Abstract
We examined the associations between a sense of purpose and all-cause mortality by gender and race/ethnicity groups. Data were from the Health and Retirement Study, a nationally representative cohort study of U.S. adults aged >50 (n = 13,159). Sense of purpose was self-reported at baseline (2006/2008), and risk of all-cause mortality was assessed over an 8-year follow-up period. We also formally tested for potential effect modification by gender and race/ethnicity. We observed the associations between higher purpose and lower all-cause mortality risk across all gender and race/ethnicity groups. There was modest evidence that the highest level of purpose (versus lowest quartile) was associated with even lower risk of all-cause mortality among women (risk ratio = 0.66, 95% confidence interval: 0.56, 0.77) compared to men (risk ratio = 0.80, 95% confidence interval: 0.69, 0.93; p-value for multiplicative effect modification =0.07). However, we observed no evidence of effect modification by race/ethnicity. Having a higher sense of purpose appears protective against all-cause mortality regardless of gender and race/ethnicity. Purpose, a potentially modifiable factor, might be a health asset across diverse populations.
Keywords: Purpose in life, Race/ethnicity, Gender, Effect modification, Interaction, Older adults, Health and retirement study
1. Introduction
The biomedical sciences and public health have focused on reducing risk factors as a way to improve health outcomes. This deficit-based approach has generated important scientific insights and interventions. Emerging research, however, has shed light on the health benefits of a strengths-based approach, which focuses on identifying and fostering resilience factors and health-promoting assets. (VanderWeele et al., 2020) Purpose in life—the extent to which people perceive their lives as having a sense of direction and goals—has emerged as a promising candidate health asset. (Ryff, 2014; McKnight and Kashdan, 2009; Ryff and Kim, 2020) Growing evidence suggests that a higher sense of purpose is associated with healthier lifestyle behaviors (e.g., increased physical activity and preventive healthcare use, as well as reduced illicit drug use and likelihood of sleep problems), (Kim et al., 2020a; Kim et al., 2020b; Kim et al., 2014; Kim et al., 2022; Kang et al., 2021) healthier biological function (e.g., reduced inflammation and allostatic load), (Zilioli et al., 2015; Steptoe and Fancourt, 2019; Hafez et al., 2018) physical function, (Kim et al., 2017) and also reduced risk of chronic disease (e.g., lower risk of cardiovascular disease and cognitive impairment), (Steptoe and Fancourt, 2019; Kim et al., 2019; Cohen et al., 2016; Sutin et al., 2021) and mortality. (Kim et al., 2022; Cohen et al., 2016; Hill and Turiano, 2014; Alimujiang et al., 2019)
It is theoretically possible that the health impacts of purpose differ by key demographic characteristics because socio-environmental factors that block the intermediate pathways linking purpose and enhanced health might be differentially distributed across the demographic groups. For example, higher purpose may lower mortality risk via promoting health behaviors (e.g., physical activity). (Kim et al., 2020a) However, without access to necessary socioenvironmental factors (e.g., adequate infrastructure that enables physical activity such as green space, a safe neighborhood, etc), one may not fully benefit from having higher purpose. A recent study in older adults evaluated three indicators of socioeconomic status (i.e., income, total wealth, educational attainment) as potential effect modifiers of the purpose-mortality association. Results showed that the highest levels of purpose appeared protective against all-cause mortality regardless of SES, while more modest levels of purpose appeared less beneficial health-wise among people with lower SES. (Shiba et al., 2021)
Similarly, gender and race/ethnicity may moderate the impacts of purpose on health because people may have differential access to risks, opportunities, and resources due to gendered and racialized social structures. Thus, people in marginalized identity groups might receive more health benefits from purpose because they have limited access to other health-promoting resources, but purpose is an alternate resource that is more accessible than others. It is important to examine potentially heterogeneous effects of purpose; finding that purpose is associated with favorable health across gender and race/ethnicity groups suggests that the health benefits of purpose are realizable in multiple demographic contexts. Alternatively, finding that purpose is associated with smaller beneficial health effects in specific subgroups would prompt further investigation into the underlying reasons for heterogeneous effects; this additional work could help identify whether and how the health benefits of purpose might become realizable for all. However, research evaluating gender and race/ethnicity as potential effect modifiers is limited.
Only two studies to date have evaluated if the purpose-mortality association is moderated by gender or race/ethnicity, and they reported mixed results. One US-based study of 1238 older adults observed that the purpose-mortality association was not moderated by either gender or race. (Boyle et al., 2009) A second study, conducted in 73,272 middle-aged and older adults in Japan evaluated Ikigai (“what makes life worth living”), which is a distinct but close conceptual cousin to purpose. This study reported that the Ikigai-mortality association was stronger in men than in women. (Tanno et al., 2009) These studies contribute substantially to the literature, but some limitations remain unaddressed. First, the only study to evaluate effect modification by race/ethnicity was conducted in a sample with a small number of case counts in the different race/ethnicity groups; thus, it might have been underpowered. Second, the study of Ikigai and mortality was conducted with a culturally and racially homogenous population, and some evidence suggests that associations between psychological factors and health outcomes are not universally observed across different cultures. (Kitayama and Park, 2021)
To fill these knowledge gaps, we analyzed a large, diverse, and nationwide longitudinal sample of U.S. older adults and examined the associations between purpose in life and all-cause mortality by gender and race/ethnicity.
