frank, L., Bayley, P. J., Boustani, M., Dean, M., Lin, P. J., ... & Ashford, J. W.
(2013). Improving dementia care: the role of screening and detection of cognitive
impairment. Alzheimer's & Dementia, 9(2), 151-159.
97
Boulware, L. E., Cooper, L. A., Ratner, L. E., LaVeist, T. A., & Powe, N. R. (2016). Race and
trust in the health care system. Public health reports.

Bradford, A., Kunik, M. E., Schulz, P., Williams, S. P., & Singh, H. (2009). Missed and delayed
diagnosis of dementia in primary care: prevalence and contributing factors. Alzheimer
disease and associated disorders, 23(4), 306.

Brooker, D., Fontaine, J. L., Evans, S., Bray, J., & Saad, K. (2014). Public health guidance to
facilitate timely diagnosis of dementia: ALzheimer's COoperative Valuation in Europe
recommendations. International journal of geriatric psychiatry, 29(7), 682-693.

Brookmeyer, R., Corrada, M. M., Curriero, F. C., & Kawas, C. (2002). Survival following a
diagnosis of Alzheimer disease. Archives of neurology, 59(11), 1764-1767.

Brown, R. M., & Brown, S. L. (2014). Informal caregiving: A reappraisal of effects on
caregivers.

Burton, P., Lethbridge, L., & Phipps, S. (2008). Children with disabilities and chronic conditions
and longer-term parental health. The Journal of Socio-Economics, 37(3), 1168-1186.

Burton, L. C., Newsom, J. T., Schulz, R., Hirsch, C. H., & German, P. S. (1997). Preventive
health behaviors among spousal caregivers. Preventive medicine, 26(2), 162-169.

Bynum, J. P., Rabins, P. V., Weller, W., Niefeld, M., Anderson, G. F., & Wu, A. W. (2004). The
relationship between a dementia diagnosis, chronic illness, medicare expenditures, and
hospital use. Journal of the American Geriatrics Society, 52(2), 187-194.

Cantarero-Prieto, D., Leon, P. L., Blazquez-Fernandez, C., Juan, P. S., & Cobo, C. S. (2020).
The economic cost of dementia: a systematic review. Dementia, 19(8), 2637-2657.

Card, D. (1999). Handbook of labor economics (No. 331 HAN v. 3).

Carpenter, B. D., Xiong, C., Porensky, E. K., Lee, M. M., Brown, P. J., Coats, M., ... & Morris,
J. C. (2008). Reaction to a dementia diagnosis in individuals with Alzheimer's disease
and mild cognitive impairment. Journal of the American Geriatrics Society, 56(3), 405-
412.

Castagné, R., Garès, V., Karimi, M., Chadeau-Hyam, M., Vineis, P., Delpierre, C., & Kelly-
Irving, M. (2018). Allostatic load and subsequent all-cause mortality: which biological
markers drive the relationship? Findings from a UK birth cohort. European journal of
epidemiology, 33(5), 441-458.
98
Castellani, R. J., Rolston, R. K., & Smith, M. A. (2010). Alzheimer disease. Disease-a-month:
DM, 56(9), 484.

Castro, C. M., King, A. C., Housemann, R., Bacak, S. J., McMullen, K. M., & Brownson, R. C.
(2007). Rural family caregivers and health behaviors: Results from an epidemiologic
survey. Journal of Aging and Health, 19(1), 87-105.

Cha, H., Farina, M. P., & Hayward, M. D. (2021). Socioeconomic status across the life course
and dementia-status life expectancy among older Americans. SSM-population health, 15,
100921.

Chen, R., Hu, Z., Wei, L., & Wilson, K. (2014). Socioeconomic status and survival among older
adults with dementia and depression. The British Journal of Psychiatry, 204(6), 436-440.

Chen, L., Reed, C., Happich, M., Nyhuis, A., & Lenox-Smith, A. (2014). Health care resource
utilisation in primary care prior to and after a diagnosis of Alzheimer’s disease: a
retrospective, matched case–control study in the United Kingdom. BMC geriatrics, 14(1),
1-9.

Chen, C. L., Liang, C. K., Yin, C. H., Lin, Y. T., Lee, C. C., & Chen, N. C. (2020a). Effects of
Socioeconomic status on Alzheimer disease mortality in Taiwan. The American Journal
of Geriatric Psychiatry, 28(2), 205-216.

Chen, C., Thunell, J., & Zissimopoulos, J. (2020b). Changes in physical and mental health of
Black, Hispanic, and White caregivers and non‐caregivers associated with onset of
spousal dementia. Alzheimer's & Dementia: Translational Research & Clinical
Interventions, 6(1), e12082.

Chen, Y., Crimmins, E., Ferido, P., & Zissimopoulos, J. M. (2022). Racial/Ethnic Disparities in
Length of Life after Dementia Diagnosis: an 18-Year Follow-up Study of Medicare
Beneficiaries. The Lancet Regional Health-Americas, 8, 100179.

Chi, W., Graf, E., Hughes, L., Hastie, J., Khatutsky, G., Shuman, S., & Lamont, H. (2019).
Community-dwelling older adults with dementia and their caregivers: Key indicators
from the national health and aging trends study. US Department of Health and Human
Services, Office of the Assistant Secretary for Planning and Evaluation, Office of
Disability, Aging and Long-Term Care Policy.

Chodosh, J., Petitti, D. B., Elliott, M., Hays, R. D., Crooks, V. C., Reuben, D. B., ... & Wenger,
N. (2004). Physician recognition of cognitive impairment: evaluating the need for
improvement. Journal of the American Geriatrics Society, 52(7), 1051-1059.

99
Coe, N. B., Skira, M. M., & Larson, E. B. (2018). A comprehensive measure of the costs of
caring for a parent: differences according to functional status. Journal of the American
Geriatrics Society, 66(10), 2003-2008.

Coe, N. B., & Van Houtven, C. H. (2009). Caring for mom and neglecting yourself? The health
effects of caring for an elderly parent. Health economics, 18(9), 991-1010.

Contador, I., Stern, Y., Bermejo-Pareja, F., Sanchez-Ferro, A., & Benito-Leon, J. (2017). Is
educational attainment associated with increased risk of mortality in people with
dementia? A population-based study. Current Alzheimer Research, 14(5), 571-576.

