University of Southern California Dissertations and Theses

Estimating survival in the face of pain: evidence from the health and retirement study

(USC Thesis Other)

Estimating survival in the face of pain: evidence from the health and retirement study

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content ESTIMATING SURVIVAL IN THE FACE OF PAIN:
EVIDENCE FROM THE HEALTH AND RETIREMENT STUDY
by
Gillian Fennell
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(GERONTOLOGY)
August 2024
Copyright 2024 Gillian Fennell

ii
Acknowledgements
I want to express my deepest gratitude for my mentors and collaborators who have gotten
me to this stage in my development as a researcher. Throughout my five years at USC, my
committee chair and co-advisor, Elizabeth Zelinski, has offered invaluable expertise, a safe place
to fail, and the motivation to continue on. I have always felt that we share a similar disposition
and approach to life, and because of that, along with their unwavering kindness, I regard her and
her husband, Jack, as a second family.
Jennifer Ailshire, another illustrious co-advisor, has pushed me in ways that have
undoubtably made me a far more thoughtful researcher. She prioritizes meaningfulness in her
brand of sophisticated, yet highly interpretable research. When I find myself emulating her
affinity for detailed descriptive work and effective scientific storytelling, I am very proud.
I am grateful for these two women, as well as Eileen Crimmins, Hanna Grol-Prokopczyk,
Anna Zajacova, Mireille Jacobson, and Corinna Löckenhoff for investing in my progress as
mentors and collaborators, and for being such strong role models. I am launching my career as a
far more decisive woman and academic in the wake of their influence.
Thank you, also, to Cary Reid, Elaine Wethington, and the folks at the Cornell
Translational Research Institute for Pain in Later Life for giving me a means of exploring my
interests in pain so early in my research career. I am still in awe of their commitment to my
success through all of the pitfalls and change-ups.
Last but not least, thank you to Calley Fisk, Eunyoung Choi, Theresa Andrasfay, Yeon
Jin Choi, and Hyungmin Cha, five brilliant biodemography post-doctoral fellows (some now new
professors!) who spent hours coaching me through new statistical methods with Expo markers in
hand and smiles on their faces.

iii
On a more personal note, there is no universe in which I would have made it across the
doctoral finish line without my family. Despite being 3,000 miles away, my mother memorized
the names and attributes of all the players in my life out west to stay informed and engaged in
my weekly rambles. She is insightful, passionate, and an exceptionally good listener. Not only
has she kept me sane, her financial safeguarding helped me successfully dodge the debt-related
pitfalls that often accompany life as an eternal student. She is my peace, my hero, and my
blueprint for bold and honest womanhood. Thank you, mom!
My father and sister brought a valuable sense of levity to these five years that I so
desperately needed. They reminded me of how much easier it is to keep your head down and
power through if you do so with humor and humility. Even when I fall, both of them still boast
on my behalf. Their support is unconditional and full; I am so incredibly grateful.
I have also had the privilege of leaning on many wonderful friends throughout this
journey, both old and new. To shout out a few: Lilly and Laura, thank you for prioritizing our
friendship over the years. Our histories are beautifully interwoven. From preteen petulance and
beyond, what a gift it is to continue to grow with both of you by my side. Stephanie, I feel like
I’ve known you for a lifetime. Thank you for picking me up when I’m down and lifting me
higher when I’m up; I couldn’t have asked for a better friend, hype woman, and couch mate.
Margarita, thank you for your exceptional kindness and keeping our one-of-a-kind cohort
connected. I am stunned by your coolness, growth mentality, and well-earned confidence. Qiao,
Erfei, and Rachel, thank you for the heart-to-hearts and laughs in and outside of the office. I love
you all.

iv
Table of Contents
Acknowledgements......................................................................................................................... ii
Abstract........................................................................................................................................... x
Chapter 1. Introduction ................................................................................................................... 1
Pain ............................................................................................................................................. 1
Subjective Survival Probabilities................................................................................................ 2
Pain & SSPs................................................................................................................................ 4
Joint Arthroplasty........................................................................................................................ 5
Joint Arthroplasty and SSPs ....................................................................................................... 6
Research Aims............................................................................................................................ 8
Chapter 2. Pain lowers subjective survival probabilities among middle-aged and ...................... 10
older adults.................................................................................................................................... 10
Introduction............................................................................................................................... 11
Methods..................................................................................................................................... 13
Data ....................................................................................................................................... 13
Measures............................................................................................................................... 13
Analytic Strategy .................................................................................................................. 16
Results....................................................................................................................................... 17
Discussion................................................................................................................................. 23
Chapter 3. A painful reality check? Examining the accuracy of subjective survival ................... 27
probabilities by pain interference and depression status............................................................... 27
Introduction............................................................................................................................... 28
Data ........................................................................................................................................... 31
Outcome Measure ................................................................................................................. 31
Independent Variables .......................................................................................................... 34
Analytic Strategy .................................................................................................................. 36
Results....................................................................................................................................... 38
Discussion................................................................................................................................. 50
Chapter 4. Identifying patient groups in total joint arthroplasty: A longitudinal study of ........... 54
need and outcome disparities........................................................................................................ 54
Introduction............................................................................................................................... 55
Data ........................................................................................................................................... 57
Measures................................................................................................................................... 59
Indicators of TJA need.......................................................................................................... 60
Outcome Measures................................................................................................................ 62

v
Analytic Strategy .................................................................................................................. 65
Results....................................................................................................................................... 66
Discussion................................................................................................................................. 79
Chapter 5. Conclusion................................................................................................................... 85
Pain and SSPs ........................................................................................................................... 85
Age-Related Susceptibility ....................................................................................................... 86
SSP Underestimation ................................................................................................................ 87
Total Joint Arthroplasty (TJA) & Longevity Expectations ...................................................... 88
Conclusion ................................................................................................................................ 89
Future Directions ...................................................................................................................... 90
References..................................................................................................................................... 92
Chapter 2 Supplementary Material ............................................................................................. 106
Chapter 3 Supplementary Material ............................................................................................. 112
Chapter 4 Supplementary Material ............................................................................................. 116

vi
List of Tables
Chapter 2
Table S1. Sample composition in the missing and analytic samples………………………….107
Table 1. Sociodemographic and health characteristics for HRS analytic sample
from their baseline interview between 2000-2018 in aggregate and stratified by pain
category………………………………………………………………………………………….18
Table 2. Pain coefficients for fully-adjusted fractional logit models estimating
subjective survival probabilities for each target age group…………………………………….21
Table S2A. Results from post-estimation tests testing the statistical significance
of pain coefficient changes model progression within each target age……………………….108
Table S2B. Results from post-estimation tests testing the statistical significance
of pain magnitude coefficient changes across the target age-stratified models………………108
Table S3. Pain coefficients for fully-adjusted fractional logit models estimating
SSPs stratified by gender for each target age group…………………………………………..109
Table S4. Results from post-estimation tests testing the statistical significance of
pain magnitude coefficient changes across gender stratified models………………………...110
Table S5. OLS regression using change in pain between two last waves in survey
predicting corresponding change in SSPs (N = 12,759)……………………………………..111
Chapter 3
Table 1. Sample characteristics in aggregate and stratified by pain and depression
status for HRS respondents…………………………………………………………………..40
Table 2. Linear regression predicting subjective survival probabilities for respondents
aged 57-89 (N = 12,835)……………………………………………………………………..43
Table 3. Cox proportional hazards regression predicting hazard ratios for mortality

vii
among respondents aged 57-89 in 2000 (N = 12,835)………………………………………46
Table 4. Multinomial logistic regression predicting accuracy of subjective survival
probabilities for respondents aged 57-89 (N = 12,835)……………………………………..48
Table S1. Multinomial logistic regression predicting accuracy of subjective
survival probabilities for respondents aged 57-89 stratified by target age
(N = 12,835)………………………………………………………………………………...112
Table S2. Multinomial logistic regression predicting accuracy of subjective
survival probabilities for respondents aged 57-89 (N = 12,835) using a more
conservative definition of "ambivalence" (i.e., SSPs of 30-70%)………………………….114
Chapter 4
Table S1. Sample characteristics of total analytic sample and respondents who
underwent TJA but were excluded due to too few observations…………………………..116
Table S2. Models estimating time-distributed fixed effects models of TJA on each
of the four outcomes by class and restricted to respondents who reported a TJA in
2010 or later………………………………………………………………………………..118
Table 1. Model fit statistics from three latent class analyses estimating one, two,
and three class solutions…………………………………………………………………….68
Table 2. Sample characteristics for HRS respondents who underwent total knee
or hip arthroplasty in aggregate and stratified by their need for the procedure…………….68
Table 3. Insurance type predicting membership into latent classes from the twoand three-class solutions……………………………………………………………………70
Table 4. Fixed effects models estimating within-individual, time-distributed effects
of TJA on pain intensity conducted separately for each latent class………………………72

viii
Table 5. Fixed effects models estimating within-individual, time-distributed effect
s of TJA on functional limitations conducted separately for each latent class…………….73
Table 6. Fixed effects models estimating within-individual, time-distributed effects
of TJA on depressive symptoms conducted separately for each latent class……………...75
Table 7. Fixed effects models estimating within-individual, time-distributed
effects of TJA on subjective survival probabilities conducted separately for
each latent class…………………………………………………………………………….76
Table S3. Models estimating time-distributed fixed effects models of TJA on
each of the four outcomes by class with a time (relative to TJA) x age
interaction term…………………………………………………………………………...119
Table S4. Sample characteristics for HRS respondents who underwent total
knee or hip arthroplasty by operation year……………………………………………….121

ix
List of Figures
Chapter 2
Figure 1. Mean predicted values of SSP by pain category in fully-adjusted models
stratified by target age………………………………………………………………………22
Chapter 3
Figure 1. Kaplan-Meier survival estimated by pain and depression………………………..47
Figure 2. Mean predicted values of SSP by pain interference and depression in a
fully adjusted model………………………………………………………………………...45
Chapter 4
Figure 1. Analytic sample selection………………………………………………………..59
Figure 2. Linear predictions of the time-distributed fixed effect of TJA receipt on
pain intensity by class……………………………………………………………………....77
Figure 3. Linear predictions of the time-distributed fixed effect of TJA receipt on
number of functional limitations by class………………………………………………….78
Figure 4. Linear predictions of the time-distributed fixed effect of TJA receipt on
number of depressive symptoms by class………………………………………………….78
Figure 5. Linear predictions of the time-distributed fixed effect of TJA receipt on
subjective survival probabilities by class…………………………………………………..79

x
Abstract
Background: The experience of severe or activity-limiting pain (i.e., high impact pain) is
increasingly common among middle-aged and older Americans. High impact pain is associated
with poor mental health, disability, and heightened mortality risk. It is not known whether high
impact pain is associated with subjective survival probabilities (SSPs), or one’s own perceived
chances of living to a given age, which meaningfully predict engagement in health-seeking
behaviors and responsible financial decisions.
Objective: I aim to test the link between high impact pain and SSPs, and assess whether reported
longevity expectations are accurate to respondents’ own lifespans. Additionally, I examine
whether the gold-standard treatment for osteoarthritis, total joint arthroplasty (TJA), increases
individuals’ SSPs.
Method: Across three studies, I use data from the Health and Retirement Study, a nationally
representative study of American adults aged 51 and older. In Study 1, I use age-stratified
fractional logit regressions on repeated cross-sectional data from 2000-2018 (N = 31,773; aged
51-89) to examine the relationship between pain and SSPs. In Study 2 (N = 12,835; aged 57-89),
I use a cox proportional hazards model and a multinomial logistic regression to assess both the
actual survival probabilities and accuracy of SSPs among respondents with pain and depression.
In Study 3 (N = 1,865 who underwent TJA; aged 51-89), I employ latent class analysis and fixed
effects models to examine the pain, physical functioning, depressive, and SSP outcomes
following a TJA procedure. These analyses are stratified by membership into high and low TJA
“need” classes.
Results: Respondents with high impact pain reported significantly lower average SSPs than
those with no or non-interfering pain. Individuals with pain or depression were more likely to

xi
underestimate SSPs relative to their actual lifespans. The SSPs of respondents with high impact
pain did not improve following TJA despite clear post-operative improvements in pain intensity,
physical functioning, and depressive symptoms.
Conclusion: Middle-aged and older adults with high impact pain have low and underestimated
longevity expectations that do not seem to be amenable to effective pain relief. I discuss the
possibility of a self-fulfilling prophecy, the consequences of underestimated SSPs, and highlight
alternative interventions to foster accurate SSPs among individuals experiencing pain.

1
Chapter 1. Introduction
Pain
Approximately 1 in 5 American adults experience chronic pain (Rikard et al., 2023), with
the majority of affected individuals being middle-aged (aged 45-64) and older (aged 65+; Zelaya
et al., 2020). As of 2019, pain prevalence among older adults was 30.8% (Zelaya et al., 2020).
This percentage is steadily increasing by 2-3% annually (Zimmer & Zajacova, 2018). In fact,
based on data from the National Health Interview Survey, the incidence rate of chronic pain
between 2019 and 2020 (52.4 cases per 1,000 person-years) was higher than that of diabetes,
depression, or hypertension (Nahin et al., 2023). Some scholars consider this sustained increase
in reported pain to be of epidemic proportions (Zimmer & Zajacova, 2018).
This “epidemic” is particularly concerning as pain has been linked to a number of poor
health outcomes. Relative to pain-free individuals, those with chronic pain report more disability
(Moore & Tumin, 2024), poorer sleep (Miettinen et al., 2022), and are more likely to experience
premature aging (Lahav et al., 2021). In fact, the inclusion of persistent pain as an indicator of
frailty—a state of comprehensive physical weakness and vulnerability—increased frail
respondents’ risk of death from a hazard ratio of 2.10 to 3.87 (Lohman et al., 2017).
Psychologically, pain is associated with cognitive decline (Horgas et al., 2022), anxiety,
depression (De La Rosa et al., 2024), and suicidal ideation (Petrosky et al., 2018). Relative to the
general population, co-occurring anxiety and depression is five times more common among
Americans experiencing chronic pain (De La Rosa et al., 2024). Although this association is
bidirectional, a 2021 longitudinal study found that individuals with chronic pain but no
depression at baseline were significantly more likely to develop depression over the following
four years relative to those with no pain (Glette et al., 2021).

2
Of course, pain varies in its intensity and impact on functioning, contributing to
differential effects on the physical, mental, and social health of the individual. As a subjective
experience, there are many different ways to measure and conceptualize pain. Currently, one of
the most well-used conceptualizations is “high impact chronic pain” (HICP), which is defined as
pain that often limits daily activities (Pitcher et al., 2019). In line with the age-specific pain
patterns cited above, high impact pain, which affects 6-7% of American adults, is more common
among middle-aged and older individuals: aged 45–64 (10.3%) and 65 and over (11.8%; Zelaya
et al., 2020).
Intuitively, high impact pain has more negative consequences than the experience of pain
itself. While the effect of pain alone on mortality risk is often sex-specific or nullified after the
inclusion of other health indicators (Roseen et al., 2021; Vela et al., 2021), high impact pain is
consistently associated with premature all-cause mortality in fully adjusted models (Glei &
Weinstein, 2023; Smith et al., 2014). With respect to mental health, individuals with HICP were
more likely to experience daily and weekly symptoms of depression, anxiety, fatigue, and
cognitive impairment than individuals who reported chronic pain without activity limitations
(Pitcher et al., 2019).
Subjective Survival Probabilities
One psychological measure that has gotten little attention in the pain field is subjective
survival probabilities (SSPs). SSPs are individuals’ perceived percent chances of living to a
given age (Hurd & McGarry, 2002). These longevity expectations are helpful as they predict
retirement expectations, individuals’ adherence to clinical recommendations, and mortality
(Khan et al., 2014; J.-H. Kim & Kim, 2017; Lu et al., 2022). With respect to retirement, older
adults who were one standard deviation more optimistic about their chances of living to age 85

3
than the mean were 24% more likely to expect to work into their 60s (Khan et al., 2014). The
effect of SSPs is stronger on these retirement intentions than on actual retirement behavior (Hurd
et al., 2004; Khan et al., 2014); in one study, only SSPs of 0% predicted a greater likelihood of
retiring and claiming social security early (Hurd et al., 2004).
Beyond intention, high SSPs underscore individuals’ likelihood to seek preventative
healthcare and adhere to prescribed treatments for chronic conditions (Biró, 2016). In a study of
middle-aged and older Chinese adults, SSPs partially mediated the relationships between good
self-perceived control of hypertension and three self-management behaviors: appropriate use of
medication, blood pressure monitoring, and physical activity (Lu et al., 2022). Individuals who
believe they still have long to live are also more inclined to seek colorectal (Kobayashi et al.,
2017) and breast cancer screenings (Wuebker, 2012).
SSPs are also touted as being fairly accurate predictors of mortality (J.-H. Kim & Kim,
2017; Papachristos et al., 2020), which is, in large part, due to individuals’ abilities to use their
personal health profiles to inform longevity expectations. For example, when older Dutch
workers were informed of new cardiovascular, psychological, and sleep disorder diagnoses, they
lowered their survival expectations (Vanajan & Gherdan, 2022). Similarly, Zacher et al. (2022)
found that older Americans reduced their SSPs following a hypertension diagnosis. Further,
individuals who smoke, classify as obese, practice poor nutrition, are depressed, have cognitive
impairment, or report poor self-rated health all report significantly lower subjective assessments
of survival (Falba & Busch, 2005; Kobayashi, Beeken, et al., 2017; Kumar et al., 2022;
Papachristos et al., 2020). This speaks to individuals’ capacity to understand the mortality risks
associated with certain health behaviors and profiles.

4
It is important to note that accuracy of reported SSPs is more important than them being
blindly optimistic. Having an appropriately low SSP can encourage behaviors that will ultimately
improve the quality of remaining life. For example, dialysis patients with lower SSPs preferred
treatments that would alleviate pain and improve quality of life, whereas those with higher
perceived life expectancies favored life-lengthening treatments that would often result in more
suffering (Beckwith et al., 2023). Additionally, lower SSPs increased the likelihood that
individuals would engage in advance care planning, thus increasing their odds of having their
end-of-life care needs met (Fleuren et al., 2021).
With respect to measurement, the Health and Retirement Study (HRS), a nationally
representative study of older Americans, asks about SSPs in the following manner: “What is the
percent chance that you will live to be x or more?” where x varies based on respondents’ age at
interview (Health and Retirement Study, 2023). Respondents are typically asked to think about
an age that is 11-15 years away; for example, a 72-year-old is asked about their chances of living
until age 85 or older. The exception is for individuals younger than 60 as the youngest “x” target
age is 75. In that case, individuals may be asked to think 16-24 years into the future as the
youngest eligible respondents in the HRS are 51.
Pain & SSPs
Pain is not widely considered life-limiting, but high impact, disabling pain does increase
mortality risk (Glei & Weinstein, 2023; Smith et al., 2014), especially in combination with
depression (Peele & Schnittker, 2022). This may explain why studies have suggested that new
onset or existing arthritis is not associated with reductions in SSPs (Vanajan et al., 2020;
Vanajan & Gherdan, 2022). However, these studies do not capture functional impact. Beyond
pain, there is evidence that individuals with work-related disability and limitations with activities

5
of daily living (ADLs; e.g., eating, toileting, bathing, etc.) report significantly lower subjective
assessments of survival. However, these SSP reductions may be exaggerated. Papachristos &
Verropoulou (2020) report that, relative to measures of “objective survival probabilities”,
individuals with ADL limitations are more likely to underestimate their life expectancies.
Joint Arthroplasty
The significant increases in pain prevalence described earlier are occurring largely
because of increases in osteoarthritic joint pain (Zajacova et al., 2021)—one of most common
pain complaints among older adults (Dagnino & Campos, 2022). Osteoarthritis (OA) is a
degenerative disease that erodes cartilage, stimulates osteophyte growth (i.e. bony growths on
the affected joint), and causes pain and stiffness (Osteoarthritis, 2023). Mild to moderate OA can
be treated through maintaining a healthy weight, taking pain medications, or getting regular
corticosteroid injections (Ayhan et al., 2014). However, once most of the cartilage is worn away
and the patient is suffering from severe pain and limited range of motion, non-surgical treatments
are no longer effective (Hsu & Siwiec, 2023). Total joint arthroplasty (TJA; i.e., joint
replacement) is the gold standard treatment for moderate to severe OA (Drummer et al., 2021).
In the months following TJA, patients report significant improvements in pain, physical
functioning, and health-related quality of life (Lützner et al., 2019; Núñez et al., 2007; Perneger
et al., 2019; Woodland et al., 2023). Compared to pre-operative measurements, pain and physical
function improved by more than a standard deviation one year following a total knee or hip
arthroplasty (TKA and THA; Perneger et al., 2019). Another study on TKA patients found that,
even within an older sample (70+) from whom we might expect age-related declines in mobility,
there were significant post-operative improvements in pain and physical functioning (12 and 8%
improvement from pre-operative assessment, respectively; Wang et al., 2023).

6
Studies examining post-operative trajectories several years after THA and TKAs report
similar findings. For example, patients’ pain and mobility levels remained significantly better
than pre-operative measurements 4 and 6 years following TKA (Sloan et al., 2013), and 7 years
following THA (Shan et al., 2014). In fact, the benefits of TJA are significant enough that
procedures have been found to reduce mortality risk in patients for more than a decade following
surgery (Maradit Kremers et al., 2016).
Although the vast majority of patients are satisfied with their TJA outcomes, not
everyone who undergoes a TJA sees post-operative benefit. Seven to 23% and 10 to 34% of
THA and TKA, respectively, report unfavorable outcomes (Beckwith et al., 2023). While there is
still an ongoing discussion about why this non-trivial minority of dissatisfied patients exists,
some prior work has suggested that there are pre-requisite indicators of TJA need that must be
met for the procedure to have the desired benefit. Several studies have found that individuals
who report low levels of pain and/or high levels of physical functioning pre-operatively show
little to no improvement post-TJA (Berliner et al., 2017; Hawker et al., 2023; Kornilov et al.,
2018).
Joint Arthroplasty and SSPs
Based on this work, identification of high impact pain (in addition to other diagnostic
factors) could help us delineate who would and would not benefit from TJA using large,
nationally representative survey data. After successfully identifying appropriate versus
inappropriate TJA candidates who ultimately underwent TJA, we can test if expected reductions
in pain and functional limitations coincide with increases in subjective assessments of survival.
As previously mentioned, prior work has examined the effect of negative health shocks
on SSPs. However, to my knowledge, no work has longitudinally assessed SSPs pre- and post-

7
receipt of an effective health treatment. As SSPs and like constructs (e.g., future time
perspective) have been shown to be amenable to emotional regulation and habit-forming
interventions (Gellert et al., 2012; Sakakibara & Ishii, 2020), one of the goals of this dissertation
work is to ascertain if SSPs are not only amenable to these psychological interventions, but also
to physical intervention. More specifically, I examine whether effective pain treatment through
TKA or THA is associated with corresponding increases in longevity expectations. This aim is
exploratory because there is mixed evidence on whether TJA receipt improves mental health, to
which SSPs are closely related (Bae et al., 2017). Papakostidou et al. (2012) and Vajapey et al.
(2021) reported post-TJA alleviation of depressive symptoms, whereas other studies reported no
different or worsened post-operative mental health at a one-year follow-up (Perneger et al., 2019;
Shan et al., 2014).
With longitudinal data on pain, functioning, depression, SSPs, and TJA receipt, I am well
positioned to investigate this exploratory aim. For this investigation, the fact that TJAs are
discrete events that, for the vast majority of people, offer dramatic reductions in pain and
significant improvements in quality of life is a distinct strength. In order to modulate SSPs
through psychological intervention, individuals must have a degree of psychological openness
that may better prime SSPs for improvement. Unlike engagement in these psychological
interventions, referral and willingness to undergo TJA is far less dependent on psychological
openness and more so on clinical presentation of OA (O’Connor et al., 2022). By virtue of
patients not having to possess this level of psychological openness, improving SSPs may require
more significant change in their physical health, which TJA has the potential to provide.
Additionally, through latent class analysis, I can isolate those who are the most appropriate

8
candidates for the procedure, thus clarifying any potential relationship between receipt of TJA
and improved longevity expectations.
Research Aims
This dissertation seeks to further explore the connections between high impact pain and
subjective survival probabilities. In light of the aforementioned markers of poor health that both
co-occur with high impact pain and inform subjective assessments of survival, Study 1 sets out to
officially establish the negative relationship between SSPs and pain that interferes with daily
activities using a repeated cross-section of middle-aged and older adults from the HRS.
Study 2 draws on the established links between high impact pain, depression, and
mortality. I seek to replicate findings suggesting that individuals with both high impact pain and
depression are at risk for premature mortality (Glei & Weinstein, 2023; Peele & Schnittker,
2022). Further, I use SSP and mortality data to examine whether middle-aged and older adults
with pain and depression significantly underestimate their longevity. While this hypothesized
underestimation has been reported in prior work, comparisons were made between SSPs and
“objective survival probabilities” derived from population life tables, not respondent-specific
mortality data (Papachristos & Verropoulou, 2020).
Lastly, Study 3 will explore longitudinal, within-individual effects of TJA on pain
intensity, number of functional limitations, number of depressive symptoms, and subjective
survival probabilities. Analyses for this study will be conducted separately for groups of
individuals with disparate pre-operative levels of TJA need. These groups will be identified
through latent class analysis informed by several indicators of pain and mobility issues in the
wave prior to reportedly receiving a TJA. I expect that individuals with a high need for TJA will
report significant and lasting post-operative improvements in pain, functioning, depression, and

9
SSPs. Individuals with low indicated TJA need will likely see no post-operative improvements in
any of the four outcomes.
If the hypotheses from these three studies are supported, this work will have highlighted
another damaging consequence of high impact pain that makes affected middle-aged and older
individuals more susceptible to making ill-informed health and financial decisions based on
inaccurate and underestimated longevity expectations. If total joint arthroplasty for knee and hip
OA improves in SSPs in response to discrete and immediate improvements in pain and
functioning, we may be more optimistic that individuals with chronic pain are not making
needlessly pessimistic decisions about their retirement and health futures as more pain
management studies gain efficacy.