2. Methods
Data are from the Health and Retirement Study (HRS)—an ongoing nationwide panel study of U.S. adults aged >50 years. HRS was initiated in 1992 with biennial follow-up surveys ongoing. In 2006, the HRS began visiting a randomly-selected 50% of study participants for an enhanced face-to-face interview. The remaining 50% of participants were assessed with the same protocol in 2008. At each time point, after the interview, respondents were given a self-administered psychosocial questionnaire to return by mail to study staff, which included items assessing sense of purpose (response rate was 88% in 2006 and 84% in 2008).
We combined respondents from both time points and considered 2006/2008 as the baseline for our current study (N = 13,770). Individuals with missing information on either purpose (n = 426) or death (n = 185) were excluded, resulting in a final analytic sample of 13,159 participants. Because the study used de-identified, publicly available data, the Harvard T.H. Chan School of Public Health IRB exempted it from review. In addition, all HRS respondents provided written informed consent to the HRS study.
2.1. Measures
Mortality.
To keep follow-up length constant across all participants, information about death was obtained up to 2014 (for the 2006 subsample) and up to 2016 (for the 2008 subsample). Thus, our study assessed death over an 8-year follow-up period. Information about death was obtained first via an exit interview conducted with next-of-kin. When subsequently confirming the exit interview deaths reported with those reported by the National Death Index, there was a 95.5% match.
Purpose in Life.
Purpose in life was assessed at baseline (2006/2008) using the validated 7-item purpose in life subscale of the Ryff Psychological Well-being Scales. (Ryff and Keyes, 1995) Each item was rated on a 6-point Likert scale, and overall scores were derived using the mean of item responses (range: 1–6), with higher scores reflecting higher sense of purpose (Cronbach α = 0.74 among those with complete purpose data). The mean purpose score was derived if respondents completed at least 5 of the 7 purpose items (96.9%). Of the study population (n = 13,159), 199 participants had one and 30 participants had two missing items. Mean purpose scores were 4.25 (SD = 0.95) among the 199 participants with one missing item and 4.24 (SD = 1.05) among the 30 participants with two missing items, which were relatively lower compared to those with complete data (n = 12,690; mean = 4.59 and SD = 0.93). To evaluate potential non-linearities in the purpose-mortality association, we created quartiles based on the baseline distribution of purpose scores in the analytic sample (Low: 1.00–3.86; Medium-low: 3.87–4.57; Medium-high: 4.58–5.29; High: 5.30–6.00).
Gender and Race/Ethnicity.
We considered gender and race/ethnicity as potential effect modifiers using self-reported data in the baseline survey (2006/2008). Gender was coded as: men or women and race/ethnicity was coded as: White, Black, Hispanic, or Other.
Covariates.
We considered a range of covariates assessed at baseline, including socioeconomic status, other demographic characteristics, baseline physical health, and depression. Depression has been identified as a risk factor for mortality and is also related to lower purpose (Kim et al., 2022; Cuijpers and Smit, 2002); thus, we adjusted for depression to reduce concerns that purpose simply reflects the absence of depression. Socioeconomic status included educational attainment (highest degree attained: <high school, high school of GED, and ≥ college), annual total household income following conventional HRS coding (1st Quartile: ≤$20,024; 2nd Quartile: $20,025–$38,321; 3rd Quartile: $38,322–$71,895; 4th Quartile: ≥$71,896), and total wealth (1st Quintile: ≤$35,000; 2nd Quintile: $35,001–$140,000; 3rd Quintile: $140,001–$311,000; 4th Quintile: $311,001–$652,500; 5th Quintile: ≥$652,501). Other demographic characteristics included: age (years, continuous), marital status (married, not married), health insurance status (covered, not covered). Baseline physical health status was assessed according to self-reported presence/absence of a doctor's diagnosis for each of the following major diseases: heart disease, cancer, stroke, diabetes, hypertension, lung disease, and arthritis. Depression was assessed using the validated 8-item Center for Epidemiological Studies Depression Scale (Cronbach α = 0.80; a score of ≥4 was categorized as depressed). (Radloff, 1977)
2.2. Statistical analysis
Because our outcome (death) was not rare (24.7% among the total sample), odds ratios from logistic regression can overestimate risk ratios (RRs). Thus, we used Poisson regression models to estimate RRs for the purpose-mortality association over the 8-year follow-up period. To evaluate potential effect modification by gender and race/ethnicity, we fit two separate models. In each model, we introduced product terms between baseline purpose in life (three dummy variables for quartiles) and the categories (dummy variables) of gender or race/ethnicity. In each model, we also adjusted for all covariates and the other effect modifier (e.g., we adjusted for race/ethnicity as a covariate in the model assessing effect modification by gender).