Cooper, C., Tandy, A. R., Balamurali, T. B., & Livingston, G. (2010). A systematic review and
meta-analysis of ethnic differences in use of dementia treatment, care, and research. The
American Journal of Geriatric Psychiatry, 18(3), 193-203.
Crimmins, E. M., Kim, J. K., Langa, K. M., & Weir, D. R. (2011). Assessment of cognition
using surveys and neuropsychological assessment: the Health and Retirement Study and
the Aging, Demographics, and Memory Study. Journals of Gerontology Series B:
Psychological Sciences and Social Sciences, 66(suppl_1), i162-i171.

Crimmins, E. M., Saito, Y., Kim, J. K., Zhang, Y. S., Sasson, I., & Hayward, M. D. (2018).
Educational differences in the prevalence of dementia and life expectancy with dementia:
Changes from 2000 to 2010. The Journals of Gerontology: Series B, 73(suppl_1), S20-
S28.

Crum, R. M., Anthony, J. C., Bassett, S. S., & Folstein, M. F. (1993). Population-based norms
for the Mini-Mental State Examination by age and educational level. Jama, 269(18),
2386-2391.

Cutler, D. M., & Lleras-Muney, A. (2010). Understanding differences in health behaviors by
education. Journal of health economics, 29(1), 1-28.

Deaton, A. (2002). Policy implications of the gradient of health and wealth. Health affairs, 21(2),
13-30.

Deaton, A. (2003). Health, inequality, and economic development. Journal of economic
literature, 41(1), 113-158.

Di Castelnuovo, A., Quacquaruccio, G., Donati, M. B., De Gaetano, G., & Iacoviello, L. (2009).
Spousal concordance for major coronary risk factors: a systematic review and meta-
analysis. American journal of epidemiology, 169(1), 1-8.
100
Dilworth-Anderson, P., Williams, I. C., & Gibson, B. E. (2002). Issues of race, ethnicity, and
culture in caregiving research: A 20-year review (1980–2000). The Gerontologist, 42(2),
237-272.

Drabo, E. F., Barthold, D., Joyce, G., Ferido, P., Chui, H. C., & Zissimopoulos, J. (2019).
Longitudinal analysis of dementia diagnosis and specialty care among racially diverse
Medicare beneficiaries. Alzheimer's & Dementia, 15(11), 1402-1411.

Dubois, B., Padovani, A., Scheltens, P., Rossi, A., & Dell’Agnello, G. (2016). Timely diagnosis
for Alzheimer’s disease: a literature review on benefits and challenges. Journal of
Alzheimer's disease, 49(3), 617-631.

Duru, O. K., Harawa, N. T., Kermah, D., & Norris, K. C. (2012). Allostatic load burden and
racial disparities in mortality. Journal of the National Medical Association, 104(1-2), 89-
95.

Ehrlich, I., & Chuma, H. (1990). A Model of the Demand for Longevity and the Value of Life
Extension. Journal of Political economy, 98(4), 761-782.

Fadlon, I., & Nielsen, T. H. (2019). Family health behaviors. American Economic Review,
109(9), 3162-91.

Fisher, G., Hassan, H., Faul, J. D., Rodgers, W. L., & Weir, D. R. (2015). Health and Retirement
Study imputation of cognitive functioning measures: 1992–2012. Ann Arbor, MI: Survey
Research Center, University of Michigan. http://hrsonline. isr. umich.
edu/modules/meta/xyear/cogimp/desc/COGIMPdd. pdf.

Fitzpatrick, A. L., Kuller, L. H., Ives, D. G., Lopez, O. L., Jagust, W., Breitner, J. C., ... &
Dulberg, C. (2004). Incidence and prevalence of dementia in the Cardiovascular Health
Study. Journal of the American Geriatrics Society, 52(2), 195-204.

Fonareva, I., & Oken, B. S. (2014). Physiological and functional consequences of caregiving for
relatives with dementia. International psychogeriatrics, 26(5), 725-747.

Fredman, L., Lyons, J. G., Cauley, J. A., Hochberg, M., & Applebaum, K. M. (2015). The
relationship between caregiving and mortality after accounting for time-varying caregiver
status and addressing the healthy caregiver hypothesis. Journals of Gerontology Series A:
Biomedical Sciences and Medical Sciences, 70(9), 1163-1168.

101
Freed, M., Biniek, J. F., Damico, A., & Neuman, T. (2021). Medicare Advantage in 2021:
enrollment update and key trends. Kaiser Family Foundation.

Freedman, V. A., Kasper, J. D., Spillman, B. C., & Plassman, B. L. (2018). Short-term changes
in the prevalence of probable dementia: An analysis of the 2011–2015 National Health
and Aging Trends Study. The Journals of Gerontology: Series B, 73(suppl_1), S48-S56.

Freels, S., Nyenhuis, D. L., & Gorelick, P. B. (2002). Predictors of survival in African American
patients with AD, VaD, or stroke without dementia. Neurology, 59(8), 1146-1153.

Gianattasio, K. Z., Prather, C., Glymour, M. M., Ciarleglio, A., & Power, M. C. (2019). Racial
disparities and temporal trends in dementia misdiagnosis risk in the United States.
Alzheimer's & Dementia: Translational Research & Clinical Interventions, 5, 891-898.

Gilden, D. M., Kubisiak, J. M., Kahle-Wrobleski, K., Ball, D. E., & Bowman, L. (2017). A
claims-based examination of health care costs among spouses of patients with
Alzheimer’s disease. Journals of Gerontology Series A: Biomedical Sciences and
Medical Sciences, 72(6), 811-817.

Goren, A., Montgomery, W., Kahle-Wrobleski, K., Nakamura, T., & Ueda, K. (2016). Impact of
caring for persons with Alzheimer’s disease or dementia on caregivers’ health outcomes:
findings from a community based survey in Japan. BMC geriatrics, 16(1), 1-12.

Grady, P. A., & Rosenbaum, L. M. (2015). The science of caregiver health. Journal of Nursing
Scholarship: An Official Publication of Sigma Theta Tau International Honor Society of
Nursing/Sigma Theta Tau, 47(3), 197.

