University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
More Knowledge, Better Plans:
A Study of Heterogeneity in Dementia Prevalence and Mid-Life Cognitive
Changes

by

Yingying Zhu

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA

In Partial Fulfillment of the
Requirements of the Degree

DOCTOR OF PHILOSOPHY
PUBLIC POLICY AND MANAGEMENT

December 2020

ii

Table of Contents
List of Tables ................................................................................................................................. iii
List of Figures ................................................................................................................................ iv
Abstract ............................................................................................................................................v
Introduction ......................................................................................................................................1
Chapter 1: Sex, Race, and Age Differences in Prevalence of Dementia in Medicare Claims and
Survey Data ......................................................................................................................................6
INTRODUCTION ..................................................................................................8
METHODS...........................................................................................................10
RESULTS.............................................................................................................17
DISCUSSION ......................................................................................................24
REFERENCE .......................................................................................................31
Chapter 2: Cognitive Changes at Middle Age .............................................................................35
INTRODUCTION ................................................................................................36
EMPIRICAL LITERATURE REVIEW ..............................................................38
DATA AND METHODS .....................................................................................41
RESULTS.............................................................................................................47
DISCUSSION ......................................................................................................55
REFERENCE .......................................................................................................59
Chapter 3: The Effect of Cognitive Decline on Labor Supply of Older Adults ...........................64
INTRODUCTION ................................................................................................65
EMPIRICAL LITERATURE ...............................................................................68
THEORETICAL FRAMEWORK .......................................................................70
EMPIRICAL STRATEGY ..................................................................................73
RESULTS.............................................................................................................77
DISCUSSION ......................................................................................................88
REFERENCE .......................................................................................................92
Chapter 4: Conclusion ..................................................................................................................96
Bibliography ...............................................................................................................................101
Appendices .................................................................................................................................112

iii

List of Tables
Table 1. Sample characteristics from ADAMS, HRS and Medicare claims data sources, ages 70
and above, 2004 ……… ……………………………………………………………… …. ………18
Table 2. Odds ratios for the presence of dementia based on cognitive tests (HRS) and diagnosis
codes (Medicare claims), ages 67 and above, 2006-2013……………………………..……..….21

Table 3. Comparison of analysis in a 2-year and a 2 to 14-year window … ………… ….....…….47
Table 4. Odds ratio of cognitive decline and improvement in 2 years, 1996-2014……… ….…..51
Table 5. Sample characteristics, 1996-2014……………………………………….……...……..80
Table 6. Correlation and FE results of the effect of short-term cognitive decline on the expected
probability of working at age 62, 1996-2014.…… ………… ………….……………..…..….…..85
Table 7. Correlation and OLS Results of The Effect of Individual-Specific Trend of Cognitive
Decline on Change in The Expected Probability of Working at Age 62, 1996-2014………..….86

iv

List of Figures
Figure 1. Dementia prevalence for the US population and by race, gender, and age in
neuropsychological assessment (ADAMS), cognitive tests (HRS), and diagnosis codes
(Medicare claims), ages 70 and above, 2004 with 95% confidence intervals ……………... ……19

Figure 2. Predicted values of dementia prevalence based on cognitive tests (HRS) and diagnosis
codes (claims) from logistic models adjusting for race, sex, age group, and wave in HRS and
claims, ages 67 and above, 2006 and 2012……… ……………………………………… ………22
Figure 3. Dementia prevalence by sex (Figure 3a), age (Figure 3b), and race (Figure 3c) based on
cognitive tests (HRS) and diagnosis codes (claims) from logistic models adjusting for sex, age
group, and waves in HRS and claims, ages 67 and above, 2006 and 2012 ……………….. …….23

Figure 4. The distribution of cognitive score among persons aged 50 to 64, 1996-2014 ………..48
Figure 5. The distribution of change in standardized cognitive score from time t-1 to time t
among cognitively normal persons, 1996-2014… …………………………………………….…49
Figure 6. Odds ratio (95%CI) of cognitive decline/improvement in the short run, 1996-2014…50
Figure 7. Odds ratio (95%CI) of cognitive decline/improvement in the long run, 1996-2014….54
Figure 8. The distribution of the expected probability of full-time working at age 62 among
workers aged 50 to 61, 1996-2014……………………………………………………………….77
Figure 9. The distribution of the change in the expected probability of full-time working at age
62 among workers aged 50 to 61 from wave t-1 to wave t, 1996-2014…… ……………………78
Figure 10. The distribution of cognitive function for workers aged 50 to 61 at time t-1, 1996-
2014………………………………………………………………………………………………79
Figure 11. The distribution of the change in cognitive scores for workers aged 50 to 61 from time
t-1 to time t, 1996-2014………………………………………………………………………….80
Figure 12. The distribution of cognitive score at the beginning of the observation window (t0),
1996-2014 ……………………………………………………………………………………..…83
Figure 13. The distribution of person-specific slopes of change in cognitive score across waves,
1996-2014. ………………………………………………………………………………………84

v

Abstract
Alzheimer’s disease and related dementia (ADRD), as one most important disabling concern for
older adults, imposes great socioeconomic burdens on individuals, families, and society. To
provide more resources and address these challenges presented by dementia, the government in
2012 established a National Plan. However, critical information is missing for implementing two
goals of the National Plan - improve data and track the current situation of dementia, and prevent
and effectively treat dementia. The missing information includes current levels and trends of
dementia among different subgroups and factors associated with rapidly declining cognitive
function at middle age, which could be the starting point of dementia.
This dissertation fills these knowledge gaps by exploring the diversity in dementia prevalence
and mid-life cognitive trajectories. It aims to quantify levels and trends of dementia prevalence
for the US population and by race, sex, and age by using three population-based data sources. It
additionally identifies persons with the most rapid rates of changes in cognitive abilities at an
early stage and factors associated with these fast cognitive changes. Moreover, this dissertation
explores how rapidly declining cognition affects the labor supply at the early retirement age (62
years old).
The dissertation ultimately provides vital information for individuals, families, health care
professionals, and policymakers to better plan for functional or pathological cognitive decline.
Additionally, it addresses policy challenges in implementing the National Plan. More
importantly, this dissertation suggests that there is disparity in the distribution of cognitive
trajectories. Ethnic minorities, those with poorer health conditions and lower SES are more likely
to experience cognitive decline and may have less access to health care services. They may need
more interventions on delaying or preventing cognitive decline and more financial and
vi

community support to buffer the consequences of dementia or middle-aged cognitive
deterioration.

1

Introduction
Alzheimer’s disease and related dementia (ADRD) is a neurodegenerative disease with
symptoms such as memory loss and language impairment at the early stage and difficulties in
walking and swallowing at later stages (Alzheimer's Association, 2019). Poor cognitive health is
one most important disabling concern for older adults (Singh-Manoux et al., 2012). Specifically,
the independence of older adults is impaired, which not only affects the life quality of persons
living with dementia (PLWDs) but also influences their families and particularly, caregivers.
These caregivers of dementia patients have higher risks of developing physical and mental
illness (Schulz & Sherwood, 2008). Particularly, dementia caregivers assist patients more and
report higher levels of stress, are more likely to give up vacations, and have more job difficulties
than other types of caregivers (Schulz & Martire, 2004). Dementia is also associated with a large
socioeconomic burden. The number of Americans ages 65 and older with Alzheimer’s disease
and other dementias was about 7 million in 2012 and is projected to increase to almost 12 million
by 2040 (Zissimopoulos, Tysinger, St Clair, & Crimmins, 2018). The annual population costs of
AD is estimated to increase from $307 billion, including $181 billion in medical costs paid out-
of-pocket and by Medicare and Medicaid, and $126 billion in the value of unpaid caregiving by
family, to $1.5 trillion by 2050 (Zissimopoulos, Crimmins, & St Clair, 2014).
In response to address this large socioeconomic burden, there is policy progress. The National
Alzheimer’s Project Act (NAPA) was signed into law in 2011, followed by a National Plan in
2012 established by the Department of Health and Human Services (United States Department of
Health and Human Services, 2012). This National Plan provided an opportunity for the
government to focus on Alzheimer’s disease and related dementia (ADRD) and set several main
goals to address many challenges presented by dementia. Two of these goals are particularly
2

related to this dissertation – Goal 5 (improve data to track progress of the National Plan) and
Goal 1 (prevent and treat dementia by 2025). The achievement of the two goals would help
individuals, families, and local policymakers to better plan for the disease (United States
Department of Health and Human Services, 2012).
Specifically, Goal 5 is to improve data that document dementia information so that the
government would better understand dementia and its impact at the population level and among
diverse subgroups. There is steady progress in quantifying and tracking trends in dementia
prevalence, suggesting a declining trend in dementia prevalence (Brookmeyer et al., 2011;
Crimmins, Saito, & Kim, 2016; Freedman, Kasper, Spillman, & Plassman, 2018; Zissimopoulos
et al., 2018). However, information about the current situation of dementia is limited. For
example, estimates of dementia prevalence for the US population vary widely, driven in large
part by differences in dementia ascertainment and study populations (Brookmeyer et al., 2011;
Chen, Tysinger, Crimmins, & Zissimopoulos, 2019).
Comparatively, Goal 1 is to prevent and treat dementia. Moreover, Goal 1 of the National Plan
was to prevent and treat dementia by 2025. One challenge for achieving this goal is that there is
no definite evidence for the effectiveness of pharmacologic and non-pharmacologic
interventions. For example, interventions such as blood pressure management, physical activity,
and cognitive training showed evidence with low-to-moderate strength on reducing dementia risk
(Blazer, Yaffe, & Liverman, 2015; National Academies of Sciences & Medicine, 2017). Recent
studies suggesting that dementia may occur 20 years or more before noticeable symptoms
(Alzheimer's Association, 2019; Beason-Held et al., 2013) may address the challenge and
facilitate the achievement of Goal 1. However, knowledge about mid-life cognitive trajectories,
3

particularly those with fast rates of cognitive decline, its risk/protective factors, and its impact is
limited.
The ultimate goal of these three chapters is to contribute to knowledge about the current situation
of dementia and cognitive changes, particularly about the diversity in dementia prevalence and
mid-life cognitive trajectories. This dissertation aims to quantify levels and trends of dementia
prevalence for the US population and by race, sex, and age by using three population-based data
sources. It additionally identifies persons with the most rapid rates of changes in cognitive
abilities at an early stage and risk factors associated with the most accelerated cognitive
trajectories. Moreover, this dissertation explores how fast speed of cognitive decline affects the
labor supply at the early retirement age. As a result, this dissertation fills existing literature gaps,
including inconsistent estimates of dementia prevalence across studies, and a lack of knowledge
about rapid cognitive changes for the middle-age group. The dissertation further provides key
information for individuals, families, health care professionals, and policymakers to better plan
for functional or pathological cognitive decline. Additionally, it addresses policy challenges in
achieving Goal 5 and Goal 1 of the National Plan: improve data and track the current situation of
dementia, and prevent and effectively treat dementia.
The first chapter compared levels and trends in dementia prevalence based on
neuropsychological assessment from ADAMS, cognitive test from HRS, and diagnosis codes
from Medicare claims for the US population and by age, sex, and race. This study found that in
2012, there remained substantial gaps in dementia prevalence based on cognitive test and
diagnosis codes among blacks and Hispanics (10.9 and 9.8 percentage points respectively). This
finding highlights the importance of understanding dementia prevalence among ethnic minorities
4

utilizing new data such as the HRS Harmonized Cognitive Assessment Protocol (HCAP) (Weir,
Langa, & Ryan, 2016).
Further, the second chapter describes the distribution of cognitive trajectories at middle age both
with a 2-year follow-up (short-term) and with a 2 to 14-year follow-up (long-term) and utilizes
the logistic regression to explore the association between baseline health and sociodemographic
risk factors and mid-life fast cognitive trajectories. This research finds that 8.4% of middle-aged
adults experienced fastest rates of short-term cognitive decline over 2 years, and 10.9% of them
experienced fast rates of short-term increase in cognitive function. The odds of cognitive
deterioration for African Americans is 1.89 times of that for whites, and the odds of cognitive
improvement for African Americans is 0.68 times of that for whites.
The third chapter utilizes the fixed-effect (FE) model and the ordinary least square model (OLS)
to examine the short-term and long-term effect of fast rates of cognitive decline on labor supply
at the early retirement age (62 years old). This study finds that a short-term (over 2 years) fast
rates of cognitive decline leads to a 2.2 percentage point decrease in the expected probability of
working at age 62. Comparatively, a long-term (over 2 to 10 years) fast speed of cognitive
decline leads to a 5.1 percentage point decrease in the expected probability of working.
Overall, these three chapters provide more information about the diversity in dementia
prevalence (both levels and trends; at the population level and by race, sex, and age) and in
middle-aged cognitive decline (both in the 2-year and in the 2 to 10/2 to 14-year window). These
findings help individuals, families, health care professionals, and policy makers to better
understand the current status of dementia and rapid cognitive trajectories at an earlier stage, thus
they can pay more attention to them, better plan to minimize their consequences, and ultimately
prevent or delay them. Specifically, the dissertation contributes to the achievement of the two
5

goals of the National Plan – improve data to better understand dementia and its impact, and
prevent and treat Alzheimer’s disease and related dementia by 2025.

6

Chapter 1:
Sex, Race, and Age Differences in Prevalence of Dementia in Medicare Claims
and Survey Data
by

Yingying Zhu, Yi Chen, Eileen Crimmins, and Julie Zissimopoulos

ABSTRACT
Objectives: This study provides the first comparison of trends in dementia prevalence in the US
population using three different dementia ascertainments/data sources: neuropsychological
assessment, cognitive tests, and diagnosis codes from Medicare claims.
Methods: We used data from the nationally representative Health and Retirement Study and
Aging, Demographics and Memory Study, and a 20% random sample of Medicare beneficiaries.
We compared dementia prevalence across the three sources by race, gender, and age. We
estimated trends in dementia prevalence from 2006 to 2013 based on cognitive tests and
diagnosis codes utilizing logistic regression.
Results: Dementia prevalence among older adults aged 70 and above in 2004 was 16.6%
(neuropsychological assessment), 15.8% (cognitive tests), and 12.2% (diagnosis codes). The
difference between dementia prevalence based on cognitive tests and diagnosis codes diminished
in 2012 (12.4% and 12.9% respectively), driven by decreasing rates of cognitive test-based and
increasing diagnosis codes-based dementia prevalence. This difference in dementia prevalence
between the two sources by sex and for age groups 75 to 79 and 90 and above vanished over
7

time. However, there remained substantial differences across measures in dementia prevalence
among blacks and Hispanics (10.9 and 9.8 percentage points respectively) in 2012.
Discussion: Our results imply that ascertainment of dementia through diagnosis may be
improving over time, but gaps across measures among racial/ethnic minorities highlight the need
for improved measurement of dementia prevalence in these populations.

8

INTRODUCTION
Rising life expectancy has led to an increasing prevalence of diseases more common at older
ages, such as Alzheimer’s disease and related dementias. The number of Americans ages 65 and
older with Alzheimer’s disease and other dementias was about 7 million in 2012 and is projected
to increase to almost 12 million by 2040 (Zissimopoulos, Tysinger, St Clair, & Crimmins, 2018).
The annual per case cost of Alzheimer’s disease (AD) is estimated to increase from $71,303 to
$140,000 from 2010 to 2050, with total population annual costs increasing from $307 billion,
including $181 billion in medical costs paid out-of-pocket and by Medicare and Medicaid, and
$126 billion in the value of unpaid caregiving by family, to $1.5 trillion (Zissimopoulos,
Crimmins, & St Clair, 2014). Accurate estimates and forecasts of dementia prevalence in the US
will aid families, policymakers and health care providers in planning for the social, economic,
and health burden of this costly and formidable disease. However, estimates of dementia
prevalence for the U.S. population vary widely, driven in large part by differences in dementia
ascertainment and study populations (Brookmeyer et al., 2011; Prince et al., 2016).
Neuropsychological assessment of dementia may provide high accuracy in ascertaining dementia
prevalence, but is typically performed only in small and non-representative samples, thus
limiting its usefulness for quantifying population levels and changes over time (Brookmeyer et
al., 2011; Demirovic et al., 2003; Gurland et al., 1999; Hebert, Weuve, Scherr, & Evans, 2013;
Rocca et al., 2011). One study, the Aging Demographics and Memory Study (ADAMS),
identified dementia using neuropsychological assessment conducted by experts including
neuropsychologists and neurologists in a sample that, when weighted, is nationally representative
of older adults in the US. Sample sizes, however, are small and the study has not been repeated
over time (Crimmins, Kim, Langa, & Weir, 2011; Plassman et al., 2007).
9

Cognitive tests of respondents in large-scale surveys such as the Health and Retirement Study
(HRS) are another source for quantifying dementia prevalence and trends in the population
(Chen & Zissimopoulos, 2018; Crimmins et al., 2011; Crimmins et al., 2018; Langa et al., 2017).
The HRS is well-suited for studying trends in dementia prevalence for two reasons. First, the
longitudinal, panel data provides multiple assessments of the same individual’s cognition over
time. Second, HRS allows for estimating population trends in dementia by observing nationally-
representative samples over time (Langa et al., 2017; Rocca et al., 2011). However, dementia
ascertainment from cognitive tests may over-estimate rates for some individuals, such as those
with low education or non-native English speakers (Crum, Anthony, Bassett, & Folstein, 1993;
Ganguli et al., 2010; Gianattasio, Wu, Glymour, & Power, 2019; Spering et al., 2012).
Medicare claims data are a potentially rich source for estimating dementia prevalence and trends
because of Medicare’s broad coverage of Americans aged 65 and older. Beneficiaries’ health
care claims, from both outpatient services and inpatient facilities, are recorded over long periods,
usually from age 65 until the beneficiary’s death. Studies, however, have found that these
records have measurement error (Amjad et al., 2018; Bradford, Kunik, Schulz, Williams, &
Singh, 2009; Chodosh et al., 2004; Taylor, Østbye, Langa, Weir, & Plassman, 2009).
Taylor et al. (2009) compared dementia diagnosis in Medicare claims from 1993 to 2005 with
neuropsychological assessment in the ADAMS 2001 to 2003 and reported 14.5% of individuals
classified as having dementia based on neuropsychological assessment were without a diagnosis
code for dementia in their Medicare claims records. Amjad et al (2018) compared cognitive test-
based dementia and dementia diagnosis using the 2011 wave of the National Health and Aging
Trends Study (NHATS) and Medicare claims and found that 39.5% of those with dementia based
on cognitive testing were undiagnosed in Medicare. However, this number may over-estimate the
10

measurement error as ascertainment of dementia based on cognitive test performance at a single
interview was found in other studies to overestimate dementia (Freedman, Kasper, Spillman, &
Plassman, 2018; Zissimopoulos et al., 2018). Chen, Tysinger, Crimmins, & Zissimopoulos
(2019) compared cognitive decline and diagnosis using HRS data linked with respondents’
Medicare claims and found 85% of respondents with incident dementia measured by cognitive
decline received a diagnosis or died within four years, with lower odds of diagnosis among
blacks and Hispanics compared to whites.
In this study, we use data from three samples, broadly representative of the U.S. population, to
estimate the level and trends over time of dementia prevalence. We quantify prevalence
differences from three measurement approaches and data sources: neuropsychological
assessment from the Aging, Demographics and Memory Study (ADAMS), cognitive tests from
the Health and Retirement Study (HRS), and diagnosis codes from Medicare claims records. We
improve upon prior comparisons of survey-based and claims-based prevalence measures with
methods that require an individual to have more than one dementia ascertainment over time to
reduce measurement error in dementia estimates. Additionally, we analyze levels in 2004 and
time trends in prevalence rates previously measured, from 2006 to 2013, and separately for
whites, blacks, and Hispanics, men and women, and persons of different ages. Quantifying the
differences in estimated prevalence across measures and sources over time improves our
understanding of strengths and weaknesses in using Medicare claims and survey-based cognitive
tests for tracking dementia trends over time and in different racial and ethnic populations.
METHODS
Data and Study Population
11

We used data from the Health and Retirement Study (HRS), the Aging, Demographics and
Memory Study (ADAMS), and a 20% random sample of Medicare beneficiaries and their
Medicare Parts A and B health care claims. We compared the prevalence of dementia at a point
in time that is common across all three data sources, 2004, and for respondents ages 70 and
older. We then analyzed trends in dementia prevalence for the subsequent years, 2006 to 2013
for the two longitudinal data sources (HRS and Medicare claims) for respondents ages 67 and
older. Internal review board approval was granted by the University of Southern California.
Health and Retirement Study (HRS)
The HRS is a biennial, nationally representative longitudinal study of adults aged 51 years and
above. Since the first study year, 1992, it has included oversamples of African Americans and
Hispanics. HRS collects data on a wide range of topics including cognition, health, family,
employment, income, and wealth. We selected respondents from the 2004 wave of the study,
aged 70 years and older (7,768 persons). Also, we selected respondents aged 67 and older from
survey waves 2006, 2008, 2010, and 2012 (13,922 persons and 40,234 person-waves). For
analyses, respondents were both community-dwelling and in nursing homes. Participants were
compensated about $80 for their participation, and verbal informed consent was obtained from
all respondents. Ethics approval was from the Health Sciences and Behavioral Sciences
institutional review board at the University of Michigan. The average participation rate across
waves among non-deceased eligible individuals in the HRS 1992-2014 sample is 86% at the
population level, and respectively 87, 87, and 85% among white, black, and Hispanic non-
deceased eligible persons (unweighted; numbers are calculated by authors).
Aging, Demographics and Memory Study (ADAMS)
12

A random subsample of 1,770 individuals, ages 70 and older, stratified based on cognitive
performance, age, and sex, were selected from HRS 2000 and 2002 interview waves for
participation in ADAMS. Assessments were completed for 856 individuals between August
2001 and December 2003 with follow-up through March 2005 (Heeringa et al., 2009). We
utilized information from the initial wave and first follow-up wave (approximately 1.5 years
later) that included reassessments of the initial diagnosis that were ambiguous or for persons with
cognitive impairment without dementia (Plassman et al., 2011). When weighted, the sample is
representative of the aged 70 and older population over the two-year data collection window.
Consent was obtained from respondents and approved by the University of Michigan
institutional review board. The participation rate in ADAMS wave A (2001 to 2003) among non-
deceased and eligible persons at the population level is 56% (Heeringa et al., 2009). The
participation rate among non-deceased and eligible whites is lower than that among non-
deceased eligible African Americans and Hispanics (54, 61, and 60% among whites, African
Americans, and Hispanic respectively; unweighted). We selected all respondents with completed
neuropsychological assessments (856 persons).
Medicare claims
We used Medicare Parts A (hospital stays) and B (outpatient) claims data from a 20% random
sample of Medicare beneficiaries enrolled in fee-for-service (FFS) in the years 2004 to 2013.
Board of the University of Southern California granted a waiver of participant consent. We
selected beneficiaries aged 70 and older in 2004 (3,649,190 persons) and beneficiaries aged 67
and older in years 2006-2013 (6,603,477 persons and 32,886,153 person-years). All selected
Medicare beneficiaries were continuously enrolled in Medicare FFS for at least three years. The
13

3-year continuous enrollment requirement excludes 15.6% (1,362,927 persons) of all Medicare
beneficiaries enrolled in FFS for at least 12 months from 2004 to 2013 at ages 65 and above.
Measurement of Dementia in ADAMS, HRS and Medicare Claims Data Sources
Dementia measured using cognitive tests (HRS)
HRS assessed cognitive functions through an adapted version of the Telephone Interview for
Cognitive Status (TICS). TICS was modeled after the Mini-Mental State Exam (MMSE) which
has been extensively used in neuropsychological assessment of cognition (Brandt, Spencer, &
Folstein, 1988; Folstein, Folstein, & McHugh, 1975). Spanish versions were developed for each
questionnaire and were administered by bilingual interviewers to Spanish-speaking respondents
(http://hrsonline.isr.umich.edu/sitedocs/surveydesign.pdf). Imputation for item non-response was
performed and described in Fisher et al. (2017). We followed prior studies on 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 (Crimmins et al., 2011; Langa et al., 2017;
Zissimopoulos et al., 2014). Scores on several questions that measured cognition determined
dementia ascertainment. Specifically, we assigned cognitive state based on scores from three
cognitive assessments for self-respondents - immediate and delayed word recall (scale 0 - 10 for
each test); counting down from 100 by 7’s (scale 0 - 5); and counting back from 20 (scale 0 - 2).
Those with a total cognitive score (scale 0 - 27) were categorized to three groups based on their
cognitive status: dementia (score 0 - 6), cognitively impaired no dementia (CIND) (score 7 - 11),
and cognitively normal (score 12 - 27). Cognitive status for respondents who had a proxy
respondent was determined by summing the following: number (0-5) of limitations with
instrumental activities of daily living (IADL); interviewer’s rating of the respondent’s difficulty
14

finishing the interview due to cognitive limitations (0 = no cognitive limitations, 1 = some
limitations, 2 = cognitive limitations); and proxy informant’s rating of the respondent’s memory
(from 0 =excellent to 4 =poor). Individuals with proxy scores were also classified into three
groups: dementia (score 6-11), CIND (score 3-5) and cognitively normal (score 0-2). Some
respondents in HRS changed dementia status over time, i.e., transitioned into or out of dementia.
For dementia ascertainment in HRS, we required an individual to be classified as having
dementia in one wave and classified as having dementia or CIND in the subsequent wave. Those
with dementia at one wave who died before the next wave were assumed to have dementia. Once
an individual was classified as having dementia, we assumed they had dementia thereafter.
Zissimopoulos et al. (2018) report the difference between dementia prevalence based on a single
assessment is five percentage points higher than based on this two-wave assessment (21%
compared to 16%) at age 85. A similar two-wave assessment was utilized and validated in
another recent study using NHATS data (Freedman et al., 2018). Although the measurement of
dementia status in HRS is based on the concordance with neuropsychological assessment in
ADAMS, dementia prevalence may differ across samples due to differences in sample
characteristics and ascertainment of dementia in the two studies.
Dementia measured using neuropsychological assessment (ADAMS)
Diagnosis of dementia in ADAMS was based on neuropsychological assessments structured in a
3 to 4-hour in-home interview, which was conducted by a neuropsychology technician and a
nurse. The final diagnosis was established by consensus conferences consisting of
neuropsychologists, a cognitive neuroscientist, neurologists, neuropsychiatrists, and internists.
Several cognitive tests were conducted including the Mini-Mental State Examination (MMSE),
Boston naming test, digit span, Symbol Digit Modality Test, animal fluency, word list three trial
15

learning, construction praxis copying, Trail Making Test, Wechsler Memory Scale, Fuld Object
Memory Test, Shipley vocabulary test and the WRAT 3 blue reading test. Proxy reports were
based on the Blessed Dementia Ratings, including, for example, questions about the ability to
accomplish household tasks, manage small amounts of money, and remembering a short list
(Blessed, Tomlinson, & Roth, 1968; Crimmins et al., 2011; Langa et al., 2005).
Dementia measured using diagnosis codes (Medicare Claims)
Providers that bill Medicare use codes for patient diagnoses, 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 International Classification of Disease, ninth revision 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. We
added additional diagnostic codes 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. The
CCW algorithm requires at least one inpatient, facility, home health, or outpatient billed claim
with one of the dementia diagnosis codes during three years. For example, a dementia diagnosis
code in 2004 utilizes claims data from 2002, 2003, and 2004. We additionally required a second
diagnosis claim over the study period to reduce measurement error from false positives (verified
dementia). Beneficiaries who died within two years after the first diagnosis were assumed to
have dementia. We compared dementia prevalence based on this verification to a different
method requiring no second claim for a dementia diagnosis (unverified dementia). Dementia
prevalence was 14.1% based on verified ascertainment and 15.3% based on unverified in 2013.
Demographic variables
16

We grouped respondents into six age groups (67 to 69, 70 to 74, 75 to 79, 80 to 84, 85 to 89, and
90 and above), and four racial/ethnic groups (non-Hispanic white, non-Hispanic black, Hispanic,
and other races). We included the distribution of education in HRS/ADAMS, and we divided the
sample into three education levels (less than high school, high school, and some college and
above). Race was self-reported in HRS/ADAMS and claims data. Race/ethnicity was determined
with the beneficiary race code in the Center for Medicare & Medicaid Services (CMS)
enrollment data using the application of a name-based identification algorithm from the Research
Triangle Institute to improve accuracy (Eicheldinger & Bonito, 2008).
Analysis
We compared dementia prevalence based on neuropsychological assessment, cognitive tests, and
diagnosis codes from Medicare claims records by race, gender, and age in 2004 - a point in time
that was comparable across all three data sources. We estimated trends over time in dementia
prevalence utilizing logistic regression adjusting for age, sex, and race, and separately for HRS
and Medicare claims data (Equation 1). We chose the years 2006 to 2013 to study trends in
dementia prevalence.
Equation (1)
𝑙𝑛
𝐷𝑒𝑚𝑒𝑛𝑡𝑖𝑎 𝑖 1 − 𝐷𝑒𝑚𝑒𝑛𝑡𝑖𝑎 𝑖
= 𝛽 0
+ 𝛽 1
𝑏𝑙𝑎𝑐𝑘 𝑖 + 𝛽 2
ℎ𝑖𝑠𝑝𝑎𝑛𝑖𝑐 𝑖 + 𝛽 3
𝑜𝑡 ℎ𝑒𝑟
𝑖 + 𝛽 4
𝑎𝑔𝑒 7074
𝑖 + 𝛽 5
𝑎𝑔𝑒 7579
𝑖
+𝛽 6
𝑎𝑔𝑒 8084
𝑖 + 𝛽 7
𝑎𝑔𝑒 8589
𝑖 + 𝛽 8
𝑎 𝑔 𝑒 90𝑎𝑏𝑜𝑣𝑒 𝑖 + 𝛽 9
𝑓𝑒𝑚𝑎𝑙𝑒 𝑖 + 𝛽 10
2007
𝑖
+𝛽 11
2008
𝑖 + 𝛽 12
2009
𝑖 + 𝛽 13
2010
𝑖 + 𝛽 14
2011
𝑖 + 𝛽 15
2012
𝑖 + 𝛽 16
2013
𝑖
The same model is estimated using data from the HRS with the exception that year indicators are
2008, 2010, 2012, reflecting the biennial nature of the HRS study. We computed predicted
17

values of dementia prevalence by age, sex, race, and year. For example, we estimated predicted
values of dementia prevalence rates among whites in 2012 with age and sex measured at their
mean values among whites in 2012. We additionally used direct standardization to standardize
the 2012 sample based on the age-, sex-, and race- distribution of the 2006 sample. We compared
standardized and non-standardized prevalence rates in 2012. We tested for differences in time
trends across data sources and measures by pooling the HRS and Medicare claims data, re-
estimating the model including an indicator variable for HRS as the data source and interacting
this indicator with the race, sex, age, and year covariates.
RESULTS
Dementia prevalence in 2004 based on neuropsychological assessment, cognitive tests, and
diagnosis codes
Table 1 shows the demographic characteristics (race, sex, age group, and education) of persons
aged 70 and older in the three data sources: ADAMS, HRS, and Medicare claims in 2004. The
racial/ethnic population distribution in ADAMS was: blacks 7.6%; Hispanics 5.3%, whites
87.1%; in HRS was: blacks 8.0%, Hispanics 5.4%, whites 84.8%, other race 1.8%; and in
Medicare claims: blacks 7.0%; Hispanics 4.3%, whites 86.3%, other race 2.3%. Individuals in
the youngest age group, those aged 70 to 74, and in the oldest, 90 and above, respectively were
26.9% and 6.7% of ADAMS respondents, 33.6% and 5.3% of HRS respondents, and 32.6% and
5.4% of the sample of Medicare beneficiaries.