Following the framework proposed by Knol and VanderWeele (2012), (Knol and VanderWeele, 2012) we computed a series of effect estimates and measures of effect modification. First, using Poisson models, we estimated RRs to evaluate the purpose-mortality association within each stratum of gender (or race/ethnicity) category. Second, we estimated RRs for the association of a joint exposure of purpose and gender (or race/ethnicity) with mortality over the 8-year follow-up period. The reference groups in these analyses were people with the lowest purpose levels and: 1) men (for gender analyses), 2) white (for race/ethnicity analyses)). Third, using the RRs from the joint purpose and gender (or race/ethnicity) estimates from the second analysis, we estimated measures of effect modification on both the additive (relative excess risk due to interaction: RERI) and multiplicative (Ratio of RRs: RRR) scales. Additive effect modification, which captures change in absolute risk difference estimates across levels of effect modifiers, is rarely reported in epidemiology despite its public health relevance. (Knol and VanderWeele, 2012) The additive effect modification measure provides insight into whether the absolute number of deaths that could be prevented via potential purpose interventions might differ across social groups. Multiplicative effect modification can be large even when the baseline prevalence of the outcome is low, and the population impact of such effect heterogeneity is minimal. See Supplementary Text for a detailed description of how to estimate and interpret both measures of effect modification. Supplementary Table 1 provides an example of how we combined the regression coefficients from the Poisson model that assesses effect modification by gender. As we performed many statistical tests for effect modification by each gender (or race/ethnicity) category, low power and multiple testing are both concerns. Thus, we conducted a sensitivity analysis where we coded purpose as a continuous variable. We then created product terms between continuous purpose and the effect modifier of interest (gender or race/ethnicity categories) and conducted an omnibus test for effect modification on the multiplicative scale.
Among the analytic sample of 13,159 individuals, 2.9% (n = 377) had missing data in at least one of the variables. We imputed missing data using multiple imputation by chained equation. (van Buuren and Groothuis-Oudshoorn, 2010) Specifically, we created 20 imputed datasets using the R package “mice” (R Foundation for Statistical Computing, Vienna, Austria), performed the analyses described above in each imputed dataset, and combined estimates across the imputed datasets based on the Rubin's rule. (Rubin, 2004) All analyses were performed using R, version 3.6.0. The data supporting the findings of this study are publicly available upon request to the Health and Retirement Study. R code to replicate the findings will be made available upon request to the corresponding author.
3. Results
3.1. Descriptive statistics
Table 1 shows the descriptive statistics of the analytic sample. Among the 13,159 people in our study sample, 3253 people (24.7%) died during the 8-year follow-up period. Overall, people with the highest baseline purpose in life showed lower mortality risk compared to those with the lowest purpose (e.g., mortality risk was 15.2% in the highest baseline purpose group and 36.5% in the lowest baseline purpose group). In a dose-response fashion, increasingly higher baseline purpose was associated with increasingly: higher socioeconomic status, better physical health, lower depression prevalence, and younger age at baseline. While mean baseline purpose scores were comparable across gender groups (4.59 in men and 4.58 in women), it was higher among Blacks (4.73), and lower among Hispanics (4.38) and people in the “Other” group (4.50), compared to the average among Whites (4.58).
Table 1.
Characteristics of Study Participants at Baseline, by Levels of Purpose in Life (n = 13,159).
| Total (n = 13,159) |
Sense of purpose in lifea | ||||
|---|---|---|---|---|---|
| Low (n = 3445) |
Medium- low (n = 3167) |
Medium- high (n = 3465) |
High (n = 3082) |
||
| Death by the end of follow-up | 3253 (24.7) | 1259 (36.5) | 801 (25.3) | 724 (20.9) | 469 (15.2) |
| Gender | |||||
| Men (%) | 5495 (41.8) | 1420 (41.2) | 1337 (42.2) | 1466 (42.3) | 1272 (41.3) |