Gräsel, E. (2002). When home care ends—changes in the physical health of informal caregivers
caring for dementia patients: a longitudinal study. Journal of the American Geriatrics
Society, 50(5), 843-849.

Grossman, M. (1972). On the Concept of Health Capital and the Demand for Health The Journal
of Political Economy 80 (2): 223–255.

Guehne, U., Matschinger, H., Angermeyer, M. C., & Riedel-Heller, S. G. (2006). Incident
dementia cases and mortality. Dementia and Geriatric Cognitive Disorders, 22(3), 185-
193.
Hamad, R., Penko, J., Kazi, D. S., Coxson, P., Guzman, D., Wei, P. C., ... & Bibbins-Domingo,
K. (2020). Association of low socioeconomic status with premature coronary heart
disease in US adults. JAMA cardiology, 5(8), 899-908.

Hebert, L. E., Bienias, J. L., Aggarwal, N. T., Wilson, R. S., Bennett, D. A., Shah, R. C., &
Evans, D. A. (2010). Change in risk of Alzheimer disease over time. Neurology, 75(9),
786-791.
102
Helzner, E. P., Scarmeas, N., Cosentino, S., Tang, M. X., Schupf, N., & Stern, Y. (2008).
Survival in Alzheimer disease: a multiethnic, population-based study of incident cases.
Neurology, 71(19), 1489-1495.

Hill, C. V., Pérez-Stable, E. J., Anderson, N. A., & Bernard, M. A. (2015). The National Institute
on Aging health disparities research framework. Ethnicity & disease, 25(3), 245.

Hurd, M. D., Martorell, P., Delavande, A., Mullen, K. J., & Langa, K. M. (2013). Monetary costs
of dementia in the United States. New England Journal of Medicine, 368(14), 1326-1334.

Jacobson, L. S., LaLonde, R. J., & Sullivan, D. G. (1993). Earnings losses of displaced workers.
The American economic review, 685-709.

Jitlal, M., Amirthalingam, G. N., Karania, T., Parry, E., Neligan, A., Dobson, R., ... & Marshall,
C. R. (2021). The Influence of Socioeconomic Deprivation on Dementia Mortality, Age
at Death, and Quality of Diagnosis: A Nationwide Death Records Study in England and
Wales 2001–2017. Journal of Alzheimer's Disease, 81(1), 321-328.

Kasper, J. D., Freedman, V. A., Spillman, B. C., & Wolff, J. L. (2015). The disproportionate
impact of dementia on family and unpaid caregiving to older adults. Health Affairs,
34(10), 1642-1649.

Katz, M. J., Lipton, R. B., Hall, C. B., Zimmerman, M. E., Sanders, A. E., Verghese, J., ... &
Derby, C. A. (2012). Age and sex specific prevalence and incidence of mild cognitive
impairment, dementia and Alzheimer’s dementia in blacks and whites: A report from the
Einstein Aging Study. Alzheimer disease and associated disorders, 26(4), 335.

Kenkel, D. S. (1991). Health behavior, health knowledge, and schooling. Journal of Political
Economy, 99(2), 287-305.

Kim, C., Kabeto, M. U., Wallace, R. B., & Langa, K. M. (2004). Quality of preventive clinical
services among caregivers in the health and retirement study. Journal of general internal
medicine, 19(8), 875-878.

Kim, D., Glazier, R. H., Zagorski, B., Kawachi, I., & Oreopoulos, P. (2018). Neighbourhood
socioeconomic position and risks of major chronic diseases and all-cause mortality: a
quasi-experimental study. BMJ open, 8(5), e018793.

Kinnear, H., Connolly, S., Rosato, M., Hall, C., Mairs, A., & O'Reilly, D. (2010). Are caregiving
responsibilities associated with non-attendance at breast screening?. BMC Public Health,
10(1), 1-6.

103
Knopman, D. S., Roberts, R. O., Geda, Y. E., Pankratz, V. S., Christianson, T. J., Petersen, R. C.,
& Rocca, W. A. (2010). Validation of the telephone interview for cognitive status-
modified in subjects with normal cognition, mild cognitive impairment, or dementia.
Neuroepidemiology, 34(1), 34-42.

Kolanowski, A. M., Fick, D., Waller, J. L., & Shea, D. (2004). Spouses of persons with
dementia: their healthcare problems, utilization, and costs. Research in nursing & health,
27(5), 296-306.

Koopmans, G. T., & Lamers, L. M. (2007). Gender and health care utilization: the role of mental
distress and help-seeking propensity. Social science & medicine, 64(6), 1216-1230.

Korhonen, K., Einiö, E., Leinonen, T., Tarkiainen, L., & Martikainen, P. (2020). Midlife
socioeconomic position and old-age dementia mortality: a large prospective register-
based study from Finland. BMJ open, 10(1), e033234.

Kotwal, A. A., Schumm, L. P., Kern, D. W., McClintock, M. K., Waite, L. J., Shega, J. W., ... &
Dale, W. (2015). Evaluation of a brief survey instrument for assessing subtle differences
in cognitive function among older adults. Alzheimer disease and associated disorders,
29(4), 317.

Laks, J., Goren, A., Dueñas, H., Novick, D., & Kahle‐Wrobleski, K. (2016). Caregiving for
patients with Alzheimer's disease or dementia and its association with psychiatric and
clinical comorbidities and other health outcomes in Brazil. International Journal of
Geriatric Psychiatry, 31(2), 176-185.

Langa, K. M., Larson, E. B., Wallace, R. B., Fendrick, A. M., Foster, N. L., Kabeto, M. U., ... &
Herzog, A. R. (2004). Out-of-pocket health care expenditures among older Americans
with dementia. Alzheimer Disease & Associated Disorders, 18(2), 90-98.

Langa, K. M., Larson, E. B., Crimmins, E. M., Faul, J. D., Levine, D. A., Kabeto, M. U., &
Weir, D. R. (2017). A comparison of the prevalence of dementia in the United States in
2000 and 2012. JAMA internal medicine, 177(1), 51-58.
Leggett, A. N., Sonnega, A. J., & Lohman, M. C. (2020). Till death do us part: intersecting
health and spousal dementia caregiving on caregiver mortality. Journal of aging and
health, 32(7-8), 871-879.