18

Table 1. Sample characteristics from ADAMS, HRS and Medicare claims data sources, ages 70
and above, 2004
ADAMS (2001 to 2005) HRS 2004 Claims 2004
Race

White 87.1% (746) 84.8% (6,589) 86.3% (31,499,03)
Black 7.6% (65) 8% (618) 7.0% (256,446)
Hispanic 5.3% (45) 5.4% (421) 4.3% (158,082)
Other race N/A 1.8% (139) 2.3% (84,759)
Gender

Male 39.3% (336) 40.2% (3,125) 39.1% (1,427,157)
Female 60.7% (520) 59.8% (4,643) 60.9% (2,222,033)
Age Group

70 to 74 26.9% (230) 33.6% (2,608) 32.6% (1,188,062)
75 to 79 31.5% (269) 28.6% (2,220) 29.4% (1,072,628)
80 to 84 22.5% (193) 21.1% (1,635) 21.3% (776,827)
85 to 89 12.4% (106) 11.5% (896) 11.3% (413,713)
90 and above 6.7% (58) 5.3% (408) 5.4% (197,960)
Education

Less than High School 34.8% (299) 32.3% (2,491) N/A
High School 28.1% (240) 33.1% (2,574) N/A
College 37.1% (317) 34.7% (2,704) N/A
Total 856 7,768 3,649,190
Note. ADAMS= the Aging Demographics and Memory Study; HRS = Health and Retirement Study; Claims =
Medicare claims. Values in ADAMS are weighted by the ADAMS sampling weights. Values in HRS are
weighted by the HRS sampling weights.

Figure 1 (and Appendix Table 1) shows dementia prevalence based on neuropsychological
assessment, cognitive tests, and diagnostic codes on claims by sex, age, and race. Dementia
prevalence in 2004 was 16.6% (neuropsychological assessment), 15.8% (cognitive tests), 12.2%
(diagnosis codes).

19

Figure 1. Dementia prevalence for the US population and by race, gender, and age in
neuropsychological assessment (ADAMS), cognitive tests (HRS), and diagnosis codes
(Medicare claims), ages 70 and above, 2004 with 95% confidence intervals

Note. ADAMS= the Aging Demographics and Memory Study; HRS = Health and Retirement Study;
Claims = Medicare claims. Values in ADAMS are weighted by the ADAMS sampling weights. Values in
HRS are weighted by the HRS sampling weights.
Table 1. Sample Characteristics from ADAMS, HRS and Medicare Claims Data Sources, Ages 70 and
above, 2004

Dementia prevalence for both men and women was higher based on neuropsychological
assessment (men, 12.3%; women, 19.4%) and cognitive tests (men, 14.0%; women,17.0%) than
based on diagnosis codes (men, 9.4%; women, 14.0%). Percentage point differences between
neuropsychological assessment and diagnosis codes were higher among women (5.4 percentage
point difference) than among men (2.9 percentage point difference).
Differences across ascertainment methods and data sources were particularly pronounced for the
oldest age group, 90 and above, with rates of 44.0% based on neuropsychological assessment,
47.9% based on cognitive tests and 34.1% based on diagnosis codes (Figure 1 and Appendix
20

Table 1). There were substantial differences across measures in dementia prevalence among
ethnic minorities. Dementia prevalence for blacks based on neuropsychological, cognitive test
and diagnosis codes was respectively 23.5% (95%CI, 16.8%-30.1%), 39.3% (95%CI, 36.1%-
42.5%), and 16.7% (95%CI, 16.6%-16.8%). Dementia prevalence among Hispanics was
respectively 24.7% (95%CI, 15.3%-34.1%), 29.3% (95%CI, 25.5%-33.1%), and 11.5% (95%CI,
11.4%-11.7%).
Dementia prevalence based on cognitive tests, and diagnosis codes over time
Figure 2 shows predicted values of dementia prevalence at the population level in years 2006 and
2012 from cognitive tests and diagnosis codes. We used estimates obtained from logistic
regressions adjusting for age, race, and sex to compute predicted values of dementia prevalence
(Table 2 reports the odds-ratio estimates). In 2006, the prevalence was higher based on cognitive
tests than that based on diagnosis codes. In 2012, however, there was no such difference in
prevalence rates. Prevalence based on cognitive tests was 13.8% (95% CI, [12.3 %-15.5%]) in
2006, and statistically significantly lower, 12.4% (95%CI [11.0%-14.0%]) in 2012 (Table 2).
Dementia prevalence in 2008 and 2010 was not statistically significantly different from that in
2006 (Table 2). Dementia prevalence based on diagnosis codes increased from 2006 to 2012
(diagnosis codes in 2006, 11.9%, 95% CI [11.9%-12.0%]; and in 2012, 12.9%, 95%CI [12.9%-
13.0%]). Time trends across data sources were statistically different (Appendix Table 5, Model
5). Standardizing the composition of the population in 2012 based on the age, sex and
race/ethnicity of the population in 2006 did not result in substantively different rates of dementia
prevalence in 2012 (12.3% and 12.7% based on cognitive test and diagnosis codes in claims
respectively; Appendix Table 4).

21

Table 2. Odds ratios for the presence of dementia based on cognitive tests (HRS) and diagnosis
codes (Medicare claims), ages 67 and above, 2006-2013
(1) (2)
Presence of Dementia
Cognitive tests (HRS) Diagnosis codes (Claims)
Odds Ratio 95%CI Odds Ratio 95%CI
Male 1 [1 1] 1 [1 1]
Female 1.05 [0.95 - 1.13] 1.28*** [1.28 - 1.28]
White 1 [1 1] 1 [1 1]
Black 4.91*** [4.41 - 5.42] 1.68*** [1.67 - 1.69]
Hispanic 4.22*** [3.69- 4.79] 1.46*** [1.45 - 1.47]
Other races 1.89 [1.52 - 2.33] 0.98*** [0.97 - 0.99]
67 to 69 1 [1 1] 1 [1 1]
70 to 74 1.44*** [1.22 - 1.71] 1.88*** [1.87 - 1.9]
75 to 79 2.43*** [2.04 - 2.85] 3.91*** [3.89 - 3.94]
80 to 84 4.36*** [3.66 – 5.14] 7.51*** [7.46 - 7.57]
85 to 89 8.86*** [7.42 - 10.51] 13.09*** [12.99 - 13.19]
90 and above 17.22*** [14.19 – 20.71] 21.91*** [21.74 - 22.08]
2006 1 [1 1] 1 [1 1]
2007

1.05*** [1.04 - 1.05]
2008 0.95 [0.84 - 1.06] 1.08*** [1.08 - 1.09]
2009

1.11*** [1.11 - 1.12]
2010 0.97 [0.86 - 1.1] 1.12*** [1.12 - 1.12]
2011

1.10*** [1.1 - 1.11]
2012 0.85*** [0.75 - 0.96] 1.083** [1.08 - 1.09]
2013

1.02*** [1.01 - 1.02]
Constant 0.04*** [0.03 - 0.04] 0.0232*** [0.02 - 0.02]
Observations 40,224

32,855,743
Pseudo R-squared 0.146 0.1194
Note. HRS = Health and Retirement Study; Claims = Medicare claims; Samples restricted to ages 67 and above.
95%CI adjusted by Bonferroni correction. ***p<0.001. **p<0.01. *p< .05.
22

Figure 2. Predicted values of dementia prevalence based on cognitive tests (HRS) and diagnosis
codes (claims) from logistic models adjusting for race, sex, age group, and wave in HRS and
claims, ages 67 and above, 2006 and 2012.
Note. HRS = Health and Retirement Study; Claims = Medicare claims; predicted values of dementia
prevalence in HRS are weighted by the HRS sampling weights. 95% confidence intervals included in the
figure. 95%CI adjusted by Bonferroni correction.

Figure 3 shows estimates of dementia prevalence based on cognitive tests and diagnosis codes
separately by sex (Figure 3a), age (Figure 3b), and race/ethnicity (Figure 3c) in 2006 and 2012
from logistic regressions (estimates for all years in Appendix Table 2). There was a 3.1
percentage point difference in dementia prevalence based on cognitive tests and diagnosis codes
among males in 2006 that reduced to 1.2 percentage points by 2012 (cognitive tests, 10.9%,
95%CI [9.6%-12.2%]; diagnosis codes, 9.7%, 95%CI [9.7%-9.7%]). Among females, the
difference across measures in 2006 was 1 percentage point and 0.6 in 2012 (cognitive tests,
13.6%, 95%CI [12.1%-15.2%]; diagnosis codes,14.2% [14.1%-14.2%]) (Figure 3a and
Appendix Table 2 report predicted values of dementia prevalence by sex).

23

Figure 3. Dementia prevalence by sex (Figure 3a), age (Figure 3b), and race (Figure 3c) based on
cognitive tests (HRS) and diagnosis codes (claims) from logistic models adjusting for sex, age
group, and waves in HRS and claims, ages 67 and above, 2006 and 2012.

Note. HRS = Health and Retirement Study; Claims = Medicare claims; predicted values of dementia
prevalence in HRS are weighted by the HRS sampling weights. 95% Confidence Intervals included in the
figure. 95%CI adjusted by Bonferroni correction.

24

The percentage difference in dementia prevalence across measures was most pronounced at ages
67 to 74 compared to ages 75 and above in years 2006 and 2012. In 2012, dementia prevalence
based on cognitive tests was statistically significantly higher than that based on diagnosis codes
among persons ages 67 to 69 (cognitive tests, 4.8%, 95%CI [4.0%-5.6%]; diagnosis codes, 3.0%,
95%CI [3.0%-3.0%]) and ages 70 to 74 (cognitive tests, 6.7%, 95%CI [5.9%-7.5%], diagnosis
codes, 5.5%, 95%CI [5.5%-5.5%]). At the oldest ages, persons 90 and above, the 4.1 percentage
point difference between these two measures in 2006 (cognitive tests, 43.4%, 95%CI [39.7%-
47.1%]; diagnosis codes, 39.3%, 95%CI [39.2%-39.4%]) reduced to 0.7 percentage points in
2012 (cognitive tests, 40.2%, 95%CI [36.7%-43.7%]; diagnosis codes, 40.9%, 95%CI [40.8%-
41.0%]) and was not statistically different (Figure 3b and Appendix Table 2).
There was a large difference in dementia prevalence from the two sources based on cognitive
tests and based on diagnosis codes among blacks and Hispanics, which decreased over time. In
2006 the difference between the two measures was 15.1 percentage points among blacks
(cognitive tests, 31.2%, 95%CI [28.3%-34.1%]; diagnosis codes, 16.1%, 95%CI [16.0%-16.1%])
and 14.0 percentage points among Hispanics (cognitive tests, 27.2%, 95%CI [24.3%-30.5%];
diagnosis codes, 13.2%, 95%CI [13.1%-13.3%]). The gap across measures was smaller in 2012:
10.9 percentage points among blacks (cognitive tests, 28.1%, 95%CI [25.4%-31.1%]; diagnosis
codes, 17.2%, 95%CI [17.1%-17.3%]) and 9.8 percentage points among Hispanics (cognitive
tests, 24.8%, 95%CI [22%-27.6%]; diagnosis codes, 15.0%, 95%CI [14.9%-15%]) (Figure 3c
and Appendix Table 2).
DISCUSSION
We analyzed the prevalence of dementia in samples broadly representative of the US population
and by sex, age, and race/ethnicity for years from 2004 through 2013 using three different
25

measurement approaches and data sources: neuropsychological assessment (ADAMS), cognitive
tests (HRS), diagnosis codes (Medicare claims). We found dementia prevalence for ages 70 and
above in 2004 was highest based on neuropsychological assessment and lowest based on
diagnosis codes from claims (16.6% (neuropsychological assessment), 15.8% (cognitive tests),
12.2% (diagnosis codes)). Dementia prevalence based on cognitive tests was lower in 2012
(12.4%) compared to 2006 (13.8%), reflecting a decline from 2010 to 2012 and no difference in
prevalence in years 2006, 2008 and 2010. This decline was also consistent with that reported in
Langa et al. (2017) and Freedman et al. (2018). In contrast, dementia prevalence based on
diagnosis codes rose over time (2006, 11.9%; 2012, 12.9%). Changes in the age, sex and
racial/ethnic composition of the population did not explain the time trends.
Subgroup analysis of the three measures in 2004 and for cognitive tests and diagnosis codes in
years 2006 to 2012 revealed several consistencies across all measures: dementia prevalence was
higher for females than males, increased with age and was higher for blacks and Hispanics than
whites.
Over time, dementia prevalence for all age groups, for men and women, and for whites, blacks
and Hispanics declined based on cognitive tests and increased based on diagnosis codes from
claims, thereby reducing the difference across measures for these subpopulations over time. The
notable difference in dementia prevalence based on cognitive tests and diagnosis codes for both
men and women in 2006, were not statistically different for either sex by 2012. Similarly, the
differences between these measures of prevalence among those aged 90 and above vanished over
time. Notably, dementia prevalence based on both cognitive tests and diagnosis codes among
people aged 90 and above in 2012 were 40.2% and 40.9%, respectively. This estimate combined
with an increase in the proportion of older Americans aged 90 and older from 5.2 percent of
26

those aged 65 and above to 9.9 percent by 2050 (Vincent & Velkoff, 2010) will drive the growth
in number of Americans with dementia and in burden of dementia in the future. Across
racial/ethnic populations, there were substantial differences in dementia prevalence across
measures and data sources and these differences reduced, but did not vanish over time. Cognitive
test-based prevalence was the highest and diagnosis code-based prevalence was the lowest
among ethnic minorities.
Our results highlight strengths and weaknesses of the different data sources and different
measures for quantifying dementia prevalence for the U.S. population and for different
subpopulations. Medicare claims data is available for a large population of older Americans and
as such has large samples of ethnic minorities. There are no other data sources with such
significant numbers of minorities from across the regions of the United States. Doctors may take
into account many factors such as cardiovascular health, education level and including but not
limited to cognition in determining a dementia diagnosis. Diagnosing dementia in racial and
ethnic populations may be improving over time. Although we do not analyze what is driving
these changes, factors may include changing social and cultural norms about stigma associated
with dementia, and/or improvement in awareness of dementia by both physicians and patients
(Amjad et al., 2018; Chodosh et al., 2004). However, there are two primary limitations. The
sample only includes older Americans enrolled in Medicare fee for service (FFS) and not those
enrolled in Medicare Advantage. The portion of beneficiaries enrolled in Medicare Advantage
has been increasing over time and these populations may have different rates of dementia.
Second, despite potential improvements in identifying dementia in racial/ethnic minority
populations, dementia may still be under-diagnosed (Chen et al., 2019; Gianattasio et al., 2019).
Increases in diagnosis-code measures of dementia prevalence over time among racial/ethnic
27

minorities that is explained by changing norms and awareness may obscure information about
changes in dementia risk.
The HRS sample includes older Americans enrolled in either fee-for-service (FFS) or Medicare
Advantage (MA) and when weighted, is nationally representative. The cognitive test based
measures of dementia in HRS were validated by neuropsychological exams and are correlated
with well-known risk factors of dementia including low education and cardiovascular risk factors
(Chen & Zissimopoulos, 2018; Langa et al., 2017). Thus, the declining trend in dementia over
time may provide the signal of declines in dementia risk in the U.S population. However, these
data and measures also have limitations for measuring prevalence of dementia in the population
and among racial/ethnic subpopulations. Several studies found a cognitive test-based
measurement approach had lower specificity among low-educated persons and non-native
English-speakers (Crum et al., 1993; Ganguli et al., 2010; Gianattasio et al., 2019; Spering et al.,
2012) and algorithmic approaches that do not adjust for educational attainment will overestimate
dementia rates among racial/ethnic minorities (Gianattasio et al., 2019). Indeed, dementia
prevalence among blacks in 2012 from NHATS, and based on criteria that included but were not
limited to cognitive tests, founds rates (15%) more similar to those based on diagnosis in
Medicare claims than cognitive test from HRS data (Freedman et al., 2018; Kasper, Freedman, &
Spillman, 2013).
Our study has limitations. We require three-years of continuous enrollment in FFS, standard
practice for measuring disease conditions based on Chronic Conditions Warehouse algorithm.
This may underestimate diagnosis codes-based prevalence by excluding individuals who die
within 3-years. This requirement will affect older beneficiaries more than younger ones and in
particular, may narrow differences in prevalence rates across measures among persons aged 80 to
28

84 (Appendix Table 2). Medicare claims data exclude beneficiaries enrolled in Medicare
Advantage (MA) plans. Consistent with other studies (St. Clair et al., 2017), respondents
enrolled in MA plans are younger, more likely to be ethnic minorities, male, and less educated
than those in FFS plans (Appendix Table 3). We used data from the HRS that included both MA
and FFS enrollees and found only small differences in dementia prevalence in 2012 among MA
beneficiaries (9.7%) and FFS beneficiaries (10.5%).
Attributing differences (or concordance) in dementia prevalence to how dementia is measured in
analyses of cognitive tests and neuropsychological assessment in 2004 was complicated by
potential differences in sample composition, and also by small sample sizes in subpopulations
that led to imprecise estimates. ADAMS is a stratified (by age, sex, cognition) random sample of
HRS self and proxy respondents ages 70 and older who contributed data to waves 2000 or 2002.
Thus, cohorts are not totally independent, yet there were compositional differences in sample
characteristics (Table 1). In supplementary analyses (results not shown), we used linked HRS
and ADAMS data to exclude variation in sample composition and found gaps in dementia
prevalence in racial/ethnic subpopulations and among persons ages 70 to 89 were due to
dementia measurement. In analyses of concordance between cognitive tests and diagnosis code
measures, we distinguished between sample composition and dementia measurement as drivers
of differences in prevalence rates from 2006 to 2012. We found differential changes in the age,
sex, and race composition of the HRS and Medicare sample populations did not explain the
reduction in differences in dementia prevalence rates between the two measures (Appendix Table
4). The improvement in diagnostic practices and/or awareness is a likely driver of the increase in
prevalence based on diagnosis codes and the convergence in dementia prevalence across data
sources. While this convergence across measures is a positive sign for measuring population
29

dementia prevalence, limitations remain for measuring dementia prevalence among racial/ethnic
populations. In these subpopulations, dementia is likely under-estimated based on diagnosis
codes and likely over-estimated based on cognitive tests. The level of under- and over-estimation
is unknown.
More recent longitudinal data sources, such as NHATS, with broadly population representative
samples and several different measures for ascertaining dementia may add to our understanding
of levels and time trends in dementia prevalence (Freedman et al., 2018; Kasper et al., 2013).
New data collection efforts may improve measurement of dementia in the U.S. population and in
subpopulations at a point in time and over time. The HRS Harmonized Cognitive Assessment
Protocol (HCAP) collected data on sample of HRS respondents in 2016 and is proposed for
follow-up in 2020 (Weir, Langa, & Ryan, 2016). The study included an in-home 1-hour battery
of cognitive tests and an informant interview. There are new opportunities for clinicians to better
detect and diagnose dementia. The Annual Wellness Visit, introduced in 2011, requires cognitive
screening among Medicare FFS beneficiaries and may improve the detection of dementia among
all Medicare beneficiaries (Chodosh, Thorpe, & Trinh-Shevrin, 2018; Ganguli, Souza,
McWilliams, & Mehrotra, 2017). Another opportunity is the collaboration of community health
workers, consisting of paraprofessionals who work in communities, share health care
information and resources among the minorities, and advocate dementia detection (Chodosh et
al., 2018). Considering this potential increasing detection of dementia among blacks and
Hispanics, continued tracking of these trends in diverse populations and using different methods
of ascertainment will aid in understanding how risk may be changing over time as well as the
extent to which high-risk populations are receiving diagnoses. Studies analyzing individual-level
agreement in dementia ascertainment based on cognitive tests and diagnosis codes will provide
30

additional insight into changes in risk of dementia over time for different racial/ethnic groups,
aid in targeting resources to populations who are underdiagnosed and assist health care providers
and policy makers prepare for future dementia burden in the US.
31

REFERENCE
Amjad, H., Roth, D., Sheehan, O., Lyketsos, C., Wolff, J., & Samus, Q. (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. doi:10.1007/s11606-
018-4377-y
Blessed, G., Tomlinson, B. E., & Roth, M. (1968). The association between quantitative
measures of dementia and of senile change in the cerebral grey matter of elderly subjects.
British Journal Of Psychiatry, 114(512), 797-811. doi:10.1192/bjp.114.512.797
Bradford, E. A., Kunik, P. M., Schulz, P. P., Williams, P. S., & Singh, P. H. (2009). Missed and
Delayed Diagnosis of Dementia in Primary Care: Prevalence and Contributing Factors.
Alzheimer Disease & Associated Disorders, 23(4), 306-314.
doi:10.1097/WAD.0b013e3181a6bebc
Brandt, J., Spencer, M., & Folstein, M. (1988). The telephone interview for cognitive status.
Neuropsychiatry, Neuropsychology and Behavioral Neurology, 1(2), 111-117.
Brookmeyer, R., Evans, D. A., Hebert, L., Langa, K. M., Heeringa, S. G., Plassman, B. L., &
Kukull, W. A. (2011). National estimates of the prevalence of Alzheimer’s disease in the
United States. Alzheimer's & dementia: the journal of the Alzheimer's Association, 7(1),
61-73. doi:10.1016/j.jalz.2010.11.007
Chen, C., & Zissimopoulos, J. M. (2018). Racial and ethnic differences in trends in dementia
prevalence and risk factors in the United States. Alzheimer's & Dementia: Translational
Research & Clinical Interventions, 4, 510-520. doi:10.1016/j.trci.2018.08.009
Chen, Y., Tysinger, B., Crimmins, E., & Zissimopoulos, J. M. (2019). Analysis of dementia in
the US population using Medicare claims: Insights from linked survey and administrative
claims data. Alzheimer's & Dementia: Translational Research & Clinical Interventions,
5(1), 197-207. doi:10.1016/j.trci.2019.04.003
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.
doi:10.1111/j.1532-5415.2004.52301.x
Chodosh, J., Thorpe, L. E., & Trinh-Shevrin, C. (2018). Changing Faces of Cognitive
Impairment in the U.S.: Detection Strategies for Underserved Communities. American
Journal of Preventive Medicine, 54(6), 842-844. doi:10.1016/j.amepre.2018.02.016
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. The journals of gerontology. Series B,
Psychological sciences and social sciences, 66 Suppl 1(1), i162-171.
doi:10.1093/geronb/gbr048
32

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. doi:10.1093/geronb/gbx135
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. Journal of the
American Medical Association, 269(18), 2386-2391. doi:10.1001/jama.269.18.2386
Demirovic, J., Prineas, R., Loewenstein, D., Bean, J., Duara, R., Sevush, S., & Szapocznik, J.
(2003). Prevalence of dementia in three ethnic groups: The south Florida program on
aging and health. Annals of Epidemiology, 13(6), 472-478. doi:10.1016/s1047-
2797(02)00437-4
Eicheldinger, C., & Bonito, A. (2008). More accurate racial and ethnic codes for Medicare
administrative data.(Report). Health Care Financing Review, 29(3), 27.
Fisher, G. G., Hassan, H., Faul, J. D., Rodgers, W. L., & Weir, D. R. (2017). Health and
Retirement Study Imputation of Cognitive Functioning Measures: 1992–2014 (Final
Release Version). http://hrsonline.isr.umich.edu/modules/meta/xyear/cogimp/desc/
COGIMPdd.pdf.
Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). "Mini-mental state". A practical
method for grading the cognitive state of patients for the clinician. Journal of Psychiatric
Research, 12(3), 189-198. doi:10.1016/0022-3956(75)90026-6
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:Psychological Sciences
and Social Sciences, 73, S48-S56. doi:10.1093/geronb/gbx144
Ganguli, I., Souza, J., McWilliams, J. M., & Mehrotra, A. (2017). Trends in Use of the US
Medicare Annual Wellness Visit, 2011-2014. Journal of the American Medical
Association, 317(21), 2233-2235. doi:10.1001/jama.2017.4342
Ganguli, M., Snitz, B. E., Lee, C.-W., Vanderbilt, J., Saxton, J. A., & Chang, C.-C. H. (2010).
Age and education effects and norms on a cognitive test battery from a population-based
cohort: The Monongahela–Youghiogheny Healthy Aging Team. Aging & Mental Health,
14(1), 100-107. doi:10.1080/13607860903071014
Gianattasio, Z. K., Wu, M. Q., Glymour, C. M., & Power, C. M. (2019). Comparison of methods
for algorithmic classification of dementia status in the Health and Retirement Study.
Epidemiology, 30(2), 291-302. doi:10.1097/EDE.0000000000000945
Gurland, B. J., Wilder, D. E., Lantigua, R., Stern, Y., Chen, J., Killeffer, E. H. P., & Mayeux, R.
(1999). Rates of dementia in three ethnoracial groups. International Journal of Geriatric
Psychiatry, 14(6), 481-493. doi:10.1002/(SICI)1099-1166(199906)14:6<481::AID-
GPS959>3.0.CO;2-5
33

Hebert, L. E., Weuve, J., Scherr, P. A., & Evans, D. A. (2013). Alzheimer disease in the United
States (2010-2050) estimated using the 2010 census. Neurology, 80(19), 1778-1783.
doi:10.1212/WNL.0b013e31828726f5
Heeringa, S. G., Fisher, G. G., Hurd, M., Langa, K. M., Ofstedal, M. B., Plassman, B. L., . . .
Weir, D. R. (2009). Aging, demographics and memory study (ADAMS). Sample design,
weighting, and analysis for ADAMS. Retrieved from
http://hrsonline.isr.umich.edu/sitedocs/userg/ADAMSSampleWeights_Jun2009.pdf.
Kasper, J. D., Freedman, V. A., & Spillman, B. C. (2013). Classification of persons by dementia
status in the National Health and Aging Trends Study. Retrieved from www.NHATS.org
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. Journal of the American Medical Association Internal Medicine, 177(1),
51-58. doi:10.1001/jamainternmed.2016.6807
Langa, K. M., Plassman, B. L., Wallace, R. B., Herzog, A. R., Heeringa, S. G., Ofstedal, M.
B., . . . Willis, R. J. (2005). The Aging, Demographics, and Memory Study: Study Design
and Methods. Neuroepidemiology, 25(4), 181-191. doi:10.1159/000087448
Plassman, B. L., Langa, K. M., Fisher, G. G., Heeringa, S. G., Weir, D. R., Ofstedal, M. B., . . .
Wallace, R. B. (2007). Prevalence of Dementia in the United States: The Aging,
Demographics, and Memory Study. Neuroepidemiology, 29(1-2), 125-132.
doi:10.1159/000109998
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. doi:10.1002/ana.22362
Prince, M., Ali, G.-C., Guerchet, M., Prina, A. M., Albanese, E., & Wu, Y.-T. (2016). Recent
global trends in the prevalence and incidence of dementia, and survival with dementia.
Alzheimer's research & therapy, 8(1), 23-23. doi:10.1186/s13195-016-0188-8
Rocca, W. A., Petersen, R. C., Knopman, D. S., Hebert, L. E., Evans, D. A., Hall, K. S., . . .
White, L. R. (2011). Trends in the incidence and prevalence of Alzheimer’s disease,
dementia, and cognitive impairment in the United States. Alzheimer's & dementia: the
journal of the Alzheimer's Association, 7(1), 80-93. doi:10.1016/j.jalz.2010.11.002
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 Alzheimerʼs
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. doi:10.1093/gerona/gls006
St. Clair, P. P., Gaudette, P. É., Zhao, P. H., Tysinger, P. B., Seyedin, P. R., & Goldman, P. D.
(2017). Using self-reports or claims to assess disease prevalence: It’s complicated.
Medical Care, 55(8), 782-788. doi:10.1097/MLR.0000000000000753
34

Taylor, 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. doi:10.3233/JAD-2009-1099
Vincent, G. K., & Velkoff, V. A. (2010). The next four decades: The older population in the
United States: 2010 to 2050. Retrieved from
https://www.census.gov/content/dam/Census/library/publications/2010/demo/p25-
1138.pdf
Weir, D., Langa, K., & Ryan, L. (2016). Harmonized cognitive assessment protocol (hcap):
Study protocol summary. Retrieved from
http://hrsonline.isr.umich.edu/index.php?p=shoavail&iyear=ZU
Zissimopoulos, J. M., Crimmins, E. M., & St Clair, P. A. (2014). The value of delaying
Alzheimer's disease onset. Forum for health economics & policy, 18(1), 25.
doi:10.1515/fhep-2014-0013
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:
Psychological Sciences and Social Sciences, 73, S38-S47. doi:10.1093/geronb/gbx147

35

Chapter 2:
Cognitive Changes at Middle Age

by

Yingying Zhu

ABSTRACT
Recent literature suggests that the actual dementia pathology may occur 20 years before its
symptoms. Therefore, interventions on cognitive decline at middle age may be more effective.
However, knowledge about the distribution of cognitive decline at middle age and its associated
risk/protective factors is limited.
This research quantifies the population with relatively fast rates of cognitive decline and
improvement. Further, it compares demographic, health, and socioeconomic factors associated
with these cognitive changes and provides implications for future policy interventions.
This study identifies persons with fast rates of cognitive decline and enhancement both in the
short run (over 2 years) and in the long run (over 2 to 14 years). Overall, 8.4% of the middle-
aged population experienced fast rates of cognitive deterioration, and 10.9% of the population
were found to rapidly increase their cognitive function over two years. Being ethnic minorities,
with low education, male, low income, low wealth, not working or working in less cognitively
demanding occupations are risk factors of experiencing sharp cognitive decline. For example, the
odds of experiencing rapid cognitive deterioration for African Americans was and Hispanics was
respectively1.89 and 1.6 times of whites over 2 years.
36