| Women (%) | 7664 (58.2) | 2025 (58.8) | 1830 (57.8) | 1999 (57.7) | 1810 (58.7) |
| Race/ethnicity | |||||
| White (%) | 10,240 (77.8) | 2615 (75.9) | 2515 (79.4) | 2758 (79.6) | 2352 (76.3) |
| Black (%) | 1645 (12.5) | 376 (10.9) | 365 (115) | 423 (12.2) | 481 (15.6) |
| Hispanic (%) | 1002 (7.6) | 371 (10.8) | 220 (6.9) | 219 (6.3) | 192 (6.2) |
| Other (%) | 271 (2.1) | 83 (2.4) | 67 (2.1) | 64 (1.8) | 57 (1.8) |
| Socioeconomic status | |||||
| Education | |||||
| <high school (%) | 2515 (19.1) | 1012 (29.4) | 616 (19.5) | 504 (14.5) | 383 (12.4) |
| High school (%) | 7223 (54.9) | 1863 (54.1) | 1806 (57.0) | 1928 (55.6) | 1626 (52.8) |
| ≥college (%) | 3395 (25.8) | 568 (16.5) | 741 (23.4) | 1022 (29.5) | 1064 (34.5) |
| Household income quartilesb,c | |||||
| Q1 (%) | 3221 (24.5%) | 1257 (36.5%) | 766 (24.2%) | 680 (19.6%) | 518 (16.8%) |
| Q2 (%) | 3303 (25.1%) | 944 (27.4%) | 879 (27.8%) | 814 (23.5%) | 666 (21.6%) |
| Q3 (%) | 3312 (25.2%) | 753 (21.9%) | 813 (25.7%) | 957 (27.6%) | 789 (25.6%) |
| Q4 (%) | 3323 (25.3%) | 491 (14.3%) | 709 (22.4%) | 1014 (29.3%) | 1109 (36.0%) |
| Wealth quintilesb,d | |||||
| Q1 (%) | 2598 (19.7) | 969 (28.1) | 651 (20.6) | 569 (16.4) | 409 (13.3) |
| Q2 (%) | 2616 (19.9) | 810 (23.5) | 659 (20.8) | 630 (18.2) | 517 (16.8) |
| Q3 (%) | 2642 (20.1) | 673 (19.5) | 623 (19.7) | 705 (20.3) | 641 (20.8) |
| Q4 (%) | 2656 (20.2) | 524 (15.3) | 651 (20.6) | 768 (22.2) | 713 (23.1) |
| Q5 (%) | 2657 (20.1) | 469 (13.6) | 583 (18.4) | 793 (22.9) | 802 (26.0) |
| Other demographic characteristics | |||||
| Mean age (SD) | 69.6 (9.58) | 71.6 (10.3) | 70.0 (9.74) | 68.9 (9.11) | 67.9 (8.60) |
| Marital status | |||||
| Not married (%) | 4885 (37.1) | 1566 (45.5) | 1232 (38.9) | 1135 (32.8) | 952 (30.9) |
| Married (%) | 8274 (62.9) | 1879 (54.5) | 1935 (61.1) | 2330 (67.2) | 2130 (69.1) |
| Health insurance | |||||
| Not covered (%) | 1432 (10.9) | 332 (9.6) | 324 (10.2) | 389 (112) | 387 (12.6) |
| Covered (%) | 11,651 (88.5) | 3083 (89.5) | 2826 (89.2) | 3055 (88.2) | 2687 (87.2) |
| Health Factors e | |||||
| Heart disease | |||||
| No (%) | 9905 (75.3) | 2362 (68.6) | 2416 (76.3) | 2630 (75.9) | 2497 (81.0) |
| Yes (%) | 3240 (24.6) | 1081 (31.4) | 746 (23.6) | 832 (24.0) | 581 (18.9) |
| Stroke | |||||
| No (%) | 12,102 (92.0) | 3035 (88.1) | 2913 (92.0) | 3229 (93.2) | 2925 (94.9) |
| Yes (%) | 1048 (8.0) | 406 (11.8) | 251 (7.9) | 235 (6.8) | 156 (5.1) |
| Cancer | |||||
| No (%) | 11,098 (84.3) | 2838 (82.4) | 2673 (84.4) | 2942 (84.9) | 2645 (85.8) |
| Yes (%) | 2037 (15.5) | 597 (17.3) | 486 (15.3) | 520 (15.0) | 434 (14.1) |
| Diabetes | |||||
| No (%) | 10,537 (80.1) | 2539 (73.7) | 2534 (80.0) | 2834 (81.8) | 2630 (85.3) |
| Yes (%) | 2609 (19.8) | 903 (26.2) | 626 (19.8) | 630 (18.2) | 450 (14.6) |
| Hypertension | |||||
| No (%) | 5580 (42.4) | 1267 (36.8) | 1310 (41.4) | 1542 (44.5) | 1461 (47.4) |
| Yes (%) | 7565 (57.5) | 2172 (63.0) | 1853 (58.5) | 1919 (55.4) | 1621 (52.6) |
| Lung diseases | |||||
| No (%) | 11,851 (90.1) | 2976 (86.4) | 2817 (88.9) | 3166 (91.4) | 2892 (93.8) |
| Yes (%) | 1291 (9.8) | 463 (13.4) | 343 (10.8) | 296 (8.5) | 189 (6.1) |
| Arthritis | |||||
| No (%) | 5119 (38.9) | 1105 (32.1) | 1230 (38.8) | 1425 (41.1) | 1359 (44.1) |
| Yes (%) | 8027 (61.0) | 2338 (67.9) | 1934 (61.1) | 2036 (58.8) | 1719 (55.8) |
| Depression f | |||||
| No (%) | 11,180 (85.0) | 2455 (71.3) | 2694 (85.1) | 3123 (90.1) | 2908 (94.4) |
| Yes (%) | 1780 (13.5) | 921 (26.7) | 426 (13.5) | 302 (8.7) | 131 (4.3) |
Purpose in life was assessed using the purpose in life subscale of the Ryff psychological well-being scales. Purpose in life quartiles: Low: 1–3.86; medium–low: 4–4.57; medium–high: 4.6–5.29; high: 5.33–6.00.
Income and wealth were imputed by health and retirement study and, thus, there was no missing in these variables in the data used.
Household income quartiles: 1st quartile: ≤$20,024; 2nd quartile: $20,025–$38,321; 3rd quartile: $38,322–$71,895; 4th quartile: ≥$71,896.
Wealth quintiles: 1st quartile: ≤$35,000; 2nd quartile: $35,001–$140,000; 3rd quartile: $140,001–$311,000; 4th quartile: $311,001–$652,500; 5th quartile: ≥$652,501.
Health factors were defined as self-reported presence/absence of having a doctor's diagnosis.
Depression was assessed using the 8-item Center for Epidemiological Studies Depression Scale with a cut-off point of ≥4.