Leibson, C. L., Long, K. H., Ransom, J. E., Roberts, R. O., Hass, S. L., Duhig, A. M., ... &
Petersen, R. C. (2015). Direct medical costs and source of cost differences across the
spectrum of cognitive decline: a population-based study. Alzheimer's & Dementia, 11(8),
917-932.

104
Lian, Y., Xiao, L. D., Zeng, F., Wu, X., Wang, Z., & Ren, H. (2017). The experiences of people
with dementia and their caregivers in dementia diagnosis. Journal of Alzheimer's Disease,
59(4), 1203-1211.

Lièvre, A., Alley, D., & Crimmins, E. M. (2008). Educational differentials in life expectancy
with cognitive impairment among the elderly in the United States. Journal of Aging and
Health, 20(4), 456-477.

Lin, P. J., & Neumann, P. J. (2013). The economics of mild cognitive impairment. Alzheimer's
& Dementia, 9(1), 58-62.

Lin, P. J., Zhong, Y., Fillit, H. M., Chen, E., & Neumann, P. J. (2016). Medicare expenditures of
individuals with Alzheimer's disease and related dementias or mild cognitive impairment
before and after diagnosis. Journal of the American Geriatrics Society, 64(8), 1549-1557.

Lin, P. J., D’Cruz, B., Leech, A. A., Neumann, P. J., Sanon Aigbogun, M., Oberdhan, D., &
Lavelle, T. A. (2019). Family and caregiver spillover effects in cost-utility analyses of
Alzheimer’s disease interventions. Pharmacoeconomics, 37(4), 597-608.

Livney, M. G., Clark, C. M., Karlawish, J. H., Cartmell, S., Negrón, M., Nuñez, J., ... & Arnold,
S. E. (2011). Ethnoracial differences in the clinical characteristics of Alzheimer's disease
at initial presentation at an urban Alzheimer's disease center. The American Journal of
Geriatric Psychiatry, 19(5), 430-439.

Llinàs-Regla, J., López-Pousa, S., Vilalta-Franch, J., Garre-Olmo, J., & Román, G. C. (2008).
Mortality after a diagnosis of dementia in a population aged 75 and over in Spain.
Neuroepidemiology, 31(2), 80-88.

Long, K. H., Moriarty, J. P., Mittelman, M. S., & Foldes, S. S. (2014). Estimating the potential
cost savings from the New York University Caregiver Intervention in Minnesota. Health
Affairs, 33(4), 596-604.

Ma, M., Dorstyn, D., Ward, L., & Prentice, S. (2018). Alzheimers’ disease and caregiving: a
meta-analytic review comparing the mental health of primary carers to controls. Aging &
mental health, 22(11), 1395-1405.
Manne, S., Kashy, D., Weinberg, D. S., Boscarino, J. A., & Bowen, D. J. (2012). Using the
interdependence model to understand spousal influence on colorectal cancer screening
intentions: A structural equation model. Annals of Behavioral Medicine, 43(3), 320-329.

Manuel, J. I. (2018). Racial/ethnic and gender disparities in health care use and access. Health
services research, 53(3), 1407-1429.

105
Martín-García, R., Martín-Avila, G., De la Rubia-Marcos, M., Maroto-Rodríguez, R., Ortega-
Angulo, C., Rodriguez, M. T. C., ... & Veiga, A. B. G. (2016). Consumption of drugs and
nonpharmacological therapies in caregivers of patients with Alzheimer's Disease: a case-
control study in Madrid. Dementia and geriatric cognitive disorders extra, 6(1), 68-77.

Massimo, L., Zee, J., Xie, S. X., McMillan, C. T., Rascovsky, K., Irwin, D. J., ... & Grossman,
M. (2015). Occupational attainment influences survival in autopsy-confirmed
frontotemporal degeneration. Neurology, 84(20), 2070-2075.

Matthews, K. A., Xu, W., Gaglioti, A. H., Holt, J. B., Croft, J. B., Mack, D., & McGuire, L. C.
(2019). Racial and ethnic estimates of Alzheimer's disease and related dementias in the
United States (2015–2060) in adults aged≥ 65 years. Alzheimer's & Dementia, 15(1), 17-
24.

Mausbach, B. T., Chattillion, E. A., Roepke, S. K., Patterson, T. L., & Grant, I. (2013). A
comparison of psychosocial outcomes in elderly Alzheimer caregivers and noncaregivers.
The American Journal of Geriatric Psychiatry, 21(1), 5-13.

Mayeda, E. R., Glymour, M. M., Quesenberry, C. P., Johnson, J. K., Pérez-Stable, E. J., &
Whitmer, R. A. (2017). Survival after dementia diagnosis in five racial/ethnic groups.
Alzheimer's & Dementia, 13(7), 761-769.

McGuire, L. C., Bouldin, E. L., Andresen, E. M., & Anderson, L. A. (2010). Examining
modifiable health behaviors, body weight, and use of preventive health services among
caregivers and non-caregivers aged 65 years and older in Hawaii, Kansas, and
Washington using 2007 BRFSS. The journal of nutrition, health & aging, 14(5), 373-379.

McGuire, T. G., Newhouse, J. P., & Sinaiko, A. D. (2011). An economic history of Medicare
Part C. The Milbank Quarterly, 89(2), 289-332.

Mead, N., & Roland, M. (2009). Understanding why some ethnic minority patients evaluate
medical care more negatively than white patients: a cross sectional analysis of a routine
patient survey in English general practices. Bmj, 339.

Mehta, K. M., & Yeo, G. W. (2017). Systematic review of dementia prevalence and incidence in
United States race/ethnic populations. Alzheimer's & Dementia, 13(1), 72-83.

Meng, X., & D’arcy, C. (2012). Education and dementia in the context of the cognitive reserve
hypothesis: a systematic review with meta-analyses and qualitative analyses. PloS one,
7(6), e38268.

106
Meng, X., D'Arcy, C., Tempier, R., Kou, C., Morgan, D., & Mousseau, D. D. (2012). Survival of
patients with incident dementia who had a pre‐existing psychiatric disorder: A
population‐based 7‐year follow‐up study. International journal of geriatric psychiatry,
27(7), 683-691.