10
Chapter 2. Pain lowers subjective survival probabilities among middle-aged and
older adults
Objective: Pain is a leading cause of disability and a limiting factor in individuals’ assessments
of their own subjective health, however its association with subjective longevity has yet to be
explored. Subjective survival probabilities (SSPs), or one’s own perceived chances of living to a
given age, can influence individuals’ behavior as they plan for their futures. This study assesses
whether pain correlates to lower SSPs.
Methods: We use repeated cross-sections of the 2000-2018 waves of the Health and Retirement
Study, a longitudinal and nationally representative survey of Americans aged 51 and older
(N=31,773). Age-stratified fractional logit regressions were estimated to examine the
relationship between pain and SSPs; fractional logit models were specifically designed to
estimate bounded probability outcomes, like SSPs.
Results: Across all age groups, respondents with severe and/or interfering pain reported
significantly lower SSPs than those with no pain (Marginal Effect (ME) = -0.03 to -0.06, p <
.05). Controlling for all covariates, mild or moderate non-interfering pain was only associated
with a significant reduction in SSPs among the youngest group reporting their chances of living
to age 75 (ME = -0.02, p < .001). Descriptively and in the model results, respondents with mild
or moderate non-interfering pain appeared to more closely resemble pain-free respondents than
those with severe or interfering pain.
Discussion: These findings highlight the importance of pain on SSPs, and contribute to the
growing evidence that pain interference is uniquely important in predicting health outcomes.

11
Introduction
Subjective survival probabilities (SSP) measure individuals’ perceived chances of living
to a given age. SSPs have important implications for individuals’ prioritization of their health
and financial futures. For example, people with higher SSPs, who are more optimistic about their
anticipated longevity, tend to retire at older ages (Khan et al., 2014), and invest more in their
retirement (Doerr, & Schulte, 2012; O’Brien, C. et al., 2005). Further, lower SSPs, signifying
pessimistic perceptions of longevity, are associated with reduced likelihoods of seeking
preventative healthcare and cancer screenings (Biró, 2016; Picone et al., 2004). In the context of
Covid-19, a recent study suggested that people with higher SSPs were more likely to heed public
health precautions such as social distancing, and were less likely to refuse medical treatments for
the virus (Celidoni et al., 2022). Although prior evidence suggests SSPs shape behaviors and
health, we still have a limited understanding of the factors that contribute to SSPs.
In this study, we focus on the role of pain in shaping SSPs as pain is of growing concern
among U.S. adults (Zimmer & Zajacova, 2018). Chronic pain affects about 20% of American
adults, and its prevalence is even higher (31%) among adults aged 65 and older (Zelaya et al.,
2020). In addition to overall pain prevalence, it is important to identify which pain profiles may
be associated with outcomes like SSPs. As in prior work, we incorporate an indicator of whether
pain interferes with daily activities (pain interference) into a traditional measure of pain intensity
(whether pain is mild, moderate, or severe) because assessments of functional disability are
strong predictors of physical and psychological health outcomes beyond the experience of pain
alone (Adams et al., 2018; Mutubuki et al., 2020). The integration of pain interference into our
assessment of pain is particularly important in the present study as middle-aged (45-64) and
older Americans (65+) report the highest rates of interfering pain (Przekop et al., 2015; Zelaya et
al., 2020).

12
As a leading cause of disability and a strong correlate of depression (GBD 2016 Disease
and Injury Incidence and Prevalence Collaborators, 2017), pain often comes with pessimistic
beliefs about one’s future, launching an unfortunate cycle of poor physical and emotional
wellbeing (Coyle, L.A & Atkinson, S., 2018; Duenas et al., 2016; Pincus & McCracken, 2013).
When asked to simply imagine experiencing chronic pain, one study found that older adults’
desire to live to advanced ages was significantly reduced (Skirbekk et al., 2021). Beyond desire
to live, interfering pain is associated with increased mortality risk (Glei & Weinstein, 2023;
Smith et al., 2014; Vartiainen et al., 2022). This work offers a basis for our expectation that pain
that limits—or is likely to limit—daily activities will contribute significantly to our estimation of
SSPs in middle-aged and older adults.
Although pain has clear consequences for expectant and realized futures, no studies have
investigated the relationship between pain and SSPs. We use the Health and Retirement Study
(HRS) to better understand this unexplored relationship. The HRS is the only large, nationally
representative survey that includes information on SSPs, pain intensity, and pain interference.
Low SSPs are important barometers for poor health and wellbeing, illustrated by documented
associations with disability (Mirowsky & Ross, 2000), poor self-rated health (Kim & Kim, 2017;
Lee et al., 2012), and higher mortality risk (Hurd & McGarry, 2002; Kim & Kim, 2017; Siegel et
al., 2003). Based on prior research, we expect that individuals with pain will report lower SSPs
relative to respondents without pain. We also expect individuals reporting severe and / or
activity-interfering pain to report the lowest SSPs. If our expectations are supported, this may
raise concerns about lifestyle decisions of adults experiencing pain; estimations of abbreviated
lifespans may contribute to maladaptive health behaviors, and thus poorer health, as individuals
enter advanced ages.

13
Methods
Data
We used data from the Health and Retirement Study (HRS; Health and Retirement Study,
2022). The HRS is a publicly-available, longitudinal survey that is nationally representative of
older adults in the United States. We used ten recent waves of biennial data (2000-2018) in a
repeated cross-sectional approach. We did not include data from 2020 because of possible
Covid-19 effects on both pain and SSPs (Kinugasa et al., 2023).
We limited our analytic sample to respondents who were asked about SSPs: 34,010
community-dwelling, non-proxy respondents aged 51-89. We further excluded individuals with
missing data on SSPs (n = 2,142), sociodemographic characteristics (n = 65), pain (n = 19), or
depression (n = 12). Our final sample consisted of 31,773 unique respondents with a total of
153,282 observations. The number of observations per respondent ranged from one to ten with
an average of 4.82 (SD = 2.9). Respondents were missing on the SSP variable if they refused to
answer, indicated that they “didn’t know,” or were unable to answer prior probability questions
requiring an answer between 0 and 100%. Respondents missing on SSP or model covariates
were significantly more likely to be older, male, less educated, non-white, and widowed, as well
as have greater relative likelihood of depression, heart disease, lung disease, diabetes, high blood
pressure, arthritis, and stroke (Table S1).
Measures
Subjective Survival Probabilities Questions about subjective survival probabilities were asked of
respondents aged 51 to 89. Respondents were asked “What is the percent chance that you will
live to be x or more?” where x varied based on the respondent’s age: x = 75 for those aged 51 to
64, x = 80 for those 65 to 69, x = 85 for those aged 70 to 74, x = 90 for those 75-79, x = 95 for
those 80-84, and x = 100 for those 85-89. It is important to note that respondents aged 51-59

14
were required to think about the probability of living another 16-24 years, which is a longer
duration than that for all other age groups. To answer this question, respondents provided a
probability between 0% (completely unlikely to reach age x) and 100% (very likely to reach age
x). The SSP variable was transformed to range between 0 and 1 in order to meet fractional logit
model assumptions.
Pain Respondents were asked “Are you often troubled with pain?” (yes/no). If a respondent said
yes, they were then asked about their pain intensity and interference with daily activities. For the
question of pain intensity, respondents were asked “How bad is the pain most of the time: mild,
moderate or severe?” They then indicated whether their pain “makes it difficult for [them] to do
[their] usual activities such as household chores or work?” (i.e., pain interference; yes/no). Prior
work has supported the use of pain measures that integrate both pain intensity and interference
as, together, they can better predict clinical and psychological outcomes (Deyo et al., 2014; Von
Korff et al., 2020). Mirroring Zimmer & Zajacova (2018) which also analyzed HRS data, we
categorized respondents into the following groups: (0) no pain, (1) mild or moderate pain that
does not limit daily activities, and (2) severe pain and / or interfering pain. This categorization is
clinically relevant as it separates people whose pain likely requires consistent treatment or
intervention from those whose pain may be managed effectively outside of a clinical setting
(Nahin, 2015; Zimmer & Zajacova, 2018). It is important to note that we cannot assess the
duration of the pain reported using cross-sectional data in the HRS, thus respondents’ reported
pain in this sample may be acute or chronic.
Demographic & social status variables Age, gender, race, educational attainment, and marital
status were included as controls in the analysis. Gender was coded as a binary variable (male or
female). Race/ethnicity was categorized as self-identified Non-Hispanic White, Non-Hispanic

15
Black, Hispanic, and Non-Hispanic Other. Educational attainment was coded as (1) less than
high school, (2) completed high school, (3) some college education, and (4) college or above.
Lastly, marital status was categorized into married/partnered, separated/divorced, widowed, or
never married.
Depression We included an indicator of depression in our analyses because of the established
positive association with pain and negative relationship with SSPs (Chen et al., 2021; Duenas et
al., 2016; Von Korff & Simon, 1996). The eight-item Center for Epidemiologic Studies
Depression (CESD-8) scale was used. CESD-8 asks respondents if they had experienced any of
eight symptoms in the past week. The items reflect respondents’ mood, motivation, energy level,
sleep quality, and life satisfaction. Higher scores indicate more depressive symptoms. We
dichotomized respondents as having symptoms indicative of depression (CESD-8 ≥3) or not
(CESD-8 score < 3) following the established threshold for determining depression with the
CESD-8 (Dang et al., 2020). As there may be concerns over the dichotomization of this measure
for reasons related to a loss of information and biased estimates, we additionally conducted
analyses using the continuous CESD-8 measure; the pain-SSP relationship remained nearly
identical.
Health Conditions Respondents reported whether a doctor had ever told them that they had heart
disease, lung disease, diabetes, high blood pressure, arthritis, cancer, or a stroke. Heart disease
included “heart attack, coronary heart disease, angina, congestive heart failure, or other heart
problems.” Lung disease included “chronic lung disease such as chronic bronchitis or
emphysema.” Cancer included “malignant tumor, excluding minor skin cancer.” The seven
conditions were analyzed separately as binary variables (1=had condition / 0=did not have
condition). We controlled for these diagnosed chronic conditions because they inform reductions

16
in individuals’ assessments of subjective survival, and could potentially confound the
relationship between pain and SSPs (Chen et al., 2021; Vanajan et al., 2020; Vanajan &
Gherdan, 2022; Zacher et al., 2022).
Analytic Strategy
We first summarized the baseline sociodemographic and health characteristics of the full
sample and also by pain category. “Baseline” refers to each respondent’s first observation in the
survey. The reference group is respondents with mild or moderate non-interfering pain in order
to better understand the intermediate differences across pain categories, not just relative to
individuals without pain. We tested differences in continuous variables with t-tests, and
differences in categorical variables with chi-square tests.
Then we estimated fractional logit regression models to examine the association between
pain and SSPs. Fractional logit models were selected because they are optimal for outcomes that
are bounded between 0 and 1 (Villadsen & Wulff, 2021). A three-stage model progression was
conducted within each target age group. In Model 1, we controlled for age, gender, race,
educational attainment, and marital status to investigate the relationship between pain and SSPs
with adjustments for basic sociodemographic characteristics. In Model 2, we further adjusted for
depressive symptoms to evaluate whether the association between pain and SSPs is due to painrelated psychological distress. Finally, we adjusted for other major chronic health conditions in
Model 3 to examine whether significant findings reflect the effects of underlying chronic
conditions that drive both pain and SSPs. These models were conducted separately for each
target age because the relationship between pain and survival probabilities may differ as
respondents estimate chances of survival to increasingly exceptional ages. Fractional logit model
results are originally provided in log odds, but we calculated average marginal effects (ME) for

17
each predictor variable so that we could interpret coefficients on the same scale as the SSP
outcome measure.
Results
Table 1 presents sociodemographic and health characteristics by pain category. Relative
to respondents with mild or moderate non-interfering pain, pain-free individuals were
significantly older, more likely to be black, highly educated, widowed, and have lower rates of
the following five conditions: depression, lung disease, diabetes, high blood pressure, and
arthritis. Relative to the same group with non-interfering lower intensity pain, respondents with
severe or interfering pain were more likely to be female, black or Hispanic, and more likely to
report depression, heart disease, lung disease, diabetes, high blood pressure, arthritis, stroke,
and/or cancer. Individuals with the worst pain also reported significantly lower levels of
education and were less likely to be married. Individuals with no pain reported the highest SSPs
(M = 60.7, SD = 31.1), followed by those with low intensity, non-interfering pain (M = 56.6, SD
= 31.4), and respondents with severe or interfering pain (M = 47.6, SD = 34.0)

18

19

20
Table 2 shows the results of the stratified fractional logit regressions on SSPs by target
age. A three-model progression was conducted over six target ages with similar results. For
brevity, we present pain-related results for the x = 75 models and describe differences across the
other target age stratified models. Model 1 controls for age, gender, race, educational attainment,
and marital status. Compared to those with no pain, SSPs were 3 and 12 percentage points lower
among respondents with mild or moderate non-interfering pain (Marginal Effect (ME) = -0.03, p
< .001) and severe or interfering pain (ME = -0.12, p < .001), respectively. The pain-SSP
relationship remained significant in Model 2, which additionally controlled for depression: mild
or moderate non-interfering pain (ME = -0.03, p < .001) and severe or interfering pain (ME = -
0.09 , p < .001). Lastly, in Model 3, which controls for chronic health conditions, both mild or
moderate non-interfering (ME = -0.02, p < .001) and interfering or severe pain (b = -0.06, p <
.001) were still significantly negatively associated with SSPs. In this fully adjusted model, mild
or moderate non-interfering pain was associated with a 2 percentage point decrease in SSPs, and
severe or interfering pain corresponded with a 6 percentage point decrease relative to no pain.
Across all other target ages, mild or moderate non-interfering pain was no longer associated with
a significant reduction in SSPs in Model 3, but severe or interfering pain remained a significant
negative predictor across all target ages. For respondents aged 84-89 who reported probabilities
of living to 100, mild or moderate non-interfering pain was not significantly associated with
SSPs across the entire three-model progression.
We conducted “Seemingly Unrelated Estimation” (SUE) tests on the fractional logit
models to assess whether the effect of pain on SSPs was significantly attenuated across the three
models within each target age (Table S2A). SUE tests evaluate whether coefficients of the

21

22
variable are statistically different across separate models (Mize et al., 2019). In this case, the
coefficients summarizing the relationship between SSPs and pain were significantly reduced
following the addition of each new block of covariates across all three models. We additionally
tested whether fully-adjusted results from each Model 3 differed significantly by target age
(Table S2B). There were no significant differences in fully-adjusted results between adjacent
target ages. Figure 1 shows the mean predicted value of SSPs at each pain level across target
ages. These values are predicted from fully-adjusted models where all covariates were held at
their means. This figure provides a condensed visual summary of our fractional logit model
results with pairwise comparisons between pain categories within each model. With the
exception of x = 100, severe or interfering pain corresponded with significantly lower mean
predicted SSPs than mild or moderate non-interfering pain across all target age stratified models.
Figure 1. Mean predicted values of SSP by pain category in fully-adjusted models stratified by target age
Additional supplemental analyses examine whether the relationship between pain and
SSPs differs significantly by gender, and whether change in pain predicts change in SSPs.

23
Detailed descriptions of these analyses and associated findings are located in the Supplementary
Material. With respect to the analysis by gender (Table S3), we found that the relationship
between pain and SSPs did not differ significantly between men and women (Table S4, p > .10).
Moving beyond a cross-sectional design in an effort to establish explanatory plausibility (Table
S5), we found that when pain decreased between two adjacent waves, SSPs increased
marginally, but non-significantly in a fully adjusted model (b = 1.30, p > .06).
All results were weighted to correct for nonresponse bias, complex survey design, and
poststratification adjustments to make the sample representative of the US population. For all
analyses, we report cluster-adjusted robust standard errors to account for correlation of multiple
within-person observations. We conducted our analysis using Stata 17. Multicollinearity checks
(Studenmund, & Johnson, 2017) indicated no concerns in the fully-adjusted models, with no
values exceeding the common VIF threshold of 5.0.
Discussion
Pain is a common debilitating health stressor that often limits activity, contributes to
emotional distress, and is associated with increased risk of mortality (GBD 2016 Disease and
Injury Incidence and Prevalence Collaborators, 2017; Pincus & McCracken, 2013; Torrance et
al., 2010), which may have implications for individual subjective assessments of survival. Our
study found that respondents with pain reported significantly lower SSPs than respondents with
no pain. Beyond descriptive work, severe or activity interfering pain was associated with
significantly lower subjective assessments of survival across the age range, even after controlling
for likely confounders. The group with severe or interfering pain was clearly distinct from those
experiencing non-interfering lower intensity pain as this less bothersome pain was only
associated with significantly lower SSPs among the youngest respondents with a target age of

24
75. Based on post-estimation analyses, respondents with lower intensity pain seemed to more
closely resemble pain-free individuals than those reporting more severe pain.
Our study builds on a growing literature that emphasizes the negative impact of painrelated activity interference on wellbeing and health-related quality of life (Adams et al., 2018;
Mutubuki et al., 2020). The finding that severe or interfering pain is associated with a reduction
in SSPs raises additional concerns as pain prevalence continues to increase in the US, especially
among middle-aged and older adults. With shorter expected lifespans, individuals with severe
and activity-interfering pain have all the more reason to forgo healthy and health-seeking
behaviors (e.g., seeing a doctor regularly, taking appropriate medications, eating well, exercising,
etc.), which may prove detrimental to their actual life expectancy (Sansbury et al., 2014;
Sarkisian et al., 2002). To mitigate these consequences of lower SSPs for individuals with pain,
clinicians may consider implementing goal-setting and emotion-focused coping practices that
have shown to successfully lengthen individuals’ perceptions of future time; these interventions
focus specifically on helping the patient accept and adapt to their pain (Arends et al., 2013;
Gellert et al., 2012; Sakakibara & Ishii, 2020).
Beyond the most disruptive pain, our findings for individuals with lower intensity noninterfering pain were consistent with work that suggests that middle-aged and older individuals
with non-interfering pain were descriptively more similar to a pain-free group than to those with
pain interference (Jordan et al., 2012). More specifically, the authors found that individuals who
were able to minimize the impact of pain on their lives, regardless of pain intensity, reported
fewer depressive symptoms and co-morbidities than those with interfering pain. Of course, not
everyone has an equal opportunity to control the impact of their pain: socioeconomic
disadvantage and older age are also significant correlates of interfering pain. Additionally, non-

25
interfering pain is a leading risk factor for future interfering pain, so some respondents may
simply be early in their trajectory to more severe pain (Jordan et al., 2008). However,
interventions that allow people to manage their pain to the extent that it does not interfere with
their daily activities help them retain more positive, expansive perceptions of the future, invest in
preventative care and health-promoting behaviors, and realize health outcomes that are more
similar to those unencumbered with pain.
Our findings should also be considered in light of two major data limitations. First, the
HRS lacks questions about pain duration, quality, and location that may have meaningful
additional associations with SSPs. For example, we could not consider whether reported pain
was acute or chronic, which could have had significant impacts on respondents’ SSP reporting.
There are mixed findings about the effect of pain duration on wellbeing and health outcomes:
some suggest that longer pain duration is associated with greater psychological adaptation to
pain, and thus more expansive perceptions of future time (Fennell et al., 2021; Khanom et al.,
2020), whereas others report that longer duration is correlated with worsened pain intensity and
interference, which translate to poorer psychological and physical functioning (Davison et al.,
2016; Peters et al., 2000). Additional research could help provide insight to this field of study.
The second major limitation is that the SSP question in the HRS is not asked among
respondents older than 89, or to those who had not provided answers to the first three probability
items in the Expectations section of the questionnaire. Therefore, the oldest old and others were
excluded from providing SSPs, leading to a sample that was younger and more educated than the
total representative sample recruited for the HRS. While our analytic sample was not
representative of the total older U.S. adult population, these systematic exclusions on the SSP
question generated more interpretable data.

26
There is ample opportunity for additional research to contribute to our understanding of
longevity expectations among adults with pain. Future work should replicate our findings using a
more robust pain assessment accounting for pain location, quality, and duration, and build on
these findings using longitudinal designs. Longitudinal studies may, first, assess how long-term
pain trajectories contribute to potential oscillations in SSP reporting. Second, longitudinal or
cross-sectional designs can be used to elucidate which additional aspects of pain (e.g. pain
presence, duration, etc.) contribute most to lower SSPs.
In conclusion, the present study highlights the significant negative association between
severe/interfering pain and subjective survival probabilities. The adjusted mean SSP for
respondents with lower intensity, non-interfering pain was significantly higher than that for
severe or interfering pain, but counter to expectations, did not significantly vary from that of the
pain-free group. These findings held for both men and women. The relationship between pain
interference and SSPs was not significant at an alpha level of .05 in a two-wave longitudinal
model, but decreased pain was marginally associated with higher SSPs at p < .10. We present
this study as further evidence that severe or interfering pain is an important predictor of lifelimiting health consequences. We urge further investigation of SSPs among middle-aged and
older adults with pain as these individuals may be at risk for making lifestyle choices that either
do not serve them adequately for the remainder of their lives or have real life-limiting
consequences.

27
Chapter 3. A painful reality check? Examining the accuracy of subjective survival
probabilities by pain interference and depression status
Background & Objective: Both pain and depression are independently associated with
heightened mortality risk and lower subjective survival probabilities (SSPs) to age-specific target
ages. We examine whether SSPs for individuals with pain and depression are accurate to their
actual lifespans.
Methods: Using data on 12,835 Health and Retirement Study respondents aged 57-89 in 2000
with follow-up mortality data through 2018, we determined whether respondents provided SSPs
that were “correct,” “underestimated,” or “overestimated” relative to their own lifespans. We
estimated a multinomial logistic regression predicting this four-category measure of SSP
accuracy from pain interference, depression, and their interaction while controlling for
sociodemographic and health-related characteristics.
Results: Severe or interfering pain (i.e., high impact pain) was associated with a 28% higher risk
of underestimating SSPs (RRR=1.28, p=.04), and individuals with depression had a 50% higher
risk of underestimating (RRR=1.50, p<.001). Reporting the experience of both high impact pain
and depression did not significantly modulate the risk of any reporting outcome. Additional
analyses supported that being doubly encumbered by both high impact pain and depression
corresponded with lower average SSPs, and that both baseline depression and pain were
associated with heightened mortality risk in our analytic sample.
Conclusion: High impact pain and depression were independently associated with higher risk of
underestimating longevity expectations. Future research should assess whether the observed
relationships between pain, depression, and SSP underestimation contributes to health and
financial decisions that interfere with older adults’ abilities to live well.

28
Introduction
The prevalence of chronic pain is increasing (Zimmer & Zajacova, 2018), as is the
evidence for its harmful effects on mental and physical health. Pain has well-established
associations with depression (Coyle, L.A & Atkinson, S., 2018; Duenas et al., 2016; Pincus &
McCracken, 2013) and mortality risk (Grol-Prokopczyk, 2017; Smith et al., 2014; Vartiainen et
al., 2022). More specifically, knee osteoarthritis, musculoskeletal, and back pain—among the
most common types of chronic pain—have been shown to predict all-cause mortality even after
controlling for potential confounders (Cleveland et al., 2019; Macfarlane et al., 2017; Roseen et
al., 2021). While there are some published exceptions after accounting for pain severity or
stratifying by gender (Roseen et al., 2021), the positive relationship between pain and mortality
is largely stable.
These findings are consistent with the prior chapter establishing that pain, and
particularly pain that limits or is likely to limit daily activities, is associated with reductions in
individuals’ subjectively perceived chances of living to advanced ages, or subjective survival
probabilities (SSPs; Fennell et al., 2024). Intuitively, SSPs are negatively associated with
mortality risk and, in fact, offer longevity predictions similar to those of population life tables
(Bago d’Uva et al., 2020; Elder, 2013; Mirowsky, 1999; O’Connell, 2011; Perozek, 2008).
Beyond their utility as a measure of mortality risk, individuals’ estimations of their own
longevity matter for decisions that may ultimately lengthen or abbreviate their lives or influence
their ability to live well. For example, low SSPs are associated with a lower likelihood of seeking
preventative healthcare (e.g., cancer screenings) and abiding by public health recommendations
(Biró, 2016; Celidoni et al., 2022; Picone et al., 2004). Additionally, individuals with low SSPs

29
tend to retire and claim Social Security benefits earlier, and are less likely to invest in their
retirements (Doerr, & Schulte, 2012; Hurd et al., 2004; O’Brien, C. et al., 2005).
In addition to pain, depression—a common co-morbidity—is also positively associated
with mortality risk and has been shown to modulate subjective assessments of survival. A 2019
meta-analysis of 61 studies found that depression among older adults was positively predictive of
both all-cause and cardiovascular mortality (Wei et al., 2019). Separate from this research
establishing depression as a risk factor for mortality, previous work also supports a negative
association between depression and SSPs (Chen et al., 2021; Fennell et al., 2024; Palgi et al.,
2019; Papachristos & Verropoulou, 2020). In fact, not only do older adults with depression seem
to report lower average SSPs than those without, but their SSPs are lower than longevity
expectations calculated based on their demographic and health profiles (Papachristos et al.,
2020). This evidence from Europe suggests that individuals with depression may be more prone
to underestimating their longevity (Papachristos et al., 2020; Papachristos & Verropoulou, 2020).
We suspect the same to be true in our US sample, not only for respondents with depression, but
also among those experiencing pain.
The experience of chronic pain in the US is increasingly common, impacting 20% of the
American adult population, including 31% of those aged 65 and older (Zelaya et al., 2020).
While chronic pain is a leading cause of disability and a positive correlate of mortality (GBD
2016 Disease and Injury Incidence and Prevalence Collaborators, 2017; Grol-Prokopczyk, 2017),
pain alone is rarely reported as a direct cause of death (Tennant, 2012). The unfortunate
possibility of suicide notwithstanding (Van Orden & Conwell, 2011), individuals with pain and
depression—who are otherwise unencumbered by additional life-threatening morbidities—may
live reasonably long lives (Zimmer & Rubin, 2016). However, the expectation of an abbreviated

30
life may actualize into shorter lifespans, or contribute to poor health through lower engagement
in preventative healthcare (Biró, 2016; Celidoni et al., 2022). Older adults with pain who report
limited perceptions of the future, or in this case, underestimated SSPs, are at risk for leading
lives that will contribute to more years spent with disability and fewer financial resources to
obtain proper care (Doerr, & Schulte, 2012; Hurd et al., 2004; Laditka & Laditka, 2017; O’Brien,
C. et al., 2005). Our study examines the accuracy of SSPs among older adults with chronic pain
to assess the validity of this large subpopulation’s often pessimistic future perceptions, and
provide a basis for further work to improve chronic pain sufferers’ abilities to plan for the future.
We assess the simultaneous effects of pain and depression because the concurrent
experience of both may have an even stronger negative influence on SSPs than would be
expected from both factors independently. Ongoing pain may be a specific stressor about which
depressed individuals ruminate and catastrophize (Quenstedt et al., 2021). This argument is
exemplified by a recent study showing that older adults reported a significantly reduced desire to
live to advanced ages when asked to imagine a hypothetical future with chronic pain (Skirbekk et
al., 2021). It is possible that, in using their health to inform their estimations of their own
longevity (Chen et al., 2021), individuals with pain and depression may estimate SSPs that are
low, but accurate to their own lifespans. However, given the associations between pain,
depression, and negative future perceptions, we suspect that individuals with both interfering
pain and depression will be more likely to underestimate their lifespans than provide accurate or
over-estimations.
To test our central hypothesis, we benefit from a longitudinal dataset that has both
subjective assessments of longevity along with follow-up mortality data for determining lifespan.
With these data, we can compare respondents’ SSPs to their own lifespans, unlike previous

31
studies that were limited by a lack of mortality follow-up data (Papachristos et al., 2020). To
provide a more holistic picture of the interplay between pain, depression, SSPs, and mortality
prior to testing the central hypothesis, we further examine both SSP reports and actual survival
probabilities among individuals who vary by pain and depression status. We expect that
individuals with both pain interference and depression will report the lowest SSPs relative to all
other defined subgroups, and that this doubly encumbered subgroup will have a higher risk of
dying at any given age during the study period.
Data
We used data from the Health and Retirement Study (HRS), a nationally representative
longitudinal study of American adults aged 51 and over. Although the survey began in 1992, our
sample includes community-dwelling, cohort eligible respondents interviewed in the 2000 wave,
as there were inconsistencies in both skip patterns and question wording prior to that year. With
follow-up mortality data through 2018, we are able to assess the accuracy of individuals’ SSPs
provided that they were between the ages of 57-89 in 2000 (N = 12,835). Respondents who did
not fall within this range were excluded for reasons related to an age-related skip pattern or target
age assignments described in the next section. Of this sample, 2,191 respondents were missing
values on four covariates (physical activity, smoking status, marital status, or race). However, we
were able to retain the full sample of 12,835 respondents by using Multiple Imputation (MI)
procedures outlined in the Measures section.
Outcome Measure
We calculated the accuracy of SSPs using the following two measures from the HRS survey:
1. Subjective survival probabilities (SSPs) HRS asked respondents aged 51-89, “what is the
percent chance you will live to be x or more?” where the target age x was dictated by

32
respondents’ age at interview: x = 75 for individuals aged 51-64, x = 80 for those aged 65-
69, x = 85 for those 70-74, x = 90 for those 75-79, and x = 95 for those 80-84, and x = 100
for those 85-89. Respondents aged 51-56 were excluded from our sample because they were
asked about living to a target age (x = 75) that was 19-25 years away. With 18 years’ worth
of follow-up data, we would not have been able to assess the accuracy of their SSPs.
Respondents provided a subjective percent chance of survival to x on a 0% (completely
unlikely) to 100% (very likely) scale. Because older target ages represent increasingly
exceptional longevity that may affect the accuracy of SSPs, we control for target age in the
final model predicting SSP accuracy.
2. Survival status at target age Using the Cross-wave Tracker File provided by the HRS, we
were able to evaluate the mortality status of respondents (i.e., whether they are alive or
dead). For respondents in our analytic sample who passed away between 2000 and 2018, we
calculated their exact age at death using data on the month and year of their birth and death.
If the respondent had not died, we simply tagged that they were still living using a separate
binary indicator. For individuals who died during the follow-up period, the HRS obtained
information on date of death from the decedent’s family members. In the case of individuals
who had dropped out of the study, mortality status was obtained either through linkage with
the National Death Index (NDI; Sonnega et al., 2014; Weir, 2016) or via imputation
informed by data on when the respondents was last known to be alive and when surveyors
learned of the respondent’s passing (2020 Tracker Final, Version 1.0 Data Description and
Usage, 2024). Searches for death information via the NDI were only conducted for
respondents whom the HRS research team was unable to contact for a request to reinterview in the most recent wave (Weir, 2016).

33
To assess SSP accuracy, we broke the 0-100% distribution of SSPs into three categories.
“Pessimistic” SSPs referred to those between 0-49%, “optimistic” SSPs were between 51-100%,
and an “ambivalent” SSP was 50%. Approximately one-fourth of the total sample reported an
SSP of 50% (n = 3,369); if we had included this group in either the "pessimistic" or "optimistic"
groups, we would risk skewing the results. Therefore, we coded individuals reporting this
specific SSP as a separate category signifying ambivalence. With such a large group endorsing
this midpoint, including them as a separate outcome category may provide insight into the
factors related to indecision. From a statistical perspective, this approach also allows us to both
retain our sample size and avoid skewing results related to under- and overestimating.
After dividing the distribution in three, we compared respondents’ SSP groupings to their
mortality status at their target ages and organized respondents into one of four groups: those who
provided (1) Correct SSPs, (2) Underestimated SSPs, (3) Overestimated SSPs, or (4) Ambivalent
SSPs. For example, a respondent who reported having an 80% chance of living to age 75 or over
would be deemed ‘correct’ if they did, in fact, live to or beyond age 75. However, if that same
respondent only lived until age 62, they were categorized as having an “overestimated” SSP. We
also made a minor exception for individuals who nearly made it to their target age, but not quite:
anyone who reported an optimistic SSP and passed away within two years prior to their target
age were deemed “correct.” Analyses conducted with and without this adjustment yield very
similar results. Overall, this categorization allows “pessimistic” respondents to be coded as either
“correct” or “underestimated” depending on their age of death; similarly, “optimistic”
respondents may be coded as “correct” or “overestimated,” while “ambivalent” respondents are
always coded as “ambivalent” regardless of their mortality status at their target ages.

34
In order to assess whether our findings are sensitive to this current categorization, we
present results from an additional model that broadened the “ambivalent” category from only
those reporting 50% to those reporting SSPs between 31-69%.
Independent Variables
Pain Respondents were asked if they are “often troubled with pain.” If the respondent
answered “no,” we coded them as experiencing no pain. If the respondent answered “yes,” they
were asked two follow-up questions about pain intensity and interference: “How bad is the pain
most of the time: mild, moderate or severe?” and “Does your pain make it difficult for you to do
your usual activities such as household chores or work? (yes/no)” In accordance with prior work
prioritizing the measurement of clinically relevant pain profiles (Zimmer & Zajacova, 2018), we
used these three questions about pain presence, intensity, and activity interference to generate a
three-category measure: (0) no reported pain, (1) mild or moderate pain that does not limit daily
activities, and (2) severe pain or pain that limits daily activities. These categories are referred to
as no pain, non-interfering mild or moderate pain (or “low impact pain”), and severe or
interfering pain (or “high impact pain”), respectively.
Depression The eight-item Center for Epidemiologic Studies Depression (CESD-8) scale
asks respondents if they had experienced eight self-reported symptoms in the past week,
including low mood, motivation, energy level, sleep quality, and life satisfaction. In accordance
with the established CESD-8 threshold (Dang et al., 2020), respondents who reported three or
more symptoms were classified as having symptoms indicative of depression and were coded as
1. Individuals endorsing two or fewer symptoms were coded as 0. Although we refer to
individuals as having depression or not throughout the manuscript, we acknowledge that
endorsing symptoms in a survey does not equate to a clinical diagnosis. In Tables 1, 2, and 3, we

35
use a categorical variable that combines pain presence, intensity, interference, and depression:
(0) no pain and no depression, (1) no pain with depression, (2) non-interfering mild or moderate
pain without depression, (3) non-interfering mild or moderate pain with depression, (4) severe or
interfering pain without depression, and (5) severe or interfering pain with depression.
Additional Covariates We also included sociodemographic and health controls in our
analyses as these may influence both SSPs and mortality risk. Sociodemographic variables
included age, gender, educational attainment (less than high school, high school, some college,
and college or above), and marital status (married/partnered, separated/divorced, widowed, and
never married). Health behavior variables included smoking status and physical activity.
Smoking status was categorized as (0) Never, (1) Former, and (2) Current smoker. Physical
activity was coded as a binary with 1 indicating that individuals engage in moderate or vigorous
physical activity with any frequency; respondents coded as 0 reported “never” engaging in such
activity. Health status variables included six self-reported chronic health conditions: heart
disease, lung disease, diabetes, high blood pressure, cancer, and stroke. For each condition, if the
respondent reported ever or currently having the disease, they were coded as 1; if not, they were
coded as 0.
To address missing values for the physical activity (n = 2,116, 15.6%), smoking status (n
= 75), marital status (n = 13), and race (n = 3) variables, we conducted multiple imputation using
chained equations (MICE; Azur et al., 2011). Given the categorical nature of these variables, we
estimated ten iterations of imputed values using multinomial logistic regression. These values
were estimated using respondents’ age, gender, educational attainment, pain experience,
depression status, drinking behavior, self-rated health, and history of six chronic health
conditions. This technique was successful in imputing all missing values on these variables.

36
Analytic Strategy
We began by describing sample characteristics for the full sample and across pain
interference and depression status categories (Table 1). For our first analysis, we used an
adjusted OLS regression to identify which group (defined by pain and depression status) reported
the lowest chances of survival to their target ages (Table 2). Specifically, we used the sixcategory variable that combines pain presence, intensity interference, and depression status as the
primary predictor. Using this variable rather than an interaction term between pain and
depression allowed us to report the specific SSP averages for each subgroup, controlling for
relevant demographic, socioeconomic, health behavior, and health status covariates. This and the
remaining three analyses were pooled over ten imputed datasets for the physical activity,
smoking, marital status, and race variables.
Our second analysis examined whether subgroups defined by pain and depression status
had increased risk of death during the study period using a Cox proportional hazards model
(Table 3). For this analysis, the time variable was the difference between respondents’ “start”
and “exit” age (either their age at death or age at last interview), ranging from 0 to 20 years. The
failure indicator was death (coded as 1). In preparation for analysis, the data were weighted using
respondent-level weights (see below for more information), and were set up to utilize the
multiply imputed values for variables previously described. The Cox proportional hazards model
yielded hazard ratios quantifying the risk of death while controlling for all covariates named in
the measures section. We additionally generated a Kaplan-Meier plot (Figure 1) to visualize
survival probabilities throughout the study period by pain-depression group, allowing us to test
our hypothesis that respondents with interfering pain and depression would have a higher risk of
dying at any given time than respondents without.

37
Third, we estimated the relative risk of underestimated, overestimated, or ambivalent
survival probabilities relative to correctly-estimated SSPs using an adjusted multinomial logistic
regression. The primary relationships of interest were between SSP accuracy, pain interference,
and depression. We assessed both the main effects of pain composite variable and depression as
well as their interaction to test whether there is an synergistic effect of experiencing both. In
supplementary materials (Table S1), we also included models stratified by target age as SSP
accuracy may differ as a function of the exceptionality of living to the assigned x. For example,
respondents may be rightfully more pessimistic about surviving to 100 than they are about
surviving to age 75, and such pessimism may be less so related to pain or depression as
respondents reach more advanced ages. Due to small cell sizes among the x = 100 group, we
analyzed the x = 95 and x = 100 respondents in the same model.
To improve confidence in our analysis, we also conducted a sensitivity check that
imposed more conservative cut-points of the 0-100% SSP scale (Table S2). Our initial
categorization of the outcome variable essentially simplifies the SSP question to “Do you think
you will live to x or older?”: “yes” (51-100%), “no” (0-49%), or “maybe” (50%). However, by
nature of the scale, respondents who reported a 51% chance of living to their target age are not as
confidently optimistic as respondents reporting 100%. Therefore, for this sensitivity check, we
widened the “ambivalent” category to include individuals reporting a 31-69% chance of living to
x. This range was informed by Wu et al. (2014) in which the authors claimed that respondents’
reports of SSPs at or exceeding 70% indicated that their perceived chances of living to their
target age were “probable” (70%), “very probable” (80%), “almost certain” (90%), and “certain”

38
(100%). On the other hand, SSPs under 30% signified respondents’ perception that there was
only a slight or no chance of living to x. Therefore, for our analysis, we consider individuals who
reported a 0-30% chance as “pessimistic,” and individuals reporting 70-100% as “optimistic.” If
this differential categorization of the outcome variable yields similar model results to our
original specification, it will bolster confidence in our findings.
Our analyses were conducted using Stata 18 and weighted to generate estimates that were
representative of the US population. Respondent-level weights adjusted for nonresponse bias,
complex survey design, and poststratification adjustments. More specifically, they controlled for
the HRS’s purposeful oversampling of Black Americans, Hispanics, and Floridians as well as the
different sampling frame used for recruiting participants for an “oldest old” subsample (Willis et
al., 2006).
Results
We first document sample characteristics in aggregate and over groups defined by
respondents’ pain interference and depression status. From our analytic sample—comprised of
12,835 adults aged 57-89—27% reported that they were troubled with pain, and 65% of those
individuals reported severe pain and/or pain-related interference with daily activities.
Approximately 22% of the total sample reported three or more depressive symptoms in the prior
week. Of respondents with suspected depression, 39.2% reported severe or interfering pain,
10.4% had non-interfering mild or moderate pain, and 50.4% reported no pain.
As shown in the left-most column of Table 1, the analytic sample was majority female
(57%), Non-Hispanic White (86%), and married (64%), with an average age of 68.6 years (SD =
8.2) in 2000. One-fifth of respondents (20.3%) had a college education or more. Roughly 60%
reported being either a former or current smoker, and 76% of respondents reportedly engaged in

39
moderate to vigorous physical activity with any frequency. Past or present chronic health
conditions were self-reported at the following rates: heart disease (22.0%), lung disease (7.8%),
diabetes (13.4%), high blood pressure (46.0%), cancer (12.4%), and stroke (6.8%). Of the total
sample, 62.1% died during the study period. With respect to the SSP accuracy outcome
categories, 52.3% respondents reported SSPs that were accurate to their own lifespans, 26.4%
reported an ambivalent 50% chance of survival to x, 11.4% underestimated their SSPs, and 10%
overestimated them.
Table 1 also highlights the significant differences in these sample characteristics when
stratified into subgroups defined by pain and depression status. When compared to respondents
without pain nor depression, individuals with either pain interference, depression, or both, were
significantly more likely to be female and more likely to die within the observation period (2000-
2020). With the exception of individuals reporting neither depression nor interfering pain, all
other subgroups were less educated, more likely to be current smokers, and more likely to have
at least three of the six measured chronic health conditions when compared to people with no
pain or depression. Individuals in the three subgroups defined (in part) by having depression
were more likely to underestimate their lifespans relative to people without pain or depression;
individuals without depression, regardless of pain interference, were not any more likely to
underestimate their chances of survival relative to the reference group (p > .05)

40

41

42

43
We then conducted an OLS regression to examine how the adjusted SSPs compare across
subgroups defined by pain and depression status (Table 2). We found that respondents who
reported both severe or interfering pain and depression reported the lowest SSPs of any group.
The mean SSP of this group was 13.3 percentage points lower than for individuals with no pain
nor depression (p < .001). Following pairwise comparisons of the marginal means from this
model, it is clear that subgroups defined by depression have significantly lower SSPs than those
without depression within each pain type / status (all p < .001).
Table 2. Linear regression predicting subjective survival probabilities for respondents aged 57-89 (N = 12,835)
Independent Predictors Composite Variable
b B SE b B SE
Pain interference
No pain
Non-interfering pain -4.24 -0.04 *** 0.89
Severe or interfering pain -6.14 -0.07 *** 0.83
Depression -7.31 -0.10 *** 0.76
Pain Interference & Depression
No pain without depression
No pain with depression -7.06 -0.07 *** 0.98
Non-interfering mild/mod pain without depression -3.73 -0.03 *** 1.02
Non-interfering mild/mod pain with depression -13.03 -0.06 *** 1.89
Interfering or severe pain without depression -6.21 -0.06 *** 0.97
Interfering or severe pain with depression -13.29 -0.12 *** 1.18
Age at interview 0.42 0.11 ** 0.16 0.41 0.11 ** 0.16
Target age
x = 75 (ref)
x = 80 -8.25 -0.10 *** 1.18 -8.24 -0.10 *** 1.18
x = 85 -15.07 -0.18 *** 1.93 -15.04 -0.18 *** 1.93
x = 90 -28.66 -0.31 *** 2.68 -28.63 -0.31 *** 2.68
x = 95 -36.94 -0.33 *** 3.40 -36.91 -0.33 *** 3.40
x = 100 -42.65 -0.26 *** 4.30 -42.59 -0.25 *** 4.30
Female 2.31 0.03 *** 0.61 2.31 0.04 *** 0.61
Race/ethnicity
White (ref)
Black 6.41 0.05 *** 1.08 6.39 0.05 *** 1.08

44
Other 0.05 0.00 2.37 0.07 0.00 2.37
Hispanic -4.11 -0.03 ** 1.46 -4.11 -0.03 ** 1.46
Education
Less than High School (ref)
High School 1.02 0.02 0.79 1.02 0.02 0.79
Some College 4.15 0.05 *** 0.87 4.15 0.05 *** 0.87
College and above 4.90 0.06 *** 0.85 4.89 0.06 *** 0.85
Marital Status
Married/Partnered (ref)
Separated/Divorced 0.57 0.01 1.03 0.57 0.01 1.03
Widowed -0.57 -0.01 0.79 -0.57 -0.01 0.79
Never Married -0.28 0.00 1.64 -0.28 0.00 1.64
Smoking Status
Never Smoker (ref)
Former Smoker 0.63 0.01 0.60 0.62 0.01 0.60
Current Smoker -6.06 -0.07 *** 0.93 -6.05 -0.07 *** 0.93
Physically active 4.25 0.05 *** 0.81 4.26 0.06 *** 0.74
Health Conditions
Heart Disease -6.27 -0.08 *** 0.74 -6.26 -0.08 *** 0.74
Lung Disease -7.31 -0.06 *** 1.14 -7.34 -0.06 *** 1.14
Diabetes -4.78 -0.05 *** 0.87 -4.78 -0.05 *** 0.87
High Blood Pressure -3.61 -0.06 *** 0.57 -3.61 -0.06 *** 0.57
Ever had Cancer -3.92 -0.04 *** 0.84 -3.91 -0.04 *** 0.84
Ever had Stroke -1.96 -0.02 1.24 -1.96 -0.02 1.24
Constant 42.08 *** 9.58 42.17 *** 9.58
† p<.10 * p<.05; ** p<.01; *** p<.001
This pattern is also observed in Figure 2, which shows adjusted mean SSPs across
groups. After calculating pairwise comparisons between group means, we see that each group
with depression reported a significantly lower SSP than the comparable group without
depression. Whether pain was non-interfering or interfering seemed to have little effect on SSPs,
e.g., the average SSP for non-depressed individuals with interfering pain (54.7%; 95% CI [52.8,
56.6]) did not differ significantly from that of those with neither pain interference nor depression
(57.3%; 95% CI [55.2, 59.3])

45
Results from the analysis predicting risk of death among respondents over the 20-year
study period suggest that both pain and depression correspond with significantly elevated
mortality risk (Table 3). However, subgroups defined by depression yielded larger hazard ratios
than those that were not. Adjusting for all other covariates and relative to the reference group,
subgroups of people with depression had the following associated elevations in mortality risk:
Individuals who had depression but were pain-free (HR = 1.20, p < .001), non-interfering mild or
moderate pain with depression (HR = 1.31, p < .001), and severe/interfering pain with depression
(HR = 1.24, p < .001). For subgroups of respondents without depression, our model yielded the
following mortality risks relative to those with no reported pain nor depression: non-interfering
mild or moderate pain with no depression (HR = 1.09, p > .05), and severe/interfering pain with
no depression (HR = 1.18, p < .001).

46
Table 3. Cox proportional hazards regression predicting hazard ratios for mortality among
respondents aged 57-89 in 2000 (N = 12,835)
HR SE
Pain & Depression
No pain without depression
No pain with depression 1.20 *** 0.04
Non-interfering mild or moderate pain without depression 1.09 † 0.04
Non-interfering mild or moderate pain with depression 1.31 *** 0.09
Severe or interfering pain without depression 1.18 *** 0.05
Severe or interfering pain with depression 1.24 *** 0.07
Age at interview 1.09 *** 0.00
Female 0.73 *** 0.02
Race/ethnicity
White (ref)
Black 0.98 0.04
Other 0.81 0.09
Hispanic 0.84 ** 0.05
Education
Less than High School (ref)
High School 0.95 † 0.03
Some College 0.89 *** 0.03
College and above 0.85 *** 0.03
Marital Status
Married/Partnered (ref)
Separated/Divorced 1.18 *** 0.05
Widowed 1.10 ** 0.04
Never Married 1.36 *** 0.10
Smoking Status
Never Smoker (ref)
Former Smoker 1.20 *** 0.03
Current Smoker 1.99 *** 0.10
Physically active 0.68 *** 0.02
Health Conditions
Heart Disease 1.30 *** 0.04
Lung Disease 1.71 *** 0.07
Diabetes 1.53 *** 0.06
High Blood Pressure 1.17 *** 0.03
Ever had Cancer 1.25 *** 0.04
Ever had Stroke 1.33 *** 0.07
† p<.10 * p<.05; ** p<.01; *** p<.001

47
Figure 1 is a Kaplan-Meier plot showing the adjusted survival curves for each group
defined by pain and depression status. Unsurprisingly, respondents with the best survival
probabilities throughout the study period are those with no pain nor depression, whereas
individuals with both severe or interfering pain and depression had the worst survival outcomes.
For example, the survival probability of those who reported no pain nor depression in 2000 was
0.75 in 2010, but only about 0.53 for those reporting both severe/interfering pain and depression
that same year.
Next, we conducted a multinomial logistic regression that estimated the relative risk of
reporting underestimated, overestimated, and ambivalent SSPs (Table 4). All results should be
interpreted net of covariates and relative to the base outcome: reporting a “correct” SSP. We
found that, while the experience of non-interfering mild or moderate pain was not significantly
associated with SSP accuracy, experiencing severe or activity-interfering pain was associated
with a 28% higher risk of underestimating SSPs (RRR = 1.28, p = .02). Severe/interfering pain
did not modulate the risk of reporting an overestimated (RRR = .84, p = .16) or ambivalent SSP

48
(RRR = 1.15, p = .09). Individuals with depression had a 50% higher risk of underestimating
(RRR = 1.50, p < .001) and a 15% lower risk of being ambivalent (RRR = 0.85, p = .048). Neither
interaction term, indicating the experience of both pain and depression, significantly modulated
the risk of any reporting outcome.
Table 4. Multinomial logistic regression predicting accuracy of subjective survival probabilities for respondents
aged 57-89 (N = 12,835)
Under-estimated
(n = 1,500)
Ambivalent
(n = 3,369)
Over-estimated
(n = 1,297)
RRR SE RRR SE RRR SE
Pain
No pain (ref)
Non-interfering mild or moderate pain 1.15 0.15 1.13 0.10 0.90 0.13
Interfering or severe pain 1.28 * 0.14 1.15 † 0.10 0.84 0.10
Depression 1.50 *** 0.15 0.85 * 0.07 0.94 0.10
Non-interfering pain * depression 0.82 0.20 0.99 0.18 0.69 0.21
Interfering or severe pain * depression 0.84 0.14 0.82 0.12 0.86 0.17
Age at interview 1.02 0.02 1.02 0.01 0.96 † 0.02
Target age
x = 75 (ref)
x = 80 1.46 ** 0.21 1.27 * 0.13 1.87 *** 0.30
x = 85 1.74 * 0.40 1.36 † 0.23 3.46 *** 0.87
x = 90 1.93 * 0.62 1.22 0.28 4.64 *** 1.64
x = 95 0.99 0.41 0.78 0.23 4.01 ** 1.79
x = 100 0.38 † 0.20 0.38 * 0.14 5.80 ** 3.21
Female 1.01 0.07 0.95 0.05 0.70 *** 0.05
Race/ethnicity
White (ref)
Black 0.85 0.10 0.75 ** 0.07 1.44 ** 0.15
Other 1.41 0.33 1.36 † 0.25 1.48 0.36
Hispanic 1.66 *** 0.20 0.80 * 0.09 0.84 0.15
Education
Less than High School (ref)
High School 0.84 * 0.07 0.98 0.06 0.90 0.08
Some College 0.66 *** 0.07 0.90 0.06 0.89 0.09
College or above 0.57 *** 0.06 0.74 *** 0.06 0.87 0.09
Marital Status
Married/Partnered (ref)

49
Separated/Divorced 0.91 0.11 0.94 0.08 1.20 0.14
Widowed 0.95 0.08 1.16 * 0.08 1.16 0.11
Never Married 0.95 0.19 1.17 0.17 1.94 *** 0.36
Smoking Status
Never Smoker (ref)
Former Smoker 0.80 ** 0.06 0.98 0.05 1.14 † 0.09
Current Smoker 1.20 † 0.13 1.36 *** 0.11 1.81 *** 0.20
Physically active 1.17 † 0.11 0.94 0.06 0.69 *** 0.07
Health Conditions
Heart Disease 1.08 0.09 1.03 0.06 1.08 0.09
Lung Disease 0.64 ** 0.09 0.90 0.08 1.04 0.13
Diabetes 1.02 0.10 1.15 * 0.08 1.49 *** 0.14
High Blood Pressure 1.19 * 0.08 1.11 * 0.05 1.10 0.08
Ever had Cancer 0.95 0.09 1.03 0.07 1.14 0.11
Ever had Stroke 0.68 ** 0.10 0.64 *** 0.07 1.07 0.13
Constant 0.05 ** 0.05 0.12 * 0.10 1.69 2.15
† p<.10 * p<.05; ** p<.01; *** p<.001
Note. Base outcome is "Correct" (n = 6,669)
When stratified by target age, high impact pain and depression seem to be associated with
pessimistic SSPs at x = 75, 80, and 85, even if non-significantly (Table S1). However, there is
variability in how pain and depression interact in these age-stratified models. While there were
no significant interactions in the larger model (Table 4), low impact pain and depression
significantly reduced the risk of overestimation in the x = 75 model (RRR = 0.31, p < .05), and
increased the risk of overestimation in the x = 80 model (RRR = 6.13, p < .01). Further, the
interaction between high impact pain and depression reduced the risk of ambivalence in the x =
80 model (RRR = 0.52, p < .05), and reduced the risk of underestimation in the x = 85 model
(RRR = 0.33, p < .01). In the models restricted to x = 90 and x = 95 / 100, there were no
significant interactions; high impact pain was associated with lower risk of overestimation for x
= 90 (RRR = 0.49, p < .05), and depression lowered the risk of ambivalence in the oldest group
(RRR = 0.56, p < .01).

50
Results from the sensitivity check (Table S2) that employed a more conservative
categorization of the outcome variable—for which the “ambivalent” category was broadened to
include respondents reporting SSPs of 30-70%—yielded nearly identical results to those shown
in Table 4. Severe or interfering pain (RRR = 1.43, p < .01) and depression (RRR = 1.62, p <
.001) were both independently associated with higher risks of SSP underestimation, and
depression was also associated with a lower risk of ambivalence (RRR = 0.83, p = .01). Similar
to the primary model, neither interaction term was significantly associated with any SSP
accuracy outcome.
Discussion
The present study used unique data with both subjective survival expectations and
follow-up information on mortality to examine the effects of pain and depression on the accuracy
of SSPs, both independently and synergistically. Considering the disruptions caused by activityinterfering pain and the pessimistic outlook that comes with pain-related depression, we
hypothesized that individuals with both severe/interfering pain and depression would both report
the lowest SSPs and be more likely to die, and die younger, during the study period. We also
expected that individuals with severe/interfering pain and/or depression would underestimate
their chances of living to a given target age for the same reasons.
As expected, individuals doubly encumbered by both severe or interfering pain and
depression report the lowest SSPs, on average. However, the difference in mean predicted SSPs
for depressed respondents with and without this high impact pain was not significant. As we
reconcile our findings with prior work suggesting that depression itself interferes with daily
activities by way of poor psychological wellbeing and lack of motivation (Edwards et al., 2011),
it is clear that depression, not pain interference, is the stronger contributing factor in reducing
individuals’ SSPs. This does not mean that experienced pain plays no role in influencing

51
peoples’ perceptions of their lifespans. In line with Fennell et al. (2024), reporting any pain,
regardless of depression status, was associated with lower average SSPs.
A somewhat similar pattern emerged when we estimated risk of death for subgroups
defined by pain and depression over the study period. Relative to individuals with no pain nor
depression, significantly larger percentages of respondents in subgroups defined by depression
died during the study period (Table 1), and similarly, had steeper survival curves regardless of
pain status (Table 3). That said, the experience of pain with or without depression also
significantly increased respondents’ risk of death over the study period. Therefore, this work
supports the growing literature on the respective links between pain and depression on mortality
(Glei & Weinstein, 2023; Macfarlane et al., 2017; Wei et al., 2019).
The primary aim of this paper was to assess whether the combined experience of
depression and high impact pain was associated with reports of inappropriately low longevity
expectations in older adults. Unexpectedly, we found no evidence for the interaction between
pain and depression in the larger model, but severe or interfering pain and depression did
independently increase the risk of SSP underestimation in our sample. This underestimation may
lead to a myriad of potential negative consequences related to financial and physical wellbeing.
Individuals who anticipate shorter lifespans tend to retire earlier and claim Social Security
benefits younger (Doerr, & Schulte, 2012; Hurd et al., 2004), increasing the likelihood that they
may prematurely exhaust retirement savings. Underestimation of one’s lifespan may also lend
itself to choices that contribute to poorer health and lower quality of life: people with lower SSPs
report lower likelihoods of seeking preventative healthcare and less willingness to abide by
public health guidelines (Biró, 2016; Celidoni et al., 2022; Picone et al., 2004). The tendency for
individuals with activity-interfering pain or depression to underestimate their lifespans may also

52
increase their need for unplanned, effortful, and expensive caregiving as they age beyond their
expectations.
Our findings should be considered alongside data and analytic limitations. First, the HRS
asks about the probability of living to a specific target age, not about the age at which a
respondent believes they will die. With data on subjective life expectancy in years, a comparison
between this subjective measure and the respondents’ objective age at death would have been
more easily made. Lacking this measure, we conducted sensitivity analyses with different
categorizations of the outcome variable; reassuringly, findings were very similar to our main
results. Other notable limitations include that our analyses are only salient to individuals aged
57-89, and that we only used measures of pain, depression, and SSPs from 2000. The experience
of pain and depression as well as subjective assessments of one’s own lifespan are subject to
change over time (Palloni & Novak, 2017; Schneider et al., 2012). It is possible that our use of
single measurements from 2000 may have contributed to our findings suggesting a nonsignificant link between pain and mortality.
Future research should further investigate the mechanisms underlying the possible
relationship between underestimated SSPs and abbreviated lifespans, including among (but not
limited to) people with pain and/or depression. Structural equation modeling and the exploitation
of longitudinal data may help elucidate whether pessimistic reports of low SSPs increase
mortality risk at younger ages through poor health behaviors. Specific to pain, it may be
clinically important to assess whether SSPs predict engagement with certain pain treatments. For
example, it may be that individuals with high SSPs are more likely to engage in pain treatments
that are uncomfortable and time-consuming in the short-term, but offer positive long-term
outcomes (e.g., physical therapy), whereas individuals with low SSPs may prefer treatments that
provide immediate relief without considering negative long-term consequences (e.g., opioid

53
therapy). Additionally, in our target age stratified models, there was mixed and inconsistent
evidence for the significant interplay between pain and depression on SSP accuracy. As
respondents age and, thus, estimate their chances of survival to more advanced ages, it is not
clear how individuals take these factors into account. Future work should continue to examine
the factors individuals consider when estimating their chances of living to exceptional ages.
In conclusion, the present study highlights that both severe/activity-interfering pain and
depression are independently associated with the underestimation of subjective assessments of
longevity. Our sensitivity test (that imposed more conservative cut points) supported these
findings. We present this study as further evidence of the negative psychological effects of high
impact pain and of depression. Subpopulations of older adults with interfering pain and/or
depression may be at risk for inadequate preparation for retirement and lack of engagement in
preventative healthcare as a consequence of underestimating their own lifespans. Further
investigation of the consequences of inaccurate SSP assessments specifically among older adults
with high impact pain and/or depression is warranted, as they represent large and vulnerable
groups that may be amenable to intervention. Within the healthcare setting, curiosity around
patient’s subjective assessments of their own longevity, as well as their own aging, may be
helpful in monitoring psychological and functional experiences of pain, depression, or other
conditions. In addition to being a signal of adaptive health and finance-related behaviors, a high
SSP is also likely a positive indicator that an individual is faring well in their everyday life.

54
Chapter 4. Identifying patient groups in total joint arthroplasty: A longitudinal study of
need and outcome disparities
Background: Total joint arthroplasty (TJA) is the gold-standard treatment for moderate to
severe osteoarthritis in the knee (TKA) and hip (THA). Despite significantly alleviating pain and
functioning limitations for most patients, 20% of TJA patients are not satisfied with their
outcomes. One possible reason is that some TJA recipients do not have a clinically indicated
need for the procedure.
Objective: To identify respondents who had and did not have an indicated pre-operative need for
a TKA/THA and assess differences in their post-operative pain, physical functioning, depressive
outcomes, and subjective longevity expectations.
Method: We examine Health and Retirement Study respondents who reportedly underwent a
TKA or THA between 2004 and 2018 (N=1,865). Latent class analysis differentiated
respondents using six indicators of TJA need (e.g., presence of pain/arthritis, treatment use,
difficulty walking, etc.). Fixed effects models estimated within-individual changes in pain
intensity, physical functioning, depressive symptoms, and subjective survival probabilities preand post-procedure for each class.
Results: From a two-class solution, 30.4% of the sample showed low indicated TJA need in the
wave prior to reporting one. These respondents reported more functional limitations in the wave
in which the TJA was reported and higher longevity expectations in two post-operative time
points. Conversely, individuals with a high TJA need reported significant and sustained
improvements in pain intensity, functioning, and depression.

55
Conclusion: TKA and THA recipients with low levels of pre-operative pain, activity limitations,
and mobility issues reported no post-operative improvements in pain, functioning, and
depressive symptoms.
Introduction
Osteoarthritis (OA) is a common chronic condition among middle-aged and older Americans
with 70% of Americans with OA older than 55 (Vos et al., 2017). Population aging and increases
in obesity are contributing to increasing incidence of OA, especially in weight-bearing lower
extremity joints (e.g., knees and hips; Neogi, 2013; Shichman et al., 2023; Singh et al., 2019).
Once patients reach the “end-stage” of this condition, characterized by severe joint pain and
mobility limitations, total joint arthroplasty (TJA) is considered the gold-standard treatment
(Drummer et al., 2021). This procedure often results in improved post-operative pain and
functioning, however, a non-trivial minority of patients see no post-operative benefit (Beswick et
al., 2012; Blackburn et al., 2024). The highest quality data estimating the prevalence of TJA
dissatisfaction are available for the two most common forms of TJA: total hip and knee
arthroplasty (THA and TKA, respectively). Seven to 23% of patients reported unfavorable
outcomes following a THA, and 10% to 34% after TKA (Beswick et al., 2012).
To date, research has identified several factors that are associated with poor post-operative
outcomes, including older age, high body-mass index (BMI), and poor mental health (e.g.,
depressive symptoms, fear avoidance, pain catastrophizing, and low self-efficacy; Olsen et al.,
2022; Sorel et al., 2019; Wylde et al., 2007). Patients are more likely to be satisfied with their
post-operative condition if they are younger, male, have a normal BMI, and had their preoperative expectations fulfilled (Lützner et al., 2019; Palazzo et al., 2014; Van Zaanen et al.,
2023). There is mixed evidence on the effect of pre-operative pain and functioning on post-

56
operative outcomes. Some studies suggest that severe symptoms prior to surgery are associated
with better post-operative outcomes (Nakano et al., 2020; Núñez et al., 2007; Olsen et al., 2022),
whereas others suggest the opposite (Palazzo et al., 2014; Rizzo et al., 2023; Wylde et al., 2007).
Building on this mixed evidence, misalignment in determined need and receipt of TJA may
partially explain post-operative disappointment. Clinicians’ abilities to determine which patients
would benefit from a TJA are variable with referring physicians and surgeons placing differential
importance on indicators of need. The most consistent need indicators are pain, functional
limitations, radiographic changes, and inability to adequately control pain with more
conservative treatments (Gademan et al., 2016; MacIntyre et al., 2015). However, even among
these agreed-upon indicators, there is variability in severity thresholds (Gademan et al., 2016;
Pacheco-Brousseau et al., 2023). For example, some physicians may consider patients with mild
pain to be eligible for TJA based on other functioning indicators (e.g., walking difficulty)
whereas others would deem this inappropriate (Dell’Isola et al., 2021). These inconsistencies in
determining who should undergo TJA have important implications for post-operative outcomes.
In fact, when predicting the probability of post-operative improvements among TKA patients,
Hawker et al. (2023) found that clinical need for the procedure (determined by radiographic
evidence of knee OA, severity of symptoms, and outcomes using other treatments) was the
strongest predictor, beyond readiness / willingness, depressive symptoms, and post-operative
expectations.
In the present study, we use latent class analysis on Health and Retirement Study (HRS) data
to identify differential classes of respondents based on established pre-operative indicators of
THA or TKA need. Given that insurance type plays a significant role in care access and quality
(Mehta et al., 2022; Wray et al., 2021), this analysis is also modeled to estimate whether

57
insurance type predicts class membership. We hypothesize that having a Medicare Health
Maintenance Organization (HMO) advantage plan will predict TJA need as these plans require
an additional screening via referral and have been shown to provide better quality care than
traditional Medicare and other public insurers (Landon et al., 2023).
Following respondents’ assignment to latent classes, we examine within-individual changes
in pain intensity, functional abilities, and mental health separately for each group. As an
exploratory aim, subjective survival expectations are also observed across the study period. In
the previous two chapters, I have established that interfering pain is associated with lower
average SSPs, and while analyses from Chapter 1 did not confirm our expectations that change in
pain would correspond to changes in SSPs, a functionally altering procedure causing nearimmediate and substantial pain relief may contribute to a significant shift towards more
expansive future perceptions. With respect to the other outcomes and in accordance with prior
work, we expect that individuals deemed to be appropriate candidates for TJA will reap
significant and sustained improvements in pain and functioning. We also hypothesize post-TJA
reductions in depressive symptoms, although this effect may not be as strong. There is mixed
evidence on whether TJA improves mental health (Papakostidou et al., 2012; Perneger et al.,
2019; Shan et al., 2014; Vajapey et al., 2021).
Data
To assess within-individual change in physical and psychological health measures before
and after older adults’ underwent THA or TKA, we use data from the HRS. The HRS is a
longitudinal, nationally representative data source of Americans aged 51 and older (Health and
Retirement Study, 2023). The survey began in 1992 and has continued biennially; its most recent
available wave was administered in 2020. For the analyses in this manuscript, we use data from

58
2002-2020 as joint replacements and non-replacement surgeries were not differentiated in the
survey until 2004. The 2002 wave was retained to offer pre-operative wave information prior to a
possible 2004 procedure. Similarly, we do not include data from individuals who underwent a
TJA in 2020 as we would be unable to evaluate their post-operative conditions.
The initial sample included only cohort-eligible respondents who reported undergoing a
knee or hip total joint arthroplasty. We compiled a dataset for these individuals using the
longitudinal tracker file, the RAND longitudinal file, the RAND detailed imputations file, and
data from the health and healthcare services sections of the core survey. Respondents were
required to have three or more survey observations with one serving as pre-operative data.
Among respondents who underwent knee or hip TJA procedures, 38.5% reported more than one
procedure throughout the observation period. Most commonly (65.3%), individuals who
underwent more than one procedure reported two TJAs in consecutive waves. However, some
respondents reported as many as seven hip or knee TJAs between 2004-2018. In all cases, we
examined the pre-operative need and post-operative outcomes from each respondents’ first
procedure. This yielded a total analytic sample of 1,865 respondents. Figure 1 provides specific
information about sample selection. Table S1 documents descriptive data and statistical
differences between individuals in the analytic sample and those who were excluded.

59
Measures
Joint replacement. In the core survey, respondents who reported a history of arthritis were
asked if they had a joint replacement or surgery in the prior two years due to their condition. If
yes, the respondents were then asked to specify whether the procedure was a joint replacement,
surgery without joint replacement, or both. The analyses in this manuscript only include
individuals who indicated that they underwent a joint replacement or had “both” a joint
replacement and non-replacement surgery at some point between 2004-2018. Respondents were
additionally asked to indicate which joint was replaced. In this study, we only examine total knee
and hip replacements as these are the most common TJA procedures (Maradit Kremers et al.,
2015).
Time to and following TJA. We coded survey waves relative to when the TJA occurred. For
example, the wave in which the respondent reported the TJA was coded as 0. The wave before
that was -1, the next wave was 1, and so on. It is important to note that because of the way that
the joint replacement question is asked (i.e., “in the prior two years”), the wave labeled “0” is
actually the first post-procedural wave, but is treated as the time of procedure in our analysis, or

60
the “event” wave. For statistical power, we recoded time to include multiple survey waves in
each time period that follows and proceeds the joint replacement wave. More specifically, time is
coded as the following for analysis: -1 (2-4 years pre-procedure), 0 (joint replacement wave), 1
(2-4 years post-procedure), and 2 (6-10 years post-procedure).
Indicators of TJA need
Indicators of TJA need were selected based on those reported in LeDoux et al. (2022) as
well as criteria that referring physicians and surgeons cite as the most important when
determining whether patients would benefit from the procedure (Dreinhöfer et al., 2006;
Gademan et al., 2016). The indicators include: (1) having an arthritis diagnosis, (2) some level of
pain, (3) having their arthritis or associated pain limit their daily activities, (4) receiving some
treatment for arthritic pain symptoms, (5) reporting either worsened or similar arthritic
symptoms from the previous wave, and (6) a self-reported difficulty walking several blocks.
Data for these six indicators of TJA need were taken from the wave prior to each respondents’
report of a TKA or THA. These data were used to predict membership into latent classes of TJA
need. One of the other most common indicators of TJA need is OA-related deterioration
identified via radiographic imaging. The HRS does not provide these data. While this does limit
our assessment of TJA need, Huynh et al. (2018) suggest that pain and functioning levels were
stronger predictors of surgeons’ decisions to recommend TJA than radiographic evidence of OA.
Arthritis and Pain. In the health section of the survey, respondents were asked if they were
“often troubled with pain,” and if they “ever had, or had a doctor ever told them that they have
arthritis or rheumatism.” These were both yes or no questions; in both cases, a “yes” represented
an indicator of TJA need.

61
Change in Arthritic Symptoms. Following a self-reported arthritis diagnosis, respondents were
asked whether their symptoms had improved, worsened, or stayed the same since the last wave.
Individuals whose arthritic symptoms remained the same or worsened were coded as 1 and those
who improved were coded as 0.
Treatment. We generated a binary indicator to establish whether respondents were receiving
treatment for their arthritis or associated pain as an indicator of TJA need. “Treatment” was
operationalized as taking over-the-counter nonsteroidal anti-inflammatory drugs (NSAIDs),
prescription medication for pain, opioids, or engaging in some sort of treatment for arthritis
(pharmacological or otherwise). All questions included in this composite variable had yes or no
response options. Interviewers provided examples of opioid (“Vicodin, OxyContin, codeine,
morphine, or similar medications”) and NSAID medications (“Over-the-counter pain
medications include such things as Advil, Aleve, Tylenol, aspirin or similar medications”) to
improve the accuracy of responses. The time frames provided in each question varied slightly.
Respondents were asked if they are “currently” and “regularly” engaging in arthritis treatment or
taking prescription pain medications, respectively. On the other hand, participants were asked if
they took opioids and NSAIDs in the prior three months. There is additional variability the
availability of these treatment data: the core survey only included the question about arthritis
treatment until 2012, respondents were only asked about prescription pain medications after
2004, and about opioid and NSAID use after 2014.
Activity limitations. Until 2012, respondents had two opportunities to report limitations with
their activities: due to arthritic symptoms and pain, separately. If they reported having arthritis,
they indicated whether their “arthritis sometimes limited their usual activities?” (yes/no). If they
reported pain, respondents then indicated whether their “pain interfered with their daily

62
activities, such as work or household chores (yes / no).” From 2014 on, the item about interfering
arthritic symptoms was no longer included in the survey. Therefore, we only used the pain
interference question to indicate interfering symptoms for more recent years.
Difficulty walking. Respondents were asked whether they have difficulty walking several
blocks, to which they could respond “yes,” “no,” or “don’t do.” Both “yes” and “don’t do” were
coded as 1 in a binary indicator; “no” was coded as 0.
Outcome Measures
Pain intensity. Respondents first indicated whether they were “often troubled with pain.”
If yes, average pain intensity was assessed: “How bad is the pain most of the time: mild,
moderate or severe?” Using these two items, we coded pain intensity as: no pain (0), mild pain
(1), moderate pain (2), and severe pain (3).
Functional limitations were assessed as the count of eleven self-reported functional
limitations (0-11). Respondents were asked whether they had any difficulty with the following
11 activities: walking several blocks, walking one block, sitting for two hours, getting up from a
chair, climbing several flights of stairs, climbing one flight of stairs, stooping, kneeling, or
crouching, extending their arms above their shoulders, pushing or pulling large objects, lifting or
carrying ten pounds, and picking up a dime. Higher scores indicate greater functional
impairment.
Depression was assessed as the count of depressive symptoms using the 8-item Center
for Epidemiologic Studies Depression (CES-D) scale. From this scale, respondents were asked
the frequency with which they experienced eight depressive symptoms over the prior week.
CES-D items inquired about deflated mood, feeling lonely, lack of motivation, and restless sleep
(Karim et al., 2015). Two of the items were reverse-coded for interpretability (“was happy” and

63
“enjoyed life”). Higher scores on the scale ranging from 0 to 8 indicated more depressive
symptoms.
Subjective survival probabilities (SSPs) were measured using the question “What is the
percent chance that you will live to be x or more?” where x varied based on the respondent’s age.
For respondents aged 51-64, x =75; for those aged 65-59, x = 80; for those aged 70-74, x = 85;
for those aged 75-79, x = 90; for those aged 80-84, x = 95; for those aged 85-89, x = 100.
Responses were provided as percentages from 0 to 100%. Respondents who were older than 89
were not asked this question, nor were people who were unable to answer the first three
probability questions in the Expectations section of the survey.
Covariates. Analyses also controlled for a set of time-varying sociodemographic and health
covariates. All selected covariates must be time-varying because we are conducting fixed effects
models (see Analytic Strategy for more information). Covariates included age, marital status,
household wealth, depression, physical activity, the number of times the respondent had visited a
doctor in the last two years, insurance type, body mass index (BMI), and six chronic health
conditions.
With the exception of age, all of these factors were coded into binary or categorical
variables. In addition to age, and only in the model estimating SSPs, we included a binary
indicator of whether the respondents’ target age (x) changed in that wave (relative to the prior
wave). Marital status was coded as “married or partnered,” “separated or divorced,” “widowed,”
or “never married.” Household wealth was quartiled with the fourth quartile representing the
highest wealth category. Depression was coded as a binary in accordance with established
thresholds for depression based on the CESD-8: a score of less than three suggested no or subclinical depression (0), and a score of three or higher indicated depression (1). Physical activity

64
was dichotomized to indicate engagement in moderate or vigorous activity at any frequency (1),
or none (0). The number of doctor visits since the last wave was categorized into 0-4, 5-10, and
11+ visits. As described in the survey, these visits exclude hospital stays, outpatient surgeries,
physical therapy appointments, and rehabilitation services. Models additionally control for
binary indicators of six chronic health conditions: stroke, heart disease, lung disease, diabetes,
high blood pressure, or cancer. Here, 1 means that the respondent has (or a doctor has told them
they have) the condition, and 0 means that they do not.
We included four categories to summarize insurance coverage: (1) Non-HMO Medicare,
(2) HMO Medicare Advantage, and (3) Private (employer-paid or other), and (4) Other (no
insurance, Medicaid, or Veterans’ Affairs (VA) insurance). This categorization was motivated by
our interest in the difference between non-HMO and HMO Medicare. HMO Medicare
Advantage plans require patients to obtain a referral from a primary care physician for
procedures like TJA (Understanding Medicare Advantage Plans, n.d.). This additional
consultation may help in screening patients’ need for TJA. In fact, a recent study showed that,
despite achieving better clinical and patient-reported quality metrics, the relative rates of TKA
and THA were 10 percent lower among patients with HMO Medicare plans than those with
traditional Medicare (Landon et al., 2023).
BMI was categorized based on both published health standards for older adults and
guideline criteria for joint replacement. Winter et al. (2014) suggests that healthy BMIs for older
adults range from 23.0 to 29.9 for optimal longevity. Additionally, while there is no established
cut-off, doctors are typically hesitant to perform joint replacement procedures of the lower
extremity on individuals whose BMI exceeds 35.0. With this information, we have categorized
respondents’ BMIs into four groups: < 23.0, 23.0 – 29.99, 30.0 – 34.99, and 35.0+.

65
Analytic Strategy
First, we conducted a latent class analysis (LCA) using the described indicators of need
from the wave prior to reported TJA. The aforementioned four-category insurance variable was
also used to predict membership into latent classes as insurance type is a known correlate of both
care delivery and quality (H. Kim et al., 2021; Mehta et al., 2022). We obtained model fit
statistics for LCAs predicting one, two, and three classes of need. We, then, described the
sociodemographic, economic, health, and healthcare use profiles of the total sample of
respondents who underwent joint replacement, and for each class. Significance testing between
subgroups was performed using t-tests for continuous variables, and chi-square tests for
categorical variables.
To assess the pre-operative condition and post-operative outcomes for individuals in each
class, we estimated fixed effects models with time preceding and following the procedure as the
primary independent variable. These models estimate respondent-specific effects and timespecific effects, while also allowing respondent-specific effects to vary across time (Gunasekara
et al., 2014). Separate models were conducted for each outcome (pain intensity, functional
limitations, depression, and SSPs) by latent class, yielding a total of eight models.
We also conducted two sensitivity analyses to explore key variables that may alter the
effects of TJAs among the respondents in our sample. First, we acknowledge that there have
been continuous technological advancements in TJA since 2004 and the outcomes from a
surgery performed at that time may not be representative of more recent procedures. For this
reason, we reconducted our analysis on a restricted sample of individuals who reported a TKA or
THA in 2010-2018 (Table S2). Second, prior work suggests that optimal TJA outcomes have an
inverted-U relationship with age so that those aged 70-80 may benefit the most with respect to

66
pain and functioning (Mota et al., 2012). To test whether different age groups reported
differential trajectories in all four outcomes, we included an interaction term between age and
time relative to TJA (Table S3). In this analysis, age was categorized as (1) 51-69, (2) 70-79, and
(3) 80+ with those aged 70-79 set as the reference group. For all models, we used each
respondents’ person-level weight from their last survey wave to adjust for the survey’s complex
sampling method (HRS Staff, 2019).
In addition to the sensitivity checks, the Appendix includes another descriptive table
documenting the sociodemographic and health profiles of respondents by year of reported TJA
(Table S4). We do this in light of speculation that recent advancements in prosthetic components
used for TJA and direct-to-consumer marketing for the procedure may be contributing to an
expansion of the eligible patient pool to younger, more active individuals (Losina et al., 2012).
For this reason, we suspect that TJAs performed earlier in the 2004-2018 study period may have
been done on individuals with more severe OA, pain, and mobility issues. For brevity, we
presented three time periods: 2004-2008, 2010-2014, and 2016-2018. Again, significance testing
between subgroups was performed using t-tests for continuous variables, and chi-square tests for
categorical variables.
Results
Sample selection procedures yielded a total sample of 1,865 respondents who report total
knee or hip replacements. There were 751 respondents who reported TJAs but did not meet other
criteria required to be included in our analytic sample. Specifically, these respondents either did
not have three waves of data including a wave preceding the TJA, or reported their first TJA in
their first or last wave in the survey. In either case, including these respondents would not have
allowed for the examination of pre-operative conditions and / or post-operative outcomes.
Commented [JA1]: I thikn a sample selection chart might
help here
Commented [GF2R1]: I have a sample selection chart
that’s referenced in the data section, but I don’t go through
sample selection in the text up there. Should I? And maybe
mention less here?

67
Missing respondents were not demographically different than the respondents in the analytic
sample. However, they were more likely to use pain medication prior to their procedure, have a
stroke history, be underweight, and be either uninsured or enrolled in Medicaid, VA, or HMO
insurance under Medicare (Table S1). Relative to respondents who underwent TJA between
2004 and 2008, those whose procedure was reported in either 2016 or 2018 had higher rates of
pre-operative pain, were younger, more educated, and were more likely enrolled in private or
HMO Medicare insurance (Table S2). Although non-significant, recent TJA patients also
reported slightly higher pre-operative rates of arthritis, pain medication use, pain-related activity
limitations, and difficulty walking several blocks. Despite these patterns, more recent TJA
patients were less likely to be assigned to the class eventually deemed the “high TJA need” class
than those who underwent procedures in 2004-2008.
We estimated models that assigned respondents to latent classes using pre-operative data
on six factors: reported pain, arthritis, pain medication use, lack of improvement in arthritic
symptoms, pain or arthritis-related limitations with daily activities, and difficulty walking several
blocks. The fit statistics for the models fitting one, two, and three latent classes are provided in
Table 1. The two-class solution had the best model fit overall, indicating greater certainty in
respondents’ membership to their allocated latent class (Weller et al., 2020). Specifically, the
two-class solution had a higher log-likelihood (LL = -4940.65), lower Bayesian Information
Criterion (BIC = 10001.56), and higher entropy (entropy = 0.93) relative to the three-class
solution.

68
Table 1. Model fit statistics from three latent class analyses estimating one,
two, and three class solutions
One Class Two Classes Three Classes
Log-Likelihood -5417.00 -4940.65 -4927.28
Degrees of freedom 6 16 26
AIC 10845.99 9913.29 9906.56
BIC 10879.09 10001.56 10050.01
Entropy `- 0.93 0.88
Of the total sample, 1,297 were assigned to Class 1 and 568 to Class 2 (Table 2). These
classes were most differentiated by their pain and functioning profiles. Only 15.4% of Class 1
reported no pain whereas 76.2% of Class 2 was pain-free two years prior to their TJAs. Class 1
also reported significantly more activity limitations (96.4% v. 9.4%) and difficulty walking
several blocks (76.8% v. 12.8%). Based on these profiles, we label Class 1 respondents as having
an high indicated “need” for TJA and Class 2 as not (i.e., “low need”). Insurance type
significantly predicted membership to these classes, so that the low need class was significantly
less likely to be uninsured, enrolled in Medicaid, or VA insurance than the high need group
(Coefficient = -1.30, p < .01; Table 3).
Table 2. Sample characteristics for HRS respondents who underwent total knee or hip arthroplasty in aggregate and
stratified by their need for the procedure.
Total sample
(N = 1,865) Class 1 (High Need)
(N=1,297)
M / % SD M / % SD M / % SD
Pain
No pain (ref) 35.3% 15.4% 76.2%
Mild pain 13.3% 12.9% 14.2% ***
Moderate pain 39.8% 54.6% 9.3% ***
Severe pain 11.7% 17.2% 0.3% ***
Arthritis 97.7% 98.6% 95.5% ***

69
Treatment 72.4% 78.1% 56.5% ***
No symptom improvement 95.2% 97.5% 89.8% ***
Pain / Arthritis limitations 71.8% 96.0% 17.4% ***
Difficulty walking several
blocks 57.2% 77.7% 0.0% ***
Age at interview 66.9 8.2 66.5 8.4 67.7 7.9
*
Female 60.8% 64.2% 53.6% **
Race/ethnicity
White 86.8% 84.9% 90.7%
Black 6.6% 7.7% 4.5% **
Other 2.4% 2.5% 2.2%
Hispanic 4.1% 4.9% 2.5%
*
Education
Less than High School 16.0% 18.3% 11.2%
High School 30.9% 32.0% 28.7%
*
Some College 25.2% 25.9% 23.8%
*
College and above 27.8% 23.7% 36.2% ***
Marital Status
Married/Partnered 70.0% 67.3% 75.6%
Separated/Divorced 11.0% 12.6% 7.7% **
Widowed 15.6% 16.3% 14.1%
Never Married 3.4% 3.9% 2.6%
Health
Depression 24.2% 31.6% 8.9% ***
Heart Disease 20.6% 23.1% 15.3% **
Lung Disease 7.9% 10.0% 3.7% ***
Diabetes 18.0% 20.3% 13.2% **
High Blood Pressure 60.7% 64.4% 53.0% ***
Ever had Cancer 14.1% 13.3% 15.7%
Ever had Stroke 6.1% 7.2% 3.9%
*
ADL 7.4% 10.9% 0.1% ***
Poor Self-rated Health 6.3% 9.1% 0.5% **
BMI
Less than 23 6.1% 5.0% 8.3%
23 to 29.99 (ref) 41.0% 36.1% 51.1%
30
-34.99 27.3% 28.0% 26.0% **
35+ 25.6% 31.0% 14.6% ***
Insurance
Non
-HMO Medicare (ref) 46.2% 47.1% 44.3%
HMO Medicare 12.5% 12.4% 12.8%
Private 36.6% 34.7% 40.5%

70
Other (no insurance, public) 4.6% 5.7% 2.3% *
Doctor visits in last 2 years
0 to 4 (ref) 26.6% 21.6% 37.0%
5 to 10 40.2% 39.8% 41.1% **
11+ 33.1% 38.6% 21.9% ***
Household Wealth Quartile
Quartile 1 (lowest; ref) 20.4% 25.9% 9.0%
Quartile 2 24.2% 25.6% 21.5% ***
Quartile 3 25.8% 24.6% 28.4% ***
Quartile 4 (highest) 29.6% 24.0% 41.1% ***
† p<.10 * p<.05; ** p<.01; *** p<.001
Note. Class 1 served as the reference group for significance testing. Testing between subgroups was performed using ttests for continuous variables, and chi-square tests for categorical variables.
Table 3. Insurance type predicting membership into latent classes from the two- and threeclass solutions.
Two-Class Solution Three-Class Solution
Coefficient SE Coefficient SE
Class 1 (ref)
Class 2
Non-HMO Medicare (ref)
HMO Medicare 0.02 0.17 0.06 0.18
Private 0.08 0.13 0.17 0.13
Other (no insurance, public) -1.30 0.39 ** -1.28 0.41 **
Constant -0.48 0.10 *** -0.68 0.11 ***
Class 3
Non-HMO Medicare (ref)
HMO Medicare 0.23 0.27
Private -0.77 0.28 **
Other (no insurance, public) -1.31 0.70 †
Constant -1.62 0.27 ***
† p<.10 * p<.05; ** p<.01; *** p<.001
Table 2 further documents additional sociodemographic characteristics, health, and
healthcare use profiles for the total sample and for both classes. All reported data is from the
wave prior to respondents’ TJA reports. The total sample had an average age of 66.9 years (SD =
8.2), was 61% female, and 87% non-Hispanic White. Over a quarter of the sample reported a

71
college education or above (28%), 70% were married or partnered, and 20% in the poorest
household wealth quartile. The sample reported chronic conditions at the following rates:
depression (24.2%), heart disease (20.6%), lung disease (7.9%), diabetes (18%), high blood
pressure (60.7%), cancer (14.1%), stroke (6.1%). Only 7% and 6% reported any ADL limitation
or poor health, respectively, and about half of respondents had BMIs above the traditionally
healthy range (52.9%). With respect to healthcare, only 4.6% were either uninsured or enrolled
in Medicaid or VA insurance, 12.5% had HMO Medicare, and 33.1% reported visiting the doctor
11 or more times in the last two years.
Intuitively, the class that had low indicated need for a TJA had a significantly more
advantaged profile. Relative to individuals with a clear TJA need, this group reported
significantly less pain and depression, fewer chronic conditions and ADL limitations, fewer
doctor visits, and generally better health. For example, whereas 71.8% of the “high need” class
reported moderate or severe pain, only 9.6% of the “low need” class reported this level of pain
intensity. The low-need group was also more likely older, male, white, wealthier, and highly
educated.
Across three of the four observed outcomes, the “high need” group improved
significantly. More specifically, relative to two to four years prior, respondents reported a 0.39
point decrease in pain intensity (Table 4) in the TJA “event” wave (b = -0.39, p < .001), a 0.53
point decrease two to four years post-TJA (b = -0.57, p < .001), and a 0.72 point decrease six to
10 years post-TJA (b = -0.72, p < .001). Similarly, individuals in need of a TJA reported
significant reductions in their number of functional limitations (Table 5) in the wave in which
they reported the TJA (b = -0.61, p < .001) and in both post-operative time points (b = -1.01, p <
. 001; b = -1.52, p < .001). With respect to depression (Table 6), “in-need” respondents only

72
reported significant reductions in their number of depressive symptoms in the TJA wave (b = -
0.22, p < .01) and two to four years after (b = -0.27, p < .05). Lastly, the SSP reports (Table 6) of
respondents with a TJA need did not significantly change over time as a function of TJA receipt.
Relative to pre-TJA, the “low need” group reported no significant within-individual
change in pain intensity (Table 4; Event: b = 0.07; 2-4 years: 0.09; 6-10 years: 0.07; all p > .05),
and depressive symptoms (Table 6; Event: b = 0.14, p < .10; 2-4 years: 0.03, p > .05; 6-10 years:
-0.01, p > .05) across any of the time points. Individuals in the low need class reported
significantly more functional limitations in the wave in which the TJA was reported relative to
pre-TJA (Table 5; Event: b = 0.50, p < .001). The only documented post-operative improvement
for the low-need group was a significantly increased subjective survival probability to x age 2-4
years and 6-10 years after the procedure (Table 7; 2-4 years: b = 5.70, p < .05; 6-10 years: b =
7.48, p < .05). All of these findings are displayed across four graphs with the marginal effects of
time relative to a TJA procedure on the four outcomes: pain intensity (Figure 2), functional
limitations (Figure 3), depression (Figure 4), and SSPs (Figure 5). Each graph displays the linear
predictions of the marginal effects for both the “high need” and “low need” groups.
Table 4. Fixed effects models estimating within-individual, time-distributed effects of TJA on pain
intensity conducted separately for each latent class.
High Need (N = 1,296) Low Need (N = 567)
Coefficient Robust SE Coefficient Robust SE
Time relative to TJA
2-4 years before (ref)
Event -0.39 0.04 *** 0.07 0.05
2-4 years after -0.53 0.06 *** 0.09 0.08
6-10 years after -0.72 0.10 *** 0.07 0.13
Depression 0.25 0.04 *** 0.19 0.07 **
Marital Status
Married or Partnered (ref)
Separated or Divorced -0.08 0.09 -0.21 0.23

73
Widowed -0.16 0.07 * -0.16 0.09 †
Never married -0.12 0.36 -0.21 0.17
Age 0.06 0.01 *** 0.02 0.01 *
Heart Disease -0.04 0.06 0.01 0.13
Lung Disease 0.06 0.08 0.13 0.22
Diabetes -0.02 0.07 * 0.06 0.15
Stroke 0.17 0.07 0.30 0.14 *
Cancer -0.08 0.07 0.06 0.12
Blood Pressure 0.01 0.08 0.03 0.09
Physical Activity 0.00 0.03 0.08 0.04 †
Body Mass Index
Less than 23 -0.06 0.09 -0.03 0.08
23 to 29.99 (ref)
30 to 34.99 0.04 0.05 0.17 0.07 *
35+ 0.13 0.07 † 0.10 0.12
Household Wealth Quartile
Quartile 1 (lowest; ref)
Quartile 2 -0.07 0.05 -0.06 0.08
Quartile 3 -0.05 0.06 -0.07 0.10
Quartile 4 (highest) 0.05 0.07 0.00 0.10
Doctors’ visits in the last two years
1 to 4 (ref)
5 to 10 0.08 0.04 * -0.04 0.04
11+ 0.14 0.04 *** 0.05 0.06
Insurance
Non-HMO Medicare (ref)
HMO Medicare 0.02 0.04 -0.02 0.06
Private 0.08 0.05 0.07 0.07
Other (no insurance, public) 0.01 0.09 0.42 0.28
Constant -2.32 0.60 *** -1.31 0.76 †
† p<.10 * p<.05; ** p<.01; *** p<.001
Table 5. Fixed effects models estimating within-individual, time-distributed effects of TJA on functional
limitations conducted separately for each latent class.
High Need (N = 1,296) Low Need (N = 567)
Coefficient Robust SE Coefficient Robust SE
Time relative to TJA
2-4 years before (ref)
Event -0.61 0.10 *** 0.50 0.10 ***

74
2-4 years after -1.01 0.14 *** 0.25 0.16
6-10 years after -1.52 0.22 *** 0.31 0.27
Depression 0.70 0.10 *** 0.58 0.18 **
Marital Status
Married or Partnered (ref)
Separated or Divorced 0.24 0.24 0.00 0.42
Widowed -0.11 0.17 0.18 0.22
Never married -0.06 0.38 0.09 0.25
Age 0.16 0.02 *** 0.06 0.03 *
Heart Disease 0.10 0.16 0.63 0.20 **
Lung Disease 0.42 0.24 † 1.26 0.41 **
Diabetes -0.18 0.16 0.05 0.31
Stroke 0.96 0.26 *** 1.05 0.34 **
Cancer 0.15 0.26 0.16 0.29
Blood Pressure 0.00 0.19 0.15 0.22
Physical Activity -0.09 0.07 -0.19 0.10 †
Body Mass Index
Less than 23 0.06 0.21 0.13 0.22
23 to 29.99 (ref)
30 to 34.99 0.06 0.11 0.40 0.14 **
35+ 0.40 0.17 * 0.62 0.21 **
Household Wealth Quartile
1 (lowest)
2 0.02 0.10 -0.07 0.16
3 0.18 0.14 -0.27 0.19
4 (higher) 0.03 0.17 -0.19 0.21
Doctors’ visits in the last two years
1 to 4 (ref)
5 to 10 0.30 0.08 *** 0.08 0.08
11+ 0.48 0.10 *** 0.08 0.10
Insurance
Non-HMO Medicare (ref)
HMO Medicare -0.04 0.10 -0.23 0.14 †
Private (employer-paid or other) 0.17 0.11 0.12 0.14
Other (no ins, public) 0.56 0.24 * 0.81 0.31 **
Constant -6.44 1.38 *** -2.49 1.77
† p<.10 * p<.05; ** p<.01; *** p<.001

75
Table 6. Fixed effects models estimating within-individual, time-distributed effects of TJA on
depressive symptoms conducted separately for each latent class.
High Need (N = 1,279) Low Need (N = 562)
Coefficient Robust SE Coefficient Robust SE
Time relative to TJA
2-4 years before (ref)
Event -0.22 0.07 ** 0.14 0.08 †
2-4 years after -0.27 0.11 * 0.03 0.10
6-10 years after -0.27 0.17 -0.01 0.16
Marital Status
Married or Partnered (ref)
Separated or Divorced 0.07 0.23 0.16 0.16 †
Widowed 0.47 0.14 *** 0.28 0.16
Never married -0.28 0.52 -0.06 0.20
Age 0.01 0.02 0.01 0.01
Heart Disease 0.21 0.11 † 0.13 0.14
Lung Disease 0.36 0.16 * -0.11 0.26
Diabetes -0.10 0.13 0.13 0.18
Stroke 0.14 0.20 0.16 0.17
Cancer 0.10 0.18 -0.11 0.15
Blood Pressure -0.13 0.13 0.06 0.11
Physical Activity -0.05 0.06 0.00 0.07
Body Mass Index
Less than 23 0.08 0.13 0.20 0.16
23 to 29.99 (ref)
30 to 34.99 -0.08 0.11 0.21 0.10 *
35+ 0.01 0.14 0.09 0.12
Household Wealth Quartile
1 (lowest)
2 -0.28 0.10 ** 0.01 0.13
3 -0.35 0.12 ** 0.10 0.15
4 (higher) -0.25 0.13 † -0.03 0.16
Doctors’ visits in the last two years
1 to 4 (ref)
5 to 10 0.05 0.06 0.02 0.06
11+ 0.15 0.07 * 0.09 0.08 †
Insurance
Non-HMO Medicare (ref)
HMO Medicare -0.11 0.08 -0.02 0.09
Private 0.02 0.09 0.06 0.08

76
Other (no insurance, public) 0.17 0.18 -0.27 0.38
Constant 0.97 1.10 0.03 1.00
† p<.10 * p<.05; ** p<.01; *** p<.001
Table 7. Fixed effects models estimating within-individual, time-distributed effects of TJA on
subjective survival probabilities conducted separately for each latent class.
High Need (N = 1,268) Low Need (N = 559)
Coefficient Robust SE Coefficient Robust SE
Time relative to TJA
2-4 years before (ref)
Event 0.28 1.06 3.03 1.66 †
2-4 years after 0.30 1.56 5.70 2.30 *
6-10 years after -1.04 2.67 7.48 3.66 *
Depression -2.69 1.00 ** -3.32 2.35
Marital Status
Married or Partnered (ref)
Separated or Divorced -5.44 3.02 † 13.06 5.85 *
Widowed -4.62 1.58 ** 3.67 2.62
Never married 8.04 10.01 15.18 7.40 *
Age -1.24 0.25 *** -2.17 0.38 ***
Heart Disease -5.24 1.84 ** 1.57 2.53
Lung Disease -5.24 2.93 1.60 4.61
Diabetes 2.34 2.08 -0.10 4.42
Stroke -2.98 2.89 -3.00 4.46
Cancer 0.66 2.03 -6.03 2.87 *
Blood Pressure 1.08 1.84 0.43 2.10
Physical Activity 1.78 0.92 † 2.69 1.47
Body Mass Index
Less than 23 3.66 2.43 -4.05 2.56
23 to 29.99 (ref)
30 to 34.99 1.19 1.23 -0.92 1.99
35+ 3.02 1.79 † 4.45 2.89
Household Wealth Quartile
1 (lowest)
2 2.97 1.47 * -2.53 2.19
3 0.71 1.87 -0.65 2.63
4 (higher) 0.51 2.11 1.18 3.00
Doctors’ visits in the last two
years
1 to 4 (ref)

77
5 to 10 0.35 1.06 1.81 1.28
11+ -0.87 1.09 0.57 1.44
Insurance
Non-HMO Medicare (ref)
Medicare HMO 2.34 1.36 † -2.39 1.41 †
Private -0.64 1.19 0.72 1.58
Other (no insurance, public) 4.86 2.92 -12.14 5.38 *
Target age changed 0.16 0.72 -0.63 0.86
Constant 135.72 16.32 *** 205.49 25.30 ***
† p<.10 * p<.05; ** p<.01; *** p<.001
0
0.5
1
1.5
2
2-4 years before Event 2-4 years after 6-10 years after
Pain intensity (0-3)
Figure 2. Linear predictions of the time-distributed fixed effect
of TJA receipt on pain intensity by class
High Need Low Need

78
0
1
2
3
4
5
6
7
2-4 years before Event 2-4 years after 6-10 years after
Number of functional limitations (0-11)
Figure 3. Linear predictions of the time-distributed fixed effect
of TJA receipt on number of functional limitations by class
High Need Low Need
0
0.5
1
1.5
2
2.5
Number of depressive symptoms (0
2-4 years before Event 2-4 years after 6-10 years after
-8)
Figure 4. Linear predictions of the time-distributed fixed effect
of TJA receipt on number of depressive symptoms by class
High Need Low Need

79
Our sensitivity analysis restricting to individuals who underwent TJAs in 2010-2018
yielded similar results to our primary models (Table S2). The major differences were in the
models conducted on “low need” respondents: SSPs were no longer significantly increased in the
two post-operative time points (2-4 years after: b = 2.67; 6-10 years after: b = 6.68, both p > .05),
and respondents reported a more pronounced increase in the number of functional limitations
post-operatively (event wave: b = 0.69, p < .001; 2-4 years after: b = 0.59, p < .05). With respect
to our interaction models exploring the age as a moderator (Table S3), we found that, in general,
individuals aged 80+ were more likely to report poorer long-term post-operative functioning and
depressive outcomes relative to those aged 70-79. On the other hand, individuals aged 51-69
reported better functioning and SSP outcomes relative to respondents aged 70-79. There only
seemed to be a moderating effect of age on the relationship between time relative to TJA receipt
and depression among the “low need” class. Age did not moderate the relationship between time
relative to TJA receipt and pain intensity for either class.
Discussion
Total knee and hip arthroplasties are typically reserved for individuals who have severe
osteoarthritis that cannot be improved through non-surgical treatment (Dreinhöfer et al., 2006;
40
45
50
55
60
65
70
2-4 years before Event 2-4 years after 6-10 years after
Subjective survival probabilities
(%)
Figure 5. Linear predictions of the time-distributed fixed effect
of TJA receipt on subjective survival probabilities by class
High Need Low Need

80
Drummer et al., 2021). These procedures aim to reduce pain and improve functioning. In the
present study, individuals who needed and underwent TJA reaped these benefits, as well as a
corresponding reduction in depressive symptoms. However, individuals with little to no preoperative pain and few mobility limitations did not report post-operative improvements in pain,
functioning, or depressive symptoms. In fact, individuals with low TJA need reported
significantly more functional limitations in the wave immediately following the procedure. This
finding only became more apparent in the model restricted to more recent TJAs reported in
2010-2018.
Our exploratory hypothesis regarding SSPs was not confirmed: individuals who reported
significant post-operative improvements in both physical and mental health (in the “high need”
group) were not any more optimistic about their longevity than they were pre-operatively. This
finding was in line with prior work suggesting that longitudinal change in pain or arthritis does
not correlate with change in SSPs (Fennell, 2024; Vanajan & Gherdan, 2022). Respondents may
have acknowledged that while their pain and mobility issues have been ameliorated, their
arthritis was never life-threatening. Interestingly, individuals with low TJA need reported more
optimistic SSPs in two post-operative time points, but this significant finding was no longer
observed in the model restricted to 2010-2018 TJAs.
Thirty percent of the total sample of TJA patients did not appear to have a strong preoperative need for the procedure. This is particularly alarming as 76% of these individuals
reported no pain at all in the wave prior to undergoing TJA despite pain being one of the most
important clinical criteria for TJA eligibility (B. L. Conner-Spady et al., 2014; Gademan et al.,
2016; Huynh et al., 2018; Pacheco-Brousseau et al., 2023; Verra et al., 2016). These data may
indicate an unwarranted expansion in TJA eligibility to healthier patient groups. However, work

81
that corroborates this suspicion has noted that these “healthier” TJA patients are also typically
younger (Losina et al., 2012). In our sample, that is not the case: the “low need” group is
significantly older than those with high TJA need. They are also more likely male, white,
wealthier, and well-educated. The low need group was also less likely to be either uninsured or
enrolled in Medicaid or VA insurance. While this is consistent with their advantaged profiles, we
did not observe any other significant associations between low TJA need and insurance type.
This goes against our expectation that individuals with HMO Medicare Advantage would be
significantly less likely to be in the low need group. The additional referral required from an
HMO did not appear to improve the accuracy of clinical need assessments for TJAs in our
sample.
The advantaged profiles of the individuals with low TJA need may hint at an unfortunate
age-old pattern. Like with many treatments, the individuals who receive TJAs may not always
represent those with the greatest need for the procedure. One of the most powerful predictors of
receiving a TJA is patient willingness, which can be surprisingly influential, especially when
certain eligibility conditions are met (Dell’Isola et al., 2021). For example, older age is often
considered an indicator of eligibility (Huynh et al., 2018; Verra et al., 2016). Additionally,
physicians are more likely to offer TJAs to men, despite similar or greater need among women
(Mota et al., 2012). With respect to patterning by social status, Mota et al. (2012) reported that
post-secondary education is associated with greater willingness to undergo TJA without a
recommendation from a physician. Highly educated patients are also more likely to switch
surgeons to reduce their time on the TJA wait list (Conner-Spady et al., 2008). When a patient is
not only willing, but advocating for surgery, some surgeons may have trouble resisting demands.

82
We also considered whether physicians may be inappropriately suggesting TJA to these
wealthier patients who may be more willing and capable of covering growing out-of-pocket costs
(Durand et al., 2022). However, findings from prior work did not substantiate this possibility.
While physicians may rely more on out-of-pocket (OOP) payments due to declining TJA
reimbursement rates from both Medicare and private insurers (Mayfield et al., 2020), ability to
pay OOP costs only seems to affect patients’ willingness to undergo the procedure, not
physicians’ willingness to perform it (Durand et al., 2022). If physicians were prioritizing
financial incentive over patients’ alignment with TJA eligibility, we might also expect that the
“low need” group would be younger and / or report significantly more private insurance
coverage as commercial insurance companies still offer higher TJA reimbursements than
Medicare (Lopez et al., 2020). While that was not the case in our sample, our table dividing the
sample by year of TJA receipt did provide preliminary evidence that mirrored this expectation.
Respondents who underwent TJAs most recently (2016-2018) were significantly less likely to be
in the “high need” class, were younger, and more likely to be enrolled in private insurance (or
non-HMO Medicare) than respondents who underwent the procedure in 2004-2008. In light of
opposing prior work, future work should investigate whether recent TJA referrals are similarly
patterned in other datasets.
Despite using comparatively more rudimentary binary indicators of pain and functioning,
we were able to successfully replicate a similar percentage of “inappropriate TJA candidates”
(30.4%) to those reported in earlier work: 31.3% in Riddle et al. (2015) and 34.3% in Riddle et
al. (2014). This calls to question why challenges associated with identifying appropriate patients
for TJA still persist. In fact, according to our data, a significantly smaller proportion of
individuals who underwent TJAs in 2016-2018 had a distinct TJA need than those who

83
underwent procedures in 2004-2008, indicating that physicians’ abilities to identify appropriate
patients may have slightly worsened over the last 20 years.
This study should be interpreted in the context of three major limitations. First,
radiographic evidence of osteoarthritis and post-operative satisfaction would have been valuable
additions as a key indicator of TJA (Gademan et al., 2016; Pacheco-Brousseau et al., 2023) and
an outcome variable, respectively. Unfortunately, the HRS does not collect this information.
Second, we included respondents who reported multiple TJAs by analyzing their first procedure.
Because of this, it is possible that the pain and functioning relief observed for respondents in the
high need class is underestimated as post-operative levels of pain and functioning may be related
to another joint still in need of the procedure. Third, there was a lack of specificity in the TJA
need indicators that informed the latent classes. While difficulty walking several blocks is an
intuitive indicator of hip or knee OA, pre-operative pain, arthritis, activity limitations, pain
treatment, and a lack of improvement in arthritic symptoms could potentially be in reference to
another joint. This possibility may have contributed to the misassignment of individuals who had
low TJA need to the “high need” class. That said, the presence of these indicators in the context
of a sample of individuals who ultimately undergo TKA or THA in the following wave bolsters
our confidence that these indicators were, in fact, in reference to hip and knee pain.
Future research should consider why individuals who do not have pain or an otherwise
apparent need for TJA willfully undergo the procedure. Insurance type, age, education, and
wealth are important factors to explore. Also, while it is important for referring physicians and
surgeons to use their expertise and discretion, the rate at which surgeons perform these
procedures that are designed to ameliorate severe pain and improve functioning on individuals
who report no pain at all should raise concern. Researchers should continue to improve efforts to

84
standardize eligibility for TJA, perhaps through the supervised use of machine learning and
artificial intelligence.
Conclusion
Consistent with prior work, we found that nearly one-third of respondents may not be
appropriate candidates for TJA, resulting in no significant post-operative improvements in pain,
physical functioning, and depression. This work highlights that self-reported binary indicators of
pain and functioning are enough to successfully identify individuals who would and would not
benefit from TJA. We also find that the “low need” group is older, more likely male, wealthier,
and more educated than respondents who clearly needed the TJA, potentially signifying that
patient advocacy may be influential beyond clinical need in predicting receipt of TJA. Although
lack of TJA need and post-operative improvement likely does not fully explain why 20% of
patients are dissatisfied with their outcomes, it may play a key role. Future research should
consider the overlap between the 20% of patients who were dissatisfied, and the 30% who did
not appear to need nor benefit from the procedure.

85
Chapter 5. Conclusion
This dissertation work sought to further examine the relationship between pain and
subjective survival probabilities (SSPs). This comes in light of recent work suggesting that the
experience of arthritic pain was not associated with older adults’ longevity expectations (Vanajan
et al., 2020; Vanajan & Gherdan, 2022). On its face, this finding seems sensical: pain is not fatal,
and thus, should not affect how long a person believes they may live. However, the effect of pain
on daily activities and functioning must be considered. A large and growing body of literature
suggests that while the experience of any pain is not consistently associated with higher mortality
risk, “high impact” pain that limits daily activity is (Cleveland et al., 2019; Glei & Weinstein,
2023; Macfarlane et al., 2017). Subjective survival probabilities are a psychological construct
that represent more than just a reasonably accurate glimpse into how long a person may live
(Hurd & McGarry, 2002). SSPs are barometers for how well individuals are faring in everyday
life, suggesting that they are likely to be affected by pain-related activity limitations.
Pain and SSPs
To examine the effect of high impact pain on SSPs in Studies 1 and 2, I mirror the
measurement approach taken in Zimmer & Zajacova (2018). Leveraging all three of the Health
and Retirement Study’s limited pain measures, I generated the following variable to capture
clinically relevant pain categories: (0) No pain, (1) Non-interfering mild or moderate pain, and
(3) Interfering or severe pain. As the core focus of this chapter, the first study provides the most
concrete evidence of the relationship between severe/interfering pain and SSPs. Here, I use a
large set of repeated cross-sectional data to examine this relationship by target age and gender
using fractional logit models, which are specifically designed to estimate outcomes like SSPs
that are reported on scales ranging from 0-1 (Villadsen & Wulff, 2021). My findings demonstrate

86
a clear negative relationship between severe or interfering pain and SSPs regardless of gender or
target age. This significant relationship remains after adjusting for sociodemographic and health
covariates.
Study 1 also assessed if the established pain-SSP link transcended cross-sectional
association. I examined whether longitudinal change in pain corresponded with change in SSPs
over two survey waves. Consistent with Vanajan & Gherdan (2022), individuals did not appear
to update their SSPs following the development of new or worsening pain. It is possible that this
non-significant result may have been due to the aggregation of all instances of worsening pain
rather than isolating to those who developed interfering pain between waves.
In Study 2, I assess the cross-sectional relationship between pain and SSPs again in order
to motivate the examination of pain and depression on SSP accuracy. As this was not the central
aim of the chapter, I opted for the use of a multivariate OLS regression for interpretability
(although fractional logit and OLS regression yield very similar results) and controlled for rather
than stratified by target age. Similar to Study 1, severe/interfering pain was associated with
significantly lower SSPs. However, contrary to all but one target-age stratified model in the prior
chapter, non-interfering mild or moderate pain was associated with SSPs as well. This is almost
certainly due to the large relative sample size of respondents estimating survival probabilities to
age 75. As seen in Study 1, the longevity expectations of this youngest group (aged 51-64)
appear to be the most susceptible to pain regardless of impact or intensity.
Age-Related Susceptibility
These age-related findings were further corroborated by the results of sensitivity check
from Studies 2 and 3. First, in Study 2, only the results from the youngest respondents somewhat
supported our central hypothesis by showing that the experience of both pain and depression

87
reduced the risk of SSP overestimation. In models estimating the fixed effect of total knee or hip
arthroplasty on SSPs using age group by time interaction terms in Study 3, it is clear that the
youngest group (aged 51-69) reported the greatest improvements in SSPs post-operatively. These
results are supported by recently published work showcasing an age-patterned relationship
between pain interference and mortality: pain interference is a more robust predictor of mortality
risk at younger ages (25-60) than at older ages (Glei & Weinstein, 2023). As Glei & Weinstein
argue, younger people may be more affected by pain than older people given their relative lack
of other chronic health conditions and mobility limitations. Our finding is further validated by
psychological work documenting that younger people are more emotionally reactive to health
stressors such as pain than older people (Piazza et al., 2007). This difference may be rooted in
that younger respondents regard persistent pain as an “off-time” experience, or something that is
abnormal for their age (Neugarten, 1968). However, it is unclear whether this reasoning can be
appropriately applied to our samples as the “youngest” respondents are 51 and nearing the end of
middle-age—a group that already experiences high rates of pain (Zelaya et al., 2020).
SSP Underestimation
Study 2 finds that individuals with severe or interfering pain and depression report lower
average SSPs and have higher mortality risks than people without. While these two findings
appear congruent, the extent of the effects differ meaningfully: respondents with high impact
pain and depression were more likely to report SSPs that were underestimated relative to their
true lifespan. This may be evidence of a self-fulfilling prophecy. Respondents encumbered by
pain or depression who anticipate shorter lifespans may be increasing the likelihood of premature
mortality through a lack of investment in health-seeking behaviors. As evidenced by the fact that

88
these SSPs were underestimated and not “correct,” we conclude that, thankfully, despite many
respondents’ initial underestimation, their actual longevity exceeded their predictions.
Total Joint Arthroplasty (TJA) & Longevity Expectations
Barring the youngest group, individuals with high TJA need who underwent a total knee
or hip replacement did not report post-operative improvements in SSPs. This was unexpected as
this group saw major improvements in pain intensity, physical functioning, and depressive
symptoms. Given that prior work has shown reductions in SSPs following negative health shocks
(Liu et al., 2007; Zacher et al., 2022), we suspected that the opposite may also be true: a discrete
improvement in physical health and functioning would lengthen an individual’s expected
longevity. A possible explanation for why this hypothesis was not substantiated is that TJArelated improvements in pain and functioning were simply not enough for respondents to expect
significant changes to their lifespan. Despite distinct functional improvement, individuals with
pain-related activity interference prior to surgery still largely reported post-operative pain
(averaging around “mild” pain) and some level of functional impairment (reporting between four
or five of 11 possible functional limitations).
Overall, the finding that TJA does not improve SSPs in the “high TJA need” class is
consistent with the non-significant longitudinal results from Study 1 and inspires a similar
question: if we restricted the analyses to include only individuals who experienced exceptional
post-operative improvement in pain and functioning, thus leaving them nearly pain-free, would
we find significant improvements in longevity expectations? Of course, this exceptional level of
improvement for older people with high impact pre-operative pain is not realistic (George et al.,
2022). However, in line with results from Study 2, it is possible that patients are underestimating
the longevity benefits of TJA. Successful knee and hip replacements are associated with

89
significant reductions in mortality risk for eight years following surgery (Maradit Kremers et al.,
2016).
Interestingly, respondents with low TJA need reported significant improvements in SSPs
in the wave that they reported the procedure despite experiencing no reductions in pain,
functional impairment, or depressive symptoms. While this finding is in direct opposition to my
original hypothesis, it may hark on the psychological nature of subjective survival probabilities.
Although this group experienced no physical post-operative improvements, respondents’
physical condition remained consistently good throughout the study period. With that in mind,
and considering that they had recently undergone a procedure intended to extend the longevity of
their joints, their higher SSPs reflected self-perceptions of good health and optimism for the
future.
Conclusion
Taken together, the three studies expanded our understanding of pain and SSPs. First, I
established that middle-aged and older adults with high impact pain report lower average SSPs
than those with no or non-interfering pain. Second, I showed that these low SSPs were, in fact,
more likely to be underestimated relative to respondents’ actual lifespans. And lastly, I
demonstrated that SSPs of respondents with high impact pain did not improve despite physical
and mental health improvements associated with TJA. This is particularly troubling as TJA is
considered the gold-standard treatment for older adults with osteoarthritis.
While it is unfortunate to report that pain-related detriments to SSPs may not be remedied
through effective pain treatment, there are extant studies showcasing the power of psychological
intervention in this space. Several studies have shown that brief psychological interventions are
successful in expanding individuals’ perceptions of future time. These increased time horizons

90
yielded lasting improvements in health-seeking and preventative self-care behaviors, such as
physical activity and diet (Gellert et al., 2012; Hall & Fong, 2003). Given the age-related results
from all three studies, middle-aged people experiencing pain may be an ideal target population.
They are more susceptible lower average SSPs in the face of pain, and may be are more likely to
report SSP improvements following intervention. Pain incidence is increasing for this middle-age
group (Zimmer & Zajacova, 2018) and intervening on pain both physically and psychologically
early in the trajectory is crucial to prevent or delay further impairment (Patel, 2021).
Regardless of the specific intervention group, it is important to bolster survival
expectations among individuals with pain to a level that is appropriately calibrated to their
estimated lifespan. Underestimated life expectancies may contribute to lesser engagement in
health screenings and a general lack of concern for one’s health that may actualize into a shorter
lifespan (Kobayashi, Von Wagner, et al., 2017; Lu et al., 2022; Wuebker, 2012). And if not,
underestimation may contribute to greater informal care burden for their loved ones and/or
financial under preparedness in their final years (Khan et al., 2014).
Future Directions
There is ample opportunity for future research to build on this work. First, researchers
should consider investigating mediating pathways been SSP underestimation and mortality.
Currently, the extent to which underestimation is a self-fulfilling prophecy is unclear. Second,
interventions to improve time perspective should be specifically tested on individuals with high
impact pain. These studies should also investigate how long associated benefits can be sustained.
Prior work assessing the effect of a time perspective intervention on physical activity levels
showed significant benefits through a 10-week follow-up period (Hall & Fong, 2003). Longer
longitudinal study designs should be implemented. For individuals with pain, using these

91
interventions to inspire greater engagement in physical activity and physical therapy could
reduce risks of developing high impact pain and cut down on healthcare expenses for procedures
like TJA.

92
References
2020 Tracker Final, Version 1.0 Data Description and Usage. (2024).
https://hrsdata.isr.umich.edu/sites/default/files/documentation/datadescriptions/1704996077/trk20_Final_V1.pdf
Ayhan, E., Kesmezacar, H., & Akgun, I. (2014). Intraarticular injections (corticosteroid,
hyaluronic acid, platelet rich plasma) for the knee osteoarthritis. World Journal of
Orthopedics, 5(3), 351. https://doi.org/10.5312/wjo.v5.i3.351
Bae, J., Kim, Y.-Y., & Lee, J.-S. (2017). Factors associated with subjective life expectancy:
Comparison with actuarial life expectancy. Journal of Preventive Medicine and Public
Health, 50(4), 240–250. https://doi.org/10.3961/jpmph.17.036
Bago d’Uva, T., O’Donnell, O., & Van Doorslaer, E. (2020). Who can predict their own demise?
Heterogeneity in the accuracy and value of longevity expectations. The Journal of the
Economics of Ageing, 17, 100135. https://doi.org/10.1016/j.jeoa.2017.10.003
Beckwith, H., Thind, A., & Brown, E. A. (2023). Perceived life expectancy among dialysis
recipients: A scoping review. Kidney Medicine, 5(8), 100687.
https://doi.org/10.1016/j.xkme.2023.100687
Berliner, J. L., Brodke, D. J., Chan, V., SooHoo, N. F., & Bozic, K. J. (2017). Can preoperative
patient-reported outcome measures be used to predict meaningful improvement in
function after TKA? Clinical Orthopaedics & Related Research, 475(1), 149–157.
https://doi.org/10.1007/s11999-016-4770-y
Beswick, A. D., Wylde, V., Gooberman-Hill, R., Blom, A., & Dieppe, P. (2012). What
proportion of patients report long-term pain after total hip or knee replacement for
osteoarthritis? A systematic review of prospective studies in unselected patients. BMJ
Open, 2(1), e000435. https://doi.org/10.1136/bmjopen-2011-000435
Biró, A. (2016). Differences between subjective and predicted survival probabilities and their
relation to preventive care use. The B.E. Journal of Economic Analysis & Policy, 16(2),
807–835. https://doi.org/10.1515/bejeap-2015-0154
Blackburn, A. Z., Feder, O., Amakiri, I., Melnic, C. M., Huddleston, J. I., Malchau, H., Kappel,
A., Troelsen, A., & Bedair, H. S. (2024). One-year postoperative patient-reported
outcome measures are associated with three-year to five-year postoperative satisfaction in
total knee arthroplasty. The Journal of Arthroplasty, 39(3), 683–688.
https://doi.org/10.1016/j.arth.2023.08.064
Celidoni, M., Costa-Font, J., & Salmasi, L. (2022). Too healthy to fall sick? Longevity
expectations and protective health behaviours during the first wave of COVID-19.
Journal of Economic Behavior & Organization, 202, 733–745.
https://doi.org/10.1016/j.jebo.2022.08.015

93
Chen, Z., Zhu, D., Hu, X., & Gao, G. (2021). Subjective life expectancy and associated factors
among cancer survivors over 45 years old: Evidence from the CHARLS. Quality of Life
Research, 30(6), 1571–1582. https://doi.org/10.1007/s11136-020-02751-8
Cleveland, R. J., Alvarez, C., Schwartz, T. A., Losina, E., Renner, J. B., Jordan, J. M., &
Callahan, L. F. (2019). The impact of painful knee osteoarthritis on mortality: A
community-based cohort study with over 24 years of follow-up. Osteoarthritis and
Cartilage, 27(4), 593–602. https://doi.org/10.1016/j.joca.2018.12.008
Conner-Spady, B. L., Marshall, D. A., Hawker, G. A., Bohm, E., Dunbar, M. J., Frank, C., &
Noseworthy, T. W. (2014). You’ll know when you’re ready: A qualitative study
exploring how patients decide when the time is right for joint replacement surgery. BMC
Health Services Research, 14(1), 454. https://doi.org/10.1186/1472-6963-14-454
Conner-Spady, B., Sanmartin, C., Johnston, G., McGurran, J., Kehler, M., & Noseworthy, T.
(2008). Willingness of patients to change surgeons for a shorter waiting time for joint
arthroplasty. Canadian Medical Association Journal, 179(4), 327–332.
https://doi.org/10.1503/cmaj.071659
Coyle, L.A & Atkinson, S. (2018). Imagined futures in living with multiple conditions:
Positivity, relationality, and hopelessness. Social Science & Medicine, 198, 53–60.
https://doi.org/10.1016/j.socscimed.2017.12.02
Dagnino, A. P. A., & Campos, M. M. (2022). Chronic pain in the elderly: Mechanisms and
perspectives. Frontiers in Human Neuroscience, 16, 736688.
https://doi.org/10.3389/fnhum.2022.736688
Dang, L., Dong, L., & Mezuk, B. (2020). Shades of blue and gray: A comparison of the Center
for Epidemiologic Studies Depression Scale and the Composite International Diagnostic
Interview for assessment of depression syndrome in later life. The Gerontologist, 60(4),
e242–e253. https://doi.org/10.1093/geront/gnz044
De La Rosa, J. S., Brady, B. R., Ibrahim, M. M., Herder, K. E., Wallace, J. S., Padilla, A. R., &
Vanderah, T. W. (2024). Co-occurrence of chronic pain and anxiety/depression
symptoms in U.S. adults: Prevalence, functional impacts, and opportunities. Pain, 165(3),
666–673. https://doi.org/10.1097/j.pain.0000000000003056
Dell’Isola, A., Turkiewicz, A., Jönsson, T., Rolfson, O., Dahlberg, L., & Englund, M. (2021).
The role of pain and walking difficulties in shaping willingness to undergo joint surgery
for osteoarthritis: Data from the Swedish BOA register. Osteoarthritis and Cartilage
Open, 3(2), 100157. https://doi.org/10.1016/j.ocarto.2021.100157
Doerr, U., & Schulte, K. (2012). Betting on a long life – the role of subjective life expectancy in
the demand for private pension insurance of German households. Schmollers Jahrbuch,
132(2), 233–263. https://doi.org/10.3790/schm.132.2.233

