Leisure engagement in older age is related to objective and subjective experiences of aging

Leisure engagement has potential to slow health and functional decline in older age. However, the benefits of different leisure domains for different aspects of aging remains unclear. In 8771 older adults from the Health and Retirement Study (a longitudinal panel study), we measured engagement in physical, creative, cognitive, and community activities. Outcome-wide analyses used 23 aging experiences across seven domains eight years later (daily functioning, physical fitness, long-term physical health problems, heart health, weight, sleep, subjective perceptions of health). Physical activity was related to more positive experiences in all domains but heart health eight years later. Creative engagement was positively related to aging experiences in four domains longitudinally. Cognitive and community engagement were less consistently related to aging experiences. Physical and creative activities may influence important aging metrics, reducing age-related decline and keeping older adults functionally independent for longer, potentially limiting increasing healthcare costs.


Participants with major chronic conditions at baseline excluded
Finally, due to concerns around reverse causation, we also limited the sample to participants without chronic health conditions at baseline (Table S9).This resulted in small sample sizes (n=395 to n=1460), meaning there was very little evidence for associations between leisure engagement and experiences of aging.Engagement in physical activities was only associated with lower odds of difficulties with mobility, chronic health conditions, and rating hearing as poor.Creative activities were only associated with lower odds of difficulties with ADLs and using sleep medication, as well as better strength and motor function.Neither cognitive nor community activities were associated with any experiences of aging.However, for all domains of leisure engagement, coefficients and confidence intervals were generally in line with the main analyses (Table S3).

Table S1 .
Sample in which each outcome was measured, and type of regression model used for analyses.

Table S2 .
Proportion of missing data on covariates and outcomes before imputation.

Table S3 .
Sample characteristics at baseline.

Table S4 .
Descriptive statistics for outcomes at baseline and follow-up.

Table S5 .
Unadjusted regression models testing longitudinal associations between leisure engagement and experiences of aging eight years later.

92 (0.88, 0.96) <0.001 0.88 (0.82, 0.95)
Note.Bold text indicates p<0.05.Results weighted and based on 20 imputed datasets.IRR: incidence rate ratio from negative binomial regression model.OR: odds ratio from ordered logistic regression model (binary logistic regression for falls).b: standardized coefficient from linear regression model.All tests were two-sided.No adjustments were made for multiple comparisons.

Table S6 .
Adjusted regression models testing longitudinal associations between leisure engagement and experiences of aging eight years later.Note.All models adjusted for age, gender, race/ethnicity, marital status, education, employment, pension status, household income, assets, household size, neighborhood safety, neighborhood physical disorder, neighborhood social cohesion, and the baseline measure of the outcome.Bold text indicates p<0.05.Results weighted and based on 20 imputed datasets.IRR: incidence rate ratio from negative binomial regression model.OR: odds ratio from ordered logistic regression model (binary logistic regression for falls and sleep medication).b: standardized coefficient from linear regression model.All tests were two-sided.No adjustments were made for multiple comparisons.

Table S7 .
Unadjusted regression models testing concurrent associations between leisure engagement and experiences of aging.Note.Bold text indicates p<0.05.Results weighted and based on 20 imputed datasets.IRR: incidence rate ratio from negative binomial regression model.
OR: odds ratio from ordered logistic regression model (binary logistic regression for falls and sleep medication).b: standardized coefficient from linear regression model.All tests were two-sided.No adjustments were made for multiple comparisons.

Table S8 .
Adjusted regression models testing concurrent associations between leisure engagement and experiences of aging.
IRR: incidence rate ratio from negative binomial regression model.OR: odds ratio from ordered logistic regression model (binary logistic regression for falls and sleep medication).b: standardized coefficient from linear regression model.All tests were two-sided.No adjustments were made for multiple comparisons.

Table S9 .
Adjusted regression models testing longitudinal associations between leisure engagement and experiences of aging eight years later, additionally adjusted for health and health behavior covariates.Note.All models adjusted for age, gender, race/ethnicity, marital status, education, employment, pension status, household income, assets, household size, neighborhood safety, neighborhood physical disorder, neighborhood social cohesion, cognition, depressive symptoms, prescription medication, psychiatric problems, self-rated health, and the outcome measured at baseline, as well as alcohol use and smoking measured at the wave prior to baseline.Bold text indicates p<0.05.Results weighted and based on 20 imputed datasets.
IRR: incidence rate ratio from negative binomial regression model.OR: odds ratio from ordered logistic regression model (binary logistic regression for falls and sleep medication).b: standardized coefficient from linear regression model.All tests were two-sided.No adjustments were made for multiple comparisons.

Table S10 .
Adjusted regression models testing longitudinal associations between leisure engagement and experiences of aging eight years later, with the baseline measure of outcome omitted.
Note.All models adjusted for age, gender, race/ethnicity, marital status, education, employment, pension status, household income, assets, household size, neighborhood safety, neighborhood physical disorder, and neighborhood social cohesion.Bold text indicates p<0.05.Results weighted and based on 20 imputed datasets.IRR: incidence rate ratio from negative binomial regression model.OR: odds ratio from ordered logistic regression model (binary logistic regression for falls and sleep medication).b: standardized coefficient from linear regression model.All tests were two-sided.No adjustments were made for multiple comparisons.