Mielke, M. M., Vemuri, P., & Rocca, W. A. (2014). Clinical epidemiology of Alzheimer’s
disease: assessing sex and gender differences. Clinical epidemiology, 6, 37.

Montano, D. (2021). Socioeconomic status, well-being and mortality: a comprehensive life
course analysis of panel data, Germany, 1984-2016. Archives of Public Health, 79(1), 1-
15.

Moore, M. J., Zhu, C. W., & Clipp, E. C. (2001). Informal costs of dementia care: estimates from
the National Longitudinal Caregiver Study. The Journals of Gerontology Series B:
Psychological Sciences and Social Sciences, 56(4), S219-S228.

Mukadam, N., Cooper, C., & Livingston, G. (2013). Improving access to dementia services for
people from minority ethnic groups. Current Opinion in Psychiatry, 26(4), 409.

Musich, S., Wang, S. S., Kraemer, S., Hawkins, K., & Wicker, E. (2017). Caregivers for older
adults: Prevalence, characteristics, and health care utilization and expenditures. Geriatric
Nursing, 38(1), 9-16.

National Center for Health Statistics. (2021). Mortality in the United States, 2020. Retrieved
from https://www.cdc.gov/nchs/data/databriefs/db427.pdf

National Institute on Aging. (2020). Care, Services, and Supports for Persons with Dementia and
Their Caregivers: Summit Virtual Meeting Series – Summary Report. Retrieved from
https://www.nia.nih.gov/2020-dementia-care-summit

O’fstedal, M. B., Fisher, G. G., & Herzog, A. R. (2005). Documentation of cognitive functioning
measures in the Health and Retirement Study. Ann Arbor, MI: University of Michigan,
10.

Ono, R., Uchida, K., Nakatsuka, K., Megumi, M., & Fukuda, H. (2022). Economic Status and
Mortality in Patients with Alzheimer's Disease in Japan: The Longevity Improvement and
Fair Evidence Study. Journal of the American Medical Directors Association, 23(1), 161-
164.

107
O'Reilly, F., Finnan, F., Allwright, S., Smith, G. D., & Ben-Shlomo, Y. (1996). The effects of
caring for a spouse with Parkinson's disease on social, psychological and physical well-
being. British Journal of General Practice, 46(410), 507-512.

Oremus, M., & Aguilar, S. (2011). A systematic review to assess the policy-making relevance of
dementia cost-of-illness studies in the US and Canada. Pharmacoeconomics, 29(2), 141-
156.

Ostbye, T., Taylor Jr, D. H., Clipp, E. C., Scoyoc, L. V., & Plassman, B. L. (2008). Identification
of dementia: agreement among national survey data, medicare claims, and death
certificates. Health services research, 43(1p1), 313-326.

Paradise, M., Cooper, C., & Livingston, G. (2009). Systematic review of the effect of education
on survival in Alzheimer's disease. International psychogeriatrics, 21(1), 25-32.

Paradies, Y., Truong, M., & Priest, N. (2014). A systematic review of the extent and
measurement of healthcare provider racism. Journal of general internal medicine, 29(2),
364-387.

Pavlik, V. N., Doody, R. S., Massman, P. J., & Chan, W. (2006). Influence of premorbid IQ and
education on progression of Alzheimer’s disease. Dementia and Geriatric Cognitive
Disorders, 22(4), 367-377.

Pearlin, L. I., Mullan, J. T., Semple, S. J., & Skaff, M. M. (1990). Caregiving and the stress
process: An overview of concepts and their measures. The gerontologist, 30(5), 583-594.

Petersen, J. D., Wehberg, S., Packness, A., Svensson, N. H., Hyldig, N., Raunsgaard, S., ... &
Waldorff, F. B. (2021). Association of socioeconomic status with dementia diagnosis
among older adults in Denmark. JAMA network open, 4(5), e2110432-e2110432.

Petrovic, D., de Mestral, C., Bochud, M., Bartley, M., Kivimäki, M., Vineis, P., ... & Stringhini,
S. (2018). The contribution of health behaviors to socioeconomic inequalities in health: a
systematic review. Preventive medicine, 113, 15-31.

Pinquart, M., & Sörensen, S. (2005). Ethnic differences in stressors, resources, and psychological
outcomes of family caregiving: A meta-analysis. The Gerontologist, 45(1), 90-106.

Piovezan, R. D., Oliveira, D., Arias, N., Acosta, D., Prince, M. J., & Ferri, C. P. (2020).
Mortality Rates and Mortality Risk Factors in Older Adults with Dementia from Low-and
Middle-Income Countries: The 10/66 Dementia Research Group Population-Based
Cohort Study. Journal of Alzheimer's Disease, 75(2), 581-593.

108
Plassman, B. L., Langa, K. M., McCammon, R. J., Fisher, G. G., Potter, G. G., Burke, J. R., ... &
Wallace, R. B. (2011). Incidence of dementia and cognitive impairment, not dementia in
the United States. Annals of neurology, 70(3), 418-426.

Pressley, J. C., Trott, C., Tang, M., Durkin, M., & Stern, Y. (2003). Dementia in community-
dwelling elderly patients: A comparison of survey data, medicare claims, cognitive
screening, reported symptoms, and activity limitations. Journal of clinical epidemiology,
56(9), 896-905.

Qian, W., Schweizer, T. A., & Fischer, C. E. (2014). Impact of socioeconomic status on initial
clinical presentation to a memory disorders clinic. International Psychogeriatrics, 26(4),
597-603.

Qiu, C., Bäckman, L., Winblad, B., Agüero-Torres, H., & Fratiglioni, L. (2001). The influence of
education on clinically diagnosed dementia incidence and mortality data from the
Kungsholmen Project. Archives of neurology, 58(12), 2034-2039.

Rahman, A., Anjum, R., & Sahakian, Y. (2019). Impact of caregiving for dementia patients on
healthcare utilization of caregivers. Pharmacy, 7(4), 138.

Reeves, K. W., Bacon, K., & Fredman, L. (2012). Caregiving associated with selected cancer
risk behaviors and screening utilization among women: cross-sectional results of the
2009 BRFSS. BMC public health, 12(1), 1-8.