94
Dreinhöfer, K. E., Dieppe, P., Stürmer, T., Gröber-Grätz, D., Flören, M., Günther, K. P., Puhl,
W., & Brenner, H. (2006). Indications for total hip replacement: Comparison of
assessments of orthopaedic surgeons and referring physicians. Annals of the Rheumatic
Diseases, 65(10), 1346–1350. https://doi.org/10.1136/ard.2005.047811
Drummer, D., McAdam, J., Seay, R., Ferrando, A., Bridges, S. L., Singh, J. A., & Bamman, M.
(2021). Osteoarthritis progression: Mitigation and rehabilitation strategies. Frontiers in
Rehabilitation Sciences, 2, 724052. https://doi.org/10.3389/fresc.2021.724052
Duenas, M., Ojeda, B., Salazar, A., Mico, J. A., & Failde, I. (2016). A review of chronic pain
impact on patients, their social environment and the health care system. Journal of Pain
Research, Volume 9, 457–467. https://doi.org/10.2147/JPR.S105892
Durand, W. M., Ortiz-Babilonia, C. D., Badin, D., Wang, K. Y., & Jain, A. (2022). Patient outof-pocket cost burden with elective orthopaedic surgery. Journal of the American
Academy of Orthopaedic Surgeons, 30(14), 669–675. https://doi.org/10.5435/JAAOS-D22-00085
Edwards, R. R., Cahalan, C., Mensing, G., Smith, M., & Haythornthwaite, J. A. (2011). Pain,
catastrophizing, and depression in the rheumatic diseases. Nature Reviews Rheumatology,
7(4), 216–224. https://doi.org/10.1038/nrrheum.2011.2
Elder, T. E. (2013). The predictive validity of subjective mortality expectations: Evidence from
the Health and Retirement Study. Demography, 50(2), 569–589.
https://doi.org/10.1007/s13524-012-0164-2
Falba, T. A., & Busch, S. H. (2005). Survival expectations of the obese: Is excess mortality
reflected in perceptions? Obesity Research, 13(4), 754–761.
https://doi.org/10.1038/oby.2005.85
Fennell, G., Osuna, M., Ailshire, J., & Zajacova, A. (2024). Pain lowers subjective survival
probabilities among middle-aged and older adults. The Journals of Gerontology, Series
B: Psychological Sciences and Social Sciences, 79(7), gbae071.
https://doi.org/10.1093/geronb/gbae071
Fleuren, N., Depla, M. F. I. A., Pasman, H. R. W., Janssen, D. J. A., Onwuteaka-Philipsen, B. D.,
Hertogh, C. M. P. M., & Huisman, M. (2021). Association between subjective remaining
life expectancy and advance care planning in older adults: A cross-sectional study.
Journal of Pain and Symptom Management, 62(4), 757–767.
https://doi.org/10.1016/j.jpainsymman.2021.02.019
Gademan, M. G. J., Hofstede, S. N., Vliet Vlieland, T. P. M., Nelissen, R. G. H. H., & Marangvan De Mheen, P. J. (2016). Indication criteria for total hip or knee arthroplasty in
osteoarthritis: A state-of-the-science overview. BMC Musculoskeletal Disorders, 17(1),
463. https://doi.org/10.1186/s12891-016-1325-z

95
GBD 2016 Disease and Injury Incidence and Prevalence Collaborators. (2017). Global, regional,
and national incidence, prevalence, and years lived with disability for 328 diseases and
injuries for 195 countries, 1990-2016: A systematic analysis for the Global Burden of
Disease Study 2016. The Lancet, 390, 1211–1259. https://doi.org/10.1016/S0140-
6736(16)31678-6
Gellert, P., Ziegelmann, J. P., Lippke, S., & Schwarzer, R. (2012). Future time perspective and
health behaviors: Temporal framing of self-regulatory processes in physical exercise and
dietary behaviors. Annals of Behavioral Medicine, 43(2), 208–218.
https://doi.org/10.1007/s12160-011-9312-y
George, S. Z., Bolognesi, M. P., Bhavsar, N. A., Penrose, C. T., & Horn, M. E. (2022). Chronic
pain prevalence and factors associated with high impact chronic pain following total joint
Aathroplasty: An observational study. The Journal of Pain, 23(3), 450–458.
https://doi.org/10.1016/j.jpain.2021.09.007
Glei, D. A., & Weinstein, M. (2023). Is the pain killing you? Could pain interference be a
warning signal for midlife mortality? SSM - Population Health, 24, 101513.
https://doi.org/10.1016/j.ssmph.2023.101513
Glette, M., Stiles, T. C., Jensen, M. P., Nilsen, T. I. L., Borchgrevink, P. C., & Landmark, T.
(2021). Impact of pain and catastrophizing on the long-term course of depression in the
general population: The HUNT pain study. Pain, 162(6), 1650–1658.
https://doi.org/10.1097/j.pain.0000000000002168
Gompertz, B. (1825). On the nature of the function expressive of the law of human mortality, and
on a new mode of determining the value of life contingencies. 313–385.
Grol-Prokopczyk, H. (2017). Sociodemographic disparities in chronic pain, based on 12-year
longitudinal data. Pain, 158(2), 313–322.
https://doi.org/10.1097/j.pain.0000000000000762
Gunasekara, F. I., Richardson, K., Carter, K., & Blakely, T. (2014). Fixed effects analysis of
repeated measures data. International Journal of Epidemiology, 43(1), 264–269.
https://doi.org/10.1093/ije/dyt221
Hall, P. A., & Fong, G. T. (2003). The effects of a brief time perspective intervention for
increasing physical activity among young adults. Psychology & Health, 18(6), 685–706.
https://doi.org/10.1080/0887044031000110447
Hawker, G. A., Bohm, E., Dunbar, M. J., Faris, P., Jones, C. A., Noseworthy, T., Ravi, B.,
Woodhouse, L. J., & Marshall, D. A. (2023). Patient appropriateness for total knee
arthroplasty and predicted probability of a good outcome. RMD Open, 9(2), e002808.
https://doi.org/10.1136/rmdopen-2022-002808

96
Health and Retirement Study. (2023). 2002-2020 public use dataset (Produced and distributed by
the University of Michigan with funding from the National Institute on Aging (grant
number NIA U01AG009740).) [dataset].
Horgas, A. L., Elliott, A. L., Yang, S., & Guo, Y. (2022). Cross-sectional relationship between
pain intensity and subjective cognitive decline among middle-aged and older adults with
arthritis or joint conditions: Results from a population-based study. SAGE Open
Medicine, 10, 205031212210959. https://doi.org/10.1177/20503121221095923
HRS Staff. (2019). Sampling Weights: Revised for Tracker 2.0 & Beyond (10018). Survey
Research Center, Institute for Social Research, University of Michigan.
https://hrs.isr.umich.edu/publications/biblio/10018
Hsu, H., & Siwiec, R. (2023). Knee Osteoarthritis.
https://www.ncbi.nlm.nih.gov/books/NBK507884/
Hurd, M. D., & McGarry, K. (2002). The predictive validity of subjective probabilities of
survival. The Economic Journal, 112(482), 966–985. https://doi.org/10.1111/1468-
0297.00065
Hurd, M. D., Smith, J. P., & Zissimopoulos, J. M. (2004). The effects of subjective survival on
retirement and Social Security claiming. Journal of Applied Econometrics, 19(6), 761–
775. https://doi.org/10.1002/jae.752
Huynh, C., Puyraimond-Zemmour, D., Maillefert, J. F., Conaghan, P. G., Davis, A. M., Gunther,
K.-P., Hawker, G., Hochberg, M. C., Kloppenburg, M., Lim, K., Lohmander, L. S.,
Mahomed, N. N., March, L., Pavelka, K., Punzi, L., Roos, E. M., Sanchez-Riera, L.,
Singh, J. A., Suarez-Almazor, M. E., … Gossec, L. (2018). Factors associated with the
orthopaedic surgeon’s decision to recommend total joint replacement in hip and knee
osteoarthritis: An international cross-sectional study of 1905 patients. Osteoarthritis and
Cartilage, 26(10), 1311–1318. https://doi.org/10.1016/j.joca.2018.06.013
Karim, J., Weisz, R., Bibi, Z., & Ur Rehman, S. (2015). Validation of the eight-item Center for
Epidemiologic Studies Depression Scale (CES-D) among older adults. Current
Psychology, 34(4), 681–692. https://doi.org/10.1007/s12144-014-9281-y
Khan, M., Rutledge, M. S., & Wu, A. Y. (2014). How do subjective longevity expectations
influence retirement plans? SSRN Electronic Journal.
https://doi.org/10.2139/ssrn.2376923
Kim, H., Meath, T. H. A., Quiñones, A. R., McConnell, K. J., & Ibrahim, S. A. (2021).
Association of Medicare mandatory bundled payment program with the receipt of
elective hip and knee replacement in white, black, and Hispanic beneficiaries. JAMA
Network Open, 4(3), e211772. https://doi.org/10.1001/jamanetworkopen.2021.1772

97
Kim, J.-H., & Kim, J.-M. (2017). Subjective life expectancy is a risk factor for perceived health
status and mortality. Health and Quality of Life Outcomes, 15(1), 190.
https://doi.org/10.1186/s12955-017-0763-0
Kobayashi, L. C., Beeken, R. J., & Meisel, S. F. (2017). Biopsychosocial predictors of perceived
life expectancy in a national sample of older men and women. PLOS ONE, 12(12),
e0189245. https://doi.org/10.1371/journal.pone.0189245
Kobayashi, L. C., Von Wagner, C., & Wardle, J. (2017). Perceived life expectancy is associated
with colorectal cancer screening in England. Annals of Behavioral Medicine, 51(3), 327–
336. https://doi.org/10.1007/s12160-016-9855-z
Kornilov, N., Lindberg, M. F., Gay, C., Saraev, A., Kuliaba, T., Rosseland, L. A., & Lerdal, A.
(2018). Higher physical activity and lower pain levels before surgery predict nonimprovement of knee pain 1 year after TKA. Knee Surgery, Sports Traumatology,
Arthroscopy, 26(6), 1698–1708. https://doi.org/10.1007/s00167-017-4713-5
Kumar, M., Srivastava, S., Muhammad, T., & Saravanakumar, P. (2022). Examining the
association between health status and subjective life expectancy among older Indian
adults based on the mindsponge approach. Humanities and Social Sciences
Communications, 9(1), 359. https://doi.org/10.1057/s41599-022-01368-1
Laditka, S. B., & Laditka, J. N. (2017). Moxie matters: Associations of future orientation with
active life expectancy. Aging & Mental Health, 21(10), 1040–1046.
https://doi.org/10.1080/13607863.2016.1199011
Lahav, Y., Levy, D., Ohry, A., Zeilig, G., Lahav, M., Golander, H., Guber, A.-C., Uziel, O., &
Defrin, R. (2021). Chronic pain and premature aging – The moderating role of physical
exercise. The Journal of Pain, 22(2), 209–218.
https://doi.org/10.1016/j.jpain.2020.08.001
Landon, B. E., Zaslavsky, A. M., Anderson, T. S., Souza, J., Curto, V., & Ayanian, J. Z. (2023).
Differences in use of services and quality of care in Medicare Advantage and Traditional
Medicare, 2010 and 2017. Health Affairs, 42(4), 459–469.
https://doi.org/10.1377/hlthaff.2022.00891
LeDoux, C. V., Lindrooth, R. C., & Stevens-Lapsley, J. E. (2022). The impact of total joint
arthroplasty on long-term physical activity: A secondary analysis of the Health and
Retirement Study. Physical Therapy, 102(1), pzab231.
https://doi.org/10.1093/ptj/pzab231
Liu, J., Tsou, M., & Hammitt, J. K. (2007). Health information and subjective survival
probability: Evidence from Taiwan. Journal of Risk Research, 10(2), 149–175.
https://doi.org/10.1080/13669870701191802

98
Lohman, M. C., Whiteman, K. L., Greenberg, R. L., & Bruce, M. L. (2017). Incorporating
persistent pain in phenotypic frailty measurement and prediction of adverse health
outcomes. The Journals of Gerontology Series A: Biological Sciences and Medical
Sciences, 72(2), 216–222. https://doi.org/10.1093/gerona/glw212
Lopez, E., Neuman, T., Jacobson, G., & Levitt, L. (2020). How much more than Medicare do
private insurers pay? A Review of the Literature. Kaiser Family Foundation.
https://www.kff.org/medicare/issue-brief/how-much-more-than-medicare-do-privateinsurers-pay-a-review-of-the-literature/
Losina, E., Thornhill, T. S., Rome, B. N., Wright, J., & Katz, J. N. (2012). The dramatic increase
in total knee replacement utilization rates in the United States cannot be fully explained
by growth in population size and the obesity epidemic. The Journal of Bone and Joint
Surgery-American Volume, 94(3), 201–207. https://doi.org/10.2106/JBJS.J.01958
Lu, J., Liu, L., Zheng, J., & Zhou, Z. (2022). Interaction between self-perceived disease control
and self-management behaviours among Chinese middle-aged and older hypertensive
patients: The role of subjective life expectancy. BMC Public Health, 22(1), 733.
https://doi.org/10.1186/s12889-022-12990-8
Lützner, C., Postler, A., Beyer, F., Kirschner, S., & Lützner, J. (2019). Fulfillment of
expectations influence patient satisfaction 5 years after total knee arthroplasty. Knee
Surgery, Sports Traumatology, Arthroscopy, 27(7), 2061–2070.
https://doi.org/10.1007/s00167-018-5320-9
Macfarlane, G. J., Barnish, M. S., & Jones, G. T. (2017). Persons with chronic widespread pain
experience excess mortality: Longitudinal results from UK Biobank and meta-analysis.
Annals of the Rheumatic Diseases, 76(11), 1815–1822.
https://doi.org/10.1136/annrheumdis-2017-211476
MacIntyre, N. J., Johnson, J., MacDonald, N., Pontarini, L., Ross, K., Zubic, G., & Majumdar, S.
S. (2015). Characteristics of people with hip or knee osteoarthritis deemed not yet ready
for total joint arthroplasty at triage. Physiotherapy Canada, 67(4), 369–377.
https://doi.org/10.3138/ptc.2014-59
Maradit Kremers, H., Larson, D. R., Crowson, C. S., Kremers, W. K., Washington, R. E.,
Steiner, C. A., Jiranek, W. A., & Berry, D. J. (2015). Prevalence of total hip and knee
replacement in the United States: The Journal of Bone and Joint Surgery-American
Volume, 97(17), 1386–1397. https://doi.org/10.2106/JBJS.N.01141
Maradit Kremers, H., Larson, D. R., Noureldin, M., Schleck, C. D., Jiranek, W. A., & Berry, D.
J. (2016). Long-term mortality trends after total hip and knee arthroplasties: A
population-based study. The Journal of Arthroplasty, 31(6), 1163–1169.
https://doi.org/10.1016/j.arth.2015.12.010

99
Mayfield, C. K., Haglin, J. M., Levine, B., Della Valle, C., Lieberman, J. R., & Heckmann, N.
(2020). Medicare reimbursement for hip and knee arthroplasty from 2000 to 2019: An
unsustainable trend. The Journal of Arthroplasty, 35(5), 1174–1178.
https://doi.org/10.1016/j.arth.2019.12.008
Mehta, B., Ho, K., Bido, J., Memtsoudis, S. G., Parks, M. L., Russell, L., Goodman, S. M., &
Ibrahim, S. (2022). Medicare/Medicaid insurance status is associated with reduced lower
bilateral knee arthroplasty utilization and higher complication rates. Journal of the
American Academy of Orthopaedic Surgeons. Global Research & Reviews, 6(4),
e21.00016. https://doi.org/10.5435/JAAOSGlobal-D-21-00016
Miettinen, T., Sverloff, J., Lappalainen, O.-P., Linton, S. J., Sipilä, K., & Kalso, E. (2022). Sleep
problems in pain patients entering tertiary pain care: The role of pain-related anxiety,
medication use, self-reported diseases, and sleep disorders. Pain, 163(7), e812–e820.
https://doi.org/10.1097/j.pain.0000000000002497
Mirowsky, J. (1999). Subjective life expectancy in the US: Correspondence to actuarial estimates
by age, sex and race. Social Science & Medicine, 49(7), 967–979.
https://doi.org/10.1016/S0277-9536(99)00193-8
Moore, A., & Tumin, D. (2024). Overlap of pain‐related and general measures of disability
among adults with chronic pain. Pain Practice, 24(1), 62–71.
https://doi.org/10.1111/papr.13281
Mota, R. E. M., Tarricone, R., Ciani, O., Bridges, J. F., & Drummond, M. (2012). Determinants
of demand for total hip and knee arthroplasty: A systematic literature review. BMC
Health Services Research, 12(1), 225. https://doi.org/10.1186/1472-6963-12-225
Nahin, R. L., Feinberg, T., Kapos, F. P., & Terman, G. W. (2023). Estimated rates of incident
and persistent chronic pain among US adults, 2019-2020. JAMA Network Open, 6(5),
e2313563. https://doi.org/10.1001/jamanetworkopen.2023.13563
Nakano, N., Shoman, H., Olavarria, F., Matsumoto, T., Kuroda, R., & Khanduja, V. (2020). Why
are patients dissatisfied following a total knee replacement? A systematic review.
International Orthopaedics, 44(10), 1971–2007. https://doi.org/10.1007/s00264-020-
04607-9
Neogi, T. (2013). The epidemiology and impact of pain in osteoarthritis. Osteoarthritis and
Cartilage, 21(9), 1145–1153. https://doi.org/10.1016/j.joca.2013.03.018
Neugarten, B., L. (1968). Adult personality: Toward a psychology of the life cycle. In Middle
age and aging: A reader in social psychology (pp. 137–147). University of Chicago
Press.

100
Núñez, M., Núñez, E., Luis Del Val, J., Ortega, R., Segur, J. M., Hernández, M. V., Lozano, L.,
Sastre, S., & Maculé, F. (2007). Health-related quality of life in patients with
osteoarthritis after total knee replacement: Factors influencing outcomes at 36 months of
follow-up. Osteoarthritis and Cartilage, 15(9), 1001–1007.
https://doi.org/10.1016/j.joca.2007.02.019
O’Brien, C., Fenn, P., & Diacon, S. (2005). How long do people expect to live? Results and
implications. Centre for Risk & Insurance Studies.
O’Connell, A. (2011). How long do we expect to live? A review of the evidence. Journal of
Population Ageing, 4(3), 185–201. https://doi.org/10.1007/s12062-011-9049-1
O’Connor, J. P., Holden, P., & Gagnier, J. J. (2022). Systematic review: Preoperative
psychological factors and total hip arthroplasty outcomes. Journal of Orthopaedic
Surgery and Research, 17(1), 457. https://doi.org/10.1186/s13018-022-03355-3
Olsen, U., Lindberg, M. F., Rose, C., Denison, E., Gay, C., Aamodt, A., Brox, J. I., Skare, Ø.,
Furnes, O., Lee, K., & Lerdal, A. (2022). Factors correlated with physical function 1 year
after total knee arthroplasty in patients with knee osteoarthritis: A systematic review and
meta-analysis. JAMA Network Open, 5(7), e2219636.
https://doi.org/10.1001/jamanetworkopen.2022.19636
Osteoarthritis. (2023). Centers for Disease Control and Prevention.
https://www.cdc.gov/arthritis/types/osteoarthritis.htm
Pacheco-Brousseau, L., Stacey, D., Desmeules, F., Ben Amor, S., Lambert, D., Tanguay, E.,
Hillaby, A., Bechiau, C., Charette, M., & Poitras, S. (2023). Instruments to assess
appropriateness of hip and knee arthroplasty: A systematic review. Osteoarthritis and
Cartilage, 31(7), 847–864. https://doi.org/10.1016/j.joca.2023.02.077
Palazzo, C., Jourdan, C., Descamps, S., Nizard, R., Hamadouche, M., Anract, P., Boisgard, S.,
Galvin, M., Ravaud, P., & Poiraudeau, S. (2014). Determinants of satisfaction 1 year
after total hip arthroplasty: The role of expectations fulfilment. BMC Musculoskeletal
Disorders, 15(1), 53. https://doi.org/10.1186/1471-2474-15-53
Palgi, Y., Shrira, A., Kavé, G., Rubinstein, T., & Shmotkin, D. (2019). Which factors do older
adults consider when estimating the time left for them to live? The Journals of
Gerontology: Series B, 74(1), 69–73. https://doi.org/10.1093/geronb/gby048
Palloni, A., & Novak, B. (2017). Subjective survival expectations and observed survival: How
consistent are they? Vienna Yearbook of Population Research, 1, 187–228.
https://doi.org/10.1553/populationyearbook2016s187
Papachristos, A., & Verropoulou, G. (2020). Subjective survival: An assessment of accuracy
according to individual social profile. Communications in Statistics: Case Studies, Data
Analysis and Applications, 6(1), 19–35. https://doi.org/10.1080/23737484.2019.1644252

101
Papachristos, A., Verropoulou, G., Ploubidis, G. B., & Tsimbos, C. (2020). Factors incorporated
into future survival estimation among Europeans. Demographic Research, 42, 15–56.
https://doi.org/10.4054/DemRes.2020.42.2
Papakostidou, I., Dailiana, Z. H., Papapolychroniou, T., Liaropoulos, L., Zintzaras, E.,
Karachalios, T. S., & Malizos, K. N. (2012). Factors affecting the quality of life after
total knee arthroplasties: A prospective study. BMC Musculoskeletal Disorders, 13(1),
116. https://doi.org/10.1186/1471-2474-13-116
Patel, N. (2021, July 7). Perspective: It’s Time to Advocate for Early Interventional Pain
Management. MedCentral.
Peele, M., & Schnittker, J. (2022). The nexus of physical and psychological pain: Consequences
for mortality and implications for medical sociology. Journal of Health and Social
Behavior, 63(2), 210–231. https://doi.org/10.1177/00221465211064533
Perneger, T. V., Hannouche, D., Miozzari, H. H., & Lübbeke, A. (2019). Symptoms of
osteoarthritis influence mental and physical health differently before and after joint
replacement surgery: A prospective study. PLOS ONE, 14(6), e0217912.
https://doi.org/10.1371/journal.pone.0217912
Perozek, M. (2008). Using subjective expectations to forecast longevity: Do survey respondents
know something we don’t know? Demography, 45(1), 95–113.
https://doi.org/10.1353/dem.2008.0010
Petrosky, E., Harpaz, R., Fowler, K. A., Bohm, M. K., Helmick, C. G., Yuan, K., & Betz, C. J.
(2018). Chronic pain among suicide decedents, 2003 to 2014: Findings from the National
Violent Death Reporting System. Annals of Internal Medicine, 169(7), 448–455.
https://doi.org/10.7326/M18-0830
Piazza, J. R., Charles, S. T., & Almeida, D. M. (2007). Living with chronic health conditions:
Age differences in affective well-being. The Journals of Gerontology: Series B, 62(6),
P313–P321. https://doi.org/10.1093/geronb/62.6.P313
Picone, G., Sloan, F., & Taylor, Jr., D. (2004). Effects of risk and time preference and expected
longevity on demand for medical tests. Journal of Risk and Uncertainty, 28(1), 39–53.
https://doi.org/10.1023/B:RISK.0000009435.11390.23
Pincus, T., & McCracken, L. M. (2013). Psychological factors and treatment opportunities in low
back pain. Best Practice & Research Clinical Rheumatology, 27(5), 625–635.
https://doi.org/10.1016/j.berh.2013.09.010
Pitcher, M. H., Von Korff, M., Bushnell, M. C., & Porter, L. (2019). Prevalence and profile of
high-impact chronic pain in the United States. The Journal of Pain, 20(2), 146–160.
https://doi.org/10.1016/j.jpain.2018.07.006

102
Riddle, D. L., Jiranek, W. A., & Hayes, C. W. (2014). Use of a validated algorithm to judge the
appropriateness of total knee arthroplasty in the United States: A multicenter longitudinal
cohort study. Arthritis & Rheumatology, 66(8), 2134–2143.
https://doi.org/10.1002/art.38685
Riddle, D. L., Perera, R. A., Jiranek, W. A., & Dumenci, L. (2015). Using surgical
appropriateness criteria to examine outcomes of total knee arthroplasty in a United States
sample. Arthritis Care & Research, 67(3), 349–357. https://doi.org/10.1002/acr.22428
Rikard, S. M., Strahan, A. E., Schmit, K. M., & Guy, G. P. (2023). Chronic pain among adults—
United States, 2019–2021. MMWR. Morbidity and Mortality Weekly Report, 72(15),
379–385. https://doi.org/10.15585/mmwr.mm7215a1
Rizzo, E. A., Phillips, R. D., Brown, J. T., Leary, E. V., & Keeney, J. A. (2023). Obesity severity
predicts patient dissatisfaction after total knee arthroplasty. The Journal of Arthroplasty,
38(12), 2492–2496. https://doi.org/10.1016/j.arth.2023.05.063
Roseen, E. J., Rajendran, I., Stein, P., Fredman, L., Fink, H. A., LaValley, M. P., & Saper, R. B.
(2021). Association of back pain with mortality: A systematic review and meta-analysis
of cohort studies. Journal of General Internal Medicine, 36(10), 3148–3158.
https://doi.org/10.1007/s11606-021-06732-6
Sakakibara, R., & Ishii, Y. (2020). Examination on how emotion regulation mediates the
relationship between future time perspective and well-being: A counter-evidence to the
socioemotional selectivity theory. European Journal of Ageing, 17(1), 21–30.
https://doi.org/10.1007/s10433-019-00522-0
Schneider, S., Junghaenel, D. U., Keefe, F. J., Schwartz, J. E., Stone, A. A., & Broderick, J. E.
(2012). Individual differences in the day-to-day variability of pain, fatigue, and wellbeing in patients with rheumatic disease: Associations with psychological variables. Pain,
153(4), 813–822. https://doi.org/10.1016/j.pain.2012.01.001
Shan, L., Shan, B., Graham, D., & Saxena, A. (2014). Total hip replacement: A systematic
review and meta-analysis on mid-term quality of life. Osteoarthritis and Cartilage, 22(3),
389–406. https://doi.org/10.1016/j.joca.2013.12.006
Shichman, I., Roof, M., Askew, N., Nherera, L., Rozell, J. C., Seyler, T. M., & Schwarzkopf, R.
(2023). Projections and epidemiology of primary hip and knee arthroplasty in Medicare
patients to 2040-2060. JBJS Open Access, 8(1).
https://doi.org/10.2106/JBJS.OA.22.00112
Singh, J. A., Yu, S., Chen, L., & Cleveland, J. D. (2019). Rates of total joint replacement in the
United States: Future projections to 2020–2040 using the national inpatient sample. The
Journal of Rheumatology, 46(9), 1134–1140. https://doi.org/10.3899/jrheum.170990