INTRODUCTION
Cognitive decline refers to the deterioration of mental abilities, including executive function,
attention, long-term memory, and knowledge (Blazer, Yaffe, & Liverman, 2015; Harada,
Natelson Love, & Triebel, 2013). This decline is a continuum and ranges in severity from
normal, age-associated diminution to pathological deterioration, such as Alzheimer's disease and
related dementia (dementia) (Daffner, 2010; National Academies of Sciences & Medicine,
2017).
Additionally, various aspects of cognitive function change differently with age. For example,
working memory, episodic memory, and the speed of information processing, begin declining as
early as one's 20s, while knowledge and experience, actually improve until one's 60s or 70s
(Finch, 2009; McArdle, Smith, & Willis, 2009; McArdle, Ferrer-Caja, Hamagami, & Woodcock,
2002; Salthouse, 2009). As a result, there is insufficient knowledge about whether cognitive
function declines at middle age. Recent studies focused on rates of cognitive changes and found
evidence of average deterioration in cognitive performance beginning from one's 40s and 50s
(Bratzke et al., 2018; Knopman, 2001; Singh-Manoux, Hillsdon, Brunner, & Marmot, 2005;
Singh-Manoux et al., 2012). However, it is worth noting that there is diversity in individual rates
of cognitive changes at middle age. Levine, Harrati, and Crimmins (2018) utilized a composite
measure to quantify cognitive aging. The authors further identified both the levels and slopes of
cognitive aging. They found that genetic, cardiometabolic, and sociodemographic factors were
associated with the levels and slopes of cognitive aging. Nevertheless, information about people
with the fastest rates of cognitive deterioration and its risk factors at middle age is missing.
Rapidly declining cognition leads to a 2 percentage point decrease in the probability of working
at age 62 over 2 years, and a 5 percentage point decrease in the probability of working over 2 to
37

10 years (see Chapter 3 for more detail). This could result in individual retirement income
insecurity and impose more socioeconomic burdens on the Social Security program. This decline
is also associated with a higher chance of developing dementia later in life (Levine et al., 2018).
Several studies suggest that dementia pathology may occur 20 years earlier than the
manifestation of severe symptoms (Alzheimer's Association, 2019; Beason-Held et al., 2013).
These findings suggest that interventions on rapidly declining cognitive function at middle age
may be more effective for preventing or delaying the onset of dementia, which addresses the
large socioeconomic burden of dementia.
Therefore, identifying people with the fastest rates of cognitive decline and associated
risk/protective factors provides information for individuals, families, and policy makers to
prevent or slow this decline thus alleviates its socioeconomic burden. As Zissimopoulos,
Crimmins, and St Clair (2014) suggest, a 5-year delay in the onset of dementia is associated with
41% lower prevalence and 40% lower cost.
Furthermore, several studies found statistical enhancement in cognition at middle age. Some
explanations of this improvement include the participation of young respondents, exclusion of
persons living with dementia, and measurement error (Cournot et al., 2006; Salthouse, 2006;
Wilson et al., 2002). Hertzog, Kramer, Wilson, and Lindenberger (2008) later argued, based on a
theoretical framework and available literature, that cognitive improvement may exist, due to
intellectual, physical, and social activities. However, evidence for this enhancement is
insufficient for a definitive conclusion. Further, little is known about factors associated with this
enhancement.
Additionally, previous studies suggest that cardiovascular risk factors, demographic factors, and
socioeconomic factors are associated with cognitive decline, but less is known about risk factors
38

associated with rapid changing cognitive function at middle age (Forte, De Pascalis, Favieri, &
Casagrande, 2019; National Academies of Sciences & Medicine, 2017; Williams & Kemper,
2010; Yaffe et al., 2013).
This paper aims to add to the current literature by asking two research questions: what is the
distribution of individual paths of cognitive changes over different time periods among middle-
aged adults? What demographic, health, and socioeconomic factors are associated with these
changes?
This research used data from the Health and Retirement Study (HRS), a widely utilized large-
sample population-based survey. This study quantified the distribution of rates of cognitive
decline and improvement for middle-aged adults both in the short run (over a 2-year follow-up)
and the long run (over a 2 to 14-year follow-up). It reduced measurement error in quantifying
fast rates of cognitive decline. This research further conducted logistic regression analysis to
explore health and socioeconomic factors associated with these changes. I found that 8.4% of
middle-aged persons experienced fast rates of cognitive deterioration, defined by 1 or more
standard deviation (SD) decrease in the standardized cognitive score. Additionally, 10.9% had a
mostly accelerated increase in cognitive function over 2 years. I further found that there was
racial disparity in the odds of experiencing rapid cognitive deterioration for African Americans
OR, 1.89) and Hispanics (OR, 1.6) relative to whites over 2 years as well as in the odds of
experiencing fast rates of enhancement in cognition among ethnic minorities (ORblack, 0.68;
ORhispanic, 0.77) compared to their white counterparts. This racial disparity remained in
cognitive decline in the long run (over 2 – 14 years) but was not found in long-term cognitive
improvement.
EMPIRICAL LITERATURE
39

Cognitive decline and cognitive improvement
Several studies found average decline in cognitive performance with age (Cournot et al., 2006;
Finch, 2009; Hertzog et al., 2008; Salthouse, 2009). For example, Wilson et al. (2002), by
summing smoothed person-specific cognitive trajectories, found that a 1-year increase in age was
associated with 0.1 to 0.45 standard deviation (SD) decline in different aspects of cognitive
function. Additionally, a recent study explored the heterogeneity in trajectories of immediate and
delayed verbal memory trajectories. Bayes-Marin et al. (2020) utilized latent class growth
analysis to examine trajectories of cognitive score at among older adults aged 50 to 64 and 65
and above in Spain and found that 24.7% of middle-aged adults experienced an accelerated
cognitive deterioration in delayed verbal memory during a 7-year period. Nevertheless, these
studies may suffer from random noise and practice effect. The latter refers to the fact that one's
improved cognitive performance in a follow-up test is due to familiarity with the test rather than
actual enhancement in cognitive function. Furthermore, knowledge about risk factors associated
with the most rapidly changing cognitive function remains insufficient. Previous studies also did
not explore the enhancement in cognitive performance. Cournot et al. (2006) suggest that this
enhancement may be due to young age, practice effect, and the exclusion of persons living with
dementia, but less is known about this cognitive improvement, and factors associated with it.
Health factors
There is extensive literature exploring the association between health risk factors (hypertension,
heart disease, diabetes, obesity, stroke, and depression) and cognitive decline or dementia. There
is a hypothesis supporting this association. One hypothesized mechanism for the association
between cardiovascular risk factors and cognitive changes is that vascular pathology may lead to
hypertension and brain atrophy (such as gray matter atrophy and white matter loss) at the same
40

time (Leritz, McGlinchey, Kellison, Rudolph, & Milberg, 2011; Viswanathan & Chabriat,
2006). Another potential explanation is that long-term hypertension may lead to the alteration of
cerebral flow that deteriorates one's cognitive performance (Forte et al., 2019). One specific
mechanism for diabetes is that diabetes-related hyperglycemia, inflammation, and insulin
resistance may contribute to plaques or neurofibrillary tangles. Further, diabetes and other
cardiovascular risk factors might lead to stroke and cortical atrophy that deteriorate one's
cognitive performance (Lu, Lin, & Kuo, 2009). Empirical findings of the association between
cardiovascular risk factors and cognitive decline may differ by age (Abell et al., 2018;
Alzheimer's Association, 2019; Levine et al., 2018). For example, recent studies suggest that
high blood pressure and extremely low blood pressure are associated with cognitive
deterioration, both at middle age and late life (Forte et al., 2019; Jefferson, 2020). Other studies
found that mid-life hypertension is statistically insignificantly associated with mid-life cognitive
deterioration, but may be associated with later-life decline (Guan & Huang, 2011; Knopman,
2001; Power et al., 2011). The relative risk of hypertension on dementia ranges from 1.03 to
1.45, varying by cognitive measures, age groups, and durations (Abell et al., 2018; Power et al.,
2011). Another example is obesity. Recent studies suggest that both low and high BMI are a risk
factor of dementia (Beydoun, Beydoun, & Wang, 2008; Levine et al., 2018; Whitmer,
Gunderson, Barrett-Connor, Quesenberry, & Yaffe, 2005). However, some scholars found that
higher BMI is a protective factor of cognitive decline later in life (Hughes, Borenstein,
Schofield, Wu, & Larson, 2009; Suemoto, Gilsanz, Mayeda, & Glymour, 2015). Overall,
knowledge about the significance and magnitude of the association between mid-life health risk
factors and fast declining cognition at middle age is limited.
Socioeconomic factors
41

The cognitive reserve hypothesis demonstrates that cognitive reserve, the brain's neural
networks, can compensate for daily cognitive function despite pathological changes. Education,
occupational attainment, physical exercise, and cognitively-stimulating leisure activities could
build cognitive reserve (Perneczky et al., 2006; Pool et al., 2016; Stern, 2002, 2012; Wang,
MacDonald, Dekhtyar, & Fratiglioni, 2017). Education affects one's cognitive function through
two pathways. Firstly, education may directly build one's cognitive reserve. Additionally, the
"use-it-or-lose-it" hypothesis suggests that one with higher educational attainment may be more
likely to get involved in cognitive-stimulating activities, which further increase one's cognitive
reserve (Lee, Buring, Cook, & Grodstein, 2006; Lee, Kawachi, Berkman, & Grodstein, 2003).
Furthermore, workers are motivated to get involved in cognitive-stimulating activities to enhance
their cognitive reserve (Bonsang, Adam, & Perelman, 2012; Mazzonna & Peracchi, 2012;
Rohwedder & Willis, 2009). Studies also found that cognitive stimulation and social
engagement, related to higher cognitively demanding occupations, may slow the rate of cognitive
aging (Andel, Finkel, & Pedersen, 2016; Finkel, Andel, Gatz, & Federsen, 2009; Lane, Windsor,
Andel, & Luszcz, 2017). As a result, education is associated with both the level of cognitive
function and the rate of cognitive decline. Furthermore, cognitive-stimulation and social
engagement related to cognitively-demanding occupations and work may slow the rate of
cognitive decline. Less is known about the correlation between socioeconomic factors and fast
mid-life rates of cognitive decline.
Additionally, it remains unknown whether these factors are associated with the observed rapid
cognitive enhancement. Thirdly, measurement error in cognitive measurement may bias the
relationship between socioeconomic factors and fast rates of cognitive changes.
DATA AND METHODS
42

This research utilized data from the Health and Retirement Study (HRS) to quantify cognitive
decline/cognitive improvement and associated factors. HRS is a biennial longitudinal survey data
among people aged 50 and above from 1992 to 2016. It includes rich information about
demographics, health, labor supply, and socioeconomic characteristics of older adults. HRS has
an over-sample of Hispanics and African Americans, yet a nationally representative estimate can
be achieved based on its sampling weight. With its longitudinal, nationally representative, and
large-sample nature, HRS is widely utilized to understand the health and socioeconomic status of
older adults.
I restricted the sample to adults aged 50 to 64 with cognitive score for at least two waves from
1996 to 2014. I conducted statistical analysis respectively over 2 years (short-term) and 2-14
years (long-term) to advance understanding about cognitive changes with different durations.
Short-term rapidly changing cognitive function reflects the starting point of divergence in
individual cognitive trajectories. Comparatively, long-term fast cognitive decline and
enhancement refers to a steadily diverging trend.
For the short-term analysis, I restricted the sample to adults aged 50 to 64 with self-reported
cognitive scores at waves t-1 and t from 1996 to 2014, and the sample size was 49,489 person-
years (at wave t). Proxy respondents were dropped in this study, accounting for 4.9% (3,746) of
the whole sample aged 50 to 64. Overall, 7.4% and 13.6% of these proxy respondents,
respectively, were with dementia and cognitive impairment no dementia (CIND). The mortality
rate for persons aged 50 to 64 in HRS 1996 - 2014 by age group was 0.3% (50 to 54), 1.2% (55
to 59), and 2% (60 to 64). The small proportion of deceased respondents indicates that survival
selection - persons with poor health conditions or lower SES who survived to the end of the
43

study period may be more resilient thus experienced slower cognitive decline - may not be an
issue for this short-term analysis.
I additionally identified persons with rapid decline and enhancement in cognition in the long run.
Specifically, I restricted the sample to people aged 50 to 64 with cognitive score for at least 2
waves from 1996 to 2014 (18,023 persons). The mortality rate by age group at the end of the
follow-up was respectively 1.1% (50 to 54), 2.3% (55 to 59), and 7.6% (60 to 64). The sample
with longer durations may be more likely to suffer from mortality selection than the sample with
a 2-year follow-up.
Cognitive Measurement
The HRS utilizes the Telephone Interview for Cognitive Status (TICS) to test self-respondents'
cognitive function. TICS was adapted from the Mini-Mental State Exam (MMSE), a clinical tool
to measure cognitive function. Particularly, the cognitive test-based measure in HRS has been
validated against the gold standard-neuropsychological assessment-based cognition from Aging,
Demographics, and Memory Study (ADAMS) (Crimmins, Kim, Langa, & Weir, 2011) . This
cognitive test-based measure among self-respondents consists of three cognitive tests –
immediate and delayed word recall (0-10 for each test), counting down from 100 by 7 (0-5), and
counting backward from 20 (0-2). The total cognitive score (0-27) can be classified into three
groups based on their cognitive status: dementia (0-6), cognitive impairment no dementia
(CIND), and cognitively normal (12-27). HRS imputed some missing values in self-reported
cognitive score based on multivariate regression analysis (Fisher, Hassan, Faul, Rodgers, &
Weir, 2017).
I standardized this composite cognitive score among each age group and wave to account for
cohort differences. This measure was used by previous scholars (Zheng, Yan, Yang, Zhong, &
44

Xie, 2018) to study the association between glucose and cognitive decline. Specifically, I defined
a dummy variable of short-term cognitive decline that equaled 1 if the respondent experienced 1
or more standard deviation (SD) decline in his/her cognitive score from wave t-1 to t, and 0
otherwise. Similarly, I generated a dummy variable that equaled 1 if one's standardized cognitive
score increased by 1 or more SD over 2 years. Furthermore, I generated an individual-specific
slope of cognitive decline and enhancement. As Model 1 below shows, I ran a linear regression
of cognitive score on wave and computed a person-specific slope of change in cognitive score,
𝛽 1𝑖 ̂
.This slope, ranging from -18 to 15, referred to one's linear trend of change in cognitive score
by wave. For example, the interpretation of a slope of -0.1 from HRS 2000 to HRS 2014 is that
from 2000 to 2014, the respondent experienced a 0.1 point decrease in his/her cognitive score
( scale 0-27) every two years on average. Then I generated long-term cognitive decline as a
dummy variable that equaled 1 if one's slope was -2 or below, indicating that person i had the
fast speed of cognitive deterioration if the respondent experienced 2 or more points decline in
his/her cognitive score in every two years. The average duration was 8.3 years and the median
was 8 years (Appendix Table 8).
𝐶𝑜𝑔𝑠𝑐𝑜𝑟 𝑒 𝑖𝑡
= 𝛽 1𝑖 𝑤𝑎𝑣𝑒 𝑡 + µ
𝑖 (1)
This research reduced two types of measurement error in the cognitive variables. Firstly, I
censored the highest 1 percent of cognitive changes, which might be outliers. Secondly, dementia
is irreversible. Therefore, those whose cognitive status changed from dementia at wave t to
cognitively normal in subsequent waves may be due to measurement error at wave t. Therefore, I
dropped 123 observations at wave t with cognitive score lower than 6 (with dementia) but higher
than 12 (cognitively normal) at time t+1 and t+2.
Independent Variables
45

This study controlled for variables in the empirical model that, hypothetically, would be
associated with cognitive decline - demographic characteristics (race, gender, age, education),
marital and socioeconomic status (SES, household annual income quartiles, household wealth
quartiles), working characteristics (occupation, working status), and health risk factors (heart
disease, diabetes, hypertension cancer, stroke, body mass index, depressive symptoms) at
baseline. The author further controlled for incident stroke.
Specifically, household income and wealth quartiles were generated based on the rank among
older adults aged 50 to 64 with cognitive score from 1996 to 2014. The occupation variables in
HRS were respectively defined by the respondent's current job occupation based on 1980 (for
HRS 1992-2014), 2000 (for HRS 2004-2014), or 2010 (for HRS 2010-2014) Census occupation
codes. HRS asked respondents whether they have ever been told by a doctor that they have any
health conditions, such as heart disease. Additionally, HRS used a mental health index, a score
on the Center for Epidemiologic Studies Depression (CESD) to measure one's depressive
symptoms. This score included eight indicators: depression, everything takes effort, not feeling
happy, not enjoying life, sleep difficulty, loneliness, sadness, and could not get going. A
respondent was classified as having depressive symptoms if his/her CESD score was four and
above.
Model Specification
I separately conducted four logistic regression models, Equation (2) to (5), to explore factors
associated with fast rates of cognitive changes in a 2-year (Equation (2) (3)) and a 2 to 14-year
(Equation (4) (5)) window. For Equation (2) and (3), the independent variable 𝐶𝑜𝑔𝑑𝑒𝑐𝑙𝑖𝑛𝑒 𝑖𝑡

(𝐶𝑜𝑔𝑖𝑚𝑝𝑟𝑜𝑣𝑒 𝑖𝑡
) represents, from time t-1 to time t, one with the most rapid declining
(increasing) cognitive abilities. Vectors 𝑑𝑒𝑚𝑜𝑔𝑟𝑎𝑝 ℎ𝑖𝑐
𝑖 represent age at time t, race, education,
46

and gender for respondent i. Vectors 𝒍𝟐𝒉𝒆𝒂𝒍𝒕𝒉 𝒊𝒕
represent health conditions at time t-1 for
respondent i (heart disease, diabetes, hypertension, cancer, stroke, BMI, and depressive
symptom) and ∆𝑠𝑡𝑟𝑜𝑘𝑒 𝑖 𝑡 stands for incident stroke. The variable 𝑐 𝑜𝑔𝑠𝑐𝑜𝑟𝑒 𝑖𝑡 −1
refers to the
cognitive score at time t-1, which indicates one's baseline level of cognition. Additionally, the
vectors 𝑺𝑬𝑺 𝒊𝒕 −𝟏 consist of one's income, wealth, and occupation at time t-1. Last, 𝛿 𝑡 indicates
the time-fixed effect. As Table 3 illustrates, Equation (4) and (5) differ from Equations (2) and
(3) on four aspects: Firstly, the unit of analysis is a person instead of a person-year. Secondly,
baseline characteristics were observed at time t0 - the beginning of the observation rather than
time t-1 . Furthermore, the average duration of the analysis was 8.3 years instead of 2 years. Last,
the analysis with a 2 to 14-year window no longer controlled for time-fixed effects compared to
the 2-year window analysis.

𝑙𝑛
𝐶𝑜𝑔𝑑𝑒𝑐𝑙𝑖𝑛𝑒 𝑖𝑡
1 − 𝐶𝑜𝑔𝑑𝑒𝑐𝑙𝑖𝑛𝑒 𝑖𝑡

= 𝛽 0
+ 𝜶 𝒅𝒆𝒎𝒐𝒈𝒓𝒂𝒑𝒉𝒊𝒄 𝒊 + 𝜷 𝒍𝟐𝒉𝒆𝒂𝒍𝒕𝒉 𝑖𝑡
+ϕ∆𝑠𝑡𝑟𝑜𝑘𝑒 𝑖𝑡
+ 𝜃 𝑐𝑜𝑔𝑠𝑐𝑜𝑟𝑒 𝑖𝑡 −1
+ 𝜸 𝑺𝑬𝑺 𝒊𝒕 −𝟏 + 𝛿 𝑡 (2)
𝑙𝑛
𝐶𝑜𝑔𝑖𝑚𝑝𝑟𝑜𝑣𝑒 𝑖𝑡
1 − 𝐶𝑜𝑔𝑖𝑚𝑝 𝑟𝑜𝑣𝑒 𝑖𝑡

= 𝛽 0
+ 𝜶 𝒅𝒆𝒎𝒐𝒈𝒓𝒂𝒑𝒉𝒊𝒄 𝒊 + 𝜷 𝒍𝟐𝒉𝒆𝒂𝒍𝒕𝒉 𝑖𝑡
+ϕ∆𝑠𝑡𝑟𝑜𝑘𝑒 𝑖𝑡
+ 𝜃 𝑐𝑜𝑔𝑠𝑐𝑜𝑟𝑒 𝑖𝑡 −1
+ 𝜸 𝑺𝑬𝑺 𝒊𝒕 −𝟏 + 𝛿 𝑡 (3)
𝑙𝑛
𝐶𝑜𝑔𝑑𝑒𝑐𝑙𝑖𝑛𝑒 𝑖 1 − 𝐶𝑜𝑔𝑑𝑒𝑐𝑙𝑖𝑛𝑒 𝑖
= 𝛽 0
+ 𝜶 𝒅𝒆𝒎𝒐𝒈𝒓𝒂𝒑𝒉𝒊𝒄 𝒊 + 𝜷 𝒍𝟐𝒉𝒆𝒂𝒍𝒕𝒉 𝑖𝑡 0
+ ϕ∆𝑠𝑡𝑟𝑜𝑘𝑒 𝑖 + 𝜃 𝑐𝑜𝑔𝑠𝑐𝑜𝑟𝑒 𝑖𝑡 0
+ 𝜸 𝑺𝑬𝑺 𝒊𝒕𝟎 (4)
𝑙𝑛
𝐶𝑜𝑔𝑖𝑚𝑝𝑟𝑜𝑣𝑒 𝑖 1 − 𝐶𝑜𝑔𝑖𝑚𝑝𝑟𝑜𝑣𝑒 𝑖
= 𝛽 0
+ 𝜶 𝒅𝒆𝒎𝒐𝒈𝒓𝒂𝒑𝒉𝒊𝒄 𝒊 + 𝜷 𝒍𝟐𝒉𝒆𝒂 𝒍𝒕𝒉 𝑖𝑡 0
+ϕ∆𝑠𝑡𝑟𝑜𝑘𝑒 𝑖 + 𝜃 𝑐𝑜𝑔𝑠 𝑐𝑜𝑟𝑒 𝑖𝑡 0
+ 𝜸 𝑺𝑬𝑺 𝒊𝒕𝟎 (5)

47

Table 3. Comparison of analysis in a 2-year and a 2 to 14-year window

2 years 2 to 14 years
Unit of analysis Person-year Person
Average duration 2 years 8.3 years
Cognitive measure Standardized cognitive score Slope of cognitive score
Cognitive
Decline/improvement
1 or more SD decline/increase
in standardized cognitive score
2 or more points decline in
cognitive score every wave
(average rate)
Beginning of the
window Time t-1 Time t0
End of the window Time t
Time tn, n refers to the last
observation
Time fixed effects Yes No
Sample size 50,913 18,013
Note. Statistics unweighted.

RESULTS
Rapid Cognitive Changes in The Short Run
Figure 4 and 5 show the distribution of cognitive score and change in standardized cognitive
score over two years.
Figure 4 suggests that 10.1 percent of middle-aged adults have dementia (score 0-6; 1.2%) or
CIND (score 7 – 11; 8.9%). Figure 5 shows that 8.4 percent of the population experienced a one
or more SD decline in cognitive score over two years, and 10.9 percent have one or more SD
increase in cognitive score. Statistical analysis demonstrates that even among cognitively normal
individuals (score 12-27, 89.9%), 7.8 percent of them experience sharp cognitive decline over
two years, and 11.2 percent of them have fast rates of cognitive enhancement over two years.
Particularly, the "cognitive decline" group has the lowest cognitive score at baseline and the
highest prevalence of health risks (e.g., hypertension, 40.2%; diabetes, 14.1%) except for the
BMI. Specifically, the "no change" group has the highest SES, then the "cognitive improvement"
group, and the last, the "cognitive decline" group (Appendix Table 6).

48

Figure 4. The distribution of cognitive score among persons aged 50 to 64, 1996-2014

Note. Outliers are excluded. Sample restricted to persons aged 50 to 64 in HRS 1996-2014, with at least
two consecutive waves of cognitive score.

49

Figure 5. The distribution of changes in standardized cognitive score from time t-1 to time t
among cognitively normal persons, 1996-2014

Note. Sample restricted to persons aged 50 to 64, with at least two waves of self-reported cognitive score,
1996-2014. I define cognitive decline/improvement by identifying individuals with 1 or more standard
deviation decrease/increase over two years in his/her standardized cognitive score (stratified by age group
and wave).

Figure 6 and Table 4 (Model 1 and 2) suggest that ethnic minorities, male, those with lower
education, income, wealth, and less cognitively demanding occupations are statistically
significantly associated with a higher chance of cognitive deterioration and a lower chance of
enhancement. For example, African Americans have 90% higher odds to experience
deteriorating cognitive function (odds ratio, 1.87; 95%CI, 1.70 – 2.05) and a 32% lower chance
for cognitive enhancement (odds ratio, 0.68; 95%CI,0.62-0.75) compared to whites controlling
for other covariates. Those from the highest income quartile, compared to the lowest income
quartile, have a 35% lower chance (odds ratio, 0.65; 95%CI, 0.56-0.74) to develop cognitive
50

dysfunction and a 53% higher chance for improvement (odds ratio, 1.53, 95%CI, 1.33-1.7) .
Additionally, the onset of stroke, baseline diabetes, hypertension, stroke, and depressive
symptoms are statistically significantly associated with a higher chance (odds ratios range from
1.07 to 1.51) of the fast rate of deterioration in cognition and a lower chance of enhancement
(odds ratios range from 0.66 to 0.9). For example, the odds of having the fast speed of cognitive
decline among diabetics is 11% higher than non-diabetics (odds ratio, 1.11, 95%CI, 1.01-1.23).
One point-increase in BMI is associated with a lower chance of a fast cognitive decline (odds
ratio 0.99, 95%CI 0.99-1.00), indicating a slight protective effect.
Figure 6. Odds ratio (95%CI) of cognitive decline/improvement in the short run, 1996-2014.

Note. Reference groups: white, less than high school, female, household income 25 percentiles at t-1,
household wealth 25 percentiles at t-1, managerial occupation at t-1. Baseline health conditions, income,
wealth, and occupation are measured at t-1. Data pooled from 1996 to 2014 among individuals aged 50 to
64.

0
0.5
1
1.5
2
2.5
3
Cognitive Decline Cognitive Improvement
51

Table 4. Odds ratio of cognitive decline and improvement in 2 years, 1996-2014

(1) (2) (3) (4)
Decline, 2 years Improvement, 2
years
Decline, 2-14
years
Improvement,
2-14 years
Black 1.89*** 0.68*** 2.30*** 0.92
[1.73 - 2.08] [0.62 - 0.75] [1.98 - 2.68] [0.78 - 1.09]
Hispanic 1.60*** 0.75*** 2.65*** 1.11
[1.37 - 1.86] [0.65 - 0.88] [2.06 - 3.42] [0.84 - 1.45]
Other race 1.55*** 0.76** 1.63** 1.28
[1.29 - 1.86] [0.63 - 0.92] [1.18 - 2.24] [0.92 - 1.76]
High school 0.70*** 1.59*** 0.73*** 1.93***
[0.64 - 0.77] [1.44 - 1.76] [0.63 - 0.85] [1.61 - 2.31]
College and above 0.48*** 2.28*** 0.63*** 2.91***
[0.44 - 0.53] [2.06 - 2.53] [0.53 - 0.74] [2.41 - 3.52]
Male 1.16*** 0.73*** 1.36*** 1
[1.08 - 1.24] [0.68 - 0.78] [1.20 - 1.53] [0.88 - 1.14]
Married (t-1) 1.05 0.89** 0.9 0.86*
[0.97 - 1.14] [0.82 - 0.97] [0.78 - 1.04] [0.74 - 0.99]
Onset of stroke 1.59** 0.63* 1.33 0.32***
[1.16 - 2.18] [0.43 - 0.93] [0.98 - 1.79] [0.19 - 0.53]
Heart disease (t-1) 1.07 1.09 1.12 1.2
[0.97 - 1.17] [0.98 - 1.20] [0.94 - 1.33] [1.00 - 1.45]
Diabetes (t-1) 1.12* 0.89* 1.45*** 1.1
[1.02 - 1.23] [0.81 - 0.98] [1.24 - 1.70] [0.92 - 1.32]
BMI (t-1) 0.99* 1 0.99* 1
[0.99 - 1.00] [1.00 - 1.01] [0.98 - 1.00] [0.99 - 1.02]
Hypertension (t-1) 1.08* 0.88*** 1.21** 1
[1.01 - 1.15] [0.82 - 0.94] [1.07 - 1.36] [0.88 - 1.14]
Stroke (t-1) 1.32** 0.71*** 1.16 0.93
[1.11 - 1.58] [0.59 - 0.86] [0.86 - 1.56] [0.69 - 1.27]
Depressive symptoms (t-1) 1.22** 0.82** 1.19 0.95
[1.05 - 1.41] [0.71 - 0.95] [0.91 - 1.55] [0.72 - 1.25]
Baseline cognitive score 1.33*** 0.75*** 1.23*** 0.77***
[1.32 - 1.35] [0.74 - 0.76] [1.21 - 1.25] [0.75 - 0.78]
Income 50 percentiles 0.84*** 1.31*** 1.01 1.22*
[0.76 - 0.92] [1.19 - 1.45] [0.86 - 1.18] [1.02 - 1.47]
Income 75 percentiles 0.81*** 1.30*** 0.91 1.34**
[0.73 - 0.90] [1.16 - 1.46] [0.75 - 1.10] [1.08 - 1.65]
Income 100 percentiles 0.69*** 1.37*** 0.9 1.46**
[0.61 - 0.78] [1.21 - 1.56] [0.72 - 1.12] [1.14 - 1.87]
Wealth 50 percentiles 0.92 1.12* 0.72*** 0.91
[0.84 - 1.01] [1.02 - 1.23] [0.62 - 0.84] [0.77 - 1.08]
Wealth 75 percentiles 0.83*** 1.25*** 0.75** 0.96
[0.75 - 0.92] [1.13 - 1.39] [0.63 - 0.89] [0.79 - 1.18]
52

Wealth 100 percentiles 0.74*** 1.33*** 0.68*** 1.01
[0.66 - 0.83] [1.18 - 1.49] [0.56 - 0.83] [0.80 - 1.27]
Professional 1.08 1.08 1.04 1.17
[0.94 - 1.23] [0.95 - 1.22] [0.80 - 1.34] [0.90 - 1.52]
Sales&admin 1.09 0.97 1.27 1
[0.96 - 1.25] [0.85 - 1.10] [0.99 - 1.63] [0.77 - 1.30]
Protection/Military 1.13 1.06 1.48 1.17
[0.82 - 1.55] [0.79 - 1.43] [0.89 - 2.44] [0.67 - 2.04]
Cleaning/Building 1.28** 0.80** 1.50** 1.14
[1.09 - 1.51] [0.69 - 0.94] [1.12 - 2.01] [0.84 - 1.53]
Production/Operation 1.57*** 0.87* 1.63*** 0.93
[1.37 - 1.80] [0.76 - 0.99] [1.27 - 2.10] [0.71 - 1.22]
Nonworking 1.57*** 0.74*** 1.86*** 1.06
[1.38 - 1.78] [0.66 - 0.84] [1.47 - 2.35] [0.82 - 1.36]
Constant 6,888.71* 0 0.00*** 0.00***
[1.07 -
44339209.29]
[0.00 - 4.30] [0.00 - 0.00] [0.00 - 0.01]
Observations 50,923 50,923 18,013 18,011
Note. "Decline" refers to the model that the outcome is a dummy that equals 1 if one has 1 or more SD
decline in cognitive score over 2 years. "Improvement" refers to the model that the outcome is a dummy
that equals 1 if one has cognitive improvement. Other covariates include age and age square, time fixed
effects, cancer, flag of missing values for each variable. Standard errors are clustered at the individual
level. ***p<0.01. **p<0.05. *p< 0.1.