Rivera-Hernandez, M., Kumar, A., Epstein-Lubow, G., & Thomas, K. S. (2019). Disparities in
nursing home use and quality among African American, Hispanic, and White Medicare
residents with Alzheimer’s disease and related dementias. Journal of aging and health,
31(7), 1259-1

Rote, S. M., Angel, J. L., Moon, H., & Markides, K. (2019). Caregiving across diverse
populations: New evidence from the national study of caregiving and Hispanic EPESE.
Innovation in aging, 3(2), igz033.

Roth, D. L., Dilworth-Anderson, P., Huang, J., Gross, A. L., & Gitlin, L. N. (2015a). Positive
aspects of family caregiving for dementia: Differential item functioning by race. Journals
of Gerontology Series B: Psychological Sciences and Social Sciences, 70(6), 813-819.

Roth, D. L., Fredman, L., & Haley, W. E. (2015b). Informal caregiving and its impact on health:
A reappraisal from population-based studies. The Gerontologist, 55(2), 309-319.

109
Rountree, S. D., Chan, W., Pavlik, V. N., Darby, E. J., & Doody, R. S. (2012). Factors that
influence survival in a probable Alzheimer disease cohort. Alzheimer's research &
therapy, 4(3), 1-6.

Ruitenberg, A., Ott, A., van Swieten, J. C., Hofman, A., & Breteler, M. M. (2001). Incidence of
dementia: does gender make a difference?. Neurobiology of aging, 22(4), 575-580.

Russ, T. C., Stamatakis, E., Hamer, M., Starr, J. M., Kivimäki, M., & Batty, G. D. (2013).
Socioeconomic status as a risk factor for dementia death: individual participant meta-
analysis of 86 508 men and women from the UK. The British journal of psychiatry,
203(1), 10-17.

Sakshaug, J. W., Weir, D. R., & Nicholas, L. H. (2014). Identifying diabetics in Medicare claims
and survey data: implications for health services research. BMC health services research,
14(1), 1-6.

Sala, E., Knies, G., & Burton, J. (2014). Propensity to consent to data linkage: experimental
evidence on the role of three survey design features in a UK longitudinal panel.
International Journal of Social Research Methodology, 17(5), 455-473.

Savva, G. M., & Arthur, A. (2015). Who has undiagnosed dementia? A cross-sectional analysis
of participants of the Aging, Demographics and Memory Study. Age and ageing, 44(4),
642-647.

Scharlach, A. E., Midanik, L. T., Runkle, M. C., & Soghikian, K. (1997). Health practices of
adults with elder care responsibilities. Preventive medicine, 26(2), 155-161.

Schoenmakers, B., Buntinx, F., & Delepeleire, J. (2010). Factors determining the impact of care-
giving on caregivers of elderly patients with dementia. A systematic literature review.
Maturitas, 66(2), 191-200.

Schulz, R., Newsom, J., Mittelmark, M., Burton, L., Hirsch, C., & Jackson, S. (1997). Health
effects of caregiving: the caregiver health effects study: an ancillary study of the
Cardiovascular Health Study. Annals of Behavioral Medicine, 19(2), 110-116.

Schulz, R., & Martire, L. M. (2004). Family caregiving of persons with dementia: prevalence,
health effects, and support strategies. The American journal of geriatric psychiatry, 12(3),
240-249.

110
Shaffer, K. M., & Nightingale, C. L. (2020). Comparison of healthcare utilization between
informal caregivers and non-caregivers: an analysis of the Health Information National
Trends Survey. Journal of aging and health, 32(5-6), 453-461.

Shaw, W. S., Patterson, T. L., Semple, S. J., Ho, S., Irwin, M. R., Hauger, R. L., & Grant, I.
(1997). Longitudinal analysis of multiple indicators of health decline among spousal
caregivers. Annals of Behavioral Medicine, 19(2), 101-109.

Sheehan, O. C., Haley, W. E., Howard, V. J., Huang, J., Rhodes, J. D., & Roth, D. L. (2021).
Stress, burden, and well-being in dementia and nondementia caregivers: Insights from the
caregiving transitions study. The Gerontologist, 61(5), 670-679.

Shepherd, S. M., Willis-Esqueda, C., Paradies, Y., Sivasubramaniam, D., Sherwood, J., &
Brockie, T. (2018). Racial and cultural minority experiences and perceptions of health
care provision in a mid-western region. International journal for equity in health, 17(1),
1-10.

Son, J., Erno, A., Shea, D. G., Femia, E. E., Zarit, S. H., & Parris Stephens, M. A. (2007). The
caregiver stress process and health outcomes. Journal of aging and health, 19(6), 871-
887.

Spering, C. C., Hobson, V., Lucas, J. A., Menon, C. V., Hall, J. R., & O’Bryant, S. E. (2012).
Diagnostic accuracy of the MMSE in detecting probable and possible Alzheimer's disease
in ethnically diverse highly educated individuals: an analysis of the NACC database.
Journals of Gerontology Series A: Biomedical Sciences and Medical Sciences, 67(8),
890-896.

St.Clair, P., Gaudette, É., Zhao, H., Tysinger, B., Seyedin, R., & Goldman, D. P. (2017). Using
self-reports or claims to assess disease prevalence: It's complicated. Medical care, 55(8),
782.

Stella, F., Radanovic, M., Balthazar, M. L., Canineu, P. R., de Souza, L. C., & Forlenza, O. V.
(2014). Neuropsychiatric symptoms in the prodromal stages of dementia. Current opinion
in psychiatry, 27(3), 230-235.

Stern, Y. (2006). Cognitive reserve and Alzheimer disease. Alzheimer Disease & Associated
Disorders, 20, S69-S74.

Stern, Y., Albert, S., Tang, M. X., & Tsai, W. Y. (1999). Rate of memory decline in AD is
related to education and occupation: cognitive reserve?. Neurology, 53(9), 1942-1942.
111
Stern, Y., Tang, M. X., Denaro, J., & Mayeux, R. (1995). Increased risk of mortality in
Alzheimer's disease patients with more advanced educational and occupational
attainment. Annals of Neurology: Official Journal of the American Neurological
Association and the Child Neurology Society, 37(5), 590-595.