3.2. Purpose and mortality within strata of gender or race/ethnicity
Supplementary Fig. 1 shows associations between purpose quartiles and 8-year mortality risk for each stratum of gender or race/ethnicity (see Supplementary Table 2 for exact estimates). Overall, there was a dose-response trend so that people with increasingly higher levels of purpose displayed decreasingly lower mortality risk across gender and race/ethnicity groups. However, the purpose-mortality point-estimates were stronger in women, than men. For example, RRs comparing the highest (versus lowest) baseline purpose group was 0.66 for women and 0.80 for men. When comparing people with the highest (versus lowest) levels of purpose, the purpose-mortality association was stronger in Blacks (RR = 0.59; 95% Confidence Interval [CI]: 0.43, 0.80), Hispanics (RR = 0.52; 95% CI: 0.32, 0.86), than in Whites (RR = 0.76; 95% CI: 0.67, 0.86). Further, in Blacks, we did not observe strong evidence of a purpose-mortality association among people with mid-range levels of purpose (i.e., “Medium-high” and “Medium-low” purpose). People in the “Other” group displayed increased mortality risk with increasing purpose but confidence intervals were wide (e.g., people with high purpose had RR = 1.31; 95% CI: 0.57, 3.05).
3.3. Evaluating potential effect modification
Tables 2 and 3 show associations for the joint exposure of purpose in life and gender (Table 2) or race/ethnicity (Table 3) with mortality, as well as corresponding measures of effect modification. Although point estimates suggested that the combination of having the highest purpose and being a woman (versus being a man) had a stronger association with lower mortality risk than the simple sum of the two factors alone, confidence intervals were wide, and evidence of effect modification was only modest (e.g., having the highest purpose and being a woman combined: RERI = 0.06, p = 0.45; RRR = 0.82, p = 0.07). However, in sensitivity analyses where we conducted an omnibus test using continuous purpose scores (Supplementary Table 3), we observed confirmatory evidence for multiplicative effect modification by gender (RRR = 0.93, p = 0.05). Point estimates for the combination of having the highest purpose levels and being Hispanic or Black (versus White) hinted at possible effect modification, but formal tests of additive and multiplicative effect modification provided little evidence for this trend. People in the “Other” group showed relatively similar estimates as Whites.
Table 2.
Relative risks of 8-year all-cause mortality for the joint exposure of purpose in life and gender (N = 13,159).a,b,c,d
| Sense of purpose in Lifeb | ||||||||
|---|---|---|---|---|---|---|---|---|
| Gender | Low | Medium-low | Medium-high | High | ||||
| N Died/alive |
RR (95% CI) |
N Died/alive |
RR (95% CI) |
N Died/alive |
RR (95% CI) |
N Died/alive |
RR (95% CI) |
|
| Men | 586/834 | 1.0 | 392/945 | 0.92 (0.80, 1.04) p = 0.18 |
368/1098 | 0.88 (0.77, 1.01) p = 0.06 |
246/1026 | 0.80 (0.69, 0.93) p < 0.01 |
| Women | 673/1352 | 0.76 (0.68, 0.86) p < 0.01 |
409/1421 | 0.65 (0.57, 0.74) p < 0.01 |
356/1643 | 0.61 (0.53, 0.70) p < 0.01 |
223/1587 | 0.50 (0.43, 0.59) p < 0.01 |
| Additive and multiplicative effect modification | ||||||||
| Low |
Medium-low
Additive c |
Multiplicative d |
Medium-high
Additive |
Multiplicative |
High
Additive |
Multiplicative | ||
| Men | Reference | Reference | Reference | Reference | Reference | Reference | Reference | |
| Women | Reference | −0.03 (−0.18, 0.12) p = 0.73 |
0.93 (0.78, 1.11) p = 0.45 |
−0.04 (−0.19, 0.11) p = 0.62 |
0.90 (0.75, 1.08) p = 0.27 |
−0.06 (−0.22, 0.10) p = 0.45 |
0.82 (0.66, 1.02) p = 0.07 |
|
Risk ratios were adjusted for: Age, race/ethnicity, marital status, education, household income, wealth, health insurance, heart disease, stroke, cancer, diabetes, hypertension, lung disease, arthritis, and depression using Poisson regression.
Purpose in life was assessed using the purpose in life subscale of the Ryff psychological well-being scales. Purpose in life quartiles: Low: 1.00–3.86; medium-low: 3.87–4.57; medium-high: 4.58–5.29; high: 5.30–6.00.
Effect modification on additive scale: Relative excess risk due to interaction (RERI; standard errors were calculated by using the delta method).
Effect modification on multiplicative scale: Ratio of risk ratio (RRR).
Table 3.
