Aging on the Job? The Association Between Occupational Characteristics and Accelerated Biological Aging

Abstract

Objectives

There is a common belief that demanding jobs can make workers age faster, but there is little empirical evidence linking occupational characteristics to accelerated biological aging. We examine how occupational categorizations and self-reported working conditions are associated with expanded biological age, which incorporates 22 biomarkers and captures physiologic dysregulation throughout several bodily systems.

Methods

Data are from 1,133 participants in the Health and Retirement Study who were aged 51–60 and working for pay in the 2010 or 2012 wave and who participated in the 2016 Venous Blood Study. We estimate associations between occupational category (professional/managerial, sales/clerical, service, and manual) and self-reported working conditions (psychosocial demands, job control, heavy lifting, and working 55 or more hours per week) and expanded biological age.

Results

Compared to same-age individuals working in professional or managerial positions, those working in service jobs appear 1.65 years older biologically even after adjusting for social and economic characteristics, self-reported working conditions, health insurance, and lifestyle-related risk factors. Low job control is associated with 1.40 years, heavy lifting with 2.08 years, and long working hours with 1.87 years of accelerated biological aging.

Discussion

Adverse occupational characteristics held at midlife, particularly service work, low job control, heavy lifting, and long work hours, are associated with accelerated biological aging. These findings suggest that work may be important for the overall aging process beyond its associations with specific diseases or risk factors.

Expressions such as “working oneself to an early grave” are indicators of a common belief that certain jobs can make workers age faster than expected and increase the risk of poor health in the process. Such sayings may have empirical support, as work is a social determinant of health that has been associated with mortality and many age-related diseases—musculoskeletal disorders, cardiovascular disease, diabetes, certain cancers—which would support the theory that working conditions can alter the underlying aging process (Andrasfay et al., 2021; Brussig & Drescher, 2022; Landsbergis, 2010; Marmot & Brunner, 2005; Townsend & Mehta, 2021).

Background on Occupational Characteristics and Aging

It is well established from research in both the United States and other countries that risks for self-assessed health and several age-related diseases vary according to the types of occupations in which individuals are employed (Al-Qaoud et al., 2011; Case & Deaton, 2005; Cutler et al., 2011; Elovainio et al., 2011; Engström et al., 2006; Fraga et al., 2015; Hotchkiss et al., 2012; Landsbergis, 2010; Marmot & Brunner, 2005; Schmitz, 2016; Tabassum et al., 2014; Townsend & Mehta, 2021). Manual work, which is often characterized by physical demands, has been associated with hypertension and biomarkers indicating elevated inflammation and increased risk of cardiovascular disease (Engström et al., 2006; Fraga et al., 2015; Hotchkiss et al., 2012; Tabassum et al., 2014). Service sector occupations, which are characterized by performing services for customers or patients and are frequently low paying, have been associated with similar or worse levels of self-assessed health compared to manual occupations (Case & Deaton, 2005; Cutler et al., 2011; Schmitz, 2016). Even within higher-status occupations, there are differences in health profiles, with sales and clerical workers reporting worse overall health and having higher levels of biomarkers indicating suboptimal kidney functioning, glucose control, and cholesterol levels than managerial or professional workers (Al-Qaoud et al., 2011; Case & Deaton, 2005; Cutler et al., 2011; Elovainio et al., 2011).

These health differences likely result in part from the working conditions experienced by individuals employed in these occupational categories (Burgard & Lin, 2013; Landsbergis, 2010). There is extensive research to suggest that common working conditions, including psychosocial stressors, physical demands, and long hours, are associated with biomarkers indicating elevated risk for several age-related diseases. As laid out in the stress process model, the experience of stress in work or other domains of life can negatively influence the aging process through chronic activation of the stress response and increased inflammation, which can result in physiologic dysregulation over the long term (Ferraro & Shippee, 2009; Turner, 2010). Apart from these direct effects, stress experienced during work hours may lead to the adoption of unhealthy coping behaviors such as smoking and drinking and lead to worse sleep quality, which can all contribute to physiologic dysregulation (Burgard & Ailshire, 2009; Burgard & Lin, 2013). Psychosocial stress at work, measured through high psychosocial demands, low control over one’s tasks and schedule, or job strain, which is defined by low control combined with high psychosocial demands, has been associated with elevated blood pressure, blood glucose, and inflammation (Clays et al., 2005; Landsbergis et al., 2013; Magnusson Hanson et al., 2019; Netterstrøm & Sjøl, 1991; Steptoe & Willemsen, 2004).