Stringhini, S., Sabia, S., Shipley, M., Brunner, E., Nabi, H., Kivimaki, M., & Singh-Manoux, A.
(2010). Association of socioeconomic position with health behaviors and mortality. Jama,
303(12), 1159-1166.

Stringhini, S., Dugravot, A., Kivimaki, M., Shipley, M., Zins, M., Goldberg, M., ... & Singh-
Manoux, A. (2011). Do different measures of early life socioeconomic circumstances
predict adult mortality? Evidence from the British Whitehall II and French GAZEL
studies. J Epidemiol Community Health, 65(12), 1097-1103.

Suehs, B. T., Davis, C. D., Alvir, J., van Amerongen, D., PharmD, N. C. P., Joshi, A. V., ... &
Shah, S. N. (2013). The clinical and economic burden of newly diagnosed Alzheimer’s
disease in a medicare advantage population. American Journal of Alzheimer's Disease &
Other Dementias®, 28(4), 384-392.

Suehs, B. T., Shah, S. N., Davis, C. D., Alvir, J., Faison, W. E., Patel, N. C., ... & Bobula, J.
(2014). Household Members of Persons with A lzheimer's Disease: Health Conditions,
Healthcare Resource Use, and Healthcare Costs. Journal of the American Geriatrics
Society, 62(3), 435-441.

Syed, S. T., Gerber, B. S., & Sharp, L. K. (2013). Traveling towards disease: transportation
barriers to health care access. Journal of community health, 38(5), 976-993.

Tang, M. X., Cross, P., Andrews, H., Jacobs, D. M., Small, S., Bell, K., ... & Mayeux, R. (2001).
Incidence of AD in African-Americans, Caribbean hispanics, and caucasians in northern
Manhattan. Neurology, 56(1), 49-56.

Taylor Jr, D. H., Østbye, T., Langa, K. M., Weir, D., & Plassman, B. L. (2009). The accuracy of
Medicare claims as an epidemiological tool: the case of dementia revisited. Journal of
Alzheimer's Disease, 17(4), 807-815.

Taylor Jr, D. H., Fillenbaum, G. G., & Ezell, M. E. (2002). The accuracy of medicare claims data
in identifying Alzheimer's disease. Journal of clinical epidemiology, 55(9), 929-937.

112
Thorpe, C. T., Fowler, N. R., Harrigan, K., Zhao, X., Kang, Y., Hanlon, J. T., ... & Thorpe, J. M.
(2016). Racial and ethnic differences in initiation and discontinuation of antidementia
drugs by Medicare beneficiaries. Journal of the American Geriatrics Society, 64(9), 1806-
1814.

Thunell, J., Ferido, P., & Zissimopoulos, J. (2019). Measuring Alzheimer’s disease and other
dementias in diverse populations using medicare claims data. Journal of Alzheimer's
Disease, 72(1), 29-33.

Tom, S. E., Hubbard, R. A., Crane, P. K., Haneuse, S. J., Bowen, J., McCormick, W. C., ... &
Larson, E. B. (2015). Characterization of dementia and Alzheimer’s disease in an older
population: updated incidence and life expectancy with and without dementia. American
journal of public health, 105(2), 408-413.

Trivedi, A. N., Moloo, H., & Mor, V. (2010). Increased ambulatory care copayments and
hospitalizations among the elderly. New England Journal of Medicine, 362(4), 320-328.

U.S. Bureau of Labor Statistics. CPI Inflation Calculator. Retrieved from
https://www.bls.gov/data/inflation_calculator.htm

U.S. Census Bureau. (2021). 2020 Census Illuminates Racial and Ethnic Composition of the
Country. Retrieved from https://www.census.gov/library/stories/2021/08/improved-race-
ethnicity-measures-reveal-united-states-population-much-more-multiracial.html

U.S. Department of Health and Human Services. (2021). National Plan to Address Alzheimer’s
Disease: 2021 Update. Retrieved from https://aspe.hhs.gov/reports/national-plan-2021-
update

Valcour, V. G., Masaki, K. H., Curb, J. D., & Blanchette, P. L. (2000). The detection of
dementia in the primary care setting. Archives of internal medicine, 160(19), 2964-2968.

Valenzuela, M., Brayne, C., Sachdev, P., Wilcock, G., & Fiona Matthews on Behalf of the
Medical Research Council Cognitive Function and Ageing Study. (2011). Cognitive
lifestyle and long-term risk of dementia and survival after diagnosis in a multicenter
population-based cohort. American journal of epidemiology, 173(9), 1004-1012.

Van De Vorst, I. E., Koek, H. L., Stein, C. E., Bots, M. L., & Vaartjes, I. (2016). Socioeconomic
disparities and mortality after a diagnosis of dementia: results from a nationwide registry
linkage study. American journal of epidemiology, 184(3), 219-226.

Van Houtven, C. H., Wilson, M. R., & Clipp, E. C. (2005). Informal care intensity and caregiver
drug utilization. Review of Economics of the Household, 3(4), 415-433.

113
Vermunt, L., Sikkes, S. A., Van Den Hout, A., Handels, R., Bos, I., Van Der Flier, W. M., ... &
AIBL Research Group. (2019). Duration of preclinical, prodromal, and dementia stages
of Alzheimer's disease in relation to age, sex, and APOE genotype. Alzheimer's &
Dementia, 15(7), 888-898.

Waring, S. C., Doody, R. S., Pavlik, V. N., Massman, P. J., & Chan, W. (2005). Survival among
patients with dementia from a large multi-ethnic population. Alzheimer Disease &
Associated Disorders, 19(4), 178-183.

Weimer, D. L., & Sager, M. A. (2009). Early identification and treatment of Alzheimer's disease:
social and fiscal outcomes. Alzheimer's & Dementia, 5(3), 215-226.

Weir, D., & Langa, K. (2018). How Well Can Medicare Records Identify Seniors with Cognitive
Impairment Needing Assistance with Financial Management?. Michigan Retirement
Research Center Research Paper, (2019-391).

Wharton, T., & Zivin, K. (2017). Relationships among caregiving, income, gender, and health: A
cross-sectional examination of a representative sample of older Americans. Journal of
social service research, 43(1), 141-148.