Rapid Cognitive Changes in The Long Run
The slope of cognitive changes over 2 to 14 years is a person-specific average rate of change in
one's cognitive score. Respectively, 8.4 and 10.9 percent of people experience rapidly declining
or enhancing cognition. Comparatively, only 1.7% of them had dementia at baseline (Appendix
Table 10).
Appendix Table 7 exhibits the distribution of person-specific demographic, health, and
socioeconomic characteristics by cognitive change over the long run. This sample, unweighted,
consisted of more African Americans (22.1%) and Hispanics (5.5%) compared to the sample for
the short-term analysis (Appendix Table 6; African Americans, 10.2%; Hispanics, 3%). This
sample was also less educated, has lower SES, but less likely to have cardiovascular risk factors.
53

One potential explanation for this contrast was the mortality selection – those with lower SES
have better health conditions if they survive to the end of the duration.

Figure 7 (and Table 4, Model 3 and 4) indicates risk and protective factors for the fast speed of
cognitive decline and improvement in 2-14 years. Hypertension is associated with sharp
cognitive decline (odds ratio, 1.21, 95%CI, 1.07-1.36). Having a degree of college and above
relative to less than high school is a protective factor of the fast rate of cognitive decline (odds
ratio, 0.48, 95%CI, 0.44-0.53), and promotes fast cognitive enhancement (odds ratio, 2.91,
95%CI, 2.41- 3.52). Higher baseline income is also statistically associated with this
improvement (ORstrokeonset 0.32, 95%CI 0.19-0.53; ORcollege 2.91, 95%CI 2.41-3.52;
ORincome 1.46 95%CI 1.14 1.87).

54

Figure 7. Odds ratio (95%CI) of cognitive decline/improvement in the long run, 1996-2014.

Note. Reference groups: white, less than high school, female, household income 25 percentiles at t-1,
household wealth 25 percentiles at t0, managerial occupation at t0. Baseline health conditions, income,
wealth, and occupation are measured at t-1. Data pooled from 1996 to 2014 among individuals aged 50 to
64.

Robustness Check
I ran the robustness check analysis and restrict the sample to those begin as cognitively normal
with short and long observational windows (Appendix Table 11). For this analysis, I excluded
the confounding effect of dementia and cognitive impairment at baseline, accounting for 11.1%
(2 years) and 13.1% (2-14 years) of the population. Since people with baseline dementia and
CIND might increase their cognitive performance over time not due to actually improved
0
0.5
1
1.5
2
2.5
3
3.5
4
Cognitive decline Cognitive Improvement
55

function but due to regression to the mean, results from this analysis is more robust. The
conclusion from the robustness analysis is not altered compared to the main results. However,
odds ratios of some cardiovascular risk factors (e.g., hypertension for a 2-year decline), although
significantly larger than 1, are no longer significant at a 10% confidence level.

DISCUSSION
This research identifies the population with the fastest rates of cognitive changes at middle age in
the short term (over 2 years) and the long term (over 2 to 14 years). It finds that around 8.4
percent of individuals experienced a fast speed of cognitive deterioration over 2 years, and 10.9
percent of them rapidly improved their cognitive abilities. Additionally, 8.7 percent of the
population had a fast cognitive decline in 2-14 years and 7.5% of them increased their cognitive
function at a fast speed. These small percentages of the population may be mostly affected by
cognitive deterioration in terms of daily functions and developing dementia later in life
(Alzheimer's Association, 2019; Beason-Held et al., 2013; Harada et al., 2013).
This paper shows that cardiovascular risk factors, demographic factors, and socioeconomic
factors are all statistically significantly associated with fast cognitive deterioration and
enhancement. For example, hypertension is a risk factor for fast cognitive decline at middle age
(2 years, odds ratio, 1.08, 95%CI, 1.01-1.15; 2-14 years, odds ratio, 1.21, 95%CI (1.07-1.36). So
is diabetes (2 years, odds ratio, 1.12, 95%CI 1.02-1.23; 2-14 years, odds ratio, 1.45, 95%CI 1.24-
1.7). A degree of college and above reduces the likelihood of fast cognitive decline and promotes
faster cognitive enhancements in 2 years compared to those with less than high school degree
(decline, odds ratio 0.48, 95%CI 0.44-0.53; improvement, odds ratio 2.28, 95%CI 2.06-2.53).
56

The author finds a significant association between ethnic minorities and rapid cognitive decline
at ages 50 to 64. For example, the odds of experiencing fastest rates of cognitive decline for
African Americans relative to whites are 1.89 (95%CI, 1.73-2.08) in the short run and 2.3
(95%CI, 1.98-2.68) in the long run. Levine et al. (2018) and Weuve et al. (2018), comparing
levels and rates of cognitive decline respectively at age 50 and above and age 65 and above,
found racial differences in levels but not slopes of cognitive decline. One potential explanation
of the divergence in findings is that cognitive trajectories by race diverge from age 50 to 65, and
the gap converges after those with the most rapid rates of cognitive decline develop dementia
later in life.
Besides affirming that cardiovascular risk factors affect cognitive dysfunction, this study adds to
current literature and suggests that socioeconomic status (SES) is a critical predictor of cognitive
changes across all models. This finding indicates the potential existence of disparities in
cognitive trajectories in addition to that in cognitive levels. Furthermore, cognitive demanding
occupations (e.g., managerial, professional) may indicate higher levels of cognitive stimulation,
social engagement that increase cognitive reserve (Lane et al., 2017). Additionally, individuals
from more cognitive demanding occupations may be more motivated to invest in their cognitive
abilities, thus more likely to have the highest rate of cognitive enhancement (Bonsang et
al. ,2012; Mazzonna and Peracchi, 2012&2017). Those with lower SES may have poorer health
conditions, retire earlier, have less access to health care services, and suffer from more severe
cognitive deterioration.
This paper has some limitations. Firstly, this research, different from the random control trials
(RCT), does not identify a causal effect of each health and SES factor on fast rates of cognitive
changes. This research, utilizing a large-sample, nationally representative, and longitudinal data,
57

identifies the at-risk population with the highest speed of cognitive decline and improvements.
Ethnic minorities, those with low education, SES, and those with cardiovascular risk factors may
be more affected by this cognitive decline in daily functions and may be more likely to develop
dementia in later life.
Additionally, this research uses self-reported diagnoses of cardiometabolic factors. Ethnic
minorities with these conditions may not get the diagnosis in time because of less awareness and
access to health care. Not controlling for this underdiagnosis (positively associated with
cognitive deterioration and negatively associated with SES), may lead to a downward bias in the
association between higher SES and cognitive decline.
Lastly, I do not control for genetic factors, lifestyle factors, specific job characteristics, and drugs
targeted at cardiovascular risk factors and antidepressant drugs, and childhood IQ (Blazer et al.,
2015; Deary et al., 2009; National Academies of Sciences & Medicine, 2017). All of these
factors may be associated with cognitive trajectories. I do not include these factors due to the
limited sample size. Additionally, this paper aims to understand the risk/protective health and
socioeconomic factors of rapid cognitive decline. The next step would be to explore the effect of
these factors further and providing further evidence of effective intervention programs.
Overall, this paper identifies the at-risk population for fast speed of cognitive decline at middle
age. Ethnic minorities, males, those with less educational attainment, the lower SES group, and
those with diabetes, hypertension, and depressive symptoms are at risk of developing rapidly
declining cognitive abilities. Notably, these subgroups, particularly ethnic minorities and those
with lower SES, are more likely to retire early and develop dementia later in life. They are also
more vulnerable in alleviating or addressing the burden of cognitive decline. Interventions on
preventing or delaying fast cognitive deterioration among these at-risk subgroups, thus, are
58

needed to ensure their life quality and financial security. Additionally, these at-risk subgroups,
being aware that they are more likely to experience high-speed decline in cognition, can better
plan for cognitive decline. One example is that they can adopt more healthy behaviors such as
physical exercise, mentally stimulating activities, and better management of cardiovascular
disease.
Furthermore, this study advances knowledge about cognitive trajectories in the short run and the
long run. The distribution of persons with rapidly declining cognitive function over two years is
similar to that over two to fourteen years. The implication of short-term cognitive changes may
differ from long-term changes. Fast declining cognition over two years reflects temporal
cognitive deterioration, some of which might be reversed. Comparatively, rapid cognitive decline
in the long run may demonstrate a steady functional loss in cognition. Persons with such a
decline may endure functional and financial difficulties over a long period, and they may need
more policy and community support. Further studies about the effectiveness of interventions on
preventing or delaying cognitive decline with different durations are needed. Moreover,
demographic and cardiovascular risk factors are statistically significantly associated with fast
rates of cognitive enhancement in the short run but not in the long run. More explorations about
cognitive enhancement in the long run will provide more insights.

59

REFERENCE
Abell, J. G., Kivimäki, M., Dugravot, A., Tabak, A. G., Fayosse, A., Shipley, M., . . . Singh-
Manoux, A. (2018). Association between systolic blood pressure and dementia in the
Whitehall II cohort study: role of age, duration, and threshold used to define
hypertension. European Heart Journal, 39(33), 3119-3125. doi:10.1093/eurheartj/ehy288
Alzheimer's Association (2019). 2019 Alzheimer's disease facts and figures. Alzheimer's &
dementia: the journal of the Alzheimer's Association, 15(3), 321-387.
doi:10.1016/j.jalz.2019.01.010
Andel, R., Finkel, D., & Pedersen, N. L. (2016). Effects of Preretirement Work Complexity and
Postretirement Leisure Activity on Cognitive Aging. The Journals of Gerontology Series
B: Psychological Sciences and Social Sciences, 71(5), 849-856.
doi:10.1093/geronb/gbv026
Bayes-Marin, I., Fernández, D., Lara, E., Martín-María, N., Miret, M., Moreno-Agostino, D., . . .
Olaya, B. (2020). Trajectories of Immediate and Delayed Verbal Memory in the Spanish
General Population of Middle-aged and Older Adults. Brain sciences, 10(4), 249.
doi:10.3390/brainsci10040249
Beason-Held, L. L., Goh, J. O., An, Y., Kraut, M. A., O'Brien, R. J., Ferrucci, L., & Resnick, S.
M. (2013). Changes in brain function occur years before the onset of cognitive
impairment. The Journal of neuroscience : the official journal of the Society for
Neuroscience, 33(46), 18008-18014. doi:10.1523/JNEUROSCI.1402-13.2013
Beydoun, M. A., Beydoun, H. A., & Wang, Y. (2008). Obesity and central obesity as risk factors
for incident dementia and its subtypes: a systematic review and meta‐analysis. Obesity
Reviews, 9(3), 204-218. doi:10.1111/j.1467-789X.2008.00473.x
Blazer, D. G., Yaffe, K., & Liverman, C. T. (2015). Cognitive aging: Progress in understanding
and opportunities for action: National Academies Press Washington, DC.
Bonsang, E., Adam, S., & Perelman, S. (2012). Does retirement affect cognitive functioning?
Journal of Health Economics, 31(3), 490-501. doi:10.1016/j.jhealeco.2012.03.005
Bratzke, L. C., Koscik, R. L., Schenning, K. J., Clark, L. R., Sager, M. A., Johnson, S. C., . . .
Hogan, K. J. (2018). Cognitive decline in the middle‐aged after surgery and anaesthesia:
results from the Wisconsin Registry for Alzheimer's Prevention cohort. Anaesthesia,
73(5), 549-555. doi:10.1111/anae.14216
Cournot, C. M., Marquié, B. J., Ansiau, B. D., Martinaud, B. C., Fonds, B. H., Ferrières, B. J., &
Ruidavets, B. J. (2006). Relation between body mass index and cognitive function in
healthy middle-aged men and women. Neurology, 67(7), 1208-1214.
doi:10.1212/01.wnl.0000238082.13860.50
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
60

the Aging, Demographics, and Memory Study. The Journals of Gerontology, Series B:
Psychological Sciences and Social Sciences, 66 Suppl 1(1), i162-171.
doi:10.1093/geronb/gbr048
Daffner, K. R. (2010). Promoting successful cognitive aging: a comprehensive review. Journal
of Alzheimer's disease : JAD, 19(4), 1101-1122. doi:10.3233/JAD-2010-1306
Finch, C. E. (2009). The neurobiology of middle-age has arrived. Neurobiology of Aging, 30(4),
515-520. doi:10.1016/j.neurobiolaging.2008.11.011
Finkel, D., Andel, R., Gatz, M., & Federsen, N. (2009). The Role of Occupational Complexity in
Trajectories of Cognitive Aging Before and After Retirement. Psychology and aging,
24(3), 563. doi:10.1037/a0015511
Fisher, G. G., Hassan, H., Faul, J. D., Rodgers, W. L., & Weir, D. R. (2017). Health and
Retirement Study Imputation of Cognitive Functioning Measures: 1992–2014 (Final
Release Version). http://hrsonline.isr.umich.edu/modules/meta/xyear/cogimp/desc/
COGIMPdd.pdf.
Forte, G., De Pascalis, V., Favieri, F., & Casagrande, M. (2019). Effects of Blood Pressure on
Cognitive Performance: A Systematic Review. Journal of clinical medicine, 9(1).
doi:10.3390/jcm9010034
Guan, J.-W., & Huang, C.-Q. (2011). No Association Between Hypertension and Risk for
Alzheimers Disease: A Meta-Analysis of Longitudinal Studies. Journal of Alzheimer's
Disease, 27(4), 799-807. doi:10.3233/JAD-2011-111160
Harada, C. N., Natelson Love, M. C., & Triebel, K. L. (2013). Normal Cognitive Aging. Clinics
in Geriatric Medicine, 29(4), 737-752. doi:10.1016/j.cger.2013.07.002
Hertzog, C., Kramer, A. F., Wilson, R. S., & Lindenberger, U. (2008). Enrichment Effects on
Adult Cognitive Development: Can the Functional Capacity of Older Adults Be
Preserved and Enhanced? Psychological Science in the Public Interest, 9(1), 1-65.
doi:10.1111/j.1539-6053.2009.01034.x
Hughes, T. F., Borenstein, A. R., Schofield, E., Wu, Y., & Larson, E. B. (2009). Association
between late-life body mass index and dementia: The Kame Project. Neurology, 72(20),
1741-1746. doi:10.1212/WNL.0b013e3181a60a58
Jefferson, L. A. (2020). Midlife Consequences of Cumulative Blood Pressure Exposure:
Importance of a Lifespan Approach. Circulation, 141(9), 725-727.
doi:10.1161/CIRCULATIONAHA.120.044447
Knopman, D. (2001). Cardiovascular Risk Factors and Cognitive Decline in Middle-Aged
Adults.(Abstract). JAMA, The Journal of the American Medical Association, 285(9),
1135.
61

Lane, A. P., Windsor, T. D., Andel, R., & Luszcz, M. A. (2017). Is Occupational Complexity
Associated with Cognitive Performance or Decline? Results from the Australian
Longitudinal Study of Ageing. Gerontology, 63(6), 550-559. doi:10.1159/000475559
Lee, S., Buring, J. E., Cook, N. R., & Grodstein, F. (2006). The Relation of Education and
Income to Cognitive Function among Professional Women. Neuroepidemiology, 26(2),
93-101. doi:10.1159/000090254
Lee, S., Kawachi, I., Berkman, L. F., & Grodstein, F. (2003). Education, Other Socioeconomic
Indicators, and Cognitive Function. American journal of epidemiology, 157(8), 712-720.
doi:10.1093/aje/kwg042
Leritz, E., McGlinchey, R., Kellison, I., Rudolph, J., & Milberg, W. (2011). Cardiovascular
Disease Risk Factors and Cognition in the Elderly. Current Cardiovascular Risk Reports,
5(5), 407-412. doi:10.1007/s12170-011-0189-x
Levine, M. E., Harrati, A., & Crimmins, E. M. (2018). Predictors and implications of accelerated
cognitive aging. Biodemography and Social Biology, 64(2), 83-101.
doi:10.1080/19485565.2018.1552513
Lu, F.-P., Lin, K.-P., & Kuo, H.-K. (2009). Diabetes and the Risk of Multi-System Aging
Phenotypes: A Systematic Review and Meta-Analysis (DM and Geriatric Syndromes).
PLoS ONE, 4(1), e4144. doi:10.1371/journal.pone.0004144
Mazzonna, F., & Peracchi, F. (2012). Ageing, cognitive abilities and retirement. European
Economic Review, 56(4), 691-710. doi:10.1016/j.euroecorev.2012.03.004
McArdle, J., Smith, J., & Willis, R. (2009). Cognition and economic outcomes in the Health and
Retirement Survey (No. w15266). In: National Bureau of economic research.
McArdle, J. J., Ferrer-Caja, E., Hamagami, F., & Woodcock, R. W. (2002). Comparative
longitudinal structural analyses of the growth and decline of multiple intellectual abilities
over the life span. Developmental Psychology, 38(1), 115.
National Academies of Sciences, E., & Medicine. (2017). Preventing cognitive decline and
dementia: A way forward: National Academies Press.
Perneczky, R., Drzezga, A., Diehl-Schmid, J., Schmid, G., Wohlschläger, A., Kars, S., . . . Kurz,
A. (2006). Schooling mediates brain reserve in Alzheimer's disease: findings of fluoro-
deoxy-glucose-positron emission tomography. Journal of neurology, neurosurgery and
psychiatry, 77(9), 1060-1063. doi:10.1136/jnnp.2006.094714
Pool, L. R., Weuve, J., Wilson, R. S., Bultmann, U., Evans, D. A., & de Leon, C. F. M. (2016).
Occupational cognitive requirements and late-life cognitive aging. Neurology, 86(15),
1386-1392. doi:10.1212/WNL.0000000000002569
Power, C. M., Weuve, J. J., Gagne, B. J., McQueen, B. M., Viswanathan, B. A., & Blacker, B.
D. (2011). The Association Between Blood Pressure and Incident Alzheimer Disease: A
62

Systematic Review and Meta-analysis. Epidemiology, 22(5), 646-659.
doi:10.1097/EDE.0b013e31822708b5
Rohwedder, S., & Willis, R. (2009). Mental Retirement. IDEAS Working Paper Series from
RePEc.
Salthouse, T. A. (2006). Robust Cognitive Change. Journal of the International
Neuropsychological Society, 18(4), 1-8. doi:10.1017/S1355617712000380
Salthouse, T. A. (2009). When does age-related cognitive decline begin? Neurobiology of Aging,
30(4), 507-514. doi:10.1016/j.neurobiolaging.2008.09.023
Singh-Manoux, A., Hillsdon, M., Brunner, E., & Marmot, M. (2005). Effects of Physical
Activity on Cognitive Functioning in Middle Age: Evidence From the Whitehall II
Prospective Cohort Study. American Journal of Public Health, 95(12), 2252-2258.
Singh-Manoux, A., Kivimaki, M., Glymour, M. M., Elbaz, A., Berr, C., Ebmeier, K. P., . . .
Dugravot, A. (2012). Timing of onset of cognitive decline: results from Whitehall II
prospective cohort study. BMJ, 344, d7622.
Stern, Y. (2002). What is cognitive reserve? Theory and research application of the reserve
concept. Journal of the International Neuropsychological Society, 8(3), 448-460.
doi:10.1017/S1355617702813248
Stern, Y. (2012). Cognitive Reserve in Ageing and Alzheimer's Disease. doi:10.7916/D87S916N
Suemoto, C. K., Gilsanz, P., Mayeda, E. R., & Glymour, M. M. (2015). Body mass index and
cognitive function: the potential for reverse causation. International Journal of Obesity,
39(9), 1383-1389. doi:10.1038/ijo.2015.83
Viswanathan, A., & Chabriat, H. (2006). Cerebral Microhemorrhage. Stroke, 37(2), 550-555.
doi:10.1161/01.STR.0000199847.96188.12
Wang, H.-X., MacDonald, S. W. S., Dekhtyar, S., & Fratiglioni, L. (2017). Association of
lifelong exposure to cognitive reserve-enhancing factors with dementia risk: A
community-based cohort study. PLoS Medicine, 14(3), e1002251.
doi:10.1371/journal.pmed.1002251
Weuve, L. J., Barnes, F. L., Mendes De Leon, B. C., Rajan, T. K., Beck, E. T., Aggarwal, A.
N., . . . Evans, A. D. (2018). Cognitive aging in black and white Americans: Cognition,
cognitive decline, and incidence of Alzheimer disease dementia. Epidemiology, 29(1),
151-159. doi:10.1097/EDE.0000000000000747
Whitmer, R. A., Gunderson, E. P., Barrett-Connor, E., Quesenberry, C. P., & Yaffe, K. (2005).
Obesity in middle age and future risk of dementia: a 27 year longitudinal population
based study. BMJ (Clinical research ed.), 330(7504), 1360-1360.
63

Williams, K. N., & Kemper, S. (2010). Interventions to Reduce Cognitive Decline in Aging.
Journal of Psychosocial Nursing & Mental Health Services, 48(5), 42-51.
doi:10.3928/02793695-20100331-03
Wilson, R. S., Beckett, L. A., Barnes, L. L., Schneider, J. A., Bach, J., Evans, D. A., & Bennett,
D. A. (2002). Individual differences in rates of change in cognitive abilities of older
persons. Psychology and aging, 17(2), 179-193. doi:10.1037/0882-7974.17.2.179
Yaffe, K., Falvey, C. M., Hamilton, N., Harris, T. B., Simonsick, E. M., Strotmeyer, E. S., . . .
Schwartz, A. V. (2013). Association Between Hypoglycemia and Dementia in a Biracial
Cohort of Older Adults With Diabetes Mellitus. Jama Internal Medicine, 173(14), 1-6.
doi:10.1001/jamainternmed.2013.6176
Zheng, F., Yan, L., Yang, Z., Zhong, B., & Xie, W. (2018). HbA 1c , diabetes and cognitive
decline: the English Longitudinal Study of Ageing. Diabetologia, 61(4), 839-848.
doi:10.1007/s00125-017-4541-7
Zissimopoulos, J. M., Crimmins, E. M., & St Clair, P. A. (2014). The value of delaying
Alzheimer's disease onset. Forum for health economics & policy, 18(1), 25.
doi:10.1515/fhep-2014-0013

64

Chapter 3:
The Effect of Cognitive Decline at Middle Age on Labor Supply of Older
Adults
by
Yingying Zhu
ABSTRACT
Current studies affirm that a negative health shock leads to a 2 to 14 percentage point decrease in
the probability of working. However, few studies specifically explore the effect of cognitive
decline on labor supply, which is an important aspect of health.
This paper fills current literature gaps by utilizing a widely-accepted survey data that measures
cognition, the Health and Retirement Study (HRS), to quantify the distribution of cognitive
decline and its effect on labor supply. This research used an individual fixed-effect model to
eliminate the omitted bias due to individual idiosyncratic factors. This study additionally reduced
measurement error by excluding extreme values and eliminated the omitted variable bias due to
physical health shocks by controlling for baseline and incident health conditions.
The author found that 12 percent of older workers aged 50 to 61 experienced fast rates of
cognitive decline over 2 years, leading to a 2.2 percentage point decrease in the expected
probability of working at age 62. The size of this negative effect was larger in the long run,
accounting for a 5.1 percentage point decrease in the probability of working.
This study suggests that cognitive decline at middle age affects labor supply of older workers.
The scientific and technology development in identifying factors associated with cognitive
decline may provide insights for delaying or preventing cognitive decline at middle age.
65

INTRODUCTION
Thorough knowledge about the effect of health on labor supply is crucial for understanding the
cost benefits of interventions on health issues. Further, this knowledge provides insights for
evaluating labor policies and social welfare policies, which moderate the relationship between
health and labor supply (Currie & Madrian, 1999). There is extensive literature that affirms the
relationship between physical or self-reported health and labor market outcomes. However, the
effect of cognitive decline – an important aspect of health –on labor supply is rarely discussed
(Gruber & Madrian, 2002) (Currie & Madrian, 1999; French, 2005; Maestas & Zissimopoulos,
2010) (Coile, 2004). The effect of health on working, thus, could be understated, leading to less
investment in health by individuals and the government. The labor market outcomes including
income and labor force participation rates could also be worse due to less health investment.
Additionally, the cost benefit of interventions on preventing or delaying cognitive deterioration
might be underestimated.
There is a general belief that people experience a decrease in cognitive function when they age.
However, the true trajectory of change in cognitive function (long- term memory, speed of
processing information, and knowledge) at middle age remains unknown. Research shows that
some aspects, such as the speed of processing information, working memory (related to decision
making and reasoning), and long-term memory, peak in one’s twenties and decline afterwards. In
contrast, knowledge and experience, another domain of cognitive function, does not decline until
one’s sixties or seventies (Finch, 2009; Harada, Natelson Love, & Triebel, 2013; McArdle,
Smith, & Willis, 2009; McArdle, Ferrer-Caja, Hamagami, & Woodcock, 2002; Salthouse, 2009;
Wilson et al., 2002). Additionally, there is heterogeneity in individual paths of cognitive change
at middle age.
66

For those with fast rates of cognitive decline, their productivity could be reduced (Harada et al.,
2013). This effect of cognitive deterioration has been acknowledged by employers, many of
whom believe that all older workers would have such a decline. This negative attitude toward
productivity among older workers may prevent them from reentering the labor force or
transitioning to other jobs. More important, however, is the fact that this general belief may not
apply to everyone considering the heterogeneity in individual cognitive trajectories (Van Dalen,
Henkens, & Schippers, 2010) (von Hippel, Kalokerinos, & Henry, 2013).
Moreover, the effect of fast rates of cognitive deterioration on labor supply in the short run and
long run may differ. Short-term cognitive decline is more likely to be temporal and sensitive to
cognitive interventions relative to such a decline in the long run. Comparatively, the effect of
long-term rapid cognitive deterioration on labor supply may be larger and longer than rapidly
declining cognition in the short run (see Chapter 2 for more detail). Therefore, exploring the
effect of cognitive decline in the short run and in the long run may provide different policy
implications.
Overall, quantifying the distribution of cognitive trajectories contributes to knowledge about the
effect of cognitive decline on labor supply. Particularly, labor supply of persons with fastest rates
of cognitive decline may be mostly affected. Additionally, the implications of the effect of short-
term and long-term cognitive decline on labor supply may differ. Identifying persons with
rapidly declining cognitive abilities in the short run and in the long run provides more insights
for the impact of mid-life cognitive deterioration.
This paper fills gaps in the literature by utilizing a widely-accepted survey data that measures
cognition, the Health and Retirement Study (HRS), to quantify the distribution of cognitive
decline at middle age. This paper contributes to knowledge about the relationship between health
67

and labor supply by examining the effect of rapidly declining cognitive function on the
subjective probability of working at age 62. Specifically, this paper uses an expected probability
of working of current workers as a proxy of actual working status. This approach addresses the
reverse causality issue and isolates the causal effect of retirement on cognitive decline that has
been discussed in previous studies (Bonsang, Adam, & Perelman, 2012; Mazzonna & Peracchi,
2012; Rohwedder & Willis, 2009). In addition, this research utilizes a fixed-effect model to
eliminate the omitted bias due to individual idiosyncratic factors. This paper builds upon
Belbase, Khan, Munnell, and Webb (2015) that documents a slope of cognitive change and
Hudomiet, Hurd, Rohwedder, and Willis (2018) that explores the effect of standardized cognitive
score on changes in self-reported probability of working. Building upon previous studies, this
research identifies persons with rapid cognitive decline and quantifies the distribution of this
decline at middle age. This paper additionally includes physical health shocks and prevalent
physical health conditions as covariates from 1996 to 2014. As a result, the author eliminates the
omitted variable bias due to other health shocks. Last, this paper compares the effect of short-
term and long-term high speed of cognitive decline on changes in the expected probability of
working at age 62.
The results indicate that a sharp cognitive decline over 2 years is associated with a 2.17
percentage point decrease in the self-reported probability of working at age 62. Comparatively,
the onset of heart disease, the number one cause of death in US, over 2 years is associated with a
5.63 percent decrease in the subjective probability of working. Although the size of the negative
effect of cognitive decline is smaller than that of heart disease, more middle-aged workers
experience rapidly deteriorating cognitive function than (12 percent) a heart disease shock (1.8
percent). Additionally, a long-term sharp cognitive decline, accounting for 12 percent of the
68

middle-aged workers, lead to a 5.14 percentage point decrease in this expected probability of
working.
EMPIRICAL LITERATURE
Cognitive Decline at Middle Age and Labor Supply of Older Adults
Explanatory factors of labor supply decisions at older ages include health status, public policy
and welfare programs such as pension and Social Security benefits, liquidity constraints, family
structures, and occupations (Gruber & Madrian, 2002; Lumsdaine & Mitchell, 1999; Maestas &
Zissimopoulos, 2010; Poterba, 2014). Many previous studies explore the effect of health on the
probability of retirement. They acknowledge this causal effect. The magnitude varies from 2 to
14 percentage points based on different health measures, empirical strategies, and measures of
labor market outcomes (Lindeboom & Kerkhofs, 2009; McGarry, 2004; Mermin, Johnson, &
Murphy, 2007). Specifically, previous health measures include subjective mortality risks, self-
reported health status, chronic illness conditions, and functional limitations. However, these
measures do not incorporate cognitive health. Therefore, the estimated effect of health could
suffer from the omitted variable bias.
A small literature focuses on the relationship between cognitive decline and labor supply and
suggest that retirement can lead to cognitive decline. The mechanism is that working provides
cognitive stimulations that may improve or maintain one’s cognitive function. Working also
improves one’s sense of self-efficacy and social engagement. Retirement for at least one year
led to a 10 to 20 percent decrease in one’s cognitive score (Bonsang et al., 2012; Mazzonna &
Peracchi, 2012, 2017; Rohwedder & Willis, 2009). On the other hand, two recent studies found
that sharp cognitive decline was associated with less labor supply. Belbase et al. (2015)
identified that a long-term sharp cognitive decline was associated with retirement 2.4 years
69

earlier than the expected retirement age. However, this research examined an association instead
of an identified effect. Hudomiet et al. (2018) captured the effect of an average change in
cognition. They found that a standard deviation decline in cognitive score was associated with
0.5 percentage point decrease in the subjective probability of working at age 65. However,
Hudomiet et al. (2018) did not explore the effect of a sharp cognitive decline, which focuses on
the heterogeneity of cognitive deterioration. Further, neither of the two studies controlled other
health shocks, which could result in an omitted-variable bias. In addition, these studies utilized
data from the Health and Retirement Study, which may be subject to measurement error in
defining cognitive decline and thus underestimate its distribution and the effect.
Cognitive Decline at Middle Age
The distribution of this sharp cognitive deterioration at middle age remains unknown. There
exists mixed evidence about the starting point of cognitive decline across the lifespan, ranging
from early forties to seventies, based on different study samples, data structures, domains of
cognitive function, and levels of potential measurement error (Salthouse, 2006, 2009; Schaie,
2012).
One explanation of the discrepancy in starting points is that different aspects of cognitive
function change differently with age. There exists a large variability in the trend of cognitive
decline across different cognitive abilities. Fluid intelligence, referring to the working memory
and speed of processing information, peaks in late twenties and declines afterwards. While
crystallized intelligence, related to the accumulation of knowledge, declines beginning from
one’s sixties or even seventies (Finch, 2009; McArdle et al., 2009; McArdle et al., 2002; Willis,
2013).
70

The second reason for the mixed evidence is measurement error. Several researches suggest that
longitudinal studies that observe the same individual repeatedly over a short period tend to
underestimate the rate of decline due to the measurement error. Respondents may have higher
cognitive scores for duration cognitive tests because of familiarity to the tests rather than the
actual improvement in cognitive function (McArdle et al., 2009; Rönnlund, Nyberg, Bäckman, &
Nilsson, 2005; Salthouse, 2006, 2009; Schaie, 2012). However, Singh-Manoux et al. (2012)
validated the use of longitudinal data to examine cognitive decline and suggest that measurement
error in longitudinal data is smaller than that in cross-sectional data that observe different
individuals at a point in time. Specifically, the authors use data from Whitehall II prospective
cohort study and suggest that cognitive decline occurs among ages 45 to 49. Both men and
women aged 45 to 49 at baseline experienced a 3.6% decline in cognitive function in ten years.
Previous studies acknowledge the variability in in individual paths of cognitive decline at middle
and late age (Finch, 2009; Wilson et al., 2002). However, few studies distinctively examine the
distribution of cognitive decline at middle age.
In sum, there are extensive studies exploring the effect of physical and mental health
deterioration (e.g., depression) on labor supply of older adults. However, few studies explored
the effect of cognitive decline at middle age. Therefore, current knowledge about the relationship
between health and working might be biased. Practically, knowing this causal relationship
provides policy implications for individuals, families, and government to plan for cognitive
decline among middle-aged workers.
THEORETICAL FRAMEWORK
The Lifecycle Model of Health and Labor Supply of Older Adults
71

This study utilizes the lifecycle model assuming that one can smooth consumption and leisure
over time. The lifecycle model predicts that an individual makes decisions about working or
retirement to maximize his/her lifetime utility subject to a budget constraint determined by
income (the opportunity cost of leisure) and lifetime wealth (purchasing power that varies across
individuals) (Heckman, 1974; MaCurdy, 1981; Yaari, 1965). Further studies extend the life cycle
model by incorporating health in the model.
Health affects one’s lifetime utility and labor supply decision through several mechanisms.
Firstly, poor health (e.g., onset of a chronic disease) is associated with higher mortality risk and
leads to the uncertainty of life expectancy. As a result, poor health reduces one’s time discount
rate for consumption of other goods and services and increases his/her time discount rate for the
bequest motive. Therefore, the marginal utility of income is lower, and the incentive for
retirement is higher. Additionally, poor health indicates the deterioration of health capital, a
source of human capital (Bartel & Taubman, 1979; Grossman, 2017). It reduces one’s
productivity by reducing the total amount of time for work, leisure, and non-market productive
activities. The productivity loss also leads to the lower wage rate. This substitute effect of lower
opportunity cost of leisure leads to early retirement. However, one’s lifetime wealth is lower as
well by summing reduced wage across periods. This income effect encourages labor supply. At
last, poor health may increase the marginal utility and the cost of consumption of other goods
and services (such as leisure time and medical resources), which encourages working.
Health interventions improve one’s health status from two aspects. Firstly, interventions on
health slow or prevent the depreciation of health at certain stage in the lifetime. Secondly, they
may reduce the cost of health investment and induce more investment on health. There are
interventions at the individual, family, and policy levels. Individual-level health interventions
72

include healthy lifestyle and behaviors. The household-level intervention can be caregiving. The
policy-level intervention include policies that facilitate support towards those with poor health
conditions and policies that provide more resources for the prevention or the early screening of
certain conditions.
Since poor health may both increase and decrease the motivation for retirement, the benefit of
policy interventions on health remain unknown. Relevant empirical research is important for
understanding the direction and magnitude of the effect of poor health on labor supply of older
adults. Further extensions in the lifecycle model predict that health, income and wealth, liquidity
constraints, family structures and occupation affect one’s labor supply decision in the lifetime.
Cognitive Decline and Cognitive Reserve Hypothesis
Cognitive reserve refers to individual-specific resilience to normal and pathological cognitive
aging such Alzheimer’s disease and dementia. It is manifested as a set of cognitive processing
approaches in respond to damage in brain. Cognitive impairment occurs when cognitive reserve
is depleted (Fritsch et al., 2007). The cognitive reserve hypothesis suggests that the lifetime
exposure to education and cognitive-stimulating occupation improves one’s cognitive reserve
and better protects an individual against dementia risk or cognitive decline (Fritsch et al., 2007;
Kohn & Schooler, 1978; Schooler & Mulatu, 2001). The magnitude of this effect at one’s early
age, middle age, and late age is inconclusive (Friedland et al., 2001; Fritsch et al., 2007).
Therefore, those with lower education or enrolled occupations with less cognitive demand are
more likely to experience cognitive decline.
The lifecycle model suggests the importance of empirical evidence in understanding the causal
effect of health on labor supply of older adults. This study focuses on the effect of one important
aspect of health, cognitive health. The cognitive reserve hypothesis implicates that education and
73

occupation affect one’s trajectory of cognitive decline. Therefore, I examine the causal effect of
cognitive decline on full-time working probability at age 62 among older adults controlling for
confounding factors such as education and occupation.
EMPIRICAL STRATEGY
Data
This study utilizes data from the Health and Retirement Study (HRS), a nationally representative
longitudinal survey. HRS is conducted biennially among older adults aged 50 and above from
1992 to 2016. The HRS sample selection is based on a multi-stage area probability sample
design. The HRS data includes rich information about respondents’ demographic characteristics,
health and cognitive status, as well as economic and working characteristics. HRS has an
oversample of African Americans, Hispanics yet a nationally representative estimate can be
achieved by utilizing the HRS sampling weight. The response rate for the HRS 1992 is 80.2% at
the household level and 81.6% at the person level.
In this research I restrict the sample to full-time workers aged 50 to 61 from HRS 1996 to HRS
2014 who are working for pay and have cognitive scores for at least two consecutive waves. The
size of the study sample is 21,839.
Cognitive Measures in HRS
HRS assessed cognitive functions through an adapted version of the Telephone Interview for
Cognitive Status (TICS). TICS was modeled after the Mini-Mental State Exam (MMSE) which
has been extensively used in neuropsychological assessment of cognition (Brandt, Spencer, &
Folstein, 1988; Folstein, Folstein, & McHugh, 1975). HRS assesses self-respondents’ cognitive
function through several cognitive tests - immediate and delayed word recall (scale 0 - 10 for
each test); counting down from 100 by 7’s (scale 0 - 5); and counting back from 20 (scale 0 - 2).
74

Those with a total cognitive score (scale 0 - 27) are categorized to three groups based on their
cognitive status: dementia (score 0 - 6), cognitively impaired no dementia (CIND) (score 7 - 11),
and cognitively normal (score 12 - 27). This classification has been validated against a
neuropsychological measurement of dementia in the Aging, Demographics and Memory Study
(ADAMS), which is regarded as the gold standard in assessing dementia (Crimmins et al., 2011).
Model Specification
The Effect of Rapid Cognitive Decline in the Short Run
This research assumes that a sharp cognitive decline is an exogenous health shock after
controlling for covariates and eliminating time-invariant individual characteristics. Therefore,
this paper used the individual fixed effect model to measure the effect of cognitive decline on the
self-reported probability of working at age 62:
∆𝑤𝑜𝑟𝑘 _𝑒𝑥𝑝 62
𝑖𝑡
= 𝛽 ∆𝑐𝑜𝑔𝑑𝑒𝑐𝑙𝑖𝑛𝑒 𝑖𝑡
+ 𝜃 𝑐𝑜𝑔𝑠𝑐𝑜𝑟𝑒 𝑖𝑡 −1
+ 𝜸 𝑿 𝑖𝑡 −1
+ 𝝓 ∆𝑿 𝑖𝑡
+ 𝛼 𝑖 +𝛿 𝑡 +𝘀 𝑖𝑡
,
where ∆𝑤𝑜𝑟𝑘 _𝑒𝑥𝑝 62
𝑖𝑡
is a continuous variable ranging from -100 to 100. ∆𝑐𝑜𝑔𝑑𝑒𝑐𝑙𝑖𝑛𝑒 𝑖𝑡
equals
1 if one experiences a one or more standard deviation - decline in his/her standardized cognitive
score from time t-1 to time t and equals 0 otherwise. The variable 𝑐𝑜𝑔𝑠𝑐𝑜𝑟𝑒 𝑖𝑡 −1
is one’s
cognitive score at time t-1, ranging from 0 to 27. The vector 𝑿 𝑖𝑡 −1
refer to all time-varying
covariates at time t-1 (marital status, heart disease, hypertension, diabetes, stroke, cancer, body
mass index (BMI), household income quartiles, household wealth quartiles, occupation). The
vector ∆𝑿 𝑖 𝑡 refer to change in these covariates from time t-1 to time t. The 𝛿 𝑡 represents the year-
fixed effects. The coefficient of cognitive decline, 𝛽 , demonstrates the effect of cognitive decline
over two years on the expected probability of full-time working at age 62.
The Effect of Rapid Cognitive Decline in the Long Run
75

I additionally examined the effect of individual-specific trend of cognitive decline on changes in
the expected probability of working. Then I examined this effect over a longer time window to
reduce the measurement error:
𝐶𝑜𝑔𝑠𝑐𝑜𝑟 𝑒 𝑖𝑡
= 𝛽 1𝑖 𝑤𝑎𝑣𝑒 𝑡 + µ
𝑖 ∆𝑤𝑜𝑟𝑘 _𝑒𝑥𝑝 62
𝑖 = 𝜃 (∆𝑐𝑜𝑔𝑑𝑒𝑐𝑙𝑖𝑛𝑒 𝑖 ) + 𝛼 𝑐𝑜 𝑔 𝑠𝑐𝑜𝑟𝑒 𝑖𝑡 0
+ 𝜸 (∆𝑿 𝑖 ) + 𝝓 𝑿 𝒊𝒕𝟎 + 𝘀 𝑖
This model is different from the fixed-effect model on three aspects. Firstly, the time interval in
this model begins from the first observation at time t0 and ends at the last observation at time t1
instead of 2 years in the fixed-effect model. The average and median duration between the start
and end of the observation is 5.0 and 4 years. This duration ranges from 2 to 10 years (Appendix
4). Secondly, the independent variable, ∆𝑐𝑜𝑔𝑑𝑒𝑐𝑙𝑖𝑛𝑒 𝑖 is defined based on an average annual rate
of cognitive decline (1 or more point decrease in cognitive score per year) over the study period
rather than on the change in the standardized cognitive score in 2 years. Last, this analysis is
based on a person-level OLS regression. The sample consists of 11,501 persons, rather than
22,001 person-years.
Dependent Variable
I used the change in the self-reported probability of working at age 62 over the time window as
the dependent variable (from time t-1 to time t for the short-term analysis, and from the first to
the last wave of observation for the long-term analysis). I utilized this variable instead of the
actual working behavior to address the reverse causal effect of retirement on cognitive
deterioration. HRS asks each respondent at various ages about the chances that he/she is still
working at age 62. It is a continuous variable ranging from 0 to 100. I generated the outcome
variables, ∆𝑤𝑜𝑟𝑘 _𝑒𝑥𝑝 62
𝑖𝑡
(2 years) 𝑎𝑛𝑑 ∆𝑤 𝑜 𝑟𝑘 _𝑒𝑥𝑝 62
𝑖 (2 − 10 years) , based on this
variable.
76

Independent Variable and Covariates
I defined a short-term sharp cognitive decline based on the change in each respondent’s cognitive
score in two years. I firstly generated a standardized cognitive score for individuals in each age
group (50 to 54, 55 to 59, 60 to 61) and wave. Then I defined the independent
variable, ∆𝑐𝑜𝑔𝑑𝑒𝑐𝑙𝑖𝑛𝑒 𝑖𝑡
, as 1 if the ith individual’s standardized cognitive score at time is 1 or
more standard deviation lower than that at time t-1. The independent variable equals 0 otherwise.
To reduce the measurement error, this study dropped 1 percent of observations with an extreme
sharp decline/increase in cognitive scores over two years.
I defined a long-term sharp cognitive decline as the dependent variable, ∆𝑐𝑜𝑔𝑑𝑒𝑐𝑙𝑖𝑛𝑒 𝑖 , referring
to the individual-specific trend of decline for individual i over the study period. This slope refers
to the average rate of biennial change in cognitive score across waves. I conducted an OLS
model that regressed individual i’s cognitive score on the tth wave. Its coefficient, 𝛽 1𝑖 , indicates
the average rate of change in individual i’s cognitive score over the study period. It is constant
for each person and varies cross-sectionally. The independent variable, ∆𝑐 𝑜 𝑔𝑑𝑒𝑐𝑙𝑖𝑛𝑒 𝑖 , is a
dummy variable that equals 1 if 𝛽 1𝑖 ̂
is smaller than -2, and 0 otherwise. In other words, if an
individual on average experiences a 1 point decline in cognitive score (scale 0-27) every year,
this individual is defined as having cognitive decline over the study period.
I chose other covariates based on the life cycle model and cognitive reserve hypothesis, which
demonstrated factors associated with cognitive decline and labor supply. I choose time-varying
factors in the fixed-effect model, including age, marital status, prevalent heart disease, diabetes,
hypertension, stroke and body mass index (BMI), income quartiles, wealth quartiles, and
occupation at time t-1. I additionally controlled for time fixed effects, incident heart disease,
diabetes, hypertension, stroke, and changes in BMI, income quartiles, wealth quartiles, and
77

occupation over the short run and the long run. I additionally control for time-invariant
covariates in the OLS model that examines the effect of long-term cognitive decline, such as
race, gender, and education.
RESULTS
Descriptive Statistics
Figure 8 and 9 illustrate the distribution of the expected probability of full-time working at age
62 and its change over 2 years. Figure 8 shows the distribution of this variable and exhibits that
people are more likely to choose certain focal points as the answer, such as 0, 50, or 100. Its
mean value is 54.7, and its standard deviation is 37.5 (Appendix Table 13). Figure 9 suggests
that the variable is continuous, similar to the normal distribution. The mean value of this
dependent variable is 1.9, and its standard deviation is 34.2.
Figure 8. The distribution of the expected probability of full-time working at age 62 among
workers aged 50 to 61, 1996-2014.

Note. Results weighted by HRS sampling weight. Sample restricted to individuals aged 50 to 61, working
for pay, HRS 1996-2014.

78

Figure 9. The distribution of the change in the expected probability of full-time working at age
62 among workers aged 50 to 61 from wave t-1 to wave t, 1996-2014.

Note. Results weighted by HRS sampling weight. Sample restricted to individuals aged 50 to 61, working
for pay, HRS 1996-2014.
Figure 10 and 11 exhibit the distribution of cognitive function and change in cognitive scores for
workers aged 50 to 61. Figure 10 illustrates that few middle-aged workers have dementia or
cognitive impairment no dementia (CIND). Specifically, 0.5% of the study sample has dementia
(scoring 0-6), and 5.7% of them has CIND (scoring 7-11). However, Figure 11 suggests the
heterogeneity of the distribution of change in cognitive scores. The mean change in cognitive
scores over 2 years is -0.1, indicating no change at the population level (Appendix Table 12).
However, 11.9% of the study population experience rapid cognitive deterioration over 2 years. It
is also worth noting that a similar proportion of the study population experiences an
improvement in their cognitive scores. Individuals with no cognitive decline are more likely to
79

be younger, white, male, more educated, healthier and have higher socioeconomic status (SES)
than the those with a decline (Table 5).
Figure 10. The distribution of cognitive function for workers aged 50 to 61 at time t-1, 1996-
2014.

Note. Results weighted by HRS sampling weight. Sample restricted to individuals aged 50 to 61, working
for pay, HRS 1996-2014.

80

Figure 11. The distribution of the change in cognitive scores for workers aged 50 to 61 from time
t-1 to time t, 1996-2014.

Note. Results weighted by HRS sampling weight. Sample restricted to individuals aged 50 to 61, working
for pay, HRS 1996-2014.

Table 5. Sample characteristics, 1996-2014.
Percent/Mean (Frequency)

No Cognitive
Decline
Cognitive
Decline
Total
Race***
non-Hispanic white 86.4% 82.6% 85.9%
non-Hispanic black 8.2% 9.8% 8.4%
Hispanic 2.6% 3.4% 2.7%
other race 2.9% 4.2% 3.0%
Education*
less than high school 10.3% 11.2% 10.5%
81

Highschool 25.8% 25.8% 25.8%
college and above 63.9% 63.0% 63.8%
Age
50 to 54 18.2% 19.8% 18.4%
55 to 59 62.1% 61.5% 62.0%
60 to 62 19.8% 18.7% 19.6%
Income Quartiles***
income 1st quartile 17.4% 20.2% 17.7%
income 2nd quartile 23.1% 23.5% 23.2%
income 3rd quartile 27.7% 27.9% 27.7%
income 4th quartile 31.8% 28.4% 31.4%
Wealth Quartiles*
wealth 1st quartile 21.3% 22.9% 21.5%
wealth 2nd quartile 24.3% 25.0% 24.4%
wealth 3rd quartile 26.3% 26.0% 26.3%
wealth 4th quartile 28.1% 26.1% 27.8%
Occupation***
Managerial 17.6% 15.0% 17.3%
Professional 25.3% 24.1% 25.2%
Sales & admin 25.6% 27.0% 25.8%
Protection/Military 1.4% 1.4% 1.4%
Cleaning/building 8.5% 9.2% 8.6%
Production/Operation 21.2% 22.6% 21.4%
Missing 0.4% 0.7% 0.4%
Male 48.6% 47.6% 48.4%
Marital status*** 74.9% 72.2% 74.6%
Heart disease 9.8% 10.3% 9.9%
Diabetes** 11.6% 13.5% 11.8%
BMI 28.4 28.4 28.4
82

Hypertension 38.3% 38.8% 38.4%
Incident heart disease 1.8% 1.9% 1.8%
Incident diabetes 2.1% 2.6% 2.1%
Change in BMI 0.194 0.156 0.19
Incident
hypertension
4.6% 4.0% 4.5%
Percent 88.1% 11.9% 100.0%
Sample Size 19,232 2,607 21,839
Note. Sample restricted to full-time workers aged 50 to 61, 1996-2014.Results weighted by the HRS
combined sampling weight (personal sampling weight and the nursing home sampling weight).

Figure 12 and 13 demonstrate the distribution of cognitive score at the beginning of the
observation window and the individual-specific trend of cognitive change. Figure 12 shows that
less than 1 percent of middle-aged workers are with dementia, and that 8 percent of them have
CIND. Figure 13 indicates that 12 percent of these persons have a sharp cognitive decline over
the long run. It should be noted that since the samples for the short-term and long-term analysis
are different, persons with a long-term cognitive decline could also experience a short-term sharp
cognitive decline at time t. They could also have no sharp cognitive decline over two years but
have a sharp decline over the whole observation window.

83

Figure 12. The distribution of cognitive score at the beginning of the observation window (t0),
1996-2014.

Note. Results unweighted. Sample restricted to full-time workers aged 50 to 61, 1996-2014.

84

Figure 13. The distribution of person-specific slopes of change in cognitive score across waves,
1996-2014.

Note. Results unweighted. Sample restricted to full-time workers aged 50 to 61, 1996-2014.

The Effect of Short-Term Cognitive Decline
Table 6 exhibits the regression output of the expected probability of working on short-term rapid
cognitive decline. Model 1 suggests that without controlling for any other covariates, this sharp
cognitive deterioration is associated with a 1.95 percentage point decrease in the self-reported
probability of working. In the FE model, a sharp decline in cognition leads to 2.17 percentage
point decrease in the working expectation, holding other covariates constant. Moreover, the onset
of heart disease is associated with a 5.63 percentage point decrease in the expected probability of
working.
85

The results indicate that a shock in one’s cognitive function is consistently associated with a
decrease in the probability of working across different model specifications. The size of this
effect in the correlation model is smaller than that in the FE model. This is consistent with the
hypothesis that the negative effect of cognitive decline can be overestimated without controlling
for covariates that are positively related to cognitive decline and negatively associated with the
working expectation (e.g., physical health shock, low SES).
Table 6. Correlation and FE results of the effect of short-term cognitive decline on the expected
probability of working at age 62, 1996-2014.

(1) (2)

Correlation FE
Cognitive Decline -1.95*** -2.17*
Married

2.5
Cognitive score, t-1

-0.03
Heart disease, t-1

-4.6
Diabetes, t-1

-3.24
Hypertension, t-1

0.46
Stroke, t-1

-1.46
BMI, t-1

0.15
Cancer, t-1

-3.9
Onset of heart disease

-5.63*
Onset of diabetes

1.28
Onset of hypertension

-0.67
Onset of stroke

-16.5**
BMI change

0.07
Onset of cancer

-6.3*
Constant -1.95*** -138.2
Observations 22,001 22,001
86

Adjusted R Square 0 0.011
Source. HRS 1996-2014. Note. Sample restricted to working older adults aged 50 to 61. Standard errors
clustered by individuals: *P<0.1, **p<0.05, ***p<0.01. Missing values included with a flag; standard
errors clustered at the individual level. Other covariates include age, education, income, wealth,
occupation, and change in income, wealth, and occupation.

The Effect of Long-Term Cognitive Decline
Table 7 suggests that for the 12 percent of the middle-aged workers who experience an annual
rate of 1 or more points of decline in their cognitive scores during the study period (Appendix
Table 16), their expected probability of working at age 62 is reduced by 5.14 percentage points.
This effect is statistically significant, and its magnitude is larger than that of a short-term shock.
However, the onset of heart disease is insignificantly associated with a 1.45 percentage point
decrease in the self-reported probability of working at age 62. One potential explanation is
survival selection – individuals who survive heart disease in the long run are less likely to
change their early retirement planning. Additionally, the magnitude of this is larger than the
short-term effect, potentially due to the fact that this measure captures more steady and severe
cognitive deterioration, which has a larger impact on people’s early retirement.

Table 7. Correlation and OLS results of the effect of individual-specific trend of cognitive
decline on changes in the expected probability of working at age 62, 1996-2014

(1) (2)
Correlation OLS
Cognitive Decline -5.67*** -5.14*
Race (ref. NH white)

NH black

1.72**
Hispanic

1.27
Other race

1.48
87

Male

-2.08**
Married

-0.25
Cognitive score, t-1

0.16
Heart disease, t-1

0.68
Diabetes, t-1

-0.93
Hypertension, t-1

-0.94
Stroke, t-1 0.55
Cancer, t-1 -3.78**
BMI, t-1

0.004
Onset of hypertension

2.89**
Onset of diabetes

0.39
Onset of heart disease

-1.45
Onset of stroke 0.37
Onset of cancer -1.56
Increase in BMI

0.18
Constant 4.44 -28.1***
Observations 11,511 11,511
Adjusted R Square 0.002 0.014
Source. HRS 1996-2014. Note. Sample restricted to working older adults aged 50 to 62. Standard errors
clustered by individuals: *P<0.1, **p<0.05, ***p<0.01. Missing values of the variable occupation are
included with a flag.

DISCUSSION
This paper quantifies the distribution of cognitive change with different durations for workers
aged 50 to 61. Most middle-aged workers had normal cognitive function - less than 1 percent of
them had dementia, and 6 percent of them had CIND. However, 11.9% of workers experienced
88

a rapid decline in their cognitive function over 2 years in mid-life. Similarly, 12% of them
experienced a long-term sharp cognitive deterioration over 2 to 14 years.
This paper advances knowledge about the relationship between health and labor supply. Fast
declining cognitive function over two years, as a type of health shock, is statistically significantly
associated with a 2.17 percentage point decrease in the probability of working at the early
retirement age, when they are at their peak earning and saving years. Additionally, middle-aged
workers with sharp cognitive decline are more likely to be ethnic minorities, male, with lower
SES, have less educational and occupational attainment, and have cardiovascular disease (see
Chapter 2 for more detail). They have less financial resources and health care access thus could
be less resilient to declining cognitive abilities. Therefore, these middle-aged workers may retire
early, save less, and collect less Social Security monthly benefits because they retire and claim
these benefits with a penalty before the normal retirement age (65 to 67). Schofield et al. (2011)
found that early retirement due to back problems is associated with 87% reduction in wealth
compared to those working for pay with no health issues. This decline may additionally impose
burdens on the Social Security Trust Fund as less workers stay in the labor force.
Comparatively, this effect of short-term decline has a smaller magnitude compared to that of the
onset of heart disease (a 5.63 percentage point decrease). However, the proportion of middle-
aged workers experiencing such a decline (11.9 percent) is significantly higher than those with
an incident heart disease (1.8 percent). It should also be noted that individuals with deteriorating
cognitive function may have longer life expectancy than those with heart disease. Therefore, the
socioeconomic burden of cognitive deterioration through its impact on early retirement can be
larger than that of heart disease.
89

There is scientific progress in understanding factors associated with dementia and cognitive
decline (Alzheimer's Association, 2019; Blazer, Yaffe, & Liverman, 2015; Livingston et al.,
2017). Studies provide encouraging although not definite evidence that physical exercise, blood
pressure management, and cognitive training may be associated with slower cognitive decline
(Blazer et al., 2015; Etgen, Sander, Bickel, & Förstl, 2011; National Academies of Sciences &
Medicine, 2017). Knowledge about the effect of cognitive decline on labor supply and potential
interventions that prevent or delay this decline help individuals and families adopt healthy life
behaviors and better plan for the cost of fast rates of cognitive deterioration.
At the meantime, such a deterioration over the long run is statistically significantly associated
with a 5.14 percentage point decrease in the subjective probability of working at age 62. The
magnitude of the effect of long-term cognitive deterioration is larger than that of a short-term
one (a 2.17 percentage point decrease). These effects of cognitive decline in the short run and in
the long run may have different implications. Fast cognitive deterioration for some people may
be temporary, and interventions on such a decline may be more effective. Comparatively, those
with long-term deterioration in cognitive abilities may be affected by it for a longer time, and
may need more support from health care institutions, families, community, and from the
government.
This study has some limitations. Firstly, there is an improvement in cognitive scores in addition
to a decline for middle-aged workers. This research does not distinguish between an actual
cognitive improvement and the measurement error. Several factors could lead to cognitive
improvement. Previous studies suggest that cognitive stimulations at work or in leisure activities
improves one’s cognitive function (Andel, Finkel, & Pedersen, 2016; Bonsang et al., 2012;
Mazzonna & Peracchi, 2017; Rohwedder & Willis, 2009). Medicine prescribed for other health
90

conditions such as cardiovascular disease or amnesia may also enhance one’s cognitive function
(Duron et al., 2009). However, there is insufficient evidence for a definite conclusion.
Understanding the distribution of cognitive improvement in the short term and long term in the
future will contribute to a more complete picture of the distribution of trajectories of cognitive
function at middle age.
Secondly, this research does not compare similarities and differences in the short-term and long-
term cognitive deterioration. Further studies are needed to focus on the comparison between a
short-term and long-term cognitive trajectory and provide advice for effective policy
interventions. These studies would additionally provide insights about effective intervention on
delaying short-term and long-term cognitive deterioration.
Thirdly, I do not control for mental health conditions (such as depression) in the main analysis.
The reason is that depression is significantly correlated with cognitive decline (Katon et al.,
2015). Empirical analysis, therefore, cannot distinguish the effect of cognitive decline and
mental health conditions on labor supply. However, I run a subgroup analysis restricting the
study sample to those with no depression. The effect of cognitive decline on labor supply is
statistically significant (Appendix Table 18). The magnitude of the effect for the mentally
healthy subsample is slightly smaller than that in the main analysis, yet not altering the
conclusion.
Last, there are no effective interventions on delaying or preventing cognitive decline at middle
age at the policy level. One main reason is that the evidence for an effective intervention is
insufficient for creating a policy intervention. Previous studies provide evidence with low or
moderate strength that include cognitive training programs, blood pressure management, and
physical exercise may prevent or delay rapidly declining cognitive abilities. More research is
91

needed to fully understand cognitive decline for the middle-aged population, its socioeconomic
burden, and effective interventions that prevent or slow the high speed of cognitive deterioration.
Overall, this study fills current literature gaps by quantifying the distribution of cognitive decline
for middle-aged workers and exploring its effect on labor supply. This paper contributes to
knowledge about the socioeconomic burden of declining cognitive function through its impact on
early retirement. One with such a decline is more likely to be ethnic minorities and have low
SES, and he/she may be less resilient to its consequences, including less lifetime wealth, Social
Security benefits, and retirement income security. At-risk individuals and families, with more
knowledge about the distribution of cognitive change and its cost, can prevent and delay such
deterioration by adopting healthy behaviors. Additionally, they need more support from families,
communities, health care professionals, and the government. Lastly, the policy implications of
the short-term and long-term cognitive decline may differ. Interventions that prevent or delay
cognitive deterioration in the short-term and financial support for those with long-term declining
cognitive abilities may be more effective. More studies about the socioeconomic burden of
cognitive deterioration and its interventions and welfare policies with different durations are
needed to address this burden.

92

REFERENCE
Andel, R., Finkel, D., & Pedersen, N. L. (2016). Effects of Preretirement Work Complexity and
Postretirement Leisure Activity on Cognitive Aging. The Journals of Gerontology Series
B: Psychological Sciences and Social Sciences, 71(5), 849-856.
doi:10.1093/geronb/gbv026
Association, A. s. (2019). 2019 Alzheimer's disease facts and figures. Alzheimer's & dementia:
the journal of the Alzheimer's Association, 15(3), 321-387.
doi:10.1016/j.jalz.2019.01.010
Bartel, A., & Taubman, P. (1979). Health and Labor Market Success: The Role of Various
Diseases. The Review of Economics and Statistics, 61(1), 1-8. doi:10.2307/1924825
Belbase, A., Khan, M., Munnell, A., & Webb, A. (2015). Slowed or Sidelined? The Effect
of'Normal'Cognitive Decline on Job Performance Among the Elderly. Center for
Retirement Research at Boston College Working Paper(2015-12).
Blazer, D. G., Yaffe, K., & Liverman, C. T. (2015). Cognitive aging: Progress in understanding
and opportunities for action: National Academies Press Washington, DC.
Bonsang, E., Adam, S., & Perelman, S. (2012). Does retirement affect cognitive functioning?
Journal of Health Economics, 31(3), 490-501. doi:10.1016/j.jhealeco.2012.03.005
Coile, C. (2004). Health Shocks and Couples' Labor Supply Decisions. In: National Bureau of
Economic Research.
Currie, J., & Madrian, B. C. (1999). Health, health insurance and the labor market. Handbook of
labor economics, 3, 3309-3416.
Duron, E., Rigaud, A.-S., Dubail, D., Mehrabian, S., Latour, F., Seux, M.-L., & Hanon, O.
(2009). Effects of Antihypertensive Therapy on Cognitive Decline in Alzheimer's
Disease. American Journal of Hypertension, 22(9), 1020-1024. doi:10.1038/ajh.2009.119
Etgen, T., Sander, D., Bickel, H., & Förstl, H. (2011). Mild cognitive impairment and dementia:
the importance of modifiable risk factors. Deutsches Arzteblatt international, 108(44),
743-750. doi:10.3238/arztebl.2011.0743
Finch, C. E. (2009). The neurobiology of middle-age has arrived. Neurobiology of Aging, 30(4),
515-520. doi:10.1016/j.neurobiolaging.2008.11.011
French, E. (2005). The Effects of Health, Wealth, and Wages on Labour Supply and Retirement
Behaviour. The Review of Economic Studies, 72(2), 395-427. doi:10.1111/j.1467-
937X.2005.00337.x
93

Friedland, R. P., Fritsch, T., Smyth, K. A., Koss, E., Lerner, A., J., Chen, C. H., . . . Debanne, S.,
M. (2001). Patients with Alzheimer's Disease Have Reduced Activities in Midlife
Compared with Healthy Control-Group Members. Proceedings of the National Academy
of Sciences of the United States of America, 98(6), 3440-3445.
doi:10.1073/pnas.061002998
Fritsch, T., McClendon, M. J., Smyth, K. A., Lerner, A. J., Friedland, R. P., & Larsen, J. D.
(2007). Cognitive functioning in healthy aging: the role of reserve and lifestyle factors
early in life. The Gerontologist, 47(3), 307-322. doi:10.1093/geront/47.3.307
Grossman, M. (2017). On the Concept of Health Capital and the Demand for Health. In (pp. 6):
Columbia University Press.
Gruber, J., & Madrian, B. C. (2002). Health insurance, labor supply, and job mobility: A critical
review of the literature (0898-2937). Retrieved from
Harada, C. N., Natelson Love, M. C., & Triebel, K. L. (2013). Normal Cognitive Aging. Clinics
in Geriatric Medicine, 29(4), 737-752. doi:10.1016/j.cger.2013.07.002
Heckman, J. (1974). Life Cycle Consumption and Labor Supply: An Explanation of the
Relationship between Income and Consumption Over the Life Cycle. The American
Economic Review, 64(1), 188-194.
Hudomiet, P., Hurd, M. D., Rohwedder, S., & Willis, R. J. (2018). The effect of physical and
cognitive decline at older ages on job mismatch and retirement (0898-2937). Retrieved
from
Katon, W., Pedersen, H. S., Ribe, A. R., Fenger-Grøn, M., Davydow, D., Waldorff, F. B., &
Vestergaard, M. (2015). Effect of Depression and Diabetes Mellitus on the Risk for
Dementia: A National Population-Based Cohort Study. JAMA psychiatry (Chicago, Ill.),
72(6), 612-619. doi:10.1001/jamapsychiatry.2015.0082
Kohn, M. L., & Schooler, C. (1978). The Reciprocal Effects of the Substantive Complexity of
Work and Intellectual Flexibility: A Longitudinal Assessment. The American journal of
sociology, 84(1), 24.
Lindeboom, M., & Kerkhofs, M. (2009). Health and Work of the Elderly: Subjective Health
Measures, Reporting Errors and Endogeneity in the Relationship between Health and
Work. Journal of Applied Econometrics, 24(6), 1024-1046. doi:10.1002/jae.1077
Livingston, G., Sommerlad, A., Orgeta, V., Costafreda, S. G., Huntley, J., Ames, D., . . . Samus,
Q. (2017). Dementia prevention, intervention, and care. The Lancet, 390(10113), 2673-
2734. doi:10.1016/S0140-6736(17)31363-6
Lumsdaine, R. L., & Mitchell, O. S. (1999). Chapter 49 New developments in the economic
analysis of retirement. In (Vol. 3, pp. 3261-3307): Elsevier B.V.
94

MaCurdy, T. E. (1981). An Empirical Model of Labor Supply in a Life-Cycle Setting. Journal of
Political Economy, 89(6), 1059-1085. doi:10.1086/261023
Maestas, N., & Zissimopoulos, J. (2010). How Longer Work Lives Ease the Crunch of
Population Aging. The Journal of Economic Perspectives, 24(1), 139-160.
doi:10.1257/jep.24.1.139
Mazzonna, F., & Peracchi, F. (2012). Ageing, cognitive abilities and retirement. European
Economic Review, 56(4), 691-710. doi:10.1016/j.euroecorev.2012.03.004
Mazzonna, F., & Peracchi, F. (2017). Unhealthy Retirement? Journal of Human Resources,
52(1), 128-151. doi:10.3368/jhr.52.1.0914-6627R1
McArdle, J., Smith, J., & Willis, R. (2009). Cognition and economic outcomes in the Health and
Retirement Survey (No. w15266). In: National Bureau of economic research.
McArdle, J. J., Ferrer-Caja, E., Hamagami, F., & Woodcock, R. W. (2002). Comparative
longitudinal structural analyses of the growth and decline of multiple intellectual abilities
over the life span.(Abstract). Developmental Psychology, 38(1), 115. doi:10.1037/0012-
1649.38.1.115
McGarry, K. (2004). Health and Retirement: Do Changes in Health Affect Retirement
Expectations? The Journal of Human Resources, 39(3), 624-648. doi:10.2307/3558990
Mermin, G. B. T., Johnson, R. W., & Murphy, D. P. (2007). Why Do Boomers Plan to Work
Longer? The journals of gerontology. Series B, Psychological sciences and social
sciences, 62(5), S286-S294. doi:10.1093/geronb/62.5.S286
National Academies of Sciences, E., & Medicine. (2017). Preventing cognitive decline and
dementia: A way forward: National Academies Press.
Poterba, J. M. (2014). Retirement Security in an Aging Population. The American Economic
Review, 104(5), 1-30. doi:10.1257/aer.104.5.1
Rohwedder, S., & Willis, R. (2009). Mental Retirement. IDEAS Working Paper Series from
RePEc.
Rönnlund, M., Nyberg, L., Bäckman, L., & Nilsson, L.-G. (2005). Stability, Growth, and Decline
in Adult Life Span Development of Declarative Memory: Cross-Sectional and
Longitudinal Data From a Population-Based Study. Psychology and aging, 20(1), 3-18.
doi:10.1037/0882-7974.20.1.3
Salthouse, T. A. (2006). Robust Cognitive Change. Journal of the International
Neuropsychological Society, 18(4), 1-8. doi:10.1017/S1355617712000380
Salthouse, T. A. (2009). When does age-related cognitive decline begin? Neurobiology of Aging,
30(4), 507-514. doi:10.1016/j.neurobiolaging.2008.09.023
95

Schaie, K. W. (2012). Studies of Cohort and Period Differences. In. New York: Oxford
University Press.
Schofield, D. J., Shrestha, R. N., Percival, R., Callander, E. J., Kelly, S. J., & Passey, M. E.
(2011). Early retirement and the financial assets of individuals with back problems.
European Spine Journal, 20(5), 731-736. doi:10.1007/s00586-010-1647-8
Schooler, C., & Mulatu, M. S. (2001). The Reciprocal Effects of Leisure Time Activities and
Intellectual Functioning in Older People: A Longitudinal Analysis. Psychology and
aging, 16(3), 466-482. doi:10.1037/0882-7974.16.3.466
Singh-Manoux, A., Kivimaki, M., Glymour, M. M., Elbaz, A., Berr, C., Ebmeier, K. P., . . .
Dugravot, A. (2012). Timing of onset of cognitive decline: results from Whitehall II
prospective cohort study. BMJ, 344, d7622.
Van Dalen, H. P., Henkens, K., & Schippers, J. (2010). Productivity of Older Workers:
Perceptions of Employers and Employees. Population and Development Review, 36(2),
309-330. doi:10.1111/j.1728-4457.2010.00331.x
von Hippel, C., Kalokerinos, E. K., & Henry, J. D. (2013). Stereotype threat among older
employees: relationship with job attitudes and turnover intentions. Psychology and aging,
28(1), 17-27. doi:10.1037/a0029825
Willis, R. J. (2013). The cognitive demands of work and the length of working life: the case of
computerization. Paper presented at the Working Longer and Retirement Conference
Stanford Institute for Economic Policy Research October.
Wilson, R. S., Beckett, L. A., Barnes, L. L., Schneider, J. A., Bach, J., Evans, D. A., & Bennett,
D. A. (2002). Individual differences in rates of change in cognitive abilities of older
persons. Psychology and aging, 17(2), 179-193. doi:10.1037/0882-7974.17.2.179
Yaari, M. E. (1965). Uncertain Lifetime, Life Insurance, and the Theory of the Consumer. The
Review of Economic Studies, 32(2), 137-150. doi:10.2307/2296058

96

Chapter 4:
Conclusion
In sum, this dissertation explores the heterogeneity in dementia prevalence and in the distribution
of mid-life cognitive trajectories.
Estimates of dementia prevalence differ across measurement/data sources and vary by age, sex,
and race. The dissertation fills current literature gaps in inconsistent estimates of dementia
prevalence at the population level and across subgroups. This chapter shows that gaps in
dementia prevalence based on cognitive tests and diagnostic codes for the US population and by
age and sex vanish over time, providing key information for dementia prevalence in US and
among different age and sex groups. More importantly, chapter 1 finds substantial gaps (African
Americans, 10.9 percentage points difference; Hispanics, 9.8 percentage points difference) in
dementia prevalence based on cognitive tests and diagnosis codes among ethnic minorities.
These gaps may reflect the potential overdiagnosis of dementia based on cognitive tests and
underdiagnosis of dementia based on diagnosis codes for blacks and Hispanics. These substantial
gaps further highlight the importance of further data collection efforts to measure dementia
among ethnic minorities, such as the HRS Harmonized Cognitive Assessment Protocal (HCAP).
Since there exists no treatment that could cure dementia, preventing or delaying the onset of
dementia can be an effective approach to address its large socioeconomic burden. Specifically,
recent literature suggests that pathological changes in the brain can occur 20 years earlier than
the manifestation of dementia symptoms and mid-life cognitive decline can be an early stage of
dementia later in life (Amieva et al., 2014;Alzheimer's Association, 2019; Beason-Held et al.,
2013; Harada, Natelson Love, & Triebel, 2013). Interventions on this mid-life declining
cognitive function, thus, may effectively prevent or delay the onset of dementia later in life. As
97

Zissimopoulos and her colleagues (2014) suggest, a 5-year delay in the onset of dementia is
associated with 41% lower dementia prevalence and 40% reduction in its cost. Additionally,
even if this decline does not develop into dementia, it could impair daily functions including
productivity. As mentioned by chapter 3, rapidly declining cognitive function is associated with
a 2 to 5 percentage point decline in the probability of working at the early retirement age over 2
years and over 2 to 10 years.
The second and third chapter fill current literature gaps in insufficient knowledge about the
diversity of cognitive decline for middle-aged adults. These two chapters quantify the
distribution of cognitive trajectories at middle age, explore risk factors for experiencing rapid
cognitive deterioration, and identify the effect of such a decline on labor supply of older adults.
The second chapter finds that 8.4 percent of individuals experienced fast speed of cognitive
deterioration over 2 years, and 8.7 percent of them rapidly decreased their cognitive abilities over
2 to 14 years. Being ethnic minorities, less educational and occupational attainment, and
cardiovascular risk factors are associated with higher risk of rapidly declining cognitive abilities.
Particularly, the odds of experiencing fastest rates of cognitive decline for African Americans
relative to whites is respectively 1.89 over 2 years and 2.3 over 2 to 14 years. Such a decline
leads to a 2.2 and 5.1 percentage point decrease in the expected probability of full-time working
at age 62 respectively in the short run and in the long run.
This dissertation suggests that there is disparity in the distribution of cognitive trajectories.
Ethnic minorities, those with poorer health conditions and lower SES are more likely to
experience fast rates of cognitive deterioration. With less access to health care services and less
social support, they and their families may also be less resilient to the negative effect of
cognitive decline on the quality of life. As a result, they may need more support from the
98

community and from the social welfare program compared to those without sharp cognitive
decline. Additionally, (Levine, Harrati, & Crimmins, 2018) and (Weuve et al., 2018) found
evidence of racial disparity in levels but not rates of cognitive decline among those aged 65 and
above. One potential explanation of the inconsistency in these studies is that racial disparity in
levels of cognitive function at age 65 and older is at least partially due to the disparity in
cognitive trajectories from age 50 to 64. Therefore, these findings suggest that interventions on
cognitive decline for middle-aged adults may be particularly effective in reducing racial
disparity.
Such a decline, in addition, is a result of the interaction between neuropathological changes such
as the accumulation of beta-amyloid and/or brain atrophy and cognitive reserve – neuron
networks in the brain that maintain or even increase one’s cognitive function (Livingston et al.,
2017; Stern, 2012). Therefore, cognitive stimulation at work may increase one’s cognitive
function by building more cognitive reserve. Specifically, workers are motivated to invest in
their cognitive function to maintain productivity through cognitive stimulation and social
engagement at work (Mazzonna & Peracchi, 2012; Rohwedder & Willis, 2009). Considering this
mutual relationship between cognitive decline and labor supply, the improvement working
characteristics such as working environment and cognitively-demanding tasks may increase
one’s motivation to invest in cognitive abilities, delay or prevent mid-life cognitive decline, and
prolong one’s working life through maintaining his/her productivity.
Together, this dissertation sheds light on the diversity in dementia prevalence and the distribution
of mid-life cognitive trajectories for individuals, families, and the government to better plan for
pathological or functional cognitive decline. The first chapter provides insights for the National
Plan about levels and trends in dementia prevalence, thus facilitate the achievement of Goal 5 –
99

improve data that measures dementia and track the progress of the National Plan. Accurate
estimates of levels and trends in dementia prevalence among subgroups also help individuals,
families, and healthcare professionals to better plan for dementia. Additionally, the second and
third chapter provide implications for individuals and policy makers. Chapter 2 explores factors
associated with fast rates of cognitive deterioration contributes to knowledge about preventing or
delaying the onset of sharp cognitive decline, which helps with the achievement of Goal 1 –
effectively prevent or treat dementia by 2025. Chapter 2 and 3 suggest that individuals and
families with lower SES or worse health conditions need to be cautious about and better plan for
rapidly declining cognitive abilities at middle age. One example is that individuals can prevent
such a decline by better managing blood pressure, taking physical exercise, and conducting
cognitive stimulating activities. Furthermore, this dissertation distinguishes cognitive decline
over different time windows, and suggests that those with short-term and long-term cognitive
decline may have different needs. Therefore, policies that address the socioeconomic burden of
cognitive decline should be tailored respectively for those with short-term and long-term
cognitive changes.
Overall, this dissertation suggests that there is disparity in dementia and in mid-life cognitive
decline. Ethnic minorities are more likely to develop dementia yet less likely to be diagnosed
with dementia in Medicare claims. For ethnic minorities and those with lower SES and worse
health conditions, they are more likely to experience high speed of cognitive decline yet less
resilient to address its burden. Both interventions on delaying or preventing cognitive
deterioration and financial and community support are needed for them. Additionally, further
studies with improved dementia ascertainment among ethnic minorities will contribute to
knowledge about the diversity in dementia prevalence. Moreover, further studies about
100

interventions on cognitive decline at middle age and its socioeconomic burden among different
subgroups and with different durations will be crucial to improve current knowledge about the
heterogeneity in cognitive trajectories.

101

Bibliography
Abell, J. G., Kivimäki, M., Dugravot, A., Tabak, A. G., Fayosse, A., Shipley, M., . . . Singh-
Manoux, A. (2018). Association between systolic blood pressure and dementia in the
Whitehall II cohort study: role of age, duration, and threshold used to define
hypertension. European Heart Journal, 39(33), 3119-3125. doi:10.1093/eurheartj/ehy288
Alzheimer's Association (2019). 2019 Alzheimer's disease facts and figures. Alzheimer's &
dementia: the journal of the Alzheimer's Association, 15(3), 321-387.
doi:10.1016/j.jalz.2019.01.010
Amieva, H., Mokri, H., Le Goff, M., Meillon, C., Jacqmin-Gadda, H., Foubert-Samier, A., . . .
Dartigues, J.-F. (2014). Compensatory mechanisms in higher-educated subjects with
Alzheimer's disease: a study of 20 years of cognitive decline. Brain (London, England :
1878), 137(Pt 4), 1167.
Amjad, H., Roth, D., Sheehan, O., Lyketsos, C., Wolff, J., & Samus, Q. (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. doi:10.1007/s11606-
018-4377-y
Andel, R., Finkel, D., & Pedersen, N. L. (2016). Effects of Preretirement Work Complexity and
Postretirement Leisure Activity on Cognitive Aging. The Journals of Gerontology Series
B: Psychological Sciences and Social Sciences, 71(5), 849-856.
doi:10.1093/geronb/gbv026
Bartel, A., & Taubman, P. (1979). Health and Labor Market Success: The Role of Various
Diseases. The Review of Economics and Statistics, 61(1), 1-8. doi:10.2307/1924825
Bayes-Marin, I., Fernández, D., Lara, E., Martín-María, N., Miret, M., Moreno-Agostino, D., . . .
Olaya, B. (2020). Trajectories of Immediate and Delayed Verbal Memory in the Spanish
General Population of Middle-aged and Older Adults. Brain sciences, 10(4), 249.
doi:10.3390/brainsci10040249
Beason-Held, L. L., Goh, J. O., An, Y., Kraut, M. A., O'Brien, R. J., Ferrucci, L., & Resnick, S.
M. (2013). Changes in brain function occur years before the onset of cognitive
impairment. The Journal of neuroscience : the official journal of the Society for
Neuroscience, 33(46), 18008-18014. doi:10.1523/JNEUROSCI.1402-13.2013
Belbase, A., Khan, M., Munnell, A., & Webb, A. (2015). Slowed or Sidelined? The Effect
of'Normal'Cognitive Decline on Job Performance Among the Elderly. Center for
Retirement Research at Boston College Working Paper(2015-12).
Beydoun, M. A., Beydoun, H. A., & Wang, Y. (2008). Obesity and central obesity as risk factors
for incident dementia and its subtypes: a systematic review and meta‐analysis. Obesity
Reviews, 9(3), 204-218. doi:10.1111/j.1467-789X.2008.00473.x
102

Blazer, D. G., Yaffe, K., & Liverman, C. T. (2015). Cognitive aging: Progress in understanding
and opportunities for action: National Academies Press Washington, DC.
Blessed, G., Tomlinson, B. E., & Roth, M. (1968). The association between quantitative
measures of dementia and of senile change in the cerebral grey matter of elderly subjects.
British Journal Of Psychiatry, 114(512), 797-811. doi:10.1192/bjp.114.512.797
Bonsang, E., Adam, S., & Perelman, S. (2012). Does retirement affect cognitive functioning?
Journal of Health Economics, 31(3), 490-501. doi:10.1016/j.jhealeco.2012.03.005
Bradford, E. A., Kunik, P. M., Schulz, P. P., Williams, P. S., & Singh, P. H. (2009). Missed and
Delayed Diagnosis of Dementia in Primary Care: Prevalence and Contributing Factors.
Alzheimer Disease & Associated Disorders, 23(4), 306-314.
doi:10.1097/WAD.0b013e3181a6bebc
Brandt, J., Spencer, M., & Folstein, M. (1988). The telephone interview for cognitive status.
Neuropsychiatry, Neuropsychology and Behavioral Neurology, 1(2), 111-117.
Bratzke, L. C., Koscik, R. L., Schenning, K. J., Clark, L. R., Sager, M. A., Johnson, S. C., . . .
Hogan, K. J. (2018). Cognitive decline in the middle‐aged after surgery and anaesthesia:
results from the Wisconsin Registry for Alzheimer's Prevention cohort. Anaesthesia,
73(5), 549-555. doi:10.1111/anae.14216
Brookmeyer, R., Evans, D. A., Hebert, L., Langa, K. M., Heeringa, S. G., Plassman, B. L., &
Kukull, W. A. (2011). National estimates of the prevalence of Alzheimer’s disease in the
United States. Alzheimer's & dementia: the journal of the Alzheimer's Association, 7(1),
61-73. doi:10.1016/j.jalz.2010.11.007
Chen, C., & Zissimopoulos, J. M. (2018). Racial and ethnic differences in trends in dementia
prevalence and risk factors in the United States. Alzheimer's & Dementia: Translational
Research & Clinical Interventions, 4, 510-520. doi:10.1016/j.trci.2018.08.009
Chen, Y., Tysinger, B., Crimmins, E., & Zissimopoulos, J. M. (2019). Analysis of dementia in
the US population using Medicare claims: Insights from linked survey and administrative
claims data. Alzheimer's & Dementia: Translational Research & Clinical Interventions,
5(1), 197-207. doi:10.1016/j.trci.2019.04.003
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.
doi:10.1111/j.1532-5415.2004.52301.x
Chodosh, J., Thorpe, L. E., & Trinh-Shevrin, C. (2018). Changing Faces of Cognitive
Impairment in the U.S.: Detection Strategies for Underserved Communities. American
Journal of Preventive Medicine, 54(6), 842-844. doi:10.1016/j.amepre.2018.02.016
Coile, C. (2004). Health Shocks and Couples' Labor Supply Decisions. In: National Bureau of
Economic Research.
103

Cournot, C. M., Marquié, B. J., Ansiau, B. D., Martinaud, B. C., Fonds, B. H., Ferrières, B. J., &
Ruidavets, B. J. (2006). Relation between body mass index and cognitive function in
healthy middle-aged men and women. Neurology, 67(7), 1208-1214.
doi:10.1212/01.wnl.0000238082.13860.50
Crimmins, E. M., Kim, J. K., Langa, K. M., & Weir, D. R. (2011a). Assessment of cognition
using surveys and neuropsychological assessment: the Health and Retirement Study and
the Aging, Demographics, and Memory Study. The journals of gerontology. Series B,
Psychological sciences and social sciences, 66 Suppl 1(1), i162-171.
doi:10.1093/geronb/gbr048
Crimmins, E. M., Kim, J. K., Langa, K. M., & Weir, D. R. (2011b). Assessment of cognition
using surveys and neuropsychological assessment: the Health and Retirement Study and
the Aging, Demographics, and Memory Study. The Journals of Gerontology, Series B:
Psychological Sciences and Social Sciences, 66 Suppl 1(1), i162-171.
doi:10.1093/geronb/gbr048
Crimmins, E. M., Saito, Y., & Kim, J. K. (2016). Change in Cognitively Healthy and Cognitively
Impaired Life Expectancy in the United States: 2000-2010. SSM - population health, 2,
793-797. doi:10.1016/j.ssmph.2016.10.007
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. doi:10.1093/geronb/gbx135
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. Journal of the
American Medical Association, 269(18), 2386-2391. doi:10.1001/jama.269.18.2386
Currie, J., & Madrian, B. C. (1999). Health, health insurance and the labor market. Handbook of
labor economics, 3, 3309-3416.
Daffner, K. R. (2010). Promoting successful cognitive aging: a comprehensive review. Journal
of Alzheimer's disease : JAD, 19(4), 1101-1122. doi:10.3233/JAD-2010-1306
Demirovic, J., Prineas, R., Loewenstein, D., Bean, J., Duara, R., Sevush, S., & Szapocznik, J.
(2003). Prevalence of dementia in three ethnic groups: The south Florida program on
aging and health. Annals of Epidemiology, 13(6), 472-478. doi:10.1016/s1047-
2797(02)00437-4
Duron, E., Rigaud, A.-S., Dubail, D., Mehrabian, S., Latour, F., Seux, M.-L., & Hanon, O.
(2009). Effects of Antihypertensive Therapy on Cognitive Decline in Alzheimer's
Disease. American Journal of Hypertension, 22(9), 1020-1024. doi:10.1038/ajh.2009.119
Eicheldinger, C., & Bonito, A. (2008). More accurate racial and ethnic codes for Medicare
administrative data.(Report). Health Care Financing Review, 29(3), 27.
104

Etgen, T., Sander, D., Bickel, H., & Förstl, H. (2011). Mild cognitive impairment and dementia:
the importance of modifiable risk factors. Deutsches Arzteblatt international, 108(44),
743-750. doi:10.3238/arztebl.2011.0743
Finch, C. E. (2009). The neurobiology of middle-age has arrived. Neurobiology of Aging, 30(4),
515-520. doi:10.1016/j.neurobiolaging.2008.11.011
Finkel, D., Andel, R., Gatz, M., & Federsen, N. (2009). The Role of Occupational Complexity in
Trajectories of Cognitive Aging Before and After Retirement. Psychology and aging,
24(3), 563. doi:10.1037/a0015511
Fisher, G. G., Hassan, H., Faul, J. D., Rodgers, W. L., & Weir, D. R. (2017). Health and
Retirement Study Imputation of Cognitive Functioning Measures: 1992–2014 (Final
Release Version). http://hrsonline.isr.umich.edu/modules/meta/xyear/cogimp/desc/
COGIMPdd.pdf.
Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). "Mini-mental state". A practical
method for grading the cognitive state of patients for the clinician. Journal of Psychiatric
Research, 12(3), 189-198. doi:10.1016/0022-3956(75)90026-6
Forte, G., De Pascalis, V., Favieri, F., & Casagrande, M. (2019). Effects of Blood Pressure on
Cognitive Performance: A Systematic Review. Journal of clinical medicine, 9(1).
doi:10.3390/jcm9010034
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:Psychological Sciences
and Social Sciences, 73, S48-S56. doi:10.1093/geronb/gbx144
French, E. (2005). The Effects of Health, Wealth, and Wages on Labour Supply and Retirement
Behaviour. The Review of Economic Studies, 72(2), 395-427. doi:10.1111/j.1467-
937X.2005.00337.x
Friedland, R. P., Fritsch, T., Smyth, K. A., Koss, E., Lerner, A., J., Chen, C. H., . . . Debanne, S.,
M. (2001). Patients with Alzheimer's Disease Have Reduced Activities in Midlife
Compared with Healthy Control-Group Members. Proceedings of the National Academy
of Sciences of the United States of America, 98(6), 3440-3445.
doi:10.1073/pnas.061002998
Fritsch, T., McClendon, M. J., Smyth, K. A., Lerner, A. J., Friedland, R. P., & Larsen, J. D.
(2007). Cognitive functioning in healthy aging: the role of reserve and lifestyle factors
early in life. The Gerontologist, 47(3), 307-322. doi:10.1093/geront/47.3.307
Ganguli, I., Souza, J., McWilliams, J. M., & Mehrotra, A. (2017). Trends in Use of the US
Medicare Annual Wellness Visit, 2011-2014. Journal of the American Medical
Association, 317(21), 2233-2235. doi:10.1001/jama.2017.4342
105

Ganguli, M., Snitz, B. E., Lee, C.-W., Vanderbilt, J., Saxton, J. A., & Chang, C.-C. H. (2010).
Age and education effects and norms on a cognitive test battery from a population-based
cohort: The Monongahela–Youghiogheny Healthy Aging Team. Aging & Mental Health,
14(1), 100-107. doi:10.1080/13607860903071014
Gianattasio, Z. K., Wu, M. Q., Glymour, C. M., & Power, C. M. (2019). Comparison of methods
for algorithmic classification of dementia status in the Health and Retirement Study.
Epidemiology, 30(2), 291-302. doi:10.1097/EDE.0000000000000945
Grossman, M. (2017). On the Concept of Health Capital and the Demand for Health. In (pp. 6):
Columbia University Press.
Gruber, J., & Madrian, B. C. (2002). Health insurance, labor supply, and job mobility: A critical
review of the literature (0898-2937). Retrieved from
Guan, J.-W., & Huang, C.-Q. (2011). No Association Between Hypertension and Risk for
Alzheimers Disease: A Meta-Analysis of Longitudinal Studies. Journal of Alzheimer's
Disease, 27(4), 799-807. doi:10.3233/JAD-2011-111160
Gurland, B. J., Wilder, D. E., Lantigua, R., Stern, Y., Chen, J., Killeffer, E. H. P., & Mayeux, R.
(1999). Rates of dementia in three ethnoracial groups. International Journal of Geriatric
Psychiatry, 14(6), 481-493. doi:10.1002/(SICI)1099-1166(199906)14:6<481::AID-
GPS959>3.0.CO;2-5
Harada, C. N., Natelson Love, M. C., & Triebel, K. L. (2013). Normal Cognitive Aging. Clinics
in Geriatric Medicine, 29(4), 737-752. doi:10.1016/j.cger.2013.07.002
Hebert, L. E., Weuve, J., Scherr, P. A., & Evans, D. A. (2013). Alzheimer disease in the United
States (2010-2050) estimated using the 2010 census. Neurology, 80(19), 1778-1783.
doi:10.1212/WNL.0b013e31828726f5
Heckman, J. (1974). Life Cycle Consumption and Labor Supply: An Explanation of the
Relationship between Income and Consumption Over the Life Cycle. The American
Economic Review, 64(1), 188-194.
Heeringa, S. G., Fisher, G. G., Hurd, M., Langa, K. M., Ofstedal, M. B., Plassman, B. L., . . .
Weir, D. R. (2009). Aging, demographics and memory study (ADAMS). Sample design,
weighting, and analysis for ADAMS. Retrieved from
http://hrsonline.isr.umich.edu/sitedocs/userg/ADAMSSampleWeights_Jun2009.pdf.
Hertzog, C., Kramer, A. F., Wilson, R. S., & Lindenberger, U. (2008). Enrichment Effects on
Adult Cognitive Development: Can the Functional Capacity of Older Adults Be
Preserved and Enhanced? Psychological Science in the Public Interest, 9(1), 1-65.
doi:10.1111/j.1539-6053.2009.01034.x
Hudomiet, P., Hurd, M. D., Rohwedder, S., & Willis, R. J. (2018). The effect of physical and
cognitive decline at older ages on job mismatch and retirement (0898-2937). Retrieved
from
106

Hughes, T. F., Borenstein, A. R., Schofield, E., Wu, Y., & Larson, E. B. (2009). Association
between late-life body mass index and dementia: The Kame Project. Neurology, 72(20),
1741-1746. doi:10.1212/WNL.0b013e3181a60a58
Jefferson, L. A. (2020). Midlife Consequences of Cumulative Blood Pressure Exposure:
Importance of a Lifespan Approach. Circulation, 141(9), 725-727.
doi:10.1161/CIRCULATIONAHA.120.044447
Kasper, J. D., Freedman, V. A., & Spillman, B. C. (2013). Classification of persons by dementia
status in the National Health and Aging Trends Study. Retrieved from www.NHATS.org
Katon, W., Pedersen, H. S., Ribe, A. R., Fenger-Grøn, M., Davydow, D., Waldorff, F. B., &
Vestergaard, M. (2015). Effect of Depression and Diabetes Mellitus on the Risk for
Dementia: A National Population-Based Cohort Study. JAMA psychiatry (Chicago, Ill.),
72(6), 612-619. doi:10.1001/jamapsychiatry.2015.0082
Knopman, D. (2001). Cardiovascular Risk Factors and Cognitive Decline in Middle-Aged
Adults.(Abstract). JAMA, The Journal of the American Medical Association, 285(9),
1135.
Kohn, M. L., & Schooler, C. (1978). The Reciprocal Effects of the Substantive Complexity of
Work and Intellectual Flexibility: A Longitudinal Assessment. The American journal of
sociology, 84(1), 24.
Lane, A. P., Windsor, T. D., Andel, R., & Luszcz, M. A. (2017). Is Occupational Complexity
Associated with Cognitive Performance or Decline? Results from the Australian
Longitudinal Study of Ageing. Gerontology, 63(6), 550-559. doi:10.1159/000475559
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. Journal of the American Medical Association Internal Medicine, 177(1),
51-58. doi:10.1001/jamainternmed.2016.6807
Langa, K. M., Plassman, B. L., Wallace, R. B., Herzog, A. R., Heeringa, S. G., Ofstedal, M.
B., . . . Willis, R. J. (2005). The Aging, Demographics, and Memory Study: Study Design
and Methods. Neuroepidemiology, 25(4), 181-191. doi:10.1159/000087448
Lee, S., Buring, J. E., Cook, N. R., & Grodstein, F. (2006). The Relation of Education and
Income to Cognitive Function among Professional Women. Neuroepidemiology, 26(2),
93-101. doi:10.1159/000090254
Lee, S., Kawachi, I., Berkman, L. F., & Grodstein, F. (2003). Education, Other Socioeconomic
Indicators, and Cognitive Function. American journal of epidemiology, 157(8), 712-720.
doi:10.1093/aje/kwg042
Leritz, E., McGlinchey, R., Kellison, I., Rudolph, J., & Milberg, W. (2011). Cardiovascular
Disease Risk Factors and Cognition in the Elderly. Current Cardiovascular Risk Reports,
5(5), 407-412. doi:10.1007/s12170-011-0189-x
107

Levine, M. E., Harrati, A., & Crimmins, E. M. (2018). Predictors and implications of accelerated
cognitive aging. Biodemography and Social Biology, 64(2), 83-101.
doi:10.1080/19485565.2018.1552513
Lindeboom, M., & Kerkhofs, M. (2009). Health and Work of the Elderly: Subjective Health
Measures, Reporting Errors and Endogeneity in the Relationship between Health and
Work. Journal of Applied Econometrics, 24(6), 1024-1046. doi:10.1002/jae.1077
Livingston, G., Sommerlad, A., Orgeta, V., Costafreda, S. G., Huntley, J., Ames, D., . . . Samus,
Q. (2017). Dementia prevention, intervention, and care. The Lancet, 390(10113), 2673-
2734. doi:10.1016/S0140-6736(17)31363-6
Lu, F.-P., Lin, K.-P., & Kuo, H.-K. (2009). Diabetes and the Risk of Multi-System Aging
Phenotypes: A Systematic Review and Meta-Analysis (DM and Geriatric Syndromes).
PLoS ONE, 4(1), e4144. doi:10.1371/journal.pone.0004144
Lumsdaine, R. L., & Mitchell, O. S. (1999). Chapter 49 New developments in the economic
analysis of retirement. In (Vol. 3, pp. 3261-3307): Elsevier B.V.
MaCurdy, T. E. (1981). An Empirical Model of Labor Supply in a Life-Cycle Setting. Journal of
Political Economy, 89(6), 1059-1085. doi:10.1086/261023
Maestas, N., & Zissimopoulos, J. (2010). How Longer Work Lives Ease the Crunch of
Population Aging. The Journal of Economic Perspectives, 24(1), 139-160.
doi:10.1257/jep.24.1.139
Mazzonna, F., & Peracchi, F. (2012). Ageing, cognitive abilities and retirement. European
Economic Review, 56(4), 691-710. doi:10.1016/j.euroecorev.2012.03.004
Mazzonna, F., & Peracchi, F. (2017). Unhealthy Retirement? Journal of Human Resources,
52(1), 128-151. doi:10.3368/jhr.52.1.0914-6627R1
McArdle, J., Smith, J., & Willis, R. (2009). Cognition and economic outcomes in the Health and
Retirement Survey (No. w15266). In: National Bureau of economic research.
McArdle, J. J., Ferrer-Caja, E., Hamagami, F., & Woodcock, R. W. (2002a). Comparative
longitudinal structural analyses of the growth and decline of multiple intellectual abilities
over the life span. Developmental Psychology, 38(1), 115.
McArdle, J. J., Ferrer-Caja, E., Hamagami, F., & Woodcock, R. W. (2002b). Comparative
longitudinal structural analyses of the growth and decline of multiple intellectual abilities
over the life span.(Abstract). Developmental Psychology, 38(1), 115. doi:10.1037/0012-
1649.38.1.115
McGarry, K. (2004). Health and Retirement: Do Changes in Health Affect Retirement
Expectations? The Journal of Human Resources, 39(3), 624-648. doi:10.2307/3558990
108

Mermin, G. B. T., Johnson, R. W., & Murphy, D. P. (2007). Why Do Boomers Plan to Work
Longer? The journals of gerontology. Series B, Psychological sciences and social
sciences, 62(5), S286-S294. doi:10.1093/geronb/62.5.S286
National Academies of Sciences, E., & Medicine. (2017). Preventing cognitive decline and
dementia: A way forward: National Academies Press.
Perneczky, R., Drzezga, A., Diehl-Schmid, J., Schmid, G., Wohlschläger, A., Kars, S., . . . Kurz,
A. (2006). Schooling mediates brain reserve in Alzheimer's disease: findings of fluoro-
deoxy-glucose-positron emission tomography. Journal of neurology, neurosurgery and
psychiatry, 77(9), 1060-1063. doi:10.1136/jnnp.2006.094714
Plassman, B. L., Langa, K. M., Fisher, G. G., Heeringa, S. G., Weir, D. R., Ofstedal, M. B., . . .
Wallace, R. B. (2007). Prevalence of Dementia in the United States: The Aging,
Demographics, and Memory Study. Neuroepidemiology, 29(1-2), 125-132.
doi:10.1159/000109998
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. doi:10.1002/ana.22362
Pool, L. R., Weuve, J., Wilson, R. S., Bultmann, U., Evans, D. A., & de Leon, C. F. M. (2016).
Occupational cognitive requirements and late-life cognitive aging. Neurology, 86(15),
1386-1392. doi:10.1212/WNL.0000000000002569
Poterba, J. M. (2014). Retirement Security in an Aging Population. The American Economic
Review, 104(5), 1-30. doi:10.1257/aer.104.5.1
Power, C. M., Weuve, J. J., Gagne, B. J., McQueen, B. M., Viswanathan, B. A., & Blacker, B.
D. (2011). The Association Between Blood Pressure and Incident Alzheimer Disease: A
Systematic Review and Meta-analysis. Epidemiology, 22(5), 646-659.
doi:10.1097/EDE.0b013e31822708b5
Prince, M., Ali, G.-C., Guerchet, M., Prina, A. M., Albanese, E., & Wu, Y.-T. (2016). Recent
global trends in the prevalence and incidence of dementia, and survival with dementia.
Alzheimer's research & therapy, 8(1), 23-23. doi:10.1186/s13195-016-0188-8
Rocca, W. A., Petersen, R. C., Knopman, D. S., Hebert, L. E., Evans, D. A., Hall, K. S., . . .
White, L. R. (2011). Trends in the incidence and prevalence of Alzheimer’s disease,
dementia, and cognitive impairment in the United States. Alzheimer's & dementia: the
journal of the Alzheimer's Association, 7(1), 80-93. doi:10.1016/j.jalz.2010.11.002
Rohwedder, S., & Willis, R. (2009). Mental Retirement. IDEAS Working Paper Series from
RePEc.
Rönnlund, M., Nyberg, L., Bäckman, L., & Nilsson, L.-G. (2005). Stability, Growth, and Decline
in Adult Life Span Development of Declarative Memory: Cross-Sectional and
109

Longitudinal Data From a Population-Based Study. Psychology and aging, 20(1), 3-18.
doi:10.1037/0882-7974.20.1.3
Salthouse, T. A. (2006). Robust Cognitive Change. Journal of the International
Neuropsychological Society, 18(4), 1-8. doi:10.1017/S1355617712000380
Salthouse, T. A. (2009). When does age-related cognitive decline begin? Neurobiology of Aging,
30(4), 507-514. doi:10.1016/j.neurobiolaging.2008.09.023
Schaie, K. W. (2012). Studies of Cohort and Period Differences. In. New York: Oxford
University Press.
Schofield, D. J., Shrestha, R. N., Percival, R., Callander, E. J., Kelly, S. J., & Passey, M. E.
(2011). Early retirement and the financial assets of individuals with back problems.
European Spine Journal, 20(5), 731-736. doi:10.1007/s00586-010-1647-8
Schooler, C., & Mulatu, M. S. (2001). The Reciprocal Effects of Leisure Time Activities and
Intellectual Functioning in Older People: A Longitudinal Analysis. Psychology and
aging, 16(3), 466-482. doi:10.1037/0882-7974.16.3.466
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. doi:10.1097/00019442-200405000-00002
Singh-Manoux, A., Hillsdon, M., Brunner, E., & Marmot, M. (2005). Effects of Physical
Activity on Cognitive Functioning in Middle Age: Evidence From the Whitehall II
Prospective Cohort Study. American Journal of Public Health, 95(12), 2252-2258.
Singh-Manoux, A., Kivimaki, M., Glymour, M. M., Elbaz, A., Berr, C., Ebmeier, K. P., . . .
Dugravot, A. (2012). Timing of onset of cognitive decline: results from Whitehall II
prospective cohort study. BMJ, 344, d7622.
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 Alzheimerʼs
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. doi:10.1093/gerona/gls006
St. Clair, P. P., Gaudette, P. É., Zhao, P. H., Tysinger, P. B., Seyedin, P. R., & Goldman, P. D.
(2017). Using self-reports or claims to assess disease prevalence: It’s complicated.
Medical Care, 55(8), 782-788. doi:10.1097/MLR.0000000000000753
Stern, Y. (2002). What is cognitive reserve? Theory and research application of the reserve
concept. Journal of the International Neuropsychological Society, 8(3), 448-460.
doi:10.1017/S1355617702813248
Stern, Y. (2012). Cognitive Reserve in Ageing and Alzheimer's Disease. doi:10.7916/D87S916N
110

Suemoto, C. K., Gilsanz, P., Mayeda, E. R., & Glymour, M. M. (2015). Body mass index and
cognitive function: the potential for reverse causation. International Journal of Obesity,
39(9), 1383-1389. doi:10.1038/ijo.2015.83
Taylor, 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. doi:10.3233/JAD-2009-1099
United States. Department of Health and Human Services, a. (2012). National plan to address
Alzheimer's disease.
Van Dalen, H. P., Henkens, K., & Schippers, J. (2010). Productivity of Older Workers:
Perceptions of Employers and Employees. Population and Development Review, 36(2),
309-330. doi:10.1111/j.1728-4457.2010.00331.x
Vincent, G. K., & Velkoff, V. A. (2010). The next four decades: The older population in the
United States: 2010 to 2050. Retrieved from
https://www.census.gov/content/dam/Census/library/publications/2010/demo/p25-
1138.pdf
Viswanathan, A., & Chabriat, H. (2006). Cerebral Microhemorrhage. Stroke, 37(2), 550-555.
doi:10.1161/01.STR.0000199847.96188.12
von Hippel, C., Kalokerinos, E. K., & Henry, J. D. (2013). Stereotype threat among older
employees: relationship with job attitudes and turnover intentions. Psychology and aging,
28(1), 17-27. doi:10.1037/a0029825
Wang, H.-X., MacDonald, S. W. S., Dekhtyar, S., & Fratiglioni, L. (2017). Association of
lifelong exposure to cognitive reserve-enhancing factors with dementia risk: A
community-based cohort study. PLoS Medicine, 14(3), e1002251.
doi:10.1371/journal.pmed.1002251
Weir, D., Langa, K., & Ryan, L. (2016). Harmonized cognitive assessment protocol (hcap):
Study protocol summary. Retrieved from
http://hrsonline.isr.umich.edu/index.php?p=shoavail&iyear=ZU
Weuve, L. J., Barnes, F. L., Mendes De Leon, B. C., Rajan, T. K., Beck, E. T., Aggarwal, A.
N., . . . Evans, A. D. (2018). Cognitive aging in black and white Americans: Cognition,
cognitive decline, and incidence of Alzheimer disease dementia. Epidemiology, 29(1),
151-159. doi:10.1097/EDE.0000000000000747
Whitmer, R. A., Gunderson, E. P., Barrett-Connor, E., Quesenberry, C. P., & Yaffe, K. (2005).
Obesity in middle age and future risk of dementia: a 27 year longitudinal population
based study. BMJ (Clinical research ed.), 330(7504), 1360-1360.
Williams, K. N., & Kemper, S. (2010). Interventions to Reduce Cognitive Decline in Aging.
Journal of Psychosocial Nursing & Mental Health Services, 48(5), 42-51.
doi:10.3928/02793695-20100331-03
111

Willis, R. J. (2013). The cognitive demands of work and the length of working life: the case of
computerization. Paper presented at the Working Longer and Retirement Conference
Stanford Institute for Economic Policy Research October.
Wilson, R. S., Beckett, L. A., Barnes, L. L., Schneider, J. A., Bach, J., Evans, D. A., & Bennett,
D. A. (2002). Individual differences in rates of change in cognitive abilities of older
persons. Psychology and aging, 17(2), 179-193. doi:10.1037/0882-7974.17.2.179
Yaari, M. E. (1965). Uncertain Lifetime, Life Insurance, and the Theory of the Consumer. The
Review of Economic Studies, 32(2), 137-150. doi:10.2307/2296058
Yaffe, K., Falvey, C. M., Hamilton, N., Harris, T. B., Simonsick, E. M., Strotmeyer, E. S., . . .
Schwartz, A. V. (2013). Association Between Hypoglycemia and Dementia in a Biracial
Cohort of Older Adults With Diabetes Mellitus. Jama Internal Medicine, 173(14), 1-6.
doi:10.1001/jamainternmed.2013.6176
Zheng, F., Yan, L., Yang, Z., Zhong, B., & Xie, W. (2018). HbA 1c , diabetes and cognitive
decline: the English Longitudinal Study of Ageing. Diabetologia, 61(4), 839-848.
doi:10.1007/s00125-017-4541-7
Zissimopoulos, J. M., Crimmins, E. M., & St Clair, P. A. (2014). The value of delaying
Alzheimer's disease onset. Forum for health economics & policy, 18(1), 25.
doi:10.1515/fhep-2014-0013
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:
Psychological Sciences and Social Sciences, 73, S38-S47. doi:10.1093/geronb/gbx147

112

Appendices
Appendix Table 1. Dementia Prevalence Based on Neuropsychological Assessments (ADAMS),
Cognitive Tests (HRS), and Diagnosis Codes (Medicare Claims) by Race, Gender, and Age
Group, ages 70 and above, 2004

ADAMS (2001 to 2005) HRS 2004 Claims 2004
Gender

Male 12.3% [8.9% 15.8%] 14% [12.8% 15.2%] 9.4% [9.3% 9.4%]
Female 19.4% [15.8% 22.8%] 17% [15.9% 18.1%] 14.0% [14% 14.1%]
Age Group

70 to 74 5.3% [1.8% 8.8%] 7.2% [6.2% 8.2%] 4.6% [4.5% 4.6%]
75 to 79 6.9% [3.2% 10.5%] 12.6% [11.2% 14.1%] 9.1% [9.0% 9.1%]
80 to 84 24% [18.3% 29.6%] 19.4% [17.5% 21.4%] 16.0% [15.9% 16.1%]
85 to 89 37.2% [29.4% 45.1%] 27.7% [24.8% 30.6%] 24.8% [24.7% 25.0%]
90+ 44% [35.5% 52.5%] 47.9% [43.6% 52.3%] 34.1% [33.9% 34.3%]
Race

White 15.5% [12.6% 18.3%] 12.6% [11.8% 13.4%] 12% [11.9% 12%]
Black 23.5% [16.8% 30.1%] 39.3% [36.1% 42.5%] 16.7% [16.6% 16.8%]
Hispanic 24.7% [15.3% 34.1%] 29.3% [25.5% 33.1%] 11.5% [11.4% 11.7%]
Total
16.6% [14.1% 19.1%] 15.8% [15.0% 16.6%] 12.2% [12.2% 12.3%]
Note. ADAMS= Aging Demographics and Memory Study; HRS = Health and Retirement Study; Claims
= Medicare claims. Values in ADAMS are weighted by the ADAMS sampling weights. Values in HRS
are weighted by the HRS sampling weights; samples restricted to older adults aged 70 and above; in
response to Figure 1.

113

Appendix Table 2. Predicted Values of Dementia Prevalence Based on Cognitive Tests (HRS)
and Diagnosis Codes (Medicare Claims) by Race, Gender, and Age Group, 2006-2012

2006 2008 2010 2012
Gender

Male
HRS 12.3% [11.0% 13.6%] 12.2% [10.8% 13.6%] 11.9% [10.5% 13.3%] 10.9% [9.6% 12.2%]
Claims 9.2% [9.2% 9.3%] 10.1% [10.1% 10.1%] 10.5% [10.5% 10.6%] 9.7% [9.7% 9.7%]
Female
HRS 14.8% [13.3% 16.6%] 14.7% [13.2% 16.2%] 14.8% [13.2% 16.5%] 13.6% [12.1% 15.2%]
Claims 13.8% [13.7% 13.8%] 14.8% [14.8% 14.9%] 15.4% [15.3% 15.4%] 14.2% [14.1% 14.2%]
Age
Group

67 to 69
HRS 5.6% [4.7% 6.6%] 5.3% [4.5% 6.3%] 5.3% [4.5% 6.3%] 4.8% [4% 5.6%]
Claims 2.8% [2.8% 2.8%] 3.0% [3.0% 3.0%] 3.1% [3.1% 3.1%] 3.0% [3.0% 3.0%]
70 to 74
HRS 7.4% [6.5% 8.5%] 7.2% [6.3% 8.3%] 7.5% [6.5% 8.5%] 6.7% [5.9% 7.5%]
Claims 5.1% [5.1% 5.2%] 5.5% [5.5% 5.5%] 5.7% [5.7% 5.7%] 5.5% [5.5% 5.5%]
75 to 79
HRS 11.5% [10.2% 12.8%] 11.2% [9.9% 12.5%] 11.2% [9.9% 12.5%] 10.2% [8.9% 11.5%]
Claims 10.1% [10.0% 10.2%] 10.8% [10.7% 10.9%] 11.2% [11.1% 11.2%] 10.8% [10.8% 10.8%]
80 to 84
HRS 17.7% [15.8% 19.7%] 17.3% [15.5% 19.4%] 17.6% [15.7% 19.7%] 16.1% [14.4% 18.1%]
Claims 17.7% [17.6% 17.8%] 18.9% [18.8% 18.9%] 19.4% [19.3% 19.5%] 18.9% [18.8% 19.0%]
85 to 89
HRS 28.6% [25.9% 31.4%] 27.7% [25.1% 30.5%] 28.5% [25.8% 31.4] 26.3% [23.6% 29.1%]
Claims 27.4% [27.3% 27.5%] 28.9% [28.8% 29.0%] 29.6% [29.5% 29.7%] 28.9% [28.9% 28.9%]
90 and
above
HRS 43.4% [39.7% 47.1%] 42.7% [39.0% 46.3%] 43.3% [39.6% 47.0%] 40.2% [36.7% 43.7%]
Claims 39.3% [39.2% 39.4%] 41% [40.9% 41.1%] 41.8% [41.7% 41.9%] 40.9% [40.8% 41.0%]
Race

White
HRS 11.1% [9.8% 12.4%] 10.8% [9.6% 12.1%] 10.7% [9.4% 12%] 9.7% [8.6% 10.8%]
Claims 11.6% [11.5% 11.6%] 12.6% [12.5% 12.6%] 13.0% [13.0% 13.1%] 12.6% [12.6% 12.6%]
Black
HRS 31.2% [28.3% 34.1%] 30.6% [27.8% 33.7%] 30.7% [27.8% 33.7%] 28.1% [25.4% 31.1%]
Claims 16.1% [16% 16.1%] 17.1% [17.0% 17.2%] 17.7% [17.6% 17.8%] 17.2% [17.1% 17.3%]
Hispanic
HRS 27.2% [24.3% 30.5%] 27.2% [24.3% 30.1%] 27.1% [24.2% 30.3%] 24.8% [22.0% 27.6%]
Claims 13.2% [13.1% 13.3%] 14.4% [14.3% 14.5%] 15.2% [15.1% 15.3%] 15% [14.9% 15.0%]
Total
HRS 13.8% [12.3% 15.5%] 13.6% [12.1% 15.3%] 13.5% [12.0% 15.2%] 12.4% [11.0% 14.0%]
Claims 11.9% [11.9% 12.0%] 12.9% [12.8% 12.9%] 13.4% [13.3% 13.4%] 12.9% [12.9% 13.0%]
Note. HRS = Health and Retirement Study; Claims = Medicare claims; predicted values of dementia
prevalence (95% Confidence Interval in parentheses) based on cognitive tests are weighted by the HRS
sampling weights; predicted values of dementia prevalence obtained from the logistic regressions
adjusting for race, sex, age group, and wave in “HRS” and “Claims” models (Table 2); 95%CI adjusted
by Bonferroni correction; in response to Figure 2 and Figure 3.

114

Appendix Table 3. Sample Characteristics and Dementia Prevalence for Medicare Advantage
(MA) and Fee-For-Service (FFS) Enrollees, 2006 and 2012

Sample Size Percent
MA FFS Total MA FFS Total
White
2006 1,073 5,844 6,917 83.8% 87.4% 86.8%
2012 1,339 4,919 6,258 80.1% 88.3% 86.6%
Black
2006 78 512 590 6.1% 7.7% 7.4%
2012 141 433 574 8.5% 7.8% 7.9%
Hispanic
2006 130 329 459 10.1% 4.9% 5.8%
2012 175 222 397 10.6% 4.0% 5.5%
Less than high school
2006 1,071 2,947 4,018 46.0% 43.2% 43.9%
2012 1,472 2,603 4,075 47.5% 45.8% 46.4%
High school
2006 930 2,775 3,705 40.0% 40.7% 40.5%
2012 1,154 2,177 3,331 37.2% 38.3% 37.9%
College and above
2006 325 1,106 1,431 14.0% 16.2% 15.6%
2012 473 906 1,380 15.3% 15.9% 15.7%
Female
2006 759 4,028 4,787 58.0% 59.3% 59.1%
2012 962 3,271 4,233 56.9% 57.7% 57.5%
Male
2006 548 2,770 3,318 41.9% 40.7% 40.9%
2012 728 2,398 3,217 43.1% 42.3% 42.5%
67 to 74
2006 334 1,939 2,273 25.5% 28.5% 28.1%
2012 457 1,160 1,617 27.1% 20.5% 22.0%
75 to 84
2006 473 2,324 2,797 36.2% 34.2% 34.5%
2012 541 1,897 2,438 32.1% 33.5% 33.1%
85 and above
2006 500 2,533 3,033 38.3% 37.3% 36.4%
2012 691 2,612 3,303 40.9% 46.1% 44.9%
Total
2006 2,326 6,828 9,154 25.4% 74.6% 100.0%
2012 3,099 5,686 8,785 35.3% 64.7% 100.0%
Total Dementia
Prevalence
2006 8.9% 12.9% 12.2%
2012 9.7% 10.5% 10.3%
Note.MA = Medicare Advantage; FFS=Fee-for-Service. Results are weighted by the HRS sampling
weights. Sample restricted ages 67 and above. Broader age group is used due to the small sample size of
narrow age bands in HRS. Values are weighted by the HRS sampling weights.

115

Appendix Table 4. Predicted values of dementia prevalence based on cognitive tests (HRS) and
diagnostic codes (Claims), 2006 to 2012, with and without standardization in 2012.

2006 2012 (standardized) 2012 (non-standardized)
Gender
Male HRS 12.3% [11.0% 13.6%] 10.9% [9.6% 12.3%] 10.9% [9.6% 12.2%]
Claims 9.2% [9.2% 9.3%] 9.9% [9.9% 9.9%] 9.7% [9.7% 9.7%]
Female

HRS 14.8% [13.3% 16.6%] 13.2% [11.8% 14.8%] 13.6% [12.1% 15.2%]
Claims 13.8% [13.7% 13.8%] 14.6% [14.6% 14.7%] 14.2% [14.1% 14.2%]
Age Group

67 to 69 HRS 5.6% [4.7% 6.6%] 4.8% [4.0% 5.8%] 4.8% [4% 5.6%]
Claims 2.8% [2.8% 2.8%] 3.0% [3.0% 3.0%] 3.0% [3.0% 3.0%]
70 to 74 HRS 7.4% [6.5% 8.5%] 6.4% [5.6% 7.4%] 6.7% [5.9% 7.5%]
Claims 5.1% [5.1% 5.2%] 5.5% [5.5% 5.6%] 5.5% [5.5% 5.5%]
75 to 79 HRS 11.5% [10.2% 12.8%] 10.1% [8.9% 11.4%] 10.2% [8.9% 11.5%]
Claims 10.1% [10.0% 10.2%] 10.8% [10.8% 10.9%] 10.8% [10.8% 10.8%]
80 to 84 HRS 17.7% [15.8% 19.7%] 15.6% [13.9% 17.5%] 16.1% [14.4% 18.1%]
Claims 17.7% [17.6% 17.8%] 18.9% [18.8% 19.0%] 18.9% [18.8% 19.0%]
85 to 89 HRS 28.6% [25.9% 31.4%] 25.6% [23.1% 28.3%] 26.3% [23.6% 29.1%]
Claims 27.4% [27.3% 27.5%] 29.0% [28.9% 29.1%] 28.9% [28.9% 28.9%]
90 and
above
HRS 43.4% [39.7% 47.1%] 39.8% [36.2% 43.4%] 40.2% [36.7% 43.7%]
Claims 39.3% [39.2% 39.4%] 41.2% [41.1% 41.3%] 40.9% [40.8% 41.0%]
Race
White HRS 11.1% [9.9% 12.5%] 9.7% [8.6% 11.0%] 9.7% [8.6% 10.8%]
Claims 11.6% [11.5 11.6%] 12.3% [12.3% 12.4%] 12.6% [12.6% 12.6%]
Black HRS 31.3% [28.3% 34.3%] 28.3% [25.5% 31.3%] 28.1% [25.4% 31.1%]
Claims 16.1% [16.0% 16.2%] 17.1% [17.0% 17.1%] 17.2% [17.1% 17.3%]
Hispanic HRS 27.3% [24.2% 30.5%] 24.5% [21.7% 27.6%] 24.8% [22.0% 27.6%]
Claims 13.2% [13.2% 13.3%] 14.1% [14.0% 14.2%] 15.0% [14.9% 15.0%]
Total HRS 13.8% [12.3% 15.5%] 12.3% [10.9% 13.8%] 12.4% [11% 14%]
Claims 11.9% [11.9% 12%] 12.7% [12.7% 12.7%] 12.9% [12.9% 13%]
Note: HRS = Health and Retirement Study; Claims = Medicare claims; predicted values of dementia
prevalence (95% Confidence Interval in parentheses) based on cognitive tests are weighted by the HRS
sampling weights. The “2012 (standardized)” prevalence is calculated based on age-, sex-, race-, and
wave- specific predicted values of dementia prevalence in 2012 and 2006 mean values for age, sex, and
race. The “2012 (non-standardized)” prevalence is calculated based on predicted values of dementia
prevalence in 2012 and 2012 mean values for age, sex, and race, allowing for changes in this
composition. Predicted values of dementia prevalence obtained from the logistic regressions adjusting for
race, sex, age group, and wave in “HRS” and “Claims” models. 95%CI adjusted by Bonferroni correction.
Values in HRS are weighted by the HRS sampling weights.

116

Appendix Table 5. Odds ratios and 95%CIs for Presence of Dementia Adjusting for Age, Sex,
Race, Wave, and Data Source in HRS and claims, 2006-2013

(1) (2) (3) (4) (5)

Data Source
Data
Source*Race
Data
Source*Age
Data
Source*Gender
Data
Source*Wave
67 to 69
1 1 1 1 1
[1 1] [1 1] [1 1] [1 1] [1 1]
70 to 74
1.88*** 1.88*** 1.88*** 1.88*** 1.88***
[1.87 - 1.89] [1.87 - 1.89] [1.87 - 1.89] [1.87 - 1.89] [1.87 - 1.89]
75 to 79
3.91*** 3.91*** 3.91*** 3.91*** 3.91***
[3.89 - 3.93] [3.89 - 3.93] [3.89 - 3.93] [3.89 - 3.93] [3.89 - 3.93]
80 to 84
7.50*** 7.51*** 7.51*** 7.50*** 7.51***
[7.47 - 7.54] [7.47 - 7.55] [7.48 - 7.55] [7.47 - 7.54] [7.47 - 7.54]
85 to 89
13.08*** 13.08*** 13.09*** 13.08*** 13.08***
[13.01-13.15] [13.01 - 13.15] [13.02 - 13.16] [13.01 - 13.15] [13.01 - 13.15]
90 and
above
21.90*** 21.90*** 21.92*** 21.90*** 21.90***
[21.77 - 22.02] [21.78 - 22.03] [21.80 - 22.05] [21.77 - 22.02] [21.77 - 22.02]
Black
1.68*** 1.68*** 1.68*** 1.68*** 1.68***
[1.68 - 1.69] [1.67 - 1.68] [1.68 - 1.69] [1.68 - 1.69] [1.68 - 1.69]
Hispanic
1.46*** 1.46*** 1.46*** 1.46*** 1.46***
[1.45 - 1.47] [1.45 - 1.47] [1.45 - 1.47] [1.45 - 1.47] [1.45 - 1.47]
Other Race
0.98*** 0.98*** 0.98*** 0.98*** 0.98***
[0.98 - 0.99] [0.98 - 0.99] [0.98 - 0.99] [0.98 - 0.99] [0.98 - 0.99]
Male
1 1 1 1 1
[1 1] [1 1] [1 1] [1 1] [1 1]
Female
1.28*** 1.28*** 1.28*** 1.28*** 1.28***
[1.28 - 1.28] [1.28 - 1.28] [1.28 - 1.28] [1.28 - 1.28] [1.28 - 1.28]
2006
1 1 1 1 1
[1 1] [1 1] [1 1] [1 1] [1 1]
2007
1.05*** 1.05*** 1.05*** 1.05*** 1.05***
[1.04 - 1.05] [1.04 - 1.05] [1.04 - 1.05] [1.04 - 1.05] [1.04 - 1.05]
2008
1.08*** 1.08*** 1.08*** 1.08*** 1.08***
[1.08 - 1.09] [1.08 - 1.09] [1.08 - 1.09] [1.08 - 1.09] [1.08 - 1.09]
2009
1.11*** 1.11*** 1.11*** 1.11*** 1.11***
[1.11 - 1.12] [1.11 - 1.12] [1.11 - 1.12] [1.11 - 1.12] [1.11 - 1.12]
2010
1.12*** 1.12*** 1.12*** 1.12*** 1.12***
[1.11 - 1.12] [1.11 - 1.12] [1.11 - 1.12] [1.11 - 1.12] [1.11 - 1.12]
2011
1.10*** 1.10*** 1.10*** 1.10*** 1.10***
[1.10 - 1.11] [1.10 - 1.11] [1.10 - 1.11] [1.10 - 1.11] [1.10 - 1.11]
2012
1.08*** 1.08*** 1.08*** 1.08*** 1.08***
[1.08 - 1.09] [1.08 - 1.09] [1.08 - 1.09] [1.08 - 1.09] [1.08 - 1.09]
117

2013
1.02*** 1.02*** 1.02*** 1.02*** 1.02***
[1.01 - 1.02] [1.01 - 1.02] [1.01 - 1.02] [1.01 - 1.02] [1.01 - 1.02]
HRS
1.16*** 0.83*** 1.99*** 1.28*** 1.30***
[1.12 - 1.19] [0.80 - 0.86] [1.80 - 2.19] [1.23 - 1.35] [1.23 - 1.38]
67 to 69
#HRS

1

[1 1]

70 to 74
#HRS

0.75***

[0.66 - 0.84]

75 to 79
#HRS

0.56***

[0.50 - 0.63]

80 to 84
#HRS

0.47***

[0.42 - 0.53]

85 to 89
#HRS

0.52***

[0.46 - 0.58]

90 and
above
#HRS

0.57***

[0.50 - 0.65]

White
#HRS

1

[1 1]

Black
#HRS

3.01***

[2.80 - 3.24]

Hispanic
#HRS

2.98***

[2.72 - 3.26]

OtherRaces
#HRS

1.89***

[1.53 - 2.35]

Male#HRS

1

[1 1]

Female
#HRS

0.85***

[0.80 - 0.90]

2006#HRS

1

[1 1]
2008#HRS

0.89**

[0.82 - 0.97]
2010#HRS

0.88**

[0.81 - 0.96]
2012#HRS

0.80***

[0.74 - 0.87]
Constant
0.02*** 0.02*** 0.02*** 0.02*** 0.02***
[0.02 - 0.02] [0.02 - 0.02] [0.02 - 0.02] [0.02 - 0.02] [0.02 - 0.02]
Observation
32,898,653 32,898,653 32,898,653 32,898,653 32,898,653
Note. HRS = the Health and Retirement Study; Claims = Medicare claims; Logistic regression pools HRS
and Medicare claims (aged 67 and above, 2006-2013), the variable “HRS” is an indicator variable
118

implicating data source from HRS. Data source indicator is interacted with race, gender, age, and wave
respectively in model 2 through 5. ***p<0.001. **p<0.01. *p< .05.