Relative risks of 8-year all-cause mortality for the joint exposure of purpose in life and race/ethnicity (N = 13,159).a,b,c,d
| Sense of purpose in lifeb | ||||||||
|---|---|---|---|---|---|---|---|---|
| Race/ethnicity | Low | Medium-low | Medium-high | High | ||||
| N Died/alive |
RR (95% CI) |
N Died/alive |
RR (95% CI) |
N Died/alive |
RR (95% CI) |
N Died/alive |
RR (95% CI) |
|
| White | 1008/1607 | 1.0 | 645/1870 | 0.89 (0.81, 0.99) p = 0.03 |
572/2186 | 0.83 (0.75, 0.93) p < 0.01 |
381/1971 | 0.76 (0.67, 0.86) p < 0.01 |
| Black | 131/245 | 1.01 (0.84, 1.22) p = 0.91 |
101/264 | 0.88 (0.71, 1.08) p = 0.21 |
110/313 | 0.93 (0.77, 1.14) p = 0.50 |
62/419 | 0.60 (0.46, 0.77) p < 0.01 |
| Hispanic | 103/268 | 0.92 (0.74, 1.13) p = 0.41 |
43/177 | 0.72 (0.53, 0.99) p = 0.04 |
32/187 | 0.59 (0.42, 0.85) p < 0.01 |
18/174 | 0.48 (0.30, 0.76) p < 0.01 |
| Others | 17/66 | 0.63 (0.39, 1.02) p = 0.06 |
12/55 | 0.69 (0.39, 1.21) p = 0.20 |
9/55 | 0.72 (0.37, 1.38) p = 0.32 |
8/49 | 0.83 (0.41, 1.67) p = 0.60 |
| Additive and multiplicative effect modification | ||||||||
| Low |
Medium-low
Additive c |
Multiplicative d |
Medium-high
Additive |
Multiplicative |
High
Additive |
Multiplicative | ||
| White | Reference | Reference | Reference | Reference | Reference | Reference | Reference | |
| Black | Reference | −0.03 (−0.29, 0.23) p = 0.84 |
0.97 (0.74, 1.28) p = 0.84 |
0.99 (−0.17, 0.35) p = 0.50 |
1.11 (0.84, 1.46) p = 0.46 |
−0.18 (−0.43, 0.07) p = 0.17 |
0.77 (0.56, 1.07) p = 0.12 |
|
| Hispanic | Reference | −0.08 (−0.38, 0.21) p = 0.59 |
0.89 (0.61, 1.29) p = 0.53 |
−0.16 (−0.44, 0.13) p = 0.29 |
0.78 (0.52, 1.17) p = 0.23 |
−0.20 (−0.50, 0.10) p = 0.20 |
0.68 (0.41, 1.14) p = 0.15 |
|
| Others | Reference | 0.16 (−0.34, 0.66) p = 0.52 |
1.22 (0.58, 2.57) p = 0.61 |
0.25 (−0.31, 0.81) p = 0.38 |
1.36 (0.60, 3.08) p = 0.46 |
0.44 (−0.22, 1.09) p = 0.19 |
1.72 (0.74, 4.03) p = 0.21 |
|
Risk ratios were adjusted for: Age, gender, marital status, education, household income, wealth, health insurance, heart disease, stroke, cancer, diabetes, hypertension, lung disease, arthritis, and depression using Poisson regression.
Purpose in life was assessed using the purpose in life subscale of the Ryff psychological well-being scales. Purpose in life quartiles: Low: 1.00–3.86; medium-low: 3.87–4.57; medium-high: 4.58–5.29; high: 5.30–6.00.
Effect modification on additive scale: Relative excess risk due to interaction (RERI; standard errors were calculated by using the delta method).
Effect modification on multiplicative scale: Ratio of risk ratio (RRR).
4. Discussion
In a national sample of U.S. adults aged >50, we examined the longitudinal association between purpose and mortality risk over the 8-year follow-up period and considered if associations differed depending on gender or race/ethnicity. We had three main findings. First, we observed an overall trend across all groups: as purpose levels increased, risk of all-cause mortality decreased. Second, although distributions of baseline purpose were similar between men and women, the purpose-mortality association was somewhat stronger among women (versus men). When formally testing measures of effect modification, there was slight evidence suggesting that the highest level of purpose may be more protective against mortality among women (versus men). The omnibus test demonstrated additional evidence for potential multiplicative effect modification by gender. Third, there was little evidence of effect modification by race/ethnicity. Point estimates from the stratum-specific purpose-mortality associations hinted that Blacks and Hispanics (versus Whites) may benefit more from protective effects of purpose concerning all-cause mortality. However, when formally testing potential effect modification by race/ethnicity, we did not observe strong statistical support for either observation. The overarching pattern of findings suggests that increasingly higher levels of purpose are beneficial for health regardless of gender or race/ethnicity.
Our findings somewhat diverged from previous evidence. One longitudinal study of older adults in the U.S. reported that the purpose-mortality association was not moderated by gender or race. (Boyle et al., 2009) We likewise observed no moderation by race; however, we also observed slight evidence that the purpose-mortality association was stronger in women than men despite the comparable mean baseline purpose scores across women (4.58) and men (4.59). The diverging results for effect modification by gender might be attributable to differences in sample size as our mortality case count was >20× larger than that of the previous study; hence we had increased power to detect potential effect modification. Further, another study of middle-aged and older adults in Japan observed that the Ikigai-mortality association was stronger in men than in women, (Tanno et al., 2009) whereas our results suggested the opposite. These diverging results might be explained by emerging evidence that associations between psychological factors and health outcomes do not consistently generalize across cultures. Although many, but not all, studies across cultures might maintain main effects (e.g., higher purpose/ikigai is associated with better health), the size of effect modification by gender was different when comparing results from our U.S.-based participants and Japanese participants from prior studies. (Kitayama and Park, 2021)
A potential reason for the weaker association between higher purpose and lower mortality among men (versus women) is that some of the pathways through which purpose provides its health benefits (e.g., promoting healthy behaviors) might be disrupted in men due to gendered social norms, expectations, responsibilities, and barriers that all cumulatively decrease men's willingness to seek health services. (Rieker and Bird, 2005; Williams, 2008; Griffith, 2016) For example, a higher sense of purpose is associated with increased use of preventive health services. (Kim et al., 2014) However, men are less likely to seek high-priority, appropriate preventive services as illustrated in a recent report showing that men were more likely than women to obtain only 0–25% of core preventive healthcare services (21.9% of men versus 11.3% of women). (Borsky et al., 2018) Future research should identify mechanisms that might be blocked among men.