103
Sloan, F. A., George, L. K., & Hu, L. (2013). Longer term effects of total knee arthroplasty from
a national longitudinal study. Journal of Aging and Health, 25(6), 982–997.
https://doi.org/10.1177/0898264313494799
Smith, D., Wilkie, R., Uthman, O., Jordan, J. L., & McBeth, J. (2014). Chronic pain and
mortality: A systematic review. PLoS ONE, 9(6), e99048.
https://doi.org/10.1371/journal.pone.0099048
Sonnega, A., Faul, J. D., Ofstedal, M. B., Langa, K. M., Phillips, J. W., & Weir, D. R. (2014).
Cohort profile: The Health and Retirement Study (HRS). International Journal of
Epidemiology, 43(2), 576–585. https://doi.org/10.1093/ije/dyu067
Sorel, J. C., Veltman, E. S., Honig, A., & Poolman, R. W. (2019). The influence of preoperative
psychological distress on pain and function after total knee arthroplasty: A systematic
review and meta-analysis. The Bone & Joint Journal, 101-B(1), 7–14.
https://doi.org/10.1302/0301-620X.101B1.BJJ-2018-0672.R1
Tennant, F. (2012). Sudden, Unexpected Death in Chronic Pain Patients.
https://www.practicalpainmanagement.com/sudden-unexpected-death-chronic-painpatients
Understanding Medicare Advantage Plans (pp. 12–13). (n.d.). U.S. Department of Health &
Human Services Centers for Medicare & Medicaid Services.
https://www.medicare.gov/Pubs/pdf/12026-Understanding-Medicare-AdvantagePlans.pdf
Vajapey, S. P., McKeon, J. F., Krueger, C. A., & Spitzer, A. I. (2021). Outcomes of total joint
arthroplasty in patients with depression: A systematic review. Journal of Clinical
Orthopaedics and Trauma, 18, 187–198. https://doi.org/10.1016/j.jcot.2021.04.028
Van Orden, K., & Conwell, Y. (2011). Suicides in Late Life. Current Psychiatry Reports, 13(3),
234–241. https://doi.org/10.1007/s11920-011-0193-3
Van Zaanen, Y., Siertsema, T., Kievit, A. J., Van Geenen, R. C. I., Pahlplatz, T. M. J.,
Hoozemans, M. J. M., Blankevoort, L., Schafroth, M. U., Haverkamp, D., Vervest, T. M.
J. S., Das, D. H. P. W., Scholtes, V. A., & Kuijer, P. P. F. M. (2023). Only low patients’
expectations are prognostic for dissatisfaction with performing work-related kneestraining activities after total knee arthroplasty: A prospective multicenter cohort study.
Archives of Physical Medicine and Rehabilitation, 104(12), 2051–2058.
https://doi.org/10.1016/j.apmr.2023.05.004
Vanajan, A., Bültmann, U., & Henkens, K. (2020). Why do older workers with chronic health
conditions prefer to retire early? Age and Ageing, 49(3), 403–410.
https://doi.org/10.1093/ageing/afz180

104
Vanajan, A., & Gherdan, C. (2022). Associations between existing and newly diagnosed chronic
health conditions and change in subjective life expectancy: Results from a panel study.
SSM - Population Health, 20, 101271. https://doi.org/10.1016/j.ssmph.2022.101271
Vartiainen, P., Roine, R. P., Kalso, E., & Heiskanen, T. (2022). Worse health‐related quality of
life, impaired functioning and psychiatric comorbidities are associated with excess
mortality in patients with severe chronic pain. European Journal of Pain, 26(5), 1135–
1146. https://doi.org/10.1002/ejp.1938
Vela, J., Cordtz, R. L., Kristensen, S., Torp-Pedersen, C., Petersen, K. K., Arendt-Nielsen, L., &
Dreyer, L. (2021). Is pain associated with premature mortality in patients with psoriatic
arthritis? A nested case–control study using the DANBIO Register. Rheumatology,
60(11), 5216–5223. https://doi.org/10.1093/rheumatology/keab192
Verra, W. C., Witteveen, K. Q., Maier, A. B., Gademan, M. G. J., Van Der Linden, H. M. J., &
Nelissen, R. G. H. H. (2016). The reason why orthopaedic surgeons perform total knee
replacement: Results of a randomised study using case vignettes. Knee Surgery, Sports
Traumatology, Arthroscopy, 24(8), 2697–2703. https://doi.org/10.1007/s00167-015-
3961-5
Villadsen, A. R., & Wulff, J. N. (2021). Are you 110% sure? Modeling of fractions and
proportions in strategy and management research. Strategic Organization, 19(2), 312–
337. https://doi.org/10.1177/1476127019854966
Vos, T., Abajobir, A. A., Abate, K. H., Abbafati, C., Abbas, K. M., Abd-Allah, F., Abdulkader,
R. S., Abdulle, A. M., Abebo, T. A., Abera, S. F., Aboyans, V., Abu-Raddad, L. J.,
Ackerman, I. N., Adamu, A. A., Adetokunboh, O., Afarideh, M., Afshin, A., Agarwal, S.
K., Aggarwal, R., … Murray, C. J. L. (2017). Global, regional, and national incidence,
prevalence, and years lived with disability for 328 diseases and injuries for 195 countries,
1990–2016: A systematic analysis for the Global Burden of Disease Study 2016. The
Lancet, 390(10100), 1211–1259. https://doi.org/10.1016/S0140-6736(17)32154-2
Wang, Y., Owen, A., Franks, A., Ackerman, I., Fitzgerald, S. M., Liew, S., Woods, R. L.,
Wluka, A. E., McNeil, J. J., & Cicuttini, F. M. (2023). Functional outcomes following
knee replacement in community-dwelling older adults. BMC Geriatrics, 23(1), 261.
https://doi.org/10.1186/s12877-023-03925-y
Wei, J., Hou, R., Zhang, X., Xu, H., Xie, L., Chandrasekar, E. K., Ying, M., & Goodman, M.
(2019). The association of late-life depression with all-cause and cardiovascular mortality
among community-dwelling older adults: Systematic review and meta-analysis. British
Journal of Psychiatry, 215(2), 449–455. https://doi.org/10.1192/bjp.2019.74
Weir, D. (2016). Validating mortality ascertainment in the Health and Retirement Study.
University of Michigan.
https://hrspubs.sites.uofmhosting.net/sites/default/files/biblio/Weir_mortality_ascertainm
ent.pdf

105
Weller, B. E., Bowen, N. K., & Faubert, S. J. (2020). Latent class analysis: A guide to best
practice. Journal of Black Psychology, 46(4), 287–311.
https://doi.org/10.1177/0095798420930932
Winter, J. E., MacInnis, R. J., Wattanapenpaiboon, N., & Nowson, C. A. (2014). BMI and allcause mortality in older adults: A meta-analysis. The American Journal of Clinical
Nutrition, 99(4), 875–890. https://doi.org/10.3945/ajcn.113.068122
Woodland, N., Takla, A., Estee, M. M., Franks, A., Bhurani, M., Liew, S., Cicuttini, F. M., &
Wang, Y. (2023). Patient-reported outcomes following total knee replacement in patients
aged 65 years and over—A systematic review. Journal of Clinical Medicine, 12(4), 1613.
https://doi.org/10.3390/jcm12041613
Wray, C. M., Khare, M., & Keyhani, S. (2021). Access to care, cost of care, and satisfaction with
care among adults with private and public health insurance in the US. JAMA Network
Open, 4(6), e2110275. https://doi.org/10.1001/jamanetworkopen.2021.10275
Wu, S., Stevens, R., & Thorp, S. (2014). Die young or live long: Modeling subjective survival
probabilities. https://www.netspar.nl/assets/uploads/E20150128_IPW-paper-thorp.pdf
Wuebker, A. (2012). Who gets a mammogram amongst European women aged 50-69 years?
Health Economics Review, 2(1), 6. https://doi.org/10.1186/2191-1991-2-6
Wylde, V., Dieppe, P., Hewlett, S., & Learmonth, I. D. (2007). Total knee replacement: Is it
really an effective procedure for all? The Knee, 14(6), 417–423.
https://doi.org/10.1016/j.knee.2007.06.001
Zacher, M., Wang, J., & Short, S. E. (2022). The contributions of hypertension diagnosis and
blood pressure control to subjective life expectancy in a representative sample of older
U.S. adults. The Journals of Gerontology: Series B, 77(2), 378–388.
https://doi.org/10.1093/geronb/gbab022
Zajacova, A., Grol-Prokopczyk, H., & Zimmer, Z. (2021). Pain trends among American adults,
2002–2018: Patterns, disparities, and correlates. Demography, 58(2), 711–738.
https://doi.org/10.1215/00703370-8977691
Zelaya, C. E., Dahlhamer, J. M., Lucas, J. W., & Connor, E. M. (2020). Chronic pain and highimpact chronic pain among U.S. adults, 2019. NCHS Data Brief, 390, 1–8.
Zimmer, Z., & Rubin, S. (2016). Life expectancy with and without pain in the U.S. elderly
population. The Journals of Gerontology Series A: Biological Sciences and Medical
Sciences, 71(9), 1171–1176. https://doi.org/10.1093/gerona/glw028
Zimmer, Z., & Zajacova, A. (2018). Persistent, consistent, and extensive: The trend of increasing
pain prevalence in older Americans. The Journals of Gerontology: Series B.
https://doi.org/10.1093/geronb/gbx162

106
Chapter 2 Supplementary Material
To assess whether the pain-SSP relationship significantly differed between men and
women, we conducted the same target-age stratified fractional logit separately for each gender
(Table S3) , then tested whether the pain coefficients were equal across the gendered models
(Table S4).
The next supplementary analysis (Table S5) tested whether change in pain predicts
change in SSPs. To do this, we conducted an OLS regression to assess whether changes in pain
across the respondents’ last two survey observations predicted corresponding change in SSPs.
We did not use a fractional logit model here as the SSP change variable ranged from -100 to 100,
not 0 to 1. Only individuals who reported pain in either or both waves were included in this
analysis (N = 12,759). Change in pain between these two waves was coded based on our original
pain categorization: (0) no change, (1) decreased pain, and (2) increased pain. Change in SSPs
was calculated as the difference in reported SSPs across the two waves. The model controlled for
baseline (wave 1) SSPs, age, age-squared, gender, race, education, marital status, depression, and
the seven chronic health conditions defined previously. Age was mean-centered for ease of
interpretation, and age-squared was derived from this mean-centered variable; the squared
covariate was included to capture the potentially strong nonlinear relationship between age and
SSPs (Gompertz, B., 1825). We additionally controlled for whether the respondent’s SSP target
age changed between the two waves.

107
Table S1. Sample composition in the missing and analytic samples
Total NonMissing
Observations
(N=153,282)
Total Missing
Observations
(N= 20,580)
Significance
test
M / % SD M / % SD
Pain
No pain (ref) 65.3% 59.0%
Non-interfering mild or moderate pain 12.2% 11.6% †
Severe and/or Interfering pain 22.5% 29.4% ***
Age at interview 65.1 9.6 68.6 10.0 ***
Female 54.7% 46.7% ***
Race/ethnicity
White (ref) 79.7% 64.5%
Black 9.3% 14.7% ***
Hispanic 7.6% 15.9% ***
Other 3.4% 4.8% ***
Education
Less than High School (ref) 18.0% 39.6%
High School 28.4% 27.7% ***
Some College 25.7% 17.6% ***
College and above 27.9% 15.1% ***
Marital Status
Married/Partnered (ref) 66.4% 65.7%
Separated/Divorced 14.4% 10.2% ***
Widowed 13.5% 19.6% ***
Never Married 5.7% 4.5% ***
Depression 20.0% 33.8% ***
Heart Disease 20.7% 26.8% ***
Lung Disease 8.7% 10.9% ***
Diabetes 19.3% 24.8% ***
High Blood Pressure 52.1% 59.5% ***
Arthritis 53.6% 57.5% ***
Ever had Cancer 13.3% 13.8% †
Ever had Stroke 6.3% 12.3% ***
Subjective Survival Probabilities 53.7 32.3 50.6 33.5 **
† p<.10 * p<.05; ** p<.01; *** p<.001
Note. Continuous variables were compared using t-tests, while categorical variables were
analyzed using chi-square tests.

108

109

110

111
Table S5. OLS regression using change in pain between two last waves in survey
predicting corresponding change in SSPs (N = 12,759)
b Robust se
Two-wave Change in Pain
Same (ref)
Decreased pain 1.30 † 0.70
Increased pain 0.56 0.66
Wave 1 SSP -0.42 *** 0.01
Target age changed in Wave 2 1.85 * 0.72
Age 0.23 *** 0.03
Age-squared -0.04 *** 0.00
Female 2.42 *** 0.57
Race/ethnicity
White (ref)
Black 7.82 *** 0.89
Hispanic -0.14 1.55
Other 2.14 1.06
Education
Less than High School (ref)
High School -0.04 0.79
Some College 1.94 * 0.81
College or above 2.61 ** 0.88
Marital Status
Married/Partnered (ref)
Separated/Divorced -1.60 * 0.80
Widowed -2.20 ** 0.77
Never Married -0.32 1.22
Health Conditions
Depression -4.16 *** 0.62
Heart Disease -1.19 † 0.62
Lung Disease -1.59 * 0.74
Diabetes -1.57 * 0.62
High Blood Pressure -1.01 † 0.60
Arthritis 0.00 0.70
Ever had Cancer -2.62 *** 0.65
Ever had Stroke -0.01 0.83
Constant 16.32 *** 1.28
R2 0.24
† p<.10 * p<.05; ** p<.01; *** p<.001

112
Chapter 3 Supplementary Material

113

114

115

116
Chapter 4 Supplementary Material
Table S1. Sample characteristics of total analytic sample and respondents who underwent TJA but
were excluded due to too few observations.
Total sample
(N =1,865)
Total Missing Sample
(N = 751)
M / % SD M / % SD
Pain
No pain 35.3% 32.5%
Mild pain 13.3% 14.7%
Moderate pain 39.8% 39.2%
Severe pain 11.7% 13.6%
Arthritis 97.7% 97.4%
Treatment 72.4% 84.6% ***
No symptom improvement 95.2% 72.8%
Pain / Arthritis limitations 71.8% 94.7%
Difficulty walking several blocks 57.2% 49.5%
Age at interview 66.9 8.2 67.5 8.7
Female 60.8% 63.7%
Race/ethnicity
White 86.8% 83.4%
Black 6.6% 9.1% †
Other 2.4% 2.4%
Hispanic 4.1% 5.1%
Education
Less than High School 16.0% 11.9%
High School 30.9% 29.2%
Some College 25.2% 30.4% †
College and above 27.8% 28.5%
Marital Status
Married/Partnered 70.0% 70.4%
Separated/Divorced 11.0% 10.8%
Widowed 15.6% 13.9%
Never Married 3.4% 5.0%
Health
Depression 24.2% 24.3%
Heart Disease 20.6% 23.5%
Lung Disease 7.9% 8.5%
Diabetes 18.0% 22.3%
High Blood Pressure 60.7% 61.3%
Ever had Cancer 14.1% 19.0% †

117
Ever had Stroke 6.1% 9.8% *
ADL 7.4% 6.4%
Poor Self-rated Health 6.3% 8.5%
BMI
Less than 23 6.1% 8.5% *
23 to 29.99 (ref) 41.0% 43.2%
30-34.99 27.3% 23.1%
35+ 25.6% 25.2%
Insurance
Non-HMO Medicare (ref) 46.2% 32.9%
HMO Medicare 12.5% 21.2% ***
Private 36.6% 38.9% †
Other (no insurance, public) 4.6% 7.0% *
Doctor visits in last 2 years
0 to 4 (ref) 26.6% 32.7%
5 to 10 40.2% 34.3% †
11+ 33.1% 33.0%
Household Wealth Quartile
Quartile 1 (lowest; ref) 20.4% 19.1%
Quartile 2 24.2% 29.3%
Quartile 3 25.8% 22.2%
Quartile 4 (highest) 29.6% 29.3%
† p<.10 * p<.05; ** p<.01; *** p<.001
Notes. The analytic sample served as the reference group for significance testing. Testing between
subgroups was performed using t-tests for continuous variables, and chi-square tests for
categorical variables.

118

119

120

121
Table S4. Sample characteristics for HRS respondents who underwent total knee or hip
arthroplasty by operation year.
2004-2008
(N = 804)
2010-2014
(N=648)
2016-2018
(N=413)
M / % SD M / % SD M / % SD
"Need" class 72.4% 69.1% 64.4% **
Pain
No pain 39.0% 35.7% 30.1%
Mild pain 9.8% 13.1% † 17.9% **
*
Moderate pain 39.2% 41.2% 38.6%
Severe pain 11.9% 9.9% 13.5%
Arthritis 95.7% 98.2% 100.0%
Treatment 66.7% 73.3% 80.7% †
No symptom improvement 93.2% 96.1% 97.3%
Pain / Arthritis limitations 72.8% 69.5% † 73.9% †
Difficulty walking several
blocks 59.5% 51.2% † 63.1% *
Age at interview 67.5 7.9 67.1 8.4 66.0 8.5 *
Female 64.2% 61.1% 56.1% †
Race/ethnicity
White 87.6% 87.7% 84.7%
Black 7.5% 5.2% † 7.4%
Other 1.6% 1.8% 4.2% *
Hispanic 3.2% 5.3% † 3.7%
Education
Less than High School 18.8% 16.5% 12.0%
High School 37.7% 30.9% 22.6%
Some College 20.7% 25.4% † 30.7% **
College and above 22.8% 27.3% 34.6% **
Marital Status
Married/Partnered 69.1% 73.9% 66.3%
Separated/Divorced 10.5% 8.9% 14.1%
Widowed 17.6% 15.0% 13.9%
Never Married 2.8% 2.2% 5.8% †
Health
Depression 27.7% 20.6% ** 24.3%
Heart Disease 20.7% 21.7% 19.0%
Lung Disease 8.6% 6.7% 8.6%
Diabetes 15.0% 18.6% 20.9% *
High Blood Pressure 59.0% 64.0% 58.7%
Ever had Cancer 12.5% 18.3% ** 10.8%

122
Ever had Stroke 6.3% 6.3% 5.7%
Poor Self-rated Health 6.4% 5.7% 7.0%
BMI
Less than 23 5.8% 6.8% 5.5%
23 to 29.99 (ref) 43.8% 40.5% 38.2%
30-34.99 26.4% 27.0% 28.8%
35+ 24.0% 25.7% 27.5%
Insurance
Non-HMO Medicare (ref) 51.5% 49.0% 36.2%
HMO Medicare 10.7% 11.7% 15.8% **
Private 33.2% 35.5% 42.2% **
Other (no insurance, public) 4.6% 3.7% 5.8%
Doctor visits in last 2 years
0 to 4 (ref) 24.6% 28.4% 27.0%
5 to 10 41.9% 39.0% 39.7%
11+ 33.6% 32.6% 33.3%
Household Wealth Quartile
Quartile 1 (lowest; ref) 20.9% 19.2% 21.2%
Quartile 2 25.3% 24.3% 22.8%
Quartile 3 26.7% 28.1% 22.0%
Quartile 4 (highest) 27.2% 28.4% 34.1%
† p<.10 * p<.05; ** p<.01; *** p<.001
Note. Individuals who underwent TJA in 2004-2008 served as the reference group for
significance testing. Testing between subgroups was performed using t-tests for continuous
variables, and chi-square tests for categorical variables.

123

124

125

126

127

128

Abstract (if available)

Abstract Background: The experience of severe or activity-limiting pain (i.e., high impact pain) is increasingly common among middle-aged and older Americans. High impact pain is associated with poor mental health, disability, and heightened mortality risk. It is not known whether high impact pain is associated with subjective survival probabilities (SSPs), or one’s own perceived chances of living to a given age, which meaningfully predict engagement in health-seeking behaviors and responsible financial decisions.
Objective: I aim to test the link between high impact pain and SSPs, and assess whether reported longevity expectations are accurate to respondents’ own lifespans. Additionally, I examine whether the gold-standard treatment for osteoarthritis, total joint arthroplasty (TJA), increases individuals’ SSPs.

Method: Across three studies, I use data from the Health and Retirement Study, a nationally representative study of American adults aged 51 and older. In Study 1, I use age-stratified fractional logit regressions on repeated cross-sectional data from 2000-2018 (N = 31,773; aged 51-89) to examine the relationship between pain and SSPs. In Study 2 (N = 12,835; aged 57-89), I use a cox proportional hazards model and a multinomial logistic regression to assess both the actual survival probabilities and accuracy of SSPs among respondents with pain and depression. In Study 3 (N = 1,865 who underwent TJA; aged 51-89), I employ latent class analysis and fixed effects models to examine the pain, physical functioning, depressive, and SSP outcomes following a TJA procedure. These analyses are stratified by membership into high and low TJA “need” classes.

Results: Respondents with high impact pain reported significantly lower average SSPs than those with no or non-interfering pain. Individuals with pain or depression were more likely to underestimate SSPs relative to their actual lifespans. The SSPs of respondents with high impact pain did not improve following TJA despite clear post-operative improvements in pain intensity, physical functioning, and depressive symptoms.

Conclusion: Middle-aged and older adults with high impact pain have low and underestimated longevity expectations that do not seem to be amenable to effective pain relief. I discuss the possibility of a self-fulfilling prophecy, the consequences of underestimated SSPs, and highlight alternative interventions to foster accurate SSPs among individuals experiencing pain.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Social determinants of physiological health and mortality in China

PDF

Age integration in late life: sociodemographic & psychosocial correlates of intergenerational-only, peer-only, and age-integrated social networks

PDF

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

PDF

Statistical algorithms for examining gene and environmental influences on human aging

PDF

Self-perceptions of Aging in the Context of Neighborhood and Their Interplay in Late-life Cognitive Health

PDF

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

PDF

Life course implications of adverse childhood experiences: impacts on elder mistreatment, subjective cognitive decline, and caregivers' health

PDF

Is stress exposure enough? Race/ethnic differences in the exposure and appraisal of chronic stressors among older adults

PDF

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

PDF

Internet communication use, psychological functioning and social connectedness at older ages

PDF

Health differences between the elderly in Japan and the United States

PDF

Mechanisms of stress effects on learning and decision making in younger and older adults

PDF

Racial disparities in advance care planning and directives completion and end-of-life outcomes in the United States

PDF

Eliciting perspectives on palliative care: outpatient visits, advance care planning, and the impact of COVID-19

PDF

Three essays on modifiable determinants of shingles: risk factors for shingles incidence and factors affecting timing of vaccine uptake

PDF

The impact of emotion, valence, and arousal on differential memory processes in younger and older people

PDF

Examination of the long-term psychosocial and functioning effects of a healthy living, life management behavior intervention for older adults

PDF

Structure and function of the locus coeruleus across the lifespan

PDF

Computational approaches to identify genetic regulators of aging and late-life mortality

PDF

Effects of sugar and fiber consumption in minority adolescents and self-tracking as a potential dietary intervention tool

Asset Metadata

Creator Fennell, Gillian Rose (author)

Core Title Estimating survival in the face of pain: evidence from the health and retirement study

School Leonard Davis School of Gerontology

Degree Doctor of Philosophy

Degree Program Gerontology

Degree Conferral Date 2024-08

Publication Date 07/12/2024

Defense Date 04/03/2024

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

Tag disability,joint replacement,mortality,OAI-PMH Harvest,osteoarthritis,subjective life expectancy

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Zelinski, Elizabeth (committee chair), Ailshire, Jennifer (committee member), Grol-Prokopczyk, Hanna (committee member), Jacobson, Mireille (committee member)

Creator Email gfennell@usc.edu,gfennell1997@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113997NAY

Unique identifier UC113997NAY

Identifier etd-FennellGil-13224.pdf (filename)

Legacy Identifier etd-FennellGil-13224

Document Type Dissertation

Format theses (aat)

Rights Fennell, Gillian Rose

Internet Media Type application/pdf

Type texts

Source 20240712-usctheses-batch-1181 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

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

Repository Email cisadmin@lib.usc.edu