White, L., Ingraham, B., Larson, E., Fishman, P., Park, S., & Coe, N. B. (2021). Observational
study of patient characteristics associated with a timely diagnosis of dementia and mild
cognitive impairment without dementia. Journal of General Internal Medicine, 1-9.

Wilkins, C. H., Wilkins, K. L., Meisel, M., Depke, M., Williams, J., & Edwards, D. F. (2007).
Dementia undiagnosed in poor older adults with functional impairment. Journal of the
American Geriatrics Society, 55(11), 1771-1776.

Wilkinson, T., Ly, A., Schnier, C., Rannikmäe, K., Bush, K., Brayne, C., ... & UK, D. P. (2018).
Identifying dementia cases with routinely collected health data: a systematic review.
Alzheimer's & Dementia, 14(8), 1038-1051.

Wolfson, C., Wolfson, D. B., Asgharian, M., M'Lan, C. E., Østbye, T., Rockwood, K., & Hogan,
D. F. (2001). A reevaluation of the duration of survival after the onset of dementia. New
England Journal of Medicine, 344(15), 1111-1116.

Zajacova, A., & Lawrence, E. M. (2018). The relationship between education and health:
reducing disparities through a contextual approach. Annual review of public health, 39,
273-289.

114
Zhao, Y., Kuo, T. C., Weir, S., Kramer, M. S., & Ash, A. S. (2008). Healthcare costs and
utilization for Medicare beneficiaries with Alzheimer's. BMC health services research,
8(1), 1-8.

Zhu, C. W., Cosentino, S., Ornstein, K., Gu, Y., Scarmeas, N., Andrews, H., & Stern, Y. (2015).
Medicare utilization and expenditures around incident dementia in a multiethnic cohort.
Journals of Gerontology Series A: Biomedical Sciences and Medical Sciences, 70(11),
1448-1453.

Zissimopoulos, J., Crimmins, E., & Clair, P. S. (2015, January). The value of delaying
Alzheimer’s disease onset. In Forum for Health Economics and Policy (Vol. 18, No. 1,
pp. 25-39). De Gruyter.

Zissimopoulos, J. M., Tysinger, B. C., St. Clair, P. A., & Crimmins, E. M. (2018). The impact of
changes in population health and mortality on future prevalence of Alzheimer’s disease
and other dementias in the United States. The Journals of Gerontology: Series B,
73(suppl_1), S38-S47.

Abstract (if available)

Abstract Amid the substantial burden of dementia on population level, certain groups of older Americans are disproportionately affected by the disease. This three-essay dissertation advances the literature of disparities in dementia burden in the U.S., by investigating the magnitude and mechanism of heterogeneity throughout the dementia care continuum, ranging from diagnosis and care of persons living with dementia (PLWD) to family support. With the guidance of health economics theories, this series of studies identify hypotheses of groups bearing a disproportionate dementia burden, use novel linkage of longitudinal representative survey and administrative claims, and apply rigorous methods to test for these hypotheses. The first chapter on diagnosis disparities incorporates a previously neglected longitudinal perspective and shows racial/ethnic minorities are more susceptible to under- and delayed- diagnosis, resulting in missed opportunities for improving well-being with timely actions. The second chapter reveals greater comorbidity burden and function/cognitive impairment at the time of incident dementia diagnosis among those with lower education and wealth. Net of these health differences, the highly educated do not outlive the lower-educated, whereas the richest survive longer than the deprived after diagnosis. The third chapter investigates whether spouses of PLWD experience disruption in health care, to complement understanding of dementia impacts on families. I show that older adults do not skimp on or spend more on health care ensuing their spouse’s incident diagnosis, yet certain groups such as males forgo ambulatory care. These findings together highlight the importance of targeted interventions throughout dementia care cycle in an aging and diversifying America, as well as inform allocation of limited resources in achieving health equities.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

More knowledge, better plans: a study of heterogeneity in dementia prevalence and mid-life cognitive changes

PDF

Health care utilization and spending of the U.S. aging population

PDF

The role of public policy in the decisions of parents and caregivers: an examination of work, fertility, and informal caregiving

PDF

A biodemographic approach to understanding sociodemographic disparities in kidney functioning on three dimensions: individual, population, and cross-national

PDF

Three essays on health, aging, and retirement

PDF

Adoption and implementation of innovative diagnostic tools for Alzheimer's Disease: challenges and barriers in primary care

PDF

Essays on work, retirement, and fostering longer working lives

PDF

Essays on health and aging with focus on the spillover of human capital

PDF

Social engagement and cognitive decline in Mexican Americans: implications for age-friendly cities

PDF

Redlining, neighborhood change, and individual outcomes: an exploration of how space shapes the landscape of inequality from past to present

PDF

Topics in U.S. healthcare

PDF

Biomarkers of age-related health changes: associations with health outcomes and disparities

PDF

Three essays on aging, wealth, and housing tenure transitions

PDF

Investigating racial and ethnic disparities in patient experiences with care and health services use following colorectal cancer diagnosis among older adults with comorbid chronic conditions

PDF

Defining the global burden of burn injuries: a 29 year review of data from 196 countries

PDF

Investigating daily effects of activity space-based built environment exposures on physical activity behaviors in Hispanic women during pregnancy and early postpartum

PDF

The disease burden among aging people living with HIV/AIDS

PDF

Residential care in Los Angeles: policy and planning for an aging population

PDF

The dollars and sense of caregiving: exploring the financial financial realities of providing informal care to an older adult

PDF

Crosscultural differences in dementia diagnosis and care-seeking in Hispanic and non-Hispanic white outpatients

Asset Metadata

Creator Chen, Yi (author)

Core Title Uncovering hidden figures: disparities in dementia burden in an aging America

School School of Policy, Planning and Development

Degree Doctor of Philosophy

Degree Program Public Policy and Management

Publication Date 03/21/2022

Defense Date 02/28/2022

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

Tag dementia,dementia burden,disparities,family spillover,OAI-PMH Harvest

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Zissimopoulos, Julie (committee chair), Chen, Alice (committee member), Crimmins, Eileen (committee member)

Creator Email chen530@usc.edu,chenyifdu@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC110816574

Unique identifier UC110816574

Document Type Dissertation

Format application/pdf (imt)

Rights Chen, Yi

Type texts

Source 20220323-usctheses-batch-917 (batch), 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 author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu