Skip to main content

Some NLM-NCBI services and products are experiencing heavy traffic, which may affect performance and availability. We apologize for the inconvenience and appreciate your patience. For assistance, please contact our Help Desk at info@ncbi.nlm.nih.gov.

NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2019 Jun 1.
Published in final edited form as: Am J Prev Med. 2018 Apr 9;54(6):786–794. doi: 10.1016/j.amepre.2018.02.022

Physical Activity, Sedentary Behavior, and Retirement: The Multi-Ethnic Study of Atherosclerosis

Sydney A Jones 1, Quefeng Li 2, Allison E Aiello 2,3, Angela M O’Rand 4, Kelly R Evenson 1
PMCID: PMC5962425  NIHMSID: NIHMS958576  PMID: 29650285

Abstract

Introduction

Physical activity and sedentary behavior are major risk factors for chronic disease. These behaviors may change at retirement with implications for health in later life. The study objective was to describe longitudinal patterns of moderate to vigorous and domain-specific physical activity and TV watching by retirement status.

Methods

Participants in the Multi-Ethnic Study of Atherosclerosis (N=6,814) were recruited from six U.S. communities and were aged 45–84 years at baseline. Retirement status and frequency and duration of domain-specific physical activity (recreational walking, transport walking, non-walking leisure activity, caregiving, household, occupational/volunteer) and TV watching were self-reported at four study exams (2000 to 2012). Fixed effect linear regression models were used to describe longitudinal patterns in physical activity and TV watching by retirement status overall and stratified by socioeconomic position. Analyses were conducted in 2017.

Results

Of 4,091 Multi-Ethnic Study of Atherosclerosis participants not retired at baseline, 1,012 (25%) retired during a median of 9 years follow-up. Retirement was associated with a 10% decrease (95% CI= −15%, −5%) in moderate to vigorous physical activity and increases of 13% to 29% in recreational walking, household activity, and TV watching. Among people of low socioeconomic position, the magnitude of association was larger for moderate to vigorous physical activity. Among people of high socioeconomic position, the magnitude of association was larger for non-walking leisure and household activity.

Conclusions

The retirement transition was associated with changes in physical activity and TV watching. To inform intervention development, future research is needed on the determinants of behavior change after retirement, particularly among individuals of low socioeconomic position.

INTRODUCTION

Retirement is a major life transition involving disruption in daily activities, time constraints, social support, and priorities.1,2 Maintenance of a healthy lifestyle after retirement can improve quality of life, delay impairment, and lower healthcare costs.35 Regular physical activity is an important component of a healthy lifestyle.3,5 However, the prevalence of physical activity among retirees is low, with 45% of American retirees reporting no physical activity.6 Understanding behavior change at retirement could inform interventions to support physical activity after retirement.7,8

Retirement has been associated with positive and negative changes in physical activity and sedentary behavior.9,10 Leisure-time physical activity and TV watching increased after retirement with inconsistent changes in overall physical activity.1013 Without measures of occupational physical activity, many prior studies could not determine whether increased leisure-time physical activity was sufficient to replace lost occupational activity.10 Utilitarian domains of physical activity, such as transportation activity, have not been well studied.10 Describing domain-specific changes in physical activity is important to guide intervention development. Interventions are more likely to be effective if targeted to specific physical activity domains.14

Patterns of behavior change at retirement may vary by socioeconomic position (SEP).12,15,16 Socioeconomically disadvantaged adults are more likely to retire because of ill health or job loss rather than voluntarily,17,18 and to live alone and with disabilities,19,20 making prevention of chronic disease a priority among disadvantaged elders. This study aims to describe longitudinal patterns in overall moderate to vigorous physical activity (MVPA) and domain-specific physical activity and TV watching among participants in the Multi-Ethnic Study of Atherosclerosis (MESA) by retirement status, overall and within strata of SEP.

METHODS

Study Population

The MESA is a prospective, longitudinal cohort study of subclinical cardiovascular disease (CVD).21 Briefly, 6,814 adults aged 45 to 84 years and free of clinical CVD were recruited at six sites: Forsyth County, North Carolina; Northern Manhattan and the Bronx, New York; Baltimore City and Baltimore County, Maryland; St. Paul, Minnesota; Chicago, Illinois; and Los Angeles County, California. Participants who were retired at baseline (n=2,584) were excluded.12 Pre-retirement physical activity and TV watching measures were not available for these participants and their date of retirement was unknown. Participants with missing data on employment status, physical activity, or covariates at all timepoints also were excluded (n=139). For the remaining 4,091 participants, data were analyzed from five study exams (2000–2002, 2002–2004, 2004–2005, 2005–2007, 2010–2012), and five follow-up phone calls (2007–2012).

Measures

MESA participants self-reported employment status at exams and follow-up calls. Participants who reported being retired and not working, retired and working, or retired and volunteering were classified as retired. All other participants were classified as not retired (Appendix Table 1). Retirement date was estimated as the midpoint between the last non-retired and first retired exam. Participants were classified as retired for all exams after the first exam at which they reported being retired.22

Physical activity and sedentary behavior were self-reported at Exams 1, 2, 3, and 5. The MESA physical activity questionnaire (www.mesa-nhlbi.org/PublicDocs/010101-011231/MESABaselineExamForms/physactivity.pdf) was adapted from the Cross-Cultural Activity Participation Study,23 which had acceptable test–retest reliability (Intra-Class Correlation coefficients 0.55 to 0.75) and validity (r= 0.45 to 0.51) among women.24 Participants reported whether they participated in multiple domains of physical activity and sedentary behavior in a typical week of the last month. Participants reported frequency (days/week) and duration (minutes/day) of activity by domain and by perceived intensity for household/yard, caregiving, conditioning, and occupational/volunteer activities.

This study analyzed one overall and six domain-specific physical activity measures. To calculate overall MVPA, MET tasks were assigned (Appendix Table 2), and MET-minutes/week were summed for moderate and vigorous walking, household/yard, caregiving, non-walking leisure activity, and occupational/volunteer activities. Domain-specific measures were: minutes/week of recreational walking, transport walking, household/yard activity, caregiving activity, non-walking leisure activity (sports, conditioning, and individual activities), and MET-minutes/week occupational/volunteer MVPA. TV watching (minutes/week) was the only sedentary behavior assessed consistently across exams.

SEP was calculated as previously in MESA25 based on self-reported education (high school or less, some college but no degree, associate/bachelor’s degree, graduate/professional degree), household income (<$25,000, $25,000–39,999, $40,000–74,999, ≥$75,000), and four indicators of wealth (ownership of a home, car, land/property, or investments). The SEP score (range, 0–10) was the sum of scores for education (0–3 from lowest to highest), income (0–3 from lowest to highest), and one point for each wealth indicator.

Self-rated health relative to others of the same age (worse, same, better) and partnership status (married/living with partner versus not) were self-reported at study exams. Partnership status was not assessed at Exam 2, so it was imputed from the closer of Exams 1 or 3. At each exam, nine chronic conditions were assessed: self-reported asthma, emphysema, or arthritis flare up in the previous 2 weeks; measured high cholesterol or hypertension; self-reported or measured diabetes; and kidney disease, cancer, and CVD ascertained from medical records and hospital billing claims.21,26

Statistical Analysis

Participant characteristics were described for the overall study population and by retirement status during follow-up. Participant characteristics were compared by retirement status using chi-square and Kruskal–Wallis tests.

Longitudinal patterns in physical activity and TV watching were described using fixed effect regression.27 Fixed effect models focus on within-person variation to control for confounding by measured and unmeasured time-fixed characteristics and account for dependence between repeated measures. The timescale for analyses was age, which is meaningful for people who did and did not retire and accommodated repeated measures taken at unequal intervals. Longitudinal models in this study had the form:

Yij=β1ageij+β2Rij+β3tij+kβkZijk+αi+εij

Age was centered at age 63 years. Rij was a time-varying indicator of retirement, and tij represented time since retirement (tij=0 if Rij=0, tij=ageij-retirement age if Rij=1). The vector Zijk was a dummy indicator variable for time-varying covariates (partnership status, self-rated health, chronic conditions). αi was an individual-specific intercept and εij was an error term. Yij represented the log-transformed outcome measure for individual i at time j. Prior to log-transformation, a small value was added to account for zeros in the data. Exponentiated model coefficients represented the percentage change in the outcome associated with retirement (β2), and percentage change in the outcome per 5-year increase in age among not retired (β1) and retired (β1 + β3) participants, conditional on fixed values of the adjustment variables. Each physical activity measure and TV watching were modeled separately for the entire sample and stratified by SEP (dichotomized at the median). There was no evidence for variation by gender or non-linear changes in outcomes over time. Crude and adjusted estimates were nearly identical so only adjusted estimates are presented. Analyses were conducted in 2017 using SAS, version 9.3.

To address potential residual confounding by health problems that could force retirement and reduce physical activity, sensitivity analyses were conducted among participants (1) in good health throughout follow-up and (2) with chronic diseases (CVD, depression, cognitive impairment, cancer, or chronic obstructive pulmonary disease; Appendix Table 3).28

To explore the robustness of findings, retirement was redefined in four sensitivity analyses (Appendix Table 1). First, retirement status was defined in three-levels: not retired (referent), retired and working for pay (n=184 people), and retired and volunteering or not working (n=828 people). Second, retirement status was reassigned at each exam to accommodate retirees who returned to work (n=141, 14% of retirees). Third, retirement was strictly defined as self-identifying as retired, reporting zero work hours, and no occupational physical activity (n=717, 71% of retirees). Fourth, individuals who ever identified as homemakers (n=866), on leave from work, or unemployed (n=855 observations) were excluded.

Additional sensitivity analyses were conducted to assess self-reported physical activity measures and attrition. People reporting ≥18 hours/day of physical activity were excluded and models were additionally adjusted for method of physical activity questionnaire administration (self versus interviewer) and season of exam. To explore potential selection bias, models were weighted by inverse probability of attrition.29

RESULTS

These analyses included 4,091 MESA participants who were not retired at baseline of the total MESA sample size of 6,814. Participants in this study were younger, more likely female and of higher SEP, and at baseline had fewer chronic conditions, watched less TV, and engaged in more MVPA compared with excluded participants. At baseline, participants had an average age of 57 years, 56% were female, 40% were non-Hispanic white, 51% had a college degree or higher, 65% lived with a partner, and 62% were employed full-time (Table 1). Participants reported a median of 10 minutes/week non-walking leisure activity, 90 minutes/week recreational walking, 180 minutes/week transport walking, 13.5 hours/week household activity, and 12 hours/week TV watching.

Table 1.

Participant Characteristics Overall by Employment Status at Follow-up, MESA (N=4,091)

Baseline characteristics Overalla
(N=4,091)
Retirement status at follow-up
Not retireda
(N=3,079)
Retireda
(N=1,012)
Age (years), mean ± SD 57.3 ± 8.9 56.9 ± 9.4 58.4 ± 6.9
Female 2,309 (56) 1,768 (57) 541 (53)
Race/ethnicity, n (%)
 Non-Hispanic white 1,628 (40) 1,193 (39) 435 (43)
 Chinese American 522 (13) 416 (14) 106 (10)
 Non-Hispanic black 975 (24) 680 (22) 295 (29)
 Hispanic 966 (24) 790 (26) 176 (17)
Education, n (%)
 <High school 684 (17) 581 (19) 103 (10)
 Some college, no degree 1,303 (32) 950 (31) 353 (35)
 ≥Bachelor’s degree 2,102 (51) 1,547 (50) 555 (55)
Low SEP, n (%) 2,003 (49) 1,568 (51) 435 (43)
Married / with partner, n (%) 2,636 (65) 1,994 (65) 642 (64)
Baseline job, n (%)
 Homemaker 749 (18) 694 (23) 55 (5)
 Employed full time 2,549 (62) 1,799 (58) 750 (74)
 Employed part time 570 (14) 411 (13) 159 (16)
 On leave or unemployed 221 (5) 174 (6) 47 (5)
Self-rated health, n (%)
 Better 2,131 (53) 1,554 (51) 577 (57)
 Same 1,612 (40) 1,261 (41) 351 (35)
 Worse 316 (8) 239 (8) 77 (8)
MESA site, n (%)
 Forsyth, NC 594 (15) 383 (12) 211 (21)
 New York City, NY 653 (16) 488 (16) 165 (16)
 Baltimore, MD 584 (14) 442 (14) 142 (14)
 St. Paul, MN 713 (17) 527 (17) 186 (18)
 Chicago, IL 743 (18) 543 (18) 200 (20)
 Los Angeles, CA 804 (20) 696 (23) 108 (11)
Physical activity, median (IQR)
 MVPA (MET-minutes/week) 4,590 (2,190, 8,475) 4,425 (2,098, 8,370) 5,220 (2,513, 9,025)
 Non-walking leisure (minutes/week) 10 (0, 180) 3 (0, 180) 15 (0, 180.0)
 Recreation walking (minutes/week) 90 (0, 225) 90 (0, 225) 90 (0, 240.0)
 Transport walking (minutes/week) 180 (45, 420) 150 (35, 375) 180 (60, 420.0)
 Household/yard (minutes/week) 810 (420, 1,370) 815 (420, 1,380) 765 (390, 1,290)
 Caregiving (minutes/week) 0 (0, 180) 0 (0, 180.0) 0 (0, 165)
 Occupational/volunteer 0 (0, 2,880) 0 (0, 2,700) 450 (0, 3,600)
 MVPA (MET-minutes/week)
 TV watching (hours/week), median (IQR) 12 (5, 18) 12 (5, 19) 12 (6, 18)

MVPA, moderate to vigorous physical activity; SEP, socioeconomic position; IQR, interquartile range; MESA, Multi-Ethnic Study of Atherosclerosis

During a median of 9 years of follow-up, 1,012 participants (25%) retired. Median retirement age was 63 years. Compared with participants who did not retire, retirees were more likely male, non-Hispanic white or black, of higher SEP, older, in better health, employed full-time, and reported more MVPA and transport walking at baseline (Table 1). There were 435 retirees with low SEP and 577 with high SEP.

Domain-specific patterns of physical activity and TV watching by retirement status, overall (Table 2) and by SEP (Table 3), are described next. Findings are presented graphically in Appendix Figures 1 and 2.

Table 2.

Changea in Physical Activity and TV Watching by Retirement Status (N=4,091)

Behavior Mean change associated with retirementb
(95% CI)
Mean 5-year changec
(95% CI)
Not retired Retired
MVPA 0.90 (0.85, 0.95) 0.94 (0.92, 0.96) 0.96 (0.90, 1.02)
Non-walking leisure 1.09 (1.03, 1.14) 1.01 (1.00, 1.03) 1.02 (0.97, 1.08)
Recreation walking 1.13 (1.07, 1.20) 1.08 (1.06, 1.10) 1.08 (1.01, 1.14)
Transport walking 0.99 (0.93, 1.06) 1.02 (0.99, 1.04) 1.01 (0.94, 1.08)
Household/yard activity 1.29 (1.22, 1.36) 1.02 (1.00, 1.04) 0.97 (0.91, 1.02)
Caregiving activity 1.00 (0.93, 1.07) 0.98 (0.96, 1.00) 0.97 (0.90, 1.05)
Occupational/volunteer 0.52 (0.48, 0.56) 0.84 (0.82, 0.87) 1.03 (0.95, 1.11)
MVPA
TV watching 1.15 (1.08, 1.21) 1.14 (1.12, 1.16) 1.14 (1.08, 1.21)
a

Values are exponentiated coefficients from linear models of log-transformed outcomes. Values can be interpreted as percentage differences, for example 1.07 represents a 7% increase in the outcome associated with retirement, conditional on adjustment variables (self-reported health, partnership status, and 9 chronic conditions: asthma, emphysema, arthritis flare up, high cholesterol, hypertension, diabetes, kidney disease, cancer, cardiovascular disease). Data presented graphically in Appendix Figure 1.

b

Exponentiated mean difference in log-transformed outcome associated with retirement, conditional on adjustment variables.

c

Exponentiated mean difference in log-transformed outcome associated with 5-year increase in age among retired and not retired participants, conditional on adjustment variables.

MVPA, moderate to vigorous physical activity

Table 3.

Changea in Physical Activity and TV Watching by Retirement Status and Socioeconomic Position (N=4,091)

Behavior Low SEP
(N=2,003)
High SEP
(N=2088)
Mean change associated with retirementb
(95% CI)
Mean 5-year changec
(95% CI)
Mean change associated with retirementb
(95% CI)
Mean 5-year changec
(95% CI)
Not retired Retired Not retired Retired
MVPA 0.76
(0.68, 0.84)
0.89
(0.86, 0.92)
0.96
(0.87, 1.07)
1.01
(0.95, 1.09)
0.98
(0.96, 1.00)
0.95
(0.89, 1.02)
Non-walking leisure 1.00
(0.92, 1.08)
1.01
(0.99, 1.04)
1.09
(1.01, 1.18)
1.16
(1.08, 1.24)
1.02
(1.00, 1.04)
0.98
(0.91, 1.05)
Recreation walking 1.17
(1.06, 1.29)
1.06
(1.03, 1.09)
1.04
(0.94, 1.15)
1.11
(1.03, 1.19)
1.09
(1.07, 1.12)
1.10
(1.02, 1.19)
Transport walking 0.99
(0.89, 1.10)
1.02
(0.99, 1.06)
1.02
(0.91, 1.13)
1.00
(0.92, 1.08)
1.01
(0.98, 1.04)
1.00
(0.92, 1.09)
Household/yard activity 1.20
(1.09, 1.32)
0.99
(0.96, 1.02)
0.97
(0.88, 1.06)
1.36
(1.27, 1.45)
1.05
(1.02, 1.07)
0.97
(0.91, 1.04)
Caregiving activity 1.08
(0.96, 1.22)
0.94
(0.91, 0.98)
0.83
(0.73, 0.95)
0.94
(0.86, 1.03)
1.01
(0.98, 1.04)
1.09
(0.99, 1.19)
Occupational/volunteer MPVA 0.43
(0.38, 0.48)
0.80
(0.77, 0.83)
1.02
(0.90, 1.16)
0.59
(0.54, 0.65)
0.88
(0.85, 0.91)
1.03
(0.94, 1.13)
TV watching 1.13
(1.03, 1.24)
1.15
(1.12, 1.19)
1.11
(1.00, 1.22)
1.16
(1.08, 1.24)
1.13
(1.11, 1.16)
1.17
(1.09, 1.26)
a

Values are exponentiated coefficients from linear models of log-transformed outcomes. Values can be interpreted as percentage differences, for example 1.07 represents a 7% increase in the outcome associated with retirement, conditional on adjustment variables (self-reported health, partnership status, and 9 chronic conditions: asthma, emphysema, arthritis flare up, high cholesterol, hypertension, diabetes, kidney disease, cancer, cardiovascular disease). Data are presented graphically in Appendix Figure 2.

b

Exponentiated mean difference in log-transformed outcome associated with retirement, conditional on adjustment variables.

c

Exponentiated mean difference in log-transformed outcome associated with 5-year increase in age among retired and not retired participants, conditional on adjustment variables.

MVPA, moderate to vigorous physical activity; SEP, socioeconomic position

Retirement was associated with a 10% decrease in MVPA (95% CI= −15%, −5%; Table 2). MVPA declined over time regardless of retirement status at a rate of −4% to −6% per 5 years. Stratified by SEP, retirement was associated with a 24% decrease in MVPA among individuals of low but not high SEP (Table 3).

Retirement was associated with a 9% increase (95% CI=3%, 14%) in non-walking leisure activity in the overall sample (Table 2) and 16% increase (95% CI=8%, 24%) among people of high SEP (Table 3). The average change in non-walking leisure activity per 5 years was small, except for an increase among low SEP retirees.

Retirement was associated with a 13% (95% CI=7%, 20%) increase in recreational walking in the overall sample (Table 2), and among individuals of low and high SEP (17% and 11% increase, respectively; Table 3). Recreational walking increased at a rate of 8% per 5 years in the overall sample with small differences between SEP strata. Retirement was not associated with changes in transport walking (Tables 2 and 3).

At retirement, household/yard activity increased by an average of 29% (95% CI=22%, 36%) in the overall sample (Table 2) and increased in both SEP strata (Table 3). Household/yard activity changed little over time. In the overall sample, retirement was not associated with changes in caregiving (Table 2). Caregiving activity declined with age among individuals of low but not high SEP (Table 3).

Retirement was associated with a 48% decrease in occupational/volunteer MVPA (95% CI= −52%, −44%) in the overall sample (Table 2) and decreases in both SEP strata (Table 3). Among non-retired participants, occupational/volunteer MVPA declined with age in all SEP strata.

Retirement was associated with a 15% increase in TV watching (95% CI=8%, 21%) in the overall sample (Table 2) and increases in both SEP strata (Table 3). TV watching increased by 11%–17% per 5 years in all retirement and SEP groups.

Among 2,085 participants who were healthy throughout follow-up, 611 retired (226 low SEP and 385 high SEP). In this healthy subset, retirement was associated with a smaller decrease in MVPA among people of low SEP and larger increase in household/yard activity in both SEP strata (Table 4). Among participants with chronic diseases (n=1,435), retirement was associated with a larger decrease in MVPA among individuals of low SEP, and smaller increase in household/yard activity in both SEP strata (Appendix Table 4).

Table 4.

Changea in Physical Activity and TV Watching by Retirement and SEP in Healthy Subsetb (N=2,085)

Behavior Low SEP
(N=828)
High SEP
(N=1,257)
Mean change associated with retirementc
(95% CI)
Mean 5-year changed
(95% CI)
Mean change associated with retirementc
(95% CI)
Mean 5-year changed
(95% CI)
Not retired Retired Not retired Retired
MVPA 0.81
(0.71, 0.93)
0.88
(0.85, 0.92)
0.93
(0.81, 1.07)
1.06
(0.97, 1.15)
0.98
(0.96, 1.01)
0.93
(0.86, 1.01)
Non-walking leisure 0.97
(0.97, 1.08)
1.00
(0.97, 1.03)
1.09
(0.97, 1.21)
1.14
(1.05, 1.24)
1.03
(1.00, 1.05)
1.00
(0.92, 1.09)
Recreation walking 1.15
(1.02, 1.33)
1.08
(1.04, 1.13)
1.02
(0.89, 1.17)
1.15
(1.05, 1.26)
1.11
(1.08, 1.14)
1.09
(1.00, 1.19)
Transport walking 1.02
(0.88, 1.18)
1.00
(0.96, 1.05)
1.01
(0.87, 1.16)
1.01
(0.91, 1.11)
1.01
(0.98, 1.04)
0.99
(0.90, 1.09)
Household/yard activity 1.33
(1.17, 1.50)
1.00
(0.97, 1.04)
0.90
(0.80, 1.02)
1.44
(1.33, 1.56)
1.06
(1.03, 1.08)
0.96
(0.89, 1.04)
Caregiving activity 1.19
(1.00, 1.42)
0.94
(0.89, 0.99)
0.75
(0.63, 0.89)
1.01
(0.91, 1.13)
1.03
(0.99, 1.06)
1.01
(0.91, 1.12)
Occupational/volunteer MVPA 0.44
(0.37, 0.52)
0.78
(0.74, 0.82)
1.10
(0.92, 1.31)
0.60
(0.54, 0.67)
0.86
(0.83, 0.89)
1.02
(0.91, 1.14)
TV watching 1.11
(0.98, 1.26)
1.16
(1.12, 1.21)
1.11
(0.97, 1.26)
1.19
(1.10, 1.30)
1.12
(1.09, 1.15)
1.11
(1.03, 1.21)
a

Values are exponentiated coefficients from linear models of log-transformed outcomes. Values can be interpreted as percentage differences, for example 1.07 represents a 7% increase in the outcome associated with retirement, conditional on adjustment variables (self-reported health, partnership status, and 9 chronic conditions: asthma, emphysema, arthritis flare up, high cholesterol, hypertension, diabetes, kidney disease, cancer, cardiovascular disease).

b

Restricted to participants present at study exam 5 and free of cardiovascular disease, depression, cognitive impairment, cancer, and chronic obstructive pulmonary disease throughout follow-up.

c

Exponentiated mean difference in log-transformed outcome associated with retirement, conditional on adjustment variables.

d

Exponentiated mean difference in log-transformed outcome associated with 5-year increase in age among retired and not retired participants, conditional on adjustment variables.

MVPA, moderate to vigorous physical activity; SEP, socioeconomic position

Employing alternative retirement definitions made little impact on the results (data not shown). Among people retired and working (n=184), retirement was not associated with change in MVPA, recreational walking, or non-walking leisure activity. Other findings were similar among retirees who were and were not working. Adjustment for mode of physical activity questionnaire administration and season of exam, exclusion of people reporting excessive activity, and weighting by inverse probability of attrition did not alter the pattern of results (data not shown).

DISCUSSION

In this U.S. cohort, retirement was associated with increased recreational walking, household/yard activities, and TV watching. Among individuals of low SEP, retirement was associated with declines in overall MVPA. Overall MVPA was stable during the retirement transition among participants of high SEP who reported more non-walking leisure activity after retirement. These findings are consistent with increased leisure-time physical activity after retirement observed among people of high SEP but not low SEP,10,12,15,16 and suggest that the retirement transition may be an important period for physical activity promotion, particularly among individuals of low SEP.30

Among retirees of low SEP, declines in occupational activity at retirement were not recouped by increases in other physical activity domains, resulting in decreased overall MVPA. Poor heath is one explanation for this finding. Illness and disability are more common reasons for retirement among people of low SEP17 and associated with limited physical activity.31 Retirement was associated with larger decreases in MVPA among individuals of low SEP with chronic diseases. However, persistence of a negative association of retirement with MVPA among healthy retirees of low SEP suggests that poor health does not entirely account for MVPA declines associated with retirement. Future studies could be strengthened by exploring the role of physical function, which was not consistently measured in MESA.

SEP is a multidimensional construct that can be measured at different points in the lifespan.32 In this study, adult SEP was measured prior to retirement as a composite of education, income, and wealth.25 Stratification by education instead of SEP yielded consistent results (data not shown).

Cumulative disadvantage was not measured in this study; however, earlier life opportunities are key determinants of retirement and physical activity.3336 Individuals with more debt or lower income are less likely to be able to retire and more likely to return to work after retirement.34,37 Low SEP can constrain earlier life physical activity through limited leisure-time, disposable income, and the demands of physical labor.36,38 After retirement, low SEP may be associated with less access to physical activity equipment and residence in a neighborhood lacking physical activity supports (e.g., few parks),39 whereas age discrimination may discourage activity in public.40 In addition to SEP, the relationship between retirement and physical activity may vary by race/ethnicity32; however, estimates stratified by race/ethnicity and SEP were imprecise due to small numbers of retirees in some subgroups (data not shown). Future research should consider race/ethnicity in addition to SEP to inform intervention development.

The feasibility of intervening to promote physical activity during the retirement transition is not well established.41 Ninety percent of large worksites offer health promotion programs; however, retirees often are excluded.42 In addition, although employers assist with financial planning, few resources are available to help workers prepare for post-retirement physical activity.34 Existing retirement planning resources could be expanded to include health promotion materials.35 However, racial/ethnic minorities and women have less access to retirement planning and are more likely to retire without the opportunity to plan (e.g., because of job loss).34 Thus, multiple strategies, including community-based interventions, are needed to avoid exacerbation of health inequities. Experience Corps was a successful intervention among low-income retirees that improved physical activity and strength after retirement.43 In addition to programming, environmental supports may enable retirees to be physically active.44 However, despite a strong theoretic rationale, it is not clear how environmental changes affect retirees’ physical activity.45,46

Retirement was associated with increased TV watching in this sample, which is consistent with findings in other studies.12,4750 There was little difference by SEP, contrary to some studies.12,50 TV watching is the most commonly reported sedentary behavior among older adults.49,51 It is unclear how retirement affects other domains of sedentary behavior not measured in MESA. For example, Internet and social media could promote sedentary behavior and be a source of information for physical activity. Also, factors that determine whether sedentary work time is replaced by active or sedentary pursuits after retirement have not been identified.13 Overall sedentary behavior increased at retirement among French48 but decreased among Australian retirees.52,53 Future work should explore correlates of sedentary behavior change at retirement.

Limitations

Limitations to this work include reliance on self-reported retirement, physical activity, and TV watching. Perceptions of what it means to be retired may vary among participants. Some retirees continued to work. Sensitivity analyses restricted to retirees reporting zero hours of work did not appreciably change results. However, physical activity differences at retirement were attenuated among retired and working participants, perhaps because retirement influences behavior partly through changes in discretionary time.16,18 Self-report typically overestimates physical activity and underestimates sedentary behavior relative to accelerometer measures.54,55 It is not clear whether measurement error in self-reported physical activity or sedentary behavior varied by retirement status. Combined use of self-report and accelerometer measures would strengthen future studies.

The generalizability of findings may be limited because at baseline MESA participants were free of clinical CVD and willing to participate in a longitudinal study, and about 25% of the sample died or was lost to follow-up.21 However, median retirement age of 63 years and 14% rate of return to work after retirement are consistent with U.S. averages.34,56 Findings were robust after statistical adjustment for attrition.

Strengths of this study include prospective follow-up of a diverse U.S. cohort with repeated measures of multiple physical activity domains and TV watching. Domain-specific changes in physical activity were explored, which may improve targeting of interventions. Also, fixed effect models controlled for confounding by time-invariant characteristics by design and were adjusted for time-varying confounders. Further, multiple sensitivity analyses explored the robustness of the findings.

CONCLUSIONS

The health of retirement age adults is of increasing public health importance given demographic trends. Almost one quarter of the American workforce is aged 55 years or older57 and 72 million Americans will be aged 65 years or older by 2030.58 Although retirement is increasingly viewed as an active phase of life,36 the prevalence of physical activity among retirees remains low.6 Interventions during the retirement transition may help support active lifestyles in later life. To inform intervention development, future research is needed on the determinants of behavior change after retirement, particularly among individuals of low SEP.

Supplementary Material

supplement

Acknowledgments

The Multi-Ethnic Study of Atherosclerosis (MESA) was supported by contracts HHSN268201500003I, N01-HC-95159, N01-HC-95160, N01-HC-95161, N01-HC-95162, N01-HC-95163, N01-HC-95164, N01-HC-95165, N01-HC-95166, N01-HC-95167, N01-HC-95168 and N01-HC-95169 from the National Heart, Lung, and Blood Institute (NHLBI) and by grants UL1-TR-000040 and UL1-TR-001079 from the National Center for Research Resources and 2R01-HL071759 from the NHLBI. The authors thank the other investigators, the staff, and the participants of the MESA study for their valuable contributions. A full list of participating MESA investigators and institutions can be found at www.mesa-nhlbi.org. SAJ was supported by the University of North Carolina Royster Society of Fellows, Gillings Dissertation Award, and the NHLBI (National Research Service Award #T32-HL007055-38). Aiello was supported by P60 MD 002249 and R01DK087864.

This research was reviewed by the Institutional Review Board of the University of North Carolina at Chapel Hill and deemed exempt (IRB #15-3296).

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

No financial disclosures were reported by the authors of this paper.

References

Associated Data

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

Supplementary Materials

supplement

RESOURCES