Physical activity at work, in contrast to leisure-time physical activity, is often characterized by repetitive motions carried out over a long period without sufficient recovery time, and can be detrimental to health over the long term (Holtermann et al., 2020). Occupational physical activity has been associated with increased blood pressure, resting heart rate, and inflammation (Barbe & Barr, 2006; Holtermann et al., 2020). Long working hours can interfere with one’s ability to maintain a healthy lifestyle, including eating a healthy diet, exercising, and getting adequate sleep, all of which may contribute to the observed association between long working hours and biomarkers indicating hypertension, unfavorable cholesterol profiles, inflammation, reduced kidney and liver functioning, and suboptimal glucose control among diabetic workers (Davila et al., 2012; Song et al., 2021; Virtanen et al., 2012, 2019).

Although this prior research suggests that occupational characteristics influence numerous bodily systems, with few exceptions studies have typically focused on only one biomarker or biomarkers of only one physiologic system at a time, so it is unclear whether occupational characteristics are associated with overall accelerated aging. To our knowledge, there has been no empirical evidence linking occupational categories or unfavorable working conditions to accelerated aging at the biological level.

Biological Age

With the increasing collection of blood samples in population-based surveys and the development of summary measures of biological age, it is now possible to study social determinants of biological age, which is an estimated age based on a set of clinical chemistry and measured biomarkers that indicate dysregulation or dysfunction across multiple bodily systems (Crimmins et al., 2021; Klemera & Doubal, 2006; Levine, 2013). Such measures of biological age can reveal variation among individuals who are the same chronological age and capture subclinical or preclinical presentations of diseases among individuals who may otherwise appear healthy (Turner, 2010). Understanding social determinants of biological age is important because accelerated biological aging—when biological age exceeds chronological age—is a strong predictor of subsequent chronic disease, disability, and mortality (Crimmins et al., 2021; Levine, 2013).

Several measures of biological age have been developed to quantify the degree of aging implied by biomarker profiles. Biological age, developed by Levine (2013), incorporated 10 biomarkers available in the third National Health and Nutrition Examination Study (conducted 1988–1994) that had the strongest associations with chronological age. Phenotypic age was later developed by Levine and colleagues as an age implied by one’s mortality risk; though some of its nine biomarkers overlap with those of biological age, phenotypic age includes several clinical chemistry measures reflecting blood cell composition obtained from a complete blood count that were not included in biological age (Liu et al., 2018).

Recently, a new measure of biological aging, expanded biological age, was developed by Crimmins et al. (2021) to examine whether a summary measure of aging based on a more comprehensive set of 22 biomarkers would outperform previous measures of biological age. In addition to the biomarkers included in biological age and phenotypic age, expanded biological age includes a set of measures called the Targeting Aging with Metformin (TAME) assays, which were identified by experts as biomarkers that should be evaluated in clinical trials assessing metformin’s potential to delay aging (Crimmins et al., 2021). With the exception of serum creatinine and fasting glucose, which are superseded by the more sensitive measures of cystatin C and glycated hemoglobin, respectively, expanded biological age includes all biomarkers from biological age, phenotypic age, and the TAME set of measures that were available in the Health and Retirement Study (HRS)’s Venous Blood Study (VBS), as well as three additional inflammatory cytokines and a measure of immune senescence, for a total of 22 biomarkers that capture cardiovascular, metabolic, kidney, lung, and immune functioning (Crimmins et al., 2021). Expanded biological age was found to perform similarly or better than biological age and phenotypic age at predicting disability, multimorbidity, cognitive dysfunction, and mortality (Crimmins et al., 2021).

Current Study

The current study takes advantage of the 2016 VBS, collected as part of the HRS, to examine how occupational characteristics at midlife are associated with accelerated biological aging, using a recently developed measure, expanded biological age, a predicted age based on 2016 levels of 22 biomarkers of aging chosen for their past links with age-related health outcomes or by experts focused on mechanisms of aging (Crimmins et al., 2021). Based on the previous literature, we hypothesize that individuals employed in service and manual work will exhibit signs of accelerated aging relative to those in professional/managerial work. We also hypothesize that occupational psychosocial stressors, physical demands at work, and long working hours will be associated with accelerated biological aging and account for part of the differences in biological age by occupational categories. While prior studies have often treated occupation as a proxy for overall socioeconomic status (SES)—making it difficult to understand whether it is the characteristics and conditions of the occupations themselves or the background characteristics of workers that account for the observed occupational differences—we take into account the sociodemographic characteristics of workers in these occupations. Additionally, we consider financial resources, access to health insurance, and lifestyle-related risk factors that are potential mechanisms through which occupational characteristics may contribute to accelerated aging.