Our study had several limitations. First, the HRS did not assess race/ethnicities beyond Whites, Blacks, and Hispanics. Although there was an “Other” category, the aggregated racial category is not a meaningful demographic because it masks nuanced trends across all other racial groups in the category (e.g., Asians of various backgrounds, Native Americans). Moreover, the category was a small subgroup (n = 271; 2.1% of the total sample) and corresponding confidence intervals were much wider than ones for other racial groups. However, we retained people in this group to spur future research in even more diverse race/ethnicity groups and highlight that larger numbers of people from diverse backgrounds are needed in our large epidemiologic studies. Second, we used a self-identified binary item (i.e., men versus women) to identify a person's gender. This binary measure of gender is crude as it does not capture the non-binary nature of gender. There are also other ways of assessing gender, but HRS does not assess this information. Future work should measure one's gender identity and expression more thoroughly and evaluate how it might influence the purpose-mortality association. Third, our primary analysis assessed the interaction between the categorical purpose variable and categorical effect modifiers, resulting in many statistical tests. Thus, some of the observed associations might be due to chance. However, we conducted omnibus tests using the continuous purpose term to mitigate this concern. Fourth, there may be cultural biases in answering the items in the purpose in life scale, which may explain some of the observed differences in the mean purpose scores across social groups. Our study also had several strengths, including: use of a large, prospective, and national sample of older U.S. adults, adjustment for a range of sociodemographic factors and health conditions which helps reduce bias from confounding, use of a widely used and validated measure of purpose in life, and use of formal interaction analyses on both the additive and multiplicative scale.
In conclusion, with some small caveats, we observed that people with increasingly higher levels of purpose displayed decreasingly lower risk of all-cause mortality, and this association persisted across gender and race/ethnicity. Although further work is needed, there was also some evidence that women might benefit more from the salubrious effects of purpose. Early randomized controlled trials, ranging from group cognitive behavioral therapy to volunteering, have explored whether a sense of purpose (and its conceptual cousin, meaning) can potentially be altered. (Breitbart et al., 2015; Friedman et al., 2019) However, further work is needed to further document the effectiveness of these interventions. Thus, with much needed additional research, purpose could emerge as an important upstream target for interventions and policies aimed at enhancing health across an increasingly diverse population and also a potential target that could improve the health of women and racial/ethnic minorities.
Supplementary Material
Conflicts of interest and source of funding
We declare no conflict of interest. This work was supported by National Institutes of Health (E.S.K, K99AG055696), the Michael Smith Foundation for Health Research (E.S.K.), and the Wellbeing for Planet Earth Foundation (K.S.).
Footnotes
CRediT authorship contribution statement
Koichiro Shiba: Conceptualization, Methodology, Software, Formal analysis, Writing – original draft, Writing – review & editing, Visualization, Project administration. Laura D. Kubzansky: Writing – review & editing, Supervision. David R. Williams: Writing – review & editing. Tyler J. VanderWeele: Writing – review & editing, Methodology. EricS. Kim: Conceptualization, Writing – original draft, Writing – review & editing, Funding acquisition.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.org/10.1016/j.ypmed.2022.107310.
Data availability
The authors do not have permission to share data.
References
- Alimujiang A, Wiensch A, Boss J, et al. , 2019. Association between life purpose and mortality among US adults older than 50 years. JAMA Netw. Open 2 (5), e194270. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Borsky A, Zhan C, Miller T, et al. , 2018. Few Americans receive all high-priority, appropriate clinical preventive services. Health Aff. 37 (6), 925–928. [DOI] [PubMed] [Google Scholar]
- Boyle PA, Barnes LL, Buchman AS, et al. , 2009. Purpose in life is associated with mortality among community-dwelling older persons. Psychosom. Med 71 (5), 574–579. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Breitbart W, Rosenfeld B, Pessin H, et al. , 2015. Meaning-centered group psychotherapy: an effective intervention for improving psychological well-being in patients with advanced cancer. J. CIin. Oncol 33 (7), 749–754. [DOI] [PMC free article] [PubMed] [Google Scholar]
- van Buuren S, Groothuis-Oudshoorn K, 2010. Mice: multivariate imputation by chained equations in R. J. Stat. Softw 1–68. [Google Scholar]
- Cohen R, Bavishi C, Rozanski A, 2016. Purpose in life and its relationship to all-cause mortality and cardiovascular events: a Meta-analysis. Psychosom. Med 78 (2), 122–133. [DOI] [PubMed] [Google Scholar]
- Cuijpers P, Smit F, 2002. Excess mortality in depression: a meta-analysis of community studies. J. Affect. Disord 72 (3), 227–236. [DOI] [PubMed] [Google Scholar]
