Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Mar 1.
Published in final edited form as: Health Psychol. 2022 Feb 14;41(3):225–234. doi: 10.1037/hea0001159

Psychological Well-being, Education, and Mortality

Jennifer Morozink Boylan 1, Justin L Tompkins 1, Patrick M Krueger 1
PMCID: PMC9901287  NIHMSID: NIHMS1864145  PMID: 35157480

Abstract

Objectives.

Research on older adults often focuses on mitigating health risks, and less is known about protective factors that contribute to longer, healthier lives. We examine longitudinal associations between psychological well-being and mortality among a national sample of older adults and test competing hypotheses about whether the education/mortality association depends on the level of psychological well-being.

Methods.

We use six waves (2006–2016) of the Health and Retirement Study, a national sample of adults over age 50 (n = 21,172), with 14 years of mortality follow-up. Psychological well-being is measured up to three times and includes positive affect, life satisfaction, purpose in life, social support, and optimism. Discrete-time survival models examine (1) the association between time-varying psychological well-being and mortality, and (2) interactions between psychological well-being and education on mortality.

Results.

Higher purpose in life, positive affect, optimism, social support, and life satisfaction predicted lower mortality. A one standard deviation increase in most measures of psychological well-being was associated with a 2–4 year increase in life expectancy at age 50. Positive affect and purpose in life moderated the education/mortality association —the inverse association between education and mortality was stronger for those with high psychological well-being.

Conclusions.

We find strong evidence that psychological well-being predicts lower mortality risk and modifies the association between education and mortality. The inverse association between education and mortality becomes stronger at higher levels of purpose in life and positive affect. Therefore, efforts to promote life satisfaction, social support, and optimism may support longer lives without widening education disparities.

Keywords: purpose in life, positive affect, life satisfaction, optimism, social support

Given aging of the U.S. population, research efforts are needed to understand factors that reduce mortality risk and support a happier and healthier population of older adults. Extant science often frames older adults as “at risk,” with fewer studies examining strengths and protective factors among older adults. We focus on psychological well-being, a positive outcome in its own right that may contribute to longer, healthier lives. In so doing, we address methodological shortcomings in the literature that limit our understanding of the importance of psychological well-being for population health, especially among older adults. Nearly all studies linking psychological well-being to mortality utilize one-time assessments of well-being and rarely examine whether distinct dimensions of psychological well-being have unique associations with mortality. Psychological well-being is multidimensional and includes feelings of happiness and life satisfaction, having quality social relationships, finding meaning and purpose in life, and being hopeful about the future. Using longitudinal data from a national sample of older adults, we advance prior work by examining associations between repeated measurements of five dimensions of psychological well-being and the risk of death.

Psychological well-being may also modify educational disparities in survival. Education has a strong, graded association with the risk of death, even among older adults (Masters et al., 2012). Despite higher mortality risks among the less educated, many adults with low education survive to older ages. Over 73% of those with less than a high school degree, and over 82% of those with a high school degree, survive from age 25 to age 65 (Krueger et al., 2019). Additionally, there is notable variability in psychological well-being within educational strata (Ryff, 2016), and interventions demonstrate that psychological well-being is modifiable, even at old ages (Weiss et al., 2016). Clarifying the joint role of psychological well-being and education for mortality can support efforts to promote population health, especially among older adults who may be more susceptible to interventions that target well-being than efforts to promote educational attainment. Thus, we further advance prior work by testing competing hypotheses (i.e., based on social structural and self-deterministic perspectives) about whether the association between education and mortality depends on levels of well-being.

Psychological Well-being and Mortality

Psychological well-being reflects the presence of wellness, which is distinct from the absence of mental illness. Research on how best to conceptualize psychological well-being draws from two broad literatures. Hedonic well-being includes subjectively pleasant life experiences and is often measured with positive affect and life satisfaction (Ryan & Deci, 2001). Eudaimonic well-being focuses on the fulfillment of human potentials, including whether one can pursue purposefully meaningful goals (purpose in life) and experience social ties characterized by warmth and trust (positive social support; (Ryan & Deci, 2001; Ryff, 1989)). Finally, optimism is a valued aspect of psychological functioning that involves having positive outlooks about the future as well as positive interpretations for the past (Scheier et al., 1994).

Psychological well-being may be associated with mortality for five reasons. First, those with greater well-being are more likely to engage in healthy behaviors, such as getting cancer screenings, smoking less, sleeping better, and exercising more (Boehm & Kubzansky, 2012; Steptoe, 2019). Second, well-being may support efforts to overcome barriers to better health by increasing self-efficacy (Rush et al., 2019) and using adaptive (e.g., problem-focused) coping strategies (Folkman, 2008). Third, psychological well-being is associated with improved cardiovascular, metabolic, and immune functioning (Boylan et al., 2020; Boylan & Ryff, 2015), which are primary biological mechanisms underlying many chronic diseases. Fourth, individuals with greater psychological well-being have lower morbidity, especially coronary heart disease and cardiovascular events (Boehm & Kubzansky, 2012; Cohen et al., 2016; Rasmussen et al., 2009; Steptoe, 2019). Finally, psychological well-being buffers individuals against the physiological toll of psychosocial stressors, which is an oft-cited mechanism contributing to educational disparities in mortality (Krueger & Chang, 2008). Consistent with these reasons, systematic reviews and meta-analyses show reduced mortality risk among those with higher hedonic well-being (Chida & Steptoe, 2008; Martín-María et al., 2017; Steptoe, 2019), higher purpose in life (Cohen et al., 2016), stronger social relationships (Holt-Lunstad et al., 2010), and higher optimism (Boehm & Kubzansky, 2012; Rasmussen et al., 2009).

We advance research on psychological well-being and mortality in four ways. First, we examine multiple dimensions of psychological well-being. Past psychometric evidence has shown independence among hedonic well-being, eudaimonic well-being, and optimism, but prior research has not examined whether they have distinct associations with mortality. Our modeling of multiple psychological well-being dimensions independently and simultaneously allows us to compare effect sizes and identify which dimensions are most important for survival. Second, we use validated, multiple-item measures of well-being, which allows us to appropriately assess complex constructs, unlike much of the literature, which relies on single-item indicators (Ryff et al., 2020). Third, we use a large, nationally representative sample of older adults. Existing studies commonly rely on convenience or clinical samples that are not representative of the broader population. Finally, we use repeated measures of psychological well-being. Existing work often includes only a single baseline assessment, which assumes that psychological well-being is stable and trait-like across follow-up. Using repeated measures is important when studying a population that is likely to experience major life transitions (e.g., retirement, deaths of loved ones, health changes) that may be associated with both psychological well-being and the risk of death.