119

Appendix Table 6. Sample characteristics by short-term cognitive change in 2 years, 1996 to
2014
No Change Cognitive Decline Cognitive
Improvement
Total
Race***

White 84.0 80.8 81.2 83.4
Black 9.9 11.9 11.3 10.2
Hispanic 2.9 3.2 3.7 3.0
other race 3.2 4.1 3.7 3.4
Gender***

Male 45.1 43.9 43.3 44.8
Education

less than high school 16.2 18.4 17.4 16.5
High school 27.3 27.8 26.7 27.3
college and above 56.6 53.8 55.9 56.2
Age Group ***

50 to 54 12.5 15.5 13.2 12.9
55 to 59 46.6 45.4 48.2 46.7
60 to 64 40.9 39.0 38.7 40.5
Married (time t-1)*** 72.1 69.4 68.8 71.5
Income quartiles***

25 percentile 18.6 21.9 20.6 19.1
50 percentile 21.8 22.9 23.3 22.1
75 percentile 27.0 27.9 26.4 27.1
100 percentile 32.5 27.4 29.7 31.7
Wealth Quartiles***

25 percentile 20.1 22.0 22.4 20.5
120

50 percentile 23.1 25.9 23.9 23.4
75 percentile 26.8 25.4 26.1 26.6
100 percentile 30.1 26.7 27.6 29.5
Occupation(t-1)***

Managerial 11.9 9.0 11.1 11.5
Professional 17.0 15.9 17.0 16.9
Sales&admin 18.2 17.1 17.5 18.0
Protection/Military 1.0 0.9 1.4 1.0
Cleaning/Building 6.3 6.3 6.5 6.3
Production/Operation 14.7 16.1 15.1 14.9
Nonworking 30.0 33.9 30.8 30.5
Other missing 0.8 0.9 0.7 0.8
Heart disease (t-1) 11.7 12.6 12.1 11.8
Diabetes (t-1)* 13.2 14.1 12.8 13.2
Hypertension (t-1) 39.8 40.2 38.6 39.7
BMI 27.9 27.5 27.8 27.8
Cognitive Score (t-1) 17.0 13.7 19.6 17.0
Stroke (t-1)* 3.2 4.0 3.1 3.2
Depressive symptoms (t-
1)
4.3 4.8 4.8 4.4
Percent of cognitive
decline/improvement
80.0% 9.7% 10.4% 100.0%
N
39,571

4,787

5,131

49,489
Note. Characteristics are weighted by the HRS personal sampling weight combined with the nursing
home weight, sample restricted to age 50 to 64, self-respondents, with at least two waves of cognitive
score in HRS 1996-2014. *** indicates that the distribution of characteristics statistically significantly
different by cognitive change from time t-1 to time t. The occupation variables in HRS are respectively
defined by the respondent’s current job occupation based on the 1980 (for HRS 1992-2014), 2000 (for
HRS 2004-2014), and 2010 (for HRS 2010-2014) Census occupation codes. There are some missing
121

values in the occupation variables based on the 1980 and 2000 Census codes. Therefore, I defined the
respondent’s current occupation based on the 2000 or 2010 census codes if the one defined by the 1980
or 2000 Census codes is missing.

122

Appendix Table 7. Sample characteristics b cognitive changes over 2 – 14 years, 1996-2014

No
Change Decline Improvement Total
Race (%)***
White 70.5 61 59.5 68.9
Black 20.9 28.6 29 22.1
Hispanic 5.3 7.1 6.8 5.5
Other race 3.4 3.4 4.6 3.5
Gender (%)***

Male 41.8 43.7 45.2 42.2
Education (%)

Less than high school 23.2 29.1 26.9 24.0
High school 28.4 29.0 28.0 28.4
College and above 48.5 41.9 45.0 47.6
Married (time t-1; %)*** 73.5 68.6 67.5 72.6
Income quartiles (%)***

25 percentile 24.0 30.8 32.0 25.2
50 percentile 24.9 29.2 26.3 25.4
75 percentile 26.9 22.8 23.5 26.3
100 percentile 24.2 17.2 18.2 23.2
Wealth Quartiles (%)***

25 percentile 26.6 33.0 34.6 27.8
50 percentile 27.8 27.3 26.3 27.6
75 percentile 25.5 23.8 22.3 25.1
100 percentile 20.1 16.0 16.9 19.5
123

Occupation(t-1; %)***

Managerial 10.3 6.7 7.6 9.8
Professional 15.5 10.7 12.8 14.9
Sales&admin 17.8 14.4 13.3 17.1
Protection/Military 1.2 1.4 1.2 1.2
Cleaning/Building 8.2 8.1 9.4 8.3
Production/Operation 16.5 17.2 15.7 16.5
Nonworking 29.3 39.4 38.3 30.8
Other missing 1.2 2.0 1.7 1.3
Heart disease (t-1; %)*** 9.5 12.7 13.0 10.1
Diabetes (t-1; %)*** 11.9 17.7 16.4 12.8
Hypertension (t-1; %)*** 36.9 43.7 42.9 38.0
BMI 27.9 27.5 27.8 27.8
Cognitive Score (t-1) 17.0 13.7 19.6 17.0
Stroke (t-1; %)*** 2.8 3.9 4.7 3.1
Cancer (t-1; %)*** 5.2 7.9 5.5 5.5
Depressive symptoms (t-1; %)** 4.3 4.9 6.0 4.5
Percent of cognitive
decline/improvement
83.8% 8.7% 7.5% 100.0%
N 15,101 1,574 1,348 18,023
Note. Characteristics unweighted. Sample restricted to individuals aged 50 to 64, self-respondents,
with at least one wave of cognitive score in HRS 1996-2014. *** indicates that the distribution of
characteristics statistically significantly different by cognitive change from time t-1 to time t.

124

Appendix Table 8. The distribution of durations between the first and last observation.
Mean 8.3
Min 2
5 percentiles 2
10 percentiles 4
25 percentiles 6
50 percentiles 8
75 percentiles 12
90 percentiles 12
95 percentiles 14
Max 14
Standard deviation 3.6
Note. Unweighted. The duration of durations is calculated as the gap between the first and last
observation for each individual. The sample is restricted to age 50 to 64, HRS 1996-2014, persons with a
slope of change in cognitive score. The sample size is 18,023 persons.

125

Appendix Table 9. The distribution of baseline cognitive score and the slope of changes in
cognitive score, 2-14 years, 1996-2014
Distribution Baseline cognitive score Slope of cognitive score
Mean 16.2 -0.1
Min 0 -18
1 percentile 5 -6
5 percentiles 9 -3
10 percentiles 10 -2
25 percentiles 14 -1
50 percentiles 17 -0.1
75 percentiles 19 0.6
90 percentiles 21 2
95 percentiles 23 3
99 percentiles 25 6
Max 27 15
Standard Deviation 4.3 2.0
Note. Unweighted. The sample is restricted to age 50 to 64, HRS 1996-2014, persons with a slope of
change in cognitive score. The sample size is 18,023 persons.

126

Appendix Table 10. The distribution of baseline cognitive impairment and cognitive decline , 2-
14 years, 1996-2014
Cognitive Impairment,
t0
Cognitive Decline, t0-tn
Category Percent (N) Category Percent (N)
Dementia 1.7% (302) Cognitive decline 8.7% (1,574)
Population CIND 11.4% (2,055) No change 83.8% (15,101)
Normal 86.9% (15,666) Cognitive improvement 7.5% (1,348)
Total 100% (18,032) Total 100% (18,032)
Cognitive decline 9.4% (1,477)
Normal No change 85% (13,316)
Cognitive improvement 5.6% (873)
Total 100% (18,032)
Note. Unweighted. The notion “t0” refers to the beginning of the observation, which is the baseline. The
notion “tn” refers to the end of the observation. Cognitive decline from t0 to tn is defined as an average of
2 or more points decline in one’s cognitive score (scale 0-27) across waves.

127

Appendix Table 11. Odds ratio of cognitive changes among persons with normal baseline
cognitive function, full table, 1996-2014
(1) (2) (3) (4)
Decline, 2 years Improvement, 2
years
Decline, 2-14 years Improvement
, 2-14 years
White 1 1 1 1

[1 1] [1 1] [1 1] [1 1]
Black 1.93*** 0.69*** 2.36*** 0.81

[1.76 - 2.13] [0.62 - 0.77] [2.02 - 2.76] [0.66 - 1.01]
Hispanic 1.64*** 0.71*** 2.77*** 0.77

[1.39 - 1.92] [0.59 - 0.86] [2.12 - 3.62] [0.52 - 1.14]
other race 1.58*** 0.84 1.63** 1.30

[1.31 - 1.90] [0.68 - 1.04] [1.17 - 2.27] [0.89 - 1.89]
less than high school 1 1 1 1

[1 1] [1 1] [1 1] [1 1]
high school 0.70*** 1.46*** 0.70*** 1.69***

[0.64 - 0.78] [1.30 - 1.64] [0.60 - 0.83] [1.33 - 2.14]
college and above 0.48*** 2.15*** 0.59*** 2.65***

[0.43 - 0.53] [1.91 - 2.42] [0.50 - 0.70] [2.09 - 3.37]
Female (ref. ) 1 1 1 1

[1 1] [1 1] [1 1] [1 1]
Male 1.16*** 0.69*** 1.33*** 0.89

[1.08 - 1.25] [0.64 - 0.75] [1.17 - 1.51] [0.76 - 1.04]
Single (ref.) 1 1 1 1

[1 1] [1 1] [1 1] [1 1]
Married 1.05 0.84*** 0.92 0.94

[0.97 - 1.14] [0.76 - 0.92] [0.78 - 1.08] [0.77 - 1.14]
128

Heart disease (t-1) 1.09 1.01 1.13 1.14

[0.99 - 1.21] [0.89 - 1.13] [0.94 - 1.36] [0.90 - 1.44]
Diabetes (t-1) 1.11* 0.92 1.39*** 1.06

[1.01 - 1.23] [0.82 - 1.03] [1.17 - 1.65] [0.85 - 1.33]
BMI (t-1) 0.99* 1.00 0.98** 1.00

[0.99 - 1.00] [1.00 - 1.01] [0.97 - 0.99] [0.99 - 1.02]
Hypertension (t-1) 1.06 0.86*** 1.22** 1.01

[0.99 - 1.14] [0.80 - 0.93] [1.08 - 1.38] [0.86 - 1.18]
Stroke (onset) 1.61** 0.59 1.28 0.63

[1.16 - 2.24] [0.35 - 1.01] [0.92 - 1.78] [0.35 - 1.15]
Stroke (t-1) 1.29** 0.89 1.11 1.11

[1.07 - 1.55] [0.71 - 1.10] [0.80 - 1.54] [0.72 - 1.72]
Depressive symptoms
(t-1)
1.23** 0.75** 1.20 0.80

[1.06 - 1.43] [0.62 - 0.91] [0.90 - 1.60] [0.54 - 1.19]
Cognitive score (t-1) 1.34*** 0.69*** 1.25*** 0.71***

[1.33 - 1.36] [0.68 - 0.70] [1.22 - 1.27] [0.69 - 0.74]
Income 25 percentile
(ref.)
1 1 1 1

[1 1] [1 1] [1 1] [1 1]
Income 50 percentile 0.85*** 1.15* 0.99 1.06

[0.77 - 0.93] [1.02 - 1.29] [0.84 - 1.16] [0.85 - 1.33]
Income 75 percentile 0.82*** 1.19** 0.90 1.16

[0.74 - 0.91] [1.05 - 1.35] [0.74 - 1.09] [0.90 - 1.50]
Income 100 percentile 0.68*** 1.35*** 0.87 1.28

[0.61 - 0.77] [1.17 - 1.55] [0.69 - 1.08] [0.96 - 1.71]
129

Wealth 25 percentile
(ref.)
1 1 1 1

[1 1] [1 1] [1 1] [1 1]
Wealth 50 percentile 0.95 1.03 0.73*** 0.78*

[0.87 - 1.05] [0.93 - 1.15] [0.63 - 0.86] [0.63 - 0.97]
Wealth 75 percentile 0.85** 1.18** 0.74*** 0.82

[0.77 - 0.94] [1.05 - 1.33] [0.62 - 0.88] [0.65 - 1.04]
Wealth 100 percentile 0.76*** 1.31*** 0.66*** 0.84

[0.68 - 0.85] [1.15 - 1.48] [0.54 - 0.81] [0.64 - 1.09]
Managerial 1 1 1 1

[1 1] [1 1] [1 1] [1 1]
Professional 1.08 1.10 1.06 1.16

[0.95 - 1.24] [0.96 - 1.26] [0.81 - 1.37] [0.86 - 1.57]
Sales&admin 1.11 0.91 1.30* 0.94

[0.97 - 1.26] [0.80 - 1.05] [1.02 - 1.68] [0.69 - 1.28]
Protection/Military 1.13 1.00 1.51 0.99

[0.82 - 1.56] [0.71 - 1.40] [0.90 - 2.53] [0.51 - 1.92]
Cleaning/Building 1.29** 0.82* 1.46* 0.99

[1.09 - 1.52] [0.69 - 0.98] [1.08 - 1.98] [0.69 - 1.43]
Production/Operation 1.60*** 0.82* 1.65*** 0.94

[1.39 - 1.84] [0.71 - 0.96] [1.28 - 2.14] [0.69 - 1.28]
Nonworking 1.56*** 0.76*** 1.80*** 1.05

[1.37 - 1.77] [0.66 - 0.87] [1.41 - 2.28] [0.78 - 1.41]
Other missing 1.77*** 0.55* 2.95*** 1.64

[1.26 - 2.48] [0.34 - 0.89] [1.87 - 4.65] [0.89 - 3.03]
Observations 45184 45168 15661 15655
130

Pseudo R-squared 0.108 0.120 0.120 0.156
Note. (1) Short-term cognitive decline, (2) Short-term cognitive improvement, (3) Long-term cognitive
decline, (4) Long-term cognitive improvement. Sample restricted to those with normal cognitive
function at the baselne. Other covariates include age and age square at baseline, time fixed effects (only
for model 1 and model 2), flag of missing values for each covariate. *** p<0.001, ** p<0.01, * p<0.05.

131

Appendix Table 12. The distribution of cognitive scores (time t-1) and the change in cognitive
scores over 2 years (time t-1 to time t), 1996-2014

Cognitive scores
at t-1
Change in cognitive
scores, t-1 to t Expected probability
of working at age 62
Change in the
expected probability
of working at age
62, time t-1 to t
Mean 17.4 -0.1 57.2 1.9
Min 0 -14 0 -100
p10 13 -4 0 -40
p25 15 -2 20 -10
Median 18 0 70 0
p75 20 2 90 20
p90 22 4 100 100
Max 27 13 100 40
Std.
Dev
3.7 3.2
37 32.7
N 21,839 21,839 21,839 21,839
Note. Statistics weighted by the HRS sampling weight.

132

Appendix Table 13. Distribution of covariates by cognitive decline status over two years, 1996-
2014.
Percent/Mean
(Frequency)

No Cognitive Decline Cognitive
Decline
Total
Race***
Non-Hispanic white 86.4% 82.7% 85.9%
Non-Hispanic black 8.2% 9.9% 8.4%
Hispanic 2.6% 3.4% 2.7%
Other race 2.9% 4.0% 3.0%
Education**
Less than high school 10.3% 11.3% 10.5%
High school 25.8% 25.7% 25.8%
College and above 63.8% 63.0% 63.8%
Age
50 to 54 18.2% 14.2% 18.4%
55 to 59 62.1% 48.2% 62.0%
60 to 62 19.7% 22.1% 19.6%
Income Quartiles***
Income 1st quartile 17.4% 20.3% 17.8%
Income 2nd quartile 23.2% 23.4% 23.2%
Income 3rd quartile 27.7% 28.0% 27.7%
Income 4th quartile 31.8% 28.4% 31.4%
Wealth Quartiles**
Wealth 1st quartile 21.4% 22.9% 21.5%
Wealth 2nd quartile 24.3% 25.1% 24.4%
Wealth 3rd quartile 26.3% 26.1% 26.3%
Wealth 4th quartile 28.1% 26.0% 27.8%
Occupation***
Managerial 17.6% 15.0% 17.3%
Professional 25.3% 24.1% 25.2%
Sales & admin 25.6% 27.0% 25.8%
Protection/Military 1.4% 1.4% 1.4%
Cleaning/building 8.5% 9.2% 8.6%
Production/Operation 21.2% 22.6% 21.4%
Male* 48.6% 47.6% 48.4%
Marital status*** 74.9% 72.2% 74.6%
Heart disease 9.8% 10.4% 9.9%
Diabetes** 11.6% 13.5% 11.8%
BMI 28.4 28.4 0.3
Hypertension 38.3% 38.9% 38.4%
Incident heart disease 1.8% 1.9% 1.8%
Incident diabetes 2.1% 2.6% 2.1%
133

Change in BMI 0.19 0.16 0.19
Incident hypertension 4.6% 4.0% 4.5%
Percent 88.1% 11.9% 100.0
%
Sample Size 19,232 2,607 21,83
9
Note. Statistics weighted by the HRS sampling weight; Chi-square test for the distribution of covariates
by cognitive decline status over 2 years; *p<0.1; **p<0.05; ***p<0.01.

134

Appendix Table 14. Full OLS and FE results of the effect of cognitive decline over 2 years on
the expected probability of working at age 62.

(1) (2) (3)
Correlation OLS FE
Cognitive decline -1.95*** -1.98*** -2.17*
Age 2.19 2.1
Age square -0.015 -0.013
1998 0 0
2000 2.30** 2.79**
2002 -0.23 0.23
2004 2.20* 2.70*
2006 1.15 0.37
2008 5.79*** 5.12***
2010 -2.93*** -4.41***
2012 0.80 0.34
2014 0.76
White 0
Black 0.81
Hispanic -0.21
Other race 1.68
Less than high school 0
High school 0.49
College and above 0.57
Male HRS -0.83**
Single
Married 0.50 2.52
Heart disease, t-1 0.48 -4.60
Diabetes, t-1 -0.84 -3.24
Hypertension, t-1 -0.22 -0.46
Stroke, t-1 0.67 -1.46
BMI, t-1 -0.0025 0.15
Cancer -1.41* -3.92
Onset of heart disease -3.99** -5.63*
Onset of diabetes 1.38 1.28
Onset of hypertension -0.26 -0.67
Onset of stroke -6.49 -16.5**
Onset of cancer -3.73* -6.31*
Change in BMI 0.026 0.070
Total cognitive score, t-1 0.028 -0.026
Decline in income 0 0
No change in income -0.053 -1.76
Increase in income -1.29 -3.49**
135

Decline in wealth 0 0
No change in wealth 0.44 1.96
Increase in wealth -0.53 0.32
transition to less cognitive demanding occupation 0 0
No change in occupation -0.18 0.67
More cognitive demanding occupation 1.01 1.91
Occupation Missing -9.98* -9.59
Income quartile, t-1, (ref. 1st) 0 0
2nd income quartile -0.059 -1.29
3rd income quartile -0.28 -3.00
4th income quartile 0.26 -2.22
Wealth quartile, t-1, (ref. 1st) 0 0
2nd wealth quartile 0.97 -0.26
3rd wealth quartile 0.69 0.26
4th wealth quartile 1.42** 2.81
Professional, t-1 0 0
Sales, t-1 -0.14 4.11
Protection/Military, t-1 -0.12 4.35
Cleaning/Building, t-1 -0.083 10.7
Production/Operation, t-1 -0.62 2.47
Occupation missing, t-1 -0.80 1.81
Constant -1.95*** -76.2 -138.2
Observations 22,011 22,011 22,011
R-squared 0 0.007 0.01
Source. HRS 1996-2014. Note. Sample restricted to working older adults aged 50 to
61. Standard errors clustered by individuals: *P<0.1, **p<0.05, ***p<0.01. Missing
values of the variable occupation are included with a flag.

136

Appendix Table 15. Distribution of the duration of the observation in the long run.

Years
Average 5.0
Minimized 2
5 percentiles 2
Median 4
95 percentiles 10
Maximized 10
Note. unweighted.

137

Appendix Table 16. Sample characteristics, 2-10 years duration, 1996-2014
Percent
Cognitive decline 12%
Race
Non-Hispanic white 65.4%
Non-Hispanic black 18.7%
Hispanic 12.7%
Other race 3.2%
Education
Less than high school 18.9%
Highschool 28.1%
College and above 53.0%
Age
50 to 54 14.9%
55 to 59 68.1%
60 to 61 17.0%
Income Quartiles
Income 1st quartile 28.2%
Income 2nd quartile 26.1%
Income 3rd quartile 24.2%
Income 4th quartile 21.5%
Wealth Quartiles
Wealth 1st quartile 29.5%
Wealth 2nd quartile 27.3%
Wealth 3rd quartile 24.3%
Wealth 4th quartile 18.9%
Occupation
Professional 22.6%
Sales 6.6%
138

Clerical 11.7%
Service 10.3%
Manual 16.9%
Missing 32.0%
Male 45.6%
Marital status 75.4%
Heart disease 7.8%
Diabetes 10.4%
BMI 28.16
Hypertension 35.0%
Stroke 1.7%
Incident heart disease 4.8%
Incident diabetes 6.0%
Change in BMI 0.42
Incident hypertension 12.6%
Incident stroke 1.2%
Percent 100.0%
Sample Size 11,511
Source. HRS 1996-2014. Note. Sample restricted to working older adults aged 50 to 61. Statistics
weighted by the HRS sampling weight.

139

Appendix Table 17. Full OLS results of the effect of individual-specific trend of cognitive
decline on change in the expected probability of working at age 62 over the study period, 1996-
2014
Variable Coefficient
Cognitive decline -5.14**
Age in years at July 1st 0.32**
Race (ref. white)

Black 1.72**
Hispanic 1.27
Other race 1.48
Education (ref. less than hs)

High school 0.53
College and above 0.41
Male HRS -2.08**
Married -0.25
Onset of heart disease -1.45
Onset of diabetes 0.39
Onset of hypertension 2.89**
Onset of stroke 0.37
Onset of cancer -1.56
Change in BMI 0.18
Heart disease, t0 0.68
Diabetes, t0 -0.93
BMI, t0 0.004
Hypertension, t0 -0.94
Stroke, t0 0.55
Cancer, t0 -3.78**
Total cognitive score, t0 0.16
Income change (ref. decline in income)

No change in income 2.26**
140

Increase in income 4.37***
Decline in wealth 0
No change in wealth -1.15
Increase in wealth -0.67
Occupation Shift (ref. transition to less
cognitive demanding occupation)

No change in occupation -2.37
More cognitive demanding occupation -0.26
Income quartile, t0, (ref. 1st)

2nd income quartile 0.74
3rd income quartile 2.52**
4th income quartile 3.63***
Wealth quartile, t0, (ref. 1st)

2nd wealth quartile 1.59*
3rd wealth quartile 2.28**
4th wealth quartile 2.31*
Occupation (ref. professional)

Sales, t0 -1.33
Clerical, t0 -1.37
Service, t0 -0.58
Manual, t0 -0.82
Occupation missing, t0 -3.91***
Constant -28.1***
Observations 11,511
R-squared 0.013
Source. HRS 1996-2014. Note. Sample restricted to working older adults aged 50 to 61. Standard errors
clustered by individuals: *P<0.1, **p<0.05, ***p<0.01. Missing values of the variable occupation are
included with a flag.

141

Appendix Table 18. FE results of the effect of cognitive decline over 2 years on the expected
probability of working at age 62 among workers with no depressive symptoms
Variable Coefficient
Cognitive decline -1.82*
Age in years at July 1st 0.85
Age Square -0.00
1998

2000 2.03
2002 -0.20
2004 2.11
2006 -0.55
2008 4.84***
2010 -5.22***
2012 0.29
2014 0
married 2.78
heart disease, t-1 -5.43
Diabetes, t-1 -2.63
Hypertension, t-1 -0.38
Stroke, t-1 -0.97
BMI, t-1 0.087
Onset of heart disease -6.33*
Onset of diabetes 1.70
Onset of hypertension -1.30
Onset of stroke -15.8**
Onset of cancer -5.67
Change in BMI 0.14
Total cognitive score, t-1 0.11
Decline in income 0
No change in income -2.57**
142

Increase in income -3.67**
Decline in wealth 0
No change in wealth 2.54*
Increase in wealth 1.47
Occupation shift (ref. transition to less cognitive
demanding occupation)

No change in occupation 1.62
More cognitive demanding occupation -0.46
Occupation Missing 3.68
Income quartile, t-1, (ref. 1st)

2nd income quartile -1.77
3rd income quartile -3.55*
4th income quartile -2.87
Wealth quartile, t-1, (ref. 1st)

2nd wealth quartile 0.12
3rd wealth quartile 1.95
4th wealth quartile 5.29*
Occupation (ref. professional, t-1)

Sales, t-1 1.16
Clerical, t-1 5.03
Service, t-1 -4.46
Manual, t-1 1.33
Occupation missing, t-1 -0.31
Constant -46.2
Observations 21,473
R-squared 0.010
Source. HRS 1996-2014. Note. Sample restricted to working older adults aged 50 to 61. Standard errors
clustered by individuals; *P<0.1, **p<0.05, ***p<0.01. Missing values of the variable occupation are
included with a flag.

Abstract (if available)

Abstract Alzheimer’s disease and related dementia (ADRD), as one most important disabling concern for older adults, imposes great socioeconomic burdens on individuals, families, and society. To provide more resources and address these challenges presented by dementia, the government in 2012 established a National Plan. However, critical information is missing for implementing two goals of the National Plan—improve data and track the current situation of dementia, and prevent and effectively treat dementia. The missing information includes current levels and trends of dementia among different subgroups and factors associated with rapidly declining cognitive function at middle age, which could be the starting point of dementia. ❧ This dissertation fills these knowledge gaps by exploring the diversity in dementia prevalence and mid-life cognitive trajectories. It aims to quantify levels and trends of dementia prevalence for the US population and by race, sex, and age by using three population-based data sources. It additionally identifies persons with the most rapid rates of changes in cognitive abilities at an early stage and factors associated with these fast cognitive changes. Moreover, this dissertation explores how rapidly declining cognition affects the labor supply at the early retirement age (62 years old). ❧ The dissertation ultimately provides vital information for individuals, families, health care professionals, and policymakers to better plan for functional or pathological cognitive decline. Additionally, it addresses policy challenges in implementing the National Plan. More importantly, this dissertation suggests that there is disparity in the distribution of cognitive trajectories. Ethnic minorities, those with poorer health conditions and lower SES are more likely to experience cognitive decline and may have less access to health care services. They may need more interventions on delaying or preventing cognitive decline and more financial and community support to buffer the consequences of dementia or middle-aged cognitive deterioration.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

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

PDF

Predicting cognitive decline and dementia in elderly twins from indicators of early life oral health

PDF

Essays on work, retirement, and fostering longer working lives

PDF

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

PDF

Three essays on aging, wealth, and housing tenure transitions

PDF

Accuray of subjective cognitive complaints in a longitudinal context: the effect of depression and dementia status

PDF

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

PDF

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

PDF

Contributions of dynamic cerebrovascular function to cognitive decline and dementia: development and validation of a novel neuroimaging approach

PDF

Associations between longitudinal loneliness, epigenetic age, and dementia risk

PDF

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

PDF

Public awareness campaign for dementia in Taiwan

PDF

Cognitive health and self-rated memory in later life: linkages to race/ethnicity, multimorbidity, and survival

PDF

Alzheimer’s disease: dysregulated genes, ethno-racial disparities, and environmental pollution

PDF

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

PDF

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

PDF

Dementia, co-morbid depressive symptoms, and mortality in older adults: a twin study

PDF

Three essays on health, aging, and retirement

PDF

Planning care with the patient in the room: a patient-focused approach to reducing heart failure readmissions

Asset Metadata

Creator Zhu, Yingying (author)

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

School School of Policy, Planning and Development

Degree Doctor of Philosophy

Degree Program Public Policy and Management

Publication Date 09/21/2020

Defense Date 05/20/2020

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

Tag cognitive decline,dementia prevalence,labor supply,OAI-PMH Harvest,racial disparity

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Zissimopoulos, Julie (committee chair), Aguila, Emma (committee member), Crimmins, Eileen (committee member), Romley, John (committee member)

Creator Email yingyingusc@gmail.com,zhuyingy@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-376833

Unique identifier UC11666565

Identifier etd-ZhuYingyin-9004.pdf (filename),usctheses-c89-376833 (legacy record id)

Legacy Identifier etd-ZhuYingyin-9004.pdf

Dmrecord 376833

Document Type Dissertation

Rights Zhu, Yingying

Type texts

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

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

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