- Friedman EM, Ruini C, Foy CR, et al. , 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. CIin. Gerontol 42 (4), 387–397. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Griffith DM, 2016. Biopsychosocial approaches to Men’s health disparities research and policy. Behav. Med 42 (3), 211–215. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hafez D, Heisler M, Choi H, et al. , 2018. Association between purpose in life and glucose control among older adults. Ann. Behav. Med 52 (4), 309–318. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hill PL, Turiano NA, 2014. Purpose in life as a predictor of mortality across adulthood. Psychol. Sci 25 (7), 1482–1486. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kang Y, Cosme D, Pei R, et al. , 2021. Purpose in life, loneliness, and protective health behaviors during the COVID-19 pandemic. Gerontologist. 61 (6), 878–887. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kim ES, Strecher VJ, Ryff CD, 2014. Purpose in life and use of preventive health care services. Proc. Natl. Acad. Sci. U. S. A 111 (46), 16331–16336. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kim ES, Kawachi I, Chen Y, et al. , 2017. Association between purpose in life and objective measures of physical function in older adults. JAMA Psychiatry 74 (10), 1039–1045. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kim ES, Delaney SW, Kubzansky LD, 2019. Sense of purpose in life and cardiovascular disease: underlying mechanisms and future directions. Curr. Cardiol. Rep 21 (11), 135. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kim ES, Shiba K, Boehm JK, et al. , 2020a. Sense of purpose in life and five health behaviors in older adults. Prev. Med 139, 106172. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kim ES, Ryff C, Hassett A, et al. , 2020b. Sense of purpose in life and likelihood of future illicit drug use or prescription medication misuse. Psychosom. Med 82 (7), 715–721. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kim ES, Chen Y, Nakamura JS, et al. , 2022. Sense of purpose in life and subsequent physical, behavioral, and psychosocial health: an outcome-wide approach. Am. J. Health Promot 36 (1), 137–147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kitayama S, Park J, 2021. Is conscientiousness always associated with better health? A U.S.-Japan cross-cultural examination of biological health risk. Personal. Soc. Psychol. Bull 47 (3), 486–498. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Knol MJ, VanderWeele TJ, 2012. Recommendations for presenting analyses of effect modification and interaction. Int. J. Epidemiol 41 (2), 514–520. [DOI] [PMC free article] [PubMed] [Google Scholar]
- McKnight PE, Kashdan TB, 2009. Purpose in life as a system that creates and sustains health and well-being: an integrative, testable theory. Rev. Gen. Psychol 13 (3), 242–251. [Google Scholar]
- Radloff LS, 1977. The CES-D scale: a self-report depression scale for research in the general population. Appl. Psychol. Meas 1 (3), 385–401. [Google Scholar]
- Rieker PP, Bird CE, 2005. Rethinking gender differences in health: why we need to integrate social and biological perspectives. J. Gerontol. B Psychol. Sci. Soc. Sci 60 Spec No 2, 40–47. [DOI] [PubMed] [Google Scholar]
- Rubin DB, 2004. Multiple Imputation for Nonresponse in Surveys. John Wiley & Sons, 258 p. [Google Scholar]
- Ryff CD, 2014. Psychological well-being revisited: advances in the science and practice of eudaimonia. Psychother. Psychosom 83 (1), 10–28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ryff CD, Keyes CL, 1995. The structure of psychological well-being revisited. J. Pers. Soc. Psychol 69 (4), 719–727. [DOI] [PubMed] [Google Scholar]
- Ryff CD, Kim ES, 2020. Extending research linking purpose in life to health: The challenges of inequality, the potential of the arts, and the imperative of virtue. In: Burrow AL, Hill PL (Eds.), The Ecology of Purposeful Living across the Lifespan: Developmental, Educational, and Social Perspectives. Springer International Publishing, Cham, pp. 29–58. [Google Scholar]
- Shiba K, Kubzansky LD, Williams DR, et al. , 2021. Associations between purpose in life and mortality by SES. Am. J. Prev. Med 61 (2), e53–e61. https://www.sciencedirect.com/science/article/pii/S0749379721001732. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Steptoe A, Fancourt D, 2019. Leading a meaningful life at older ages and its relationship with social engagement, prosperity, health, biology, and time use. Proc. Natl. Acad. Sci. U. S. A 116 (4), 1207–1212. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sutin AR, Aschwanden D, Luchetti M, et al. , 2021. Sense of purpose in life is associated with lower risk of incident dementia: a Meta-analysis. J. Alzheimers Dis 83 (1), 249–258. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Tanno K, Sakata K, Ohsawa M, et al. , 2009. Associations of ikigai as a positive psychological factor with all-cause mortality and cause-specific mortality among middle-aged and elderly Japanese people: findings from the Japan collaborative cohort study. J. Psychosom. Res 67 (1), 67–75. [DOI] [PubMed] [Google Scholar]
- VanderWeele TJ, Chen Y, Long K, et al. , 2020. Positive epidemiology? Epidemiology. 31 (2), 189–193. [DOI] [PubMed] [Google Scholar]
- Williams DR, 2008. The health of men: structured inequalities and opportunities. Am. J. Public Health 98 (9 Suppl), S150–S157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Zilioli S, Slatcher RB, Ong AD, et al. , 2015. Purpose in life predicts allostatic load ten years later. J. Psychosom. Res 79 (5), 451–457. [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
Data Availability Statement
The authors do not have permission to share data.