Hypothesis 1:

Each psychological well-being dimension (i.e., positive affect, life satisfaction, purpose in life, positive social support, and optimism) will have an inverse association with the risk of death.

Hypothesis 2:

Each psychological well-being dimension will have independent, inverse associations with the risk of death when adjusting for other well-being measures.

Psychological Well-being, Education, and Mortality

We examine competing theoretical perspectives about how psychological well-being may modify the association between education and survival. Education predicts survival through mechanisms including healthier behaviors, greater pecuniary resources, and cognitive skills that allow individuals to manage health risks (Pampel et al., 2010). Evidence increasingly supports a causal association between education and mortality (Kawachi et al., 2010). By examining well-being as a moderator of the education/mortality association, we test whether the salubrious effects of education on mortality are enhanced by psychological well-being, or whether psychological well-being mitigates the mortality risk associated with low education.

First, social structural perspectives suggest that the inverse association between education and mortality will be stronger among older adults with higher psychological well-being. That is, the mortality reductions associated with higher well-being will be greater among those with high education. According to fundamental cause theory (Link & Phelan, 1995), adults with more education have power, prestige, and resources that may allow them to reap greater mortality benefits from that well-being. Less educated adults, in contrast, experience structural barriers, including limited power, fewer financial resources, and lower levels of human capital. Those barriers make it more difficult for less educated adults to benefit from the well-being they have (Turner, 1988). This parallels the logic of the “Blaxter hypothesis” suggesting that the benefits of healthy lifestyles (e.g., non-smoking)—and perhaps psychological well-being—accrue to individuals with higher socioeconomic status (Blaxter, 1990). As such, the benefits of healthy lifestyles (and potentially well-being) may be too small to overcome the structural disadvantages associated with low levels of education.

Hypothesis 3:

Social structural perspectives predict that the inverse association between education and mortality will be stronger among those with higher psychological well-being.

In contrast, self-deterministic perspectives suggest that the inverse association between education and mortality will be weaker among older adults with higher psychological well-being. That is, the mortality reductions associated with higher well-being will be greater among those with low education. Drawing on the Reserve Capacity Model (Gallo & Matthews, 2003), individual differences in psychological well-being may mitigate deleterious health effects associated with low educational attainment by combatting the adverse effects of stress and encouraging healthy behaviors or better physiological functioning. Self-deterministic perspectives imply that individuals have some capacity to mitigate mortality risks associated with their low education via protective psychological profiles. Adults with more education, however, have diverse cognitive and socioeconomic advantages that they can consistently draw on to support better health and longevity, even if they have low levels of psychological well-being. Therefore, the influence of psychological well-being on survival may be limited among those with higher education.

Hypothesis 4:

Self-deterministic perspectives predict that the inverse association between education and mortality will be weaker among those with higher psychological well-being.

Methods

Sample

We used data from six waves (2006–2016) of the Health and Retirement Study (HRS), an ongoing, biennial, nationally representative panel study of adults over age 50 in the United States (Wallace & Herzog, 1995). Beginning in 2006, a random half of the HRS core sample (born in 1953 or earlier) was queried about their psychological well-being (as part of a Psychosocial and Lifestyle Questionnaire) in alternating waves (Smith et al., 2017). These respondents were eligible to provide psychological well-being data for three waves—either 2006, 2010, and 2014, or 2008, 2012, and 2016. In 2010, the core sample was refreshed to additionally include adults born between 1954 and 1959 to ensure that the HRS remained representative of older adults in the U.S. population. These respondents were eligible to provide psychological well-being data for two waves—either 2010 and 2014, or 2012 and 2016. In 2016 the sample was refreshed again to include adults born between 1960 and 1965. Half of these respondents were eligible to provide psychological well-being data for just one wave, 2016. The response rate for the Psychosocial and Lifestyle Questionnaire ranged from 61.9% to 87.7% across waves (Smith et al., 2017). We excluded proxy responses to preserve the validity of the psychosocial data. Those who provided psychosocial data had fewer activity limitations and better self-rated health than eligible respondents who did not provide psychosocial data. The HRS sampling weights adjusted for these non-response differences, which are described below. Our final sample included 21,149 respondents, of whom 4,934 died by the end of the 14-year follow-up period. Non-psychosocial variables came from the HRS files (Health and Retirement Study, 2020) and the RAND Longitudinal file 2016 (Bugliari et al., 2019). All respondents provided written, informed consent to participate in the study.

Measures

Vital Status.

Our outcome was all-cause mortality through May 2019. HRS determined vital status from exit interviews with proxy respondents—a strategy that captures 98.8% of deaths recorded in the National Death Index (Weir, 2016).

Psychological Well-being.

Hedonic well-being measures included positive affect and life satisfaction. We measured positive affect with the Positive and Negative Affective Schedule (Watson et al., 1988). Respondents indicated how much (5-point scale; very much to not at all), they felt each of 13 adjectives (e.g., “enthusiastic,” “happy”) during the last 30 days (α = .92-.93 across waves). HRS used a different positive affect scale in 2006, so data come from 2008–2016 waves. We measured life satisfaction with the Satisfaction with Life Scale (Diener et al., 1985). Respondents indicated their agreement (strongly disagree to strongly agree) with five items (e.g., “The conditions of my life are excellent”) on a 6-point scale in 2006 and a 7-point scale thereafter (α = .88-.89 across waves). We standardized life satisfaction as z-scores at each wave to address the differing response options across waves.

Eudaimonic well-being measures included purpose in life and positive social support. Purpose in life was based on Ryff’s theoretical framework (Ryff, 1989). Respondents indicated their agreement (7-point scale; strongly agree to strongly disagree) with each of seven items (e.g., “I have a sense of direction and purpose in life;” α = .74-.77 across waves). We assessed positive social support from four sources: spouses/partners, children, other immediate family, and friends. Respondents answered three questions about each applicable source of support (e.g., “How much do they really understand the way you feel about things?”) on 4-point scales (a lot to not at all; Schuster et al., 1990). Items were averaged to create a scale across all applicable relationship categories (α = .80-.87 across waves).

We assessed optimism with the Life Orientation Test-Revised scale (Scheier et al., 1994). Respondents indicated the extent to which they agree (6-point scale; strongly agree to strongly disagree) with six items (e.g., “In uncertain times, I usually expect the best;” α = .74-.76 across waves). In regression models, all psychological well-being scales had a mean of 0 and a standard deviation of 1, where higher scores indicated higher psychological well-being.

Educational Attainment.

Education was continuous and ranged from zero to 17+ years.

Covariates.

Covariates included age, gender, race/ethnicity, marital status, and childhood socioeconomic status. We measured age in years, centered at 50. We coded gender as women (referent) or men. Race/ethnic categories included non-Hispanic White (referent), non-Hispanic Black, non-Hispanic others, and Hispanic. Marital status categories included married or partnered (referent), previously married (i.e., divorced, separated, or widowed), and never married. To account for childhood socioeconomic status, we adjusted for the maximum of maternal and paternal education (measured continuously from 0 to 17+ years).

Supplemental models adjusted for health behaviors, activity limitations, and health conditions, each of which could plausibly confound or mediate the association between well-being and mortality (Boehm & Kubzansky, 2012; Vanderweele et al., 2020). We measure these variables at the first wave that psychological well-being is assessed for each respondent, to ensure that these variables are plausibly exogenous (i.e., most likely to work as confounders) to the association between well-being and mortality. Health behaviors included smoking (categorized as never (referent), former, and current smoker), alcohol use (categorized as 0, 1–7, and 8 or more drinks per week), vigorous and moderate physical activity (categorized as at least weekly vs. not for each intensity), and receipt of flu shot. Activity limitations were measured as a count of limitations to activities of daily living. Health conditions were measured with six dichotomous indicators of physician diagnosed high blood pressure or hypertension, diabetes or high blood sugar, cancer or malignant tumor (except skin cancer), chronic lung disease, heart problems, and stroke or transient ischemic attack, respectively. Separate models also adjusted for time-varying depressive symptoms from the Centers for Epidemiologic Studies Depression scale (excluding the positively-worded items; (Radloff, 1977). The presence of well-being is independent from the absence of depression, and depressive symptoms are not theorized to mediate well-being and mortality associations (Boehm & Kubzansky, 2012; Steptoe, 2019).

Statistical Analyses

We applied the sampling weights provided by HRS for each wave of the Psychosocial and Lifestyle Questionnaire from 2006–2016. These weights account for nursing home status, a non-response adjustment factor from a propensity score model that included sociodemographic, health status, and psychosocial predictors, and a post-stratification to the weighted HRS core sample (Smith et al., 2017). We used multiple imputation to deal with missing data among those who were eligible and returned the psychosocial questionnaire. Respondents who were eligible but did not return the psychosocial questionnaire were excluded from analyses. Multiple imputation methods allowed us to draw multiple sets of values from a distribution of likely values for each variable with missing data, conditional on other variables in the model. Drawing multiple plausible values for each piece of missing data allowed us to represent our uncertainty about those imputed values (Graham, 2009). We imputed psychological well-being at the item level as recommended by simulation studies (Eekhout et al., 2014). We created 15 imputed datasets, approximately one imputed dataset for each percent we would have dropped (von Hippel, 2007). We would have dropped 11.6% of person-year observations if we used listwise deletion of missing data. Compared to results from models that used listwise deletion, results from imputed data had similar coefficients but consistently smaller standard errors.

We used discrete-time survival models to examine the association between psychological well-being and mortality (Singer & Willett, 1993). Unlike continuous-time survival models (e.g., Cox proportional hazard models), discrete-time models allowed us to directly model the baseline hazard function so that we could calculate life tables from fitted mortality rates (see Rogers et al., 2005) and provide stable estimates given the numerous respondents with tied event times (Cox, 1972). We created a person-year dataset, where each respondent contributed one record for each year between interview and the time of censor or death. These person-years, then, represent the years that respondents are at risk of death during our study. Our final sample included 199,098 person-years for life satisfaction, purpose in life, and optimism models. Positive affect models included 160,103 person-years, due to the lack of data from the 2006 wave, and social support models included 197,240 person-years, due to some individuals lacking any of the relevant relationship categories. Age updated in each person-year record. Our models assumed that age was the time metric that drives mortality, consistent with recommendations (Korn et al., 1997). The best fitting model allowed for a linear association between age and the risk of death. Psychological well-being and marital status updated at each wave where well-being is measured (i.e., every four years) in the person-years dataset. Modeling repeated measures of psychological well-being advances existing research that includes only a baseline measure of well-being; our work does not rely on the assumption that psychological well-being is stable across time or that changes in psychological well-being are unrelated to the risk of death. Time-invariant variables (i.e., education, gender, race/ethnicity, parental education) were fixed across person-year records. We used logistic regression, in the person-year data, to fit discrete-time survival models.

After testing the direct association between psychological well-being and mortality, we used the fitted mortality rates to calculate life tables (Preston et al., 2001; Rogers et al., 2005). Life tables express the magnitude of the well-being/mortality association in an intuitive metric (i.e., expected years of remaining life), while adjusting for gender and race/ethnicity, and at specific ages. Further, we tested whether psychological well-being moderated the association between education and mortality with interaction terms between education and each well-being dimension. If an interaction was significant, we tested the associations between education and risk of death at the mean and ± one standard deviation from the mean on that well-being dimension. We incorporated survey weights, and accounted for the stratified, multi-stage sampling frame used by HRS to ensure appropriate point estimates and standard errors. We conducted all analyses with Stata version 17.0 (StataCorp, 2021).

Results

Stata code to reproduce all results is available at https://osf.io/84ru3/. Table 1 presents descriptive results by vital status. Compared to those who survived, those who died had lower positive affect, life satisfaction, purpose in life, and optimism but no differences in social support. Those who died also had fewer years of education, were older, more likely to be men, non-Hispanic White, non-Hispanic Black, previously or never married, and had lower parental education than those who survived the follow-up period. Supplemental Table 1 shows descriptive statistics in the unweighted, non-imputed dataset.

Table 1.

Sample descriptives by vital status in imputed person-year dataset.

Survived Died Significant difference by vital status t or F
Positive Affect, M 0.01 −0.43 21.76***
Life Satisfaction, M −0.02 −0.27 12 74***
Purpose in Life, M 0.03 −0.52 29.80***
Social Support, M −0.04 −0.02 0.99
Optimism, M −0.001 −0.27 18.05***
Education in years, M 13.1 11.9 13.34***
Age in years, M 68.2 79.6 56.72***
Gender, prop. men .46 .48 2.67**
Race/ethnicity, prop. 10.03***
 Non-Hispanic White .78 .80
 Non-Hispanic Black .10 .12
 Hispanic .09 .07
 Non-Hispanic Other .03 .02
Marital status, prop. 287.81***
 Married or partnered .66 .48
 Previously married .28 .47
 Never married .06 .05
Maximum of parental education in years, M 11.1 9.6 12 93***
N person-years 193,981 4,934
Open in a new tab

Note. Standard deviations are unavailable in imputed data. For means and standard deviations in unweighted, non-imputed dataset, see Table S1. Test of differences by vital status from logistic regression with each variable predicting vital status. All psychological well-being dimensions are z-scored with a mean of zero and standard deviation of one.

***

p<001

**

p<.01.

Table 2 shows odds ratios from discrete-time survival models that examine associations between psychological well-being and mortality when adjusting for age, gender, race/ethnicity, marital status, education, and parental education. Supporting hypothesis 1, Models 1 through 5 showed that each well-being dimension was inversely associated with the risk of death (all p < .01). Effect sizes varied across measures of well-being. A one standard deviation increase in psychological well-being was associated with between 0.70 times the odds of death (purpose in life) and 0.95 times the odds of death (social support). Model 6 in Table 2 included all five well-being measures and covariates in the same model. Providing partial support for hypothesis 2, we found inverse associations between positive affect, life satisfaction, and purpose in life, respectively, and mortality. Social support, however, was positively associated with mortality, and optimism was not associated with mortality in Model 6.

Table 2.

Odds ratios (95% confidence intervals) for the association between psychological well-being and the risk of death

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Positive Affect 0 74*** (0.70, 0.77) 0 90*** (0.86, 0.95)
Life Satisfaction 0.75*** (0.72, 0.78) 0.81*** (0.77, 0.85)
Purpose in Life 0 70*** (0.67, 0.72) 0 78*** (0.74, 0.83)
Social Support 0.95*** (0.91, 0.98) 1 09*** (1.03, 1.14)
Optimism 0.80*** (0.77, 0.83) 0.99 (0.94, 1.04)
N person-years 160,103 199,098 199,098 197,240 199,098 158,427
Open in a new tab

Note. All psychological well-being dimensions are z-scored. Models include age, gender, race/ethnicity, years of education, marital status, and parental education as covariates.

***

p<.001.

We conduct five sets of sensitivity analyses on the models from Table 2. First, we stratified our models by gender and found similar associations between psychological well-being and mortality for men and women (Supplemental Table 2). Second, we adjusted for smoking, alcohol consumption, flu shot receipt, and moderate and vigorous activity in first wave of well-being assessments. Third, we adjusted for self-reported health conditions or activity limitations, respectively, in the first wave of well-being assessments. Fourth, we adjusted for time-varying depressive symptoms. Finally, we included all aforementioned sets of covariates. For all sets of analyses, the protective associations between well-being and mortality were substantively unchanged, with one exception: the association between social support and mortality fell from significance (Supplemental Tables 3–7).

To illustrate the magnitude of the association between psychological well-being and mortality, Table 3 shows gender- and race-adjusted life expectancies, at age 50, for specific values of well-being, educational attainment, and smoking status. At age 50, a person could expect to live 33.2 additional years if they were one standard deviation below the mean on positive affect, 36.6 years if they had average positive affect, and 40.1 years if they were one standard deviation above the mean. The difference in life expectancy for those who were one standard deviation above the mean and those who were one standard deviation below the mean was 6.9 years for positive affect, 6.1 years for life satisfaction, 8.3 years for purpose in life, 1.2 years for social support, and 5.2 years for optimism. By comparison, at age 50, and after adjusting for gender and race, those who had a college degree could expect to live 7.3 years longer than those who had less than a high school degree, and never smokers could expect to live 9.8 years longer than current smokers.

Table 3.

Life expectancy at age 50 (in years) for low, mean, and high psychological well-being, education, and smoking status

Panel A
Psychological Well-being −1 SD Mean +1 SD
Positive Affect 33.2 36.6 40.1
Life Satisfaction 32.7 35.7 38.8
Purpose in Life 32.5 36.5 40.8
Social Support 34.9 35.5 36.1
Optimism 33.2 35.7 38.4
Panel B
Educational Attainment < HS degree HS degree College degree
32.0 34.9 39.3
Panel C
Smoking status Current Former Never
28.4 35.3 38.2
Open in a new tab

Note. Each psychological well-being dimension was modeled in separate logistic regression models. All models adjust for gender and race/ethnicity. Educational attainment was modeled categorically in this table for illustrative purposes only.

Table 4 shows results from models that examined whether psychological well-being moderated the association between education and mortality. We found partial support for hypothesis 3, that the inverse association between education and mortality would be stronger among those with higher psychological well-being. Positive affect and purpose in life significantly moderated the association between education and the risk of death in models that adjusted for age, gender, race, marital status, and parental education. For example, Model 1 shows that each additional year of education was associated with 0.95 times the odds of death when positive affect equaled 0 (i.e., the mean). The interaction term showed that the association between education and mortality decreased by 0.97 times the odds of death with each standard deviation increase in positive affect. Model 3 shows similar results for purpose in life. Interactions between education and life satisfaction, social support, and optimism, respectively, were not significant. Ancillary analyses show that the interactions between education and positive affect, and between education and purpose in life, remained significantly associated with mortality in models that adjust for health behaviors, self-reported health conditions depressive symptoms, and activity limitations (Supplemental Tables 8–11).

Table 4.

Odds ratios (95% confidence intervals) for the association between education, psychological well-being, their interaction, and the risk of death

Model 1 Model 2 Model 3 Model 4 Model 5
Years of Education 0.95*** (0.93, 0.97) 0.95*** (0.94, 0.97) 0.95*** (0.94, 0.97) 0.95*** (0.93, 0.96) 0.95*** (0.94, 0.97)
Positive Affect 1.04 (0.88, 1.22)
Education × Positive Affect 0.97*** (0.96, 0.98)
Life Satisfaction 0.77*** (0.67, 0.88)
Education × Life Satisfaction 1.00 (0.99, 1.01)
Purpose in Life 0.86* (0.74, 1.00)
Education × Purpose in Life 0 98*** (0.97, 0.99)
Social Support 0.88* (0.76, 1.01)
Education × Social Support 1.01 (0.99, 1.02)
Optimism 0.90 (0.77, 1.06)
Education × Optimism 0.99 (0.98, 1.00)
Open in a new tab

Note. All psychological well-being dimensions are z-scored. Models include age, gender, race/ethnicity, marital status, and parental education as covariates.

***

p<.001

**

p<.01

*

p<.05.

Figure 1 graphs the results for positive affect and purpose in life from Table 4, albeit using logit coefficients to depict a linear association between predictors and the risk of death [logit = ln(odds)]. The association between education and the logged odds of death was significantly different from zero when positive affect (top figure) and purpose in life (bottom figure) are at the mean or ±1 standard deviation, respectively (all p < .01). However, the association between education and mortality becomes more negative at higher levels of positive affect and purpose in life. Alternately, the left side of each figure shows that positive affect and purpose in life are not associated with mortality at very low levels of education (where the lines for different levels of well-being converge or nearly converge), but those associations become more negative as education increases (the distance between the lines for different levels of well-being becomes larger as education increases).

Figure 1.

Figure 1.

Open in a new tab

Positive affect (top) and purpose in life (bottom) moderate the association between education and mortality. The fitted values represent the logged odds of mortality for the association between education and mortality at different levels of psychological well-being and at the mean age, gender, race/ethnicity, marital status, and parental education. Solid black lines present high psychological well-being (i.e., 1 standard deviation above mean), dashed lines represent average psychological well-being, and dotted lines represent low psychological well-being (i.e., 1 standard deviation below mean).

Discussion

Psychological well-being is an important protective factor that is associated with reduced mortality among older adults. The magnitude of that association is substantial—a one standard deviation increase in positive affect, life satisfaction, purpose in life, or optimism was associated with 2.8 to 4.2 additional years of life expectancy at age 50. Effect sizes for someone with high, relative to low, well-being are comparable to the increased life expectancy from having a college degree rather than less than a high school degree. Given the importance of psychological well-being for population health, large-scale efforts are underway to surveil psychological well-being and to consider how government policies affect quality of life beyond economic indicators (e.g., UK National Wellbeing Programme; OECD Better Life Initiative). Importantly, psychological well-being is modifiable, even among older adults. Interventions to improve psychological well-being have successfully reduced recurrence of depression and anxiety and improved subjective health among older adults (Friedman et al., 2019; Weiss et al., 2016).

We advanced prior work by testing five dimensions of psychological well-being as predictors of mortality. Supporting hypothesis 1, we found that positive affect, life satisfaction, purpose in life, positive social support, and optimism were each associated with lower mortality risk. In a model that tested the five dimensions simultaneously (hypothesis 2), we found that positive affect, life satisfaction, and purpose in life remained associated with lower mortality risk. Thus, the psychological and philosophical distinctions between hedonic (i.e., positive affect and life satisfaction) and eudaimonic (i.e., purpose in life) well-being translated into independent effects with mortality, suggesting multiple targets for promoting health among older adults.

Positive social support was weakly associated with mortality in this study, and even became positively associated with mortality when adjusting for other well-being measures. Our findings contrast with research that documents health and longevity-promoting effects of social relationships (Holt-Lunstad et al., 2010). Our measure of social support averaged support from spouses/partners, children, other family, and friends, but ancillary analyses suggest that disaggregating support from those sources may be necessary to identify strong associations with mortality among older adults (Supplemental Table 13). Further, integrating relationship strain with the supportive aspects of relationships might also clarify the connection between social support and mortality (Slatcher, 2010). Whereas optimism predicted lower mortality risk in independent models, optimism was not associated with decreased mortality when other well-being measures were in the model. Thus, optimism overlaps with hedonic and eudaimonic well-being in its association with mortality. A recent meta-analysis suggests that the absence of pessimism (not assessed here) is more strongly linked to physical health than is the presence of optimism (Scheier et al., 2020).

Positive affect and purpose in life moderated the association between education and mortality. Consistent with social structural perspectives (hypothesis 3), the inverse association between education and mortality was strongest among those with high positive affect and purpose in life. More educated adults may have power, prestige, and resources to reap the survival benefits of psychological well-being, and less educated adults may have fewer financial and human capital resources to benefit from the well-being they have (Link & Phelan, 1995; Turner, 1988). We did not find support for self-deterministic perspectives (hypothesis 4), although prior research supports that perspective with outcomes including chronic inflammation or cardiovascular responses to acute stress (Boylan et al., 2016, 2020). Those biological mechanisms are just a few of many factors that affect mortality among older adults. Importantly, life satisfaction, positive social support, and optimism did not moderate the education/mortality association, suggesting that the lower mortality risks associated with higher levels of these well-being dimensions are consistent across all levels of education.

Positive affect and purpose in life—more than the other well-being measures—may support healthier stress processes (Folkman, 1997, 2008). Those with higher positive affect and purpose in life may be exposed to fewer stressors, appraise stressors more positively, and have more coping resources at their disposal. The nature of stress exposure and appraisals may also differ across education levels, where the least educated may experience more frequent stressors of greater severity and longer durations (Cundiff et al., 2020). Thus, positive affect and purpose in life may be more effective at mitigating the deleterious effects of stress among those with higher education. An important implication of these results is that efforts to promote positive affect and purpose in life in the population may widen educational disparities in survival. Therefore, policy efforts to increase educational attainment in future cohorts remains a critically important long-term solution to promoting survival among older adults (Krueger et al., 2019). However, with limited opportunity to increase educational attainment among older adults, psychological well-being remains a promising target for population health.

Given the strength of the associations between psychological well-being and mortality, future work should interrogate the mechanisms that link well-being to longer lives. Our supplementary analyses include several measures (e.g., health behaviors, health conditions, and activity limitations) that could confound or mediate the association between well-being and mortality (Boehm & Kubzansky, 2012; Vanderweele et al., 2020). It is also plausible that health behaviors, health conditions, and disability have bi-directional associations with psychological well-being (e.g., Steinmo et al., 2014; Steptoe, 2019). We fixed health behaviors, health conditions, and activity limitations at the wave of first psychological well-being assessment (rather than allowing them to vary over time) to limit their role as mediators of our key associations, but future work could examine those variables as potential mediators. We also exclude measures of income, occupation status, and wealth, even though they are important predictors of psychological well-being (Latif, 2011; Lewis & Hill, 2020; Oishi et al., 2011), because they are established mediators of the association between education and mortality, a key association in our manuscript. We focus on education rather than other measures of socioeconomic status because education is fixed early in life and evidence suggests it has a causal association with mortality (Kawachi et al., 2010). Evidence also suggests that income and wealth have associations with mortality that are, in part, causal, but they are also sensitive to health shocks in early and later life (Deaton, 2002; Haas, 2006).

Study Limitations

Several limitations warrant consideration. First, the calculated life expectancies at age 50 were slightly higher in our sample than published estimates from population data (Arias & Xu, 2020), suggesting that our sample may be healthier than the general population. This may be because the HRS excludes nursing home residents at baseline, the exit interviews might under-report deaths, or more healthy respondents being more likely to complete the Psychosocial Questionnaire. We account for selection on health and demographic characteristics by using the HRS-provided sample weights, although they might understate the better health among participants providing psychosocial data. Notably, the HRS was designed to be a representative sample of older adults in the United States, which improves on prior studies of psychological well-being and mortality that often use smaller community or clinical samples.

Second, we do not account for the life-course processes that give rise to high educational attainment and psychological well-being in older ages. Inequalities across the life-course shape opportunities to pursue education and develop high psychological well-being (Kuh et al., 2003). We also know that life transitions (e.g., retirement, widowhood) can affect dimensions of well-being differently. Preliminary analyses finds substantial variability in well-being in this sample—nearly 40% of the change in well-being across 12 years occurs within-individuals. Our analytic strategy accounted for changes in psychological well-being by using a repeated measures approach, but an important next step is to model changes in psychological well-being across the life-course, consider the early-life antecedents to later life psychological well-being trajectories, and understand the health and mortality consequences of trajectories in well-being.

Third, many important measures of well-being (e.g., autonomy, gratitude, competence) were not assessed in HRS. Positive affect was likewise assessed as a global trait, despite evidence that specific positive emotions may be more closely linked to health outcomes than others (Pressman et al., 2019). Given that well-being is a target for intervention and policy, identifying the aspects most important for health and mortality is imperative in future research. Finally, the HRS data lack information on the underlying cause of death, which could provide more insights into the biological and behavioral pathways that link psychological well-being and education to mortality. While prior research documents associations between well-being and cardiovascular and inflammation outcomes, other causes of death may also be important.

Conclusions

Research has proliferated around the salubrious associations between psychological well-being and mortality. We find that high, relative to low, positive affect, life satisfaction, purpose in life, and optimism are associated with 5 to 8 additional years of life at age 50, which is comparable to the reduced mortality associated with having a college degree rather than less than a high school degree. Psychological well-being is a positive outcome in its own right, is modifiable through interventions, and is strongly protective against mortality among older adults. Both positive affect and purpose in life modify the education/mortality association, suggesting that individual well-being can alter the association between structural factors and mortality. In short, psychological well-being focuses on existing strengths of older adults and supports our ultimate goal of promoting longer, healthier, and more fulfilling lives in our population.

Supplementary Material

Supplemental Material

Acknowledgments

The Health and Retirement Study is supported by the National Institute on Aging (grant number U01AG009740) and is conducted by the University of Michigan. P.M.K. is supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Development of the National Institutes of Health (P2CHD066613).

Footnotes

The authors declare no conflicts of interest. The authors would like to thank Carol Ryff, Richard Rogers, and Ryan O’Connell for their contributions to this project.

References

  1. Arias E, & Xu J (2020). United States life tables, 2018. National vital statistics reports; vol 69 no. 12. National Center for Health Statistics. [PubMed] [Google Scholar]
  2. Blaxter M (1990). Health and Lifestyles. Tavistock. [Google Scholar]
  3. Boehm JK, & Kubzansky LD (2012). The heart’s content: The association between positive psychological well-being and cardiovascular health. Psychological Bulletin, 138(4), 655–691. 10.1037/a0027448 [DOI] [PubMed] [Google Scholar]
  4. Boylan JM, Cundiff JM, Fuller-Rowell TE, & Ryff CD (2020). Childhood socioeconomic status and inflammation: psychological moderators among Black and White Americans. Health Psychology, 39(6), 497–508. 10.1037/hea0000866 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Boylan JM, Jennings JR, & Matthews KA (2016). Childhood socioeconomic status and cardiovascular reactivity and recovery among black and white men: mitigating effects of psychological resources. Health Psychology, 35(9), 957–966. 10.1037/hea0000355 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Boylan JM, & Ryff CD (2015). Psychological well-being and metabolic syndrome: Findings from the Midlife in the United States national sample. Psychosomatic Medicine, 77(5), 548–558. 10.1097/PSY.0000000000000192 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bugliari D, Campbell N, Chan C, Hayden O, Hayes J, & Hurd M (2019). RAND HRS longitudinal file 2016 (V1) documentation. https://www.rand.org/well-being/social-and-behavioral-policy/centers/aging/dataprod/hrs-data.html
  8. Chida Y, & Steptoe A (2008). Positive psychological well-being and mortality: a quantitative review of prospective observational studies. Psychosomatic Medicine, 70(7), 741–756. 10.1097/PSY.0b013e31818105ba [DOI] [PubMed] [Google Scholar]
  9. Cohen R, Bavishi C, & Rozanski A (2016). Purpose in life and its relationship to all-cause mortality and cardiovascular events: A meta-analysis. Psychosomatic Medicine, 78, 122–133. 10.1097/PSY.0000000000000274 [DOI] [PubMed] [Google Scholar]
  10. Cox DR (1972). Regression models and life-tables. Journal of the Royal Statistical Society. Series B (Methodological), 34(2), 187–220. http://www.jstor.org/stable/2985181 [Google Scholar]
  11. Cundiff JM, Boylan JM, & Muscatell KA (2020). The pathway from social status to physical health: taking a closer look at stress as a mediator. Current Directions in Psychological Science, 29(2), 147–153. 10.1177/0963721420901596 [DOI] [Google Scholar]
  12. Deaton A (2002). Policy implications of the gradient of health and wealth. Health Affairs, 21(2), 13–30. 10.1377/HLTHAFF.21.2.13 [DOI] [PubMed] [Google Scholar]
  13. Diener E, Emmons RA, Larsen RJ, & Griffin S (1985). The Satisfaction With Life Scale. Journal of Personality Assessment, 49(1), 71–75. 10.1207/s15327752jpa4901_13 [DOI] [PubMed] [Google Scholar]
  14. Eekhout I, De Vet HCW, Twisk JWR, Brand JPL, De Boer MR, & Heymans MW (2014). Missing data in a multi-item instrument were best handled by multiple imputation at the item score level. Journal of Clinical Epidemiology, 67(3), 335–342. 10.1016/j.jclinepi.2013.09.009 [DOI] [PubMed] [Google Scholar]
  15. Folkman S (1997). Positive psychological states and coping with severe stress. Social Science and Medicine, 45(8), 1207–1221. 10.1016/S0277-9536(97)00040-3 [DOI] [PubMed] [Google Scholar]
  16. Folkman S (2008). The case for positive emotions in the stress process. Anxiety, Stress, & Coping, 21(1), 3–14. 10.1080/10615800701740457 [DOI] [PubMed] [Google Scholar]
  17. Friedman EM, Ruini C, Foy CR, Jaros L, Love G, & Ryff CD (2019). Lighten UP! A community-based group intervention to promote eudaimonic well-being in older adults: a multi-site replication with 6 month follow-up. Clinical Gerontologist, 42(4), 387–397. 10.1080/07317115.2019.1574944 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Gallo LC, & Matthews KA (2003). Understanding the association between socioeconomic status and physical health: Do negative emotions play a role? Psychological Bulletin, 129(1), 10–51. 10.1037/0033-2909.129.1.10 [DOI] [PubMed] [Google Scholar]
  19. Graham JW (2009). Missing data analysis: making it work in the real world. Annual Review of Psychology, 60(1), 549–576. 10.1146/annurev.psych.58.110405.085530 [DOI] [PubMed] [Google Scholar]
  20. Haas SA (2006). Health selection and the process of social stratification: the effect of childhood health on socioeconomic attainment. Journal of Health and Social Behavior, 47(4), 339–354. 10.1177/002214650604700403 [DOI] [PubMed] [Google Scholar]
  21. Health and Retirement Study. (2020). The Health and Retirement Study. http://hrsonline.isr.umich.edu/
  22. Holt-Lunstad J, Smith TB, & Layton JB (2010). Social relationships and mortality risk: a meta-analytic review. PLoS Medicine, 7(7), e1000316. 10.1371/journal.pmed.1000316 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Kawachi I, Adler NE, & Dow WH (2010). Money, schooling, and health: Mechanisms and causal evidence. Annals of the New York Academy of Sciences, 1186(1), 56–68. 10.1111/j.1749-6632.2009.05340.x [DOI] [PubMed] [Google Scholar]
  24. Korn EL, Graubard BI, & Midthune D (1997). Time-to-event analysis of longitudinal follow-up of a survey: choice of the time-scale. American Journal of Epidemiology, 145(1), 72–80. 10.1093/oxfordjournals.aje.a009034 [DOI] [PubMed] [Google Scholar]
  25. Krueger PM, & Chang VW (2008). Being poor and coping with stress: health behaviors and the risk of death. American Journal of Public Health, 98(5), 889–896. 10.2105/AJPH.2007.114454 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Krueger PM, Dehry IA, & Chang VW (2019). The economic value of education for longer lives and reduced disability. The Milbank Quarterly, 97(1), 48–73. 10.1111/1468-0009.12372 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Kuh D, Ben-Shlomo Y, Lynch J, Hallqvist J, & Power C (2003). Life course epidemiology. Journal of Epidemiology & Community Health, 57(10), 778–783. 10.1136/jech.57.10.778 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Latif E (2011). The impact of retirement on psychological well-being in Canada. The Journal of Socio-Economics, 40(4), 373–380. 10.1016/j.socec.2010.12.011 [DOI] [Google Scholar]
  29. Lewis NA, & Hill PL (2020). Does being active mean being purposeful in older adulthood? Examining the moderating role of retirement. Psychology and Aging, 35(7), 1050–1057. 10.1037/PAG0000568 [DOI] [PubMed] [Google Scholar]
  30. Link BG, & Phelan J (1995). Social conditions as fundamental causes of disease. Journal of Health and Social Behavior, Spec No, 80–94. [PubMed] [Google Scholar]
  31. Martín-María N, Miret M, Caballero FF, Rico-Uribe LA, Steptoe A, Chatterji S, & Ayuso-Mateos JL (2017). The impact of subjective well-being on mortality. Psychosomatic Medicine, 79(5), 565–575. 10.1097/PSY.0000000000000444 [DOI] [PubMed] [Google Scholar]
  32. Masters RK, Hummer RA, & Powers DA (2012). Educational differences in U.S. adult mortality: a cohort perspective. American Sociological Review, 77(4), 548–572. 10.1177/0003122412451019 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Oishi S, Kesebir S, & Diener E (2011). Income inequality and happiness. Psychological Science, 22(9), 1095–1100. 10.1177/0956797611417262 [DOI] [PubMed] [Google Scholar]
  34. Pampel FC, Krueger PM, & Denney JT (2010). Socioeconomic disparities in health behaviors. Annual Review of Sociology, 36, 349–370. 10.1146/annurev.soc.012809.102529 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Pressman SD, Jenkins BN, & Moskowitz JT (2019). Positive affect and health: what do we know and where next should we go? Annual Review of Psychology, 70(1), 627–650. 10.1146/annurev-psych-010418-102955 [DOI] [PubMed] [Google Scholar]
  36. Preston S, Heuveline P, & Guillot M (2001). Demography: measuring and modeling population processes. Blackwell. [Google Scholar]
  37. Radloff LS (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1(3), 385–401. [Google Scholar]
  38. Rasmussen HN, Scheier MF, & Greenhouse JB (2009). Optimism and physical health: a meta-analytic review. Annals of Behavioral Medicine, 37(3), 239–256. 10.1007/s12160-009-9111-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Rogers RG, Hummer RA, Krueger PM, & Pampel FC (2005). Mortality attributable to cigarette smoking in the United States. Population and Development Review, 31(2), 259. 10.1111/J.1728-4457.2005.00065.X [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Rush CL, Hooker SA, Ross KM, Frers AK, Peters JC, & Masters KS (2019). Brief report: Meaning in life is mediated by self-efficacy in the prediction of physical activity. Journal of Health Psychology, 26(5), 753–757. 10.1177/1359105319828172 [DOI] [PubMed] [Google Scholar]
  41. Ryan RM, & Deci EL (2001). On happiness and human potentials: A review of research on hedonic and eudaimonic well-being. Annual Review of Psychology, 52, 141–166. [DOI] [PubMed] [Google Scholar]
  42. Ryff CD (1989). Happiness is everything, or is it? Explorations on the meaning of psychological well-being. Journal of Personality and Social Psychology, 57(6), 1069–1081. 10.1037//0022-3514.57.6.1069 [DOI] [Google Scholar]
  43. Ryff CD (2016). Eudaimonic well-being and education: probing the connections. In Harward D (Ed.), Well-being and higher education: a strategy for change and the realization of education’s greater purposes. Bringing Theory to Practice (pp. 37–48). [Google Scholar]
  44. Ryff CD, Boylan JM, & Kirsch JA (2020). Disagreement about recommendations for measurement of well-being. Preventive Medicine, 139, 106049. 10.1016/j.ypmed.2020.106049 [DOI] [PubMed] [Google Scholar]
  45. Scheier MF, Carver CS, & Bridges MW (1994). Distinguishing optimism from neuroticism (and trait anxiety, self-mastery, and self-esteem): a reevaluation of the Life Orientation Test. Journal of Personality and Social Psychology, 67(6), 1063–1078. [DOI] [PubMed] [Google Scholar]
  46. Scheier MF, Swanson JD, Barlow MA, Greenhouse JB, Wrosch C, & Tindle HA (2020). Optimism versus pessimism as predictors of physical health: A comprehensive reanalysis of dispositional optimism research. American Psychologist. 10.1037/amp0000666 [DOI] [PubMed] [Google Scholar]
  47. Schuster TL, Kessler RC, & Aseltine RH (1990). Supportive interactions, negative interactions, and depressed mood. American Journal of Community Psychology, 18(3), 423–438. 10.1007/BF00938116 [DOI] [PubMed] [Google Scholar]
  48. Singer JD, & Willett JB (1993). It’s about time: using discrete-time survival analysis to study duration and the timing of events. Journal of Educational Statistics, 18(2), 155. 10.2307/1165085 [DOI] [Google Scholar]
  49. Slatcher RB (2010). Marital functioning and physical health: implications for social and personality psychology. Social and Personality Psychology Compass, 4(7), 455–469. 10.1111/j.1751-9004.2010.00273.x [DOI] [Google Scholar]
  50. Smith J, Ryan LH, Fisher GG, Sonnega A, & Weir DR (2017). HRS Psychosocial and Lifestyle Questionnaire 2006–2016.
  51. StataCorp. (2021). Stata Statistical Software: Release 17.
  52. Steinmo S, Hagger-Johnson G, & Shahab L (2014). Bidirectional association between mental health and physical activity in older adults: Whitehall II prospective cohort study. Preventive Medicine, 66, 74–79. 10.1016/j.ypmed.2014.06.005 [DOI] [PubMed] [Google Scholar]
  53. Steptoe A (2019). Happiness and health. Annual Review of Public Health, 40(1), 339–359. 10.1146/annurev-publhealth-040218-044150 [DOI] [PubMed] [Google Scholar]
  54. Turner JH (1988). A behavioral theory of social structure. Journal for the Theory of Social Behaviour, 18(4), 355–372. 10.1111/j.1468-5914.1988.tb00505.x [DOI] [Google Scholar]
  55. Vanderweele TJ, Chen Y, Long K, Kim ES, Trudel-Fitzgerald C, & Kubzansky LD (2020). Positive Epidemiology? Epidemiology, 31(2), 189–193. 10.1097/EDE.0000000000001147 [DOI] [PubMed] [Google Scholar]
  56. von Hippel PT (2007). Regression with missing Ys: an improved strategy for analyzing multiply imputed data. Sociological Methodology, 37, 83–117. [Google Scholar]
  57. Wallace R, & Herzog R (1995). Overview of the health measures in the Health and Retirement Survey. The Journal of Human Resources, 30(Supplement 1995). [Google Scholar]
  58. Watson D, Clark LA, & Tellegen A (1988). Development and validation of brief measures of positive and negative affect: the PANAS scales. Journal of Personality and Social Psychology, 54(6), 1063–1070. [DOI] [PubMed] [Google Scholar]
  59. Weir DR (2016). Validating mortality ascertainment in the Health and Retirement Study.
  60. Weiss LA, Westerhof GJ, & Bohlmeijer ET (2016). Can we increase psychological well-being? The effects of interventions on psychological well-being: a meta-analysis of randomized controlled trials. PloS One, 11(6), e0158092. 10.1371/journal.pone.0158092 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Material
NIHMS1864145-supplement-Supplemental_Material.pdf (204.1KB, pdf)

RESOURCES