Data and Methods

Data

This study utilizes data from the HRS, a nationally representative longitudinal survey of American adults over 50 (Sonnega et al., 2014). The HRS began in 1992 as a sample of older American adults and their spouses, has included follow-up waves every 2 years and the introduction of younger refresher cohorts every 6 years, and administers a more-extensive psychosocial and lifestyle leave-behind questionnaire to alternating halves of the sample each wave (Sonnega et al., 2014). In the 2016 wave the HRS introduced the VBS to measure an extensive set of blood-based biomarkers (Crimmins et al., 2017). The 2016 VBS only included participants who had already entered the HRS panel prior to the 2016 wave and who participated in the 2016 core interview, and it is representative of adults aged 56 and older (Crimmins et al., 2017).

Because there is attrition from the labor force as individuals age, and because exiting the labor force or switching to less demanding jobs is more likely for individuals in poor health, we measure occupational characteristics prior to the measurement of expanded biological age to limit the effects of selective attrition out of employment or out of career occupations by the time individuals participated in the VBS. We take occupational characteristics from either the 2010 or 2012 wave, depending on when individuals were eligible for the leave-behind questionnaire, and limit our analysis to those who were aged 51–60 at these waves, as these individuals were not yet eligible for early retirement benefits. Figure 1 details the sample selection process, which results in a final analytic sample of 1,133 individuals.

Figure 1.

Sample selection diagram. Data are from the Health and Retirement Study. VBS = Venous Blood Study.

Measures

The outcome measure is expanded biological age, which is estimated from 22 biomarkers, including systolic blood pressure (mmHg), total cholesterol (mg/dL), cytomegalovirus seroprevalence, alkaline phosphatase (U/L), blood urea nitrogen (mg/dL), peak flow (L/min), albumin (g/dL), log C-reactive protein (mg/L), lymphocyte percentage, mean cell volume, red cell distribution width, white blood cell count, log interleukin-6 (pg/mL), tumor necrosis factor receptor 1 (pg/mL), insulin-like growth factor-1 (ng/mL), cystatin C (mg/L), log N-terminal pro-b-type natriuretic peptide (pg/mL), glycated hemoglobin (HbA1c; %), interleukin-10 (pg/mL), interleukin-1 receptor antagonist (pg/mL), transforming growth factor beta (pg/mL), and the ratio of CD4 to CD8 cell counts (Crimmins et al., 2021; see Author Note 1). We construct a composite biological age for each individual following the same method used by Levine (2013), resulting in an estimate that indicates the chronological age at which his/her levels of these 22 biomarkers would be typical. The calculations of biological age are done separately by sex on the full set of VBS participants with valid measurements on these biomarkers, before the sample restrictions specific to this study.

The independent variables considered as risk factors for accelerated aging include category of occupation and three self-reported working conditions measured at the 2010 or 2012 wave for the respondent’s main job. Occupation is categorized as professional/managerial, sales/clerical, service, and manual according to the two-digit occupation codes available in public-use HRS files that are based on the job titles reported by respondents (see Author Note 2). We create indices of psychosocial job demands and job control from items reported in the psychosocial leave-behind questionnaire. The psychosocial job demands index is the average of respondent reports of time pressure, needing to work fast, and experience of conflicting demands. The job control index is the average of respondent reports of freedom over one’s work, control over what happens in most situations, and the opportunity to develop new skills. These items have been used in prior research examining the effects of job demands and control using data from the HRS (Lunau et al., 2018). Following previous literature, we dichotomize psychosocial job demands and job control into high and low at their weighted sample medians, and we consider the combination of high psychosocial demands and low job control as an indicator of high job strain, which is thought to identify workers at particularly high risk of adverse psychological consequences (Fransson et al., 2012; Karasek, 1979). We also include measures of the frequency of heavy lifting and usual hours worked from the core HRS surveys. We dichotomize these into heavy lifting (1 = all/almost all the time; 0 if most of the time, some of the time, or none/almost none of the time), and long hours (1 = 55+ hr per week; 0 if less than 55 hr per week). The 55+-hr threshold for long working hours has been shown in previous studies to be associated with adverse outcomes including cardiovascular disease and stroke (Kivimäki et al., 2015).

Covariates

Demographic covariates include chronological age and gender. We theorize that race/ethnicity and educational attainment may confound the associations between occupational characteristics and biological age because systemic racism constrains educational opportunities, which in turn constrains employment opportunities. Additionally, prior work has shown that Black, Hispanic, and less educated older adults exhibit accelerated biological aging (Farina et al., 2022). We categorize race/ethnicity as non-Hispanic White, non-Hispanic Black, Hispanic, and non-Hispanic Other, and educational attainment as college or more, some college, high school or equivalent, and less than high school. We theorize that, apart from their associations with social background characteristics of workers, occupational characteristics may influence biological age through their associations with financial resources and access to health care as well as through health behaviors outside of work. As our measure of financial resources, we include the quartile of household financial wealth in 2016, based on the RAND-constructed variable of total nonhousing assets. To proxy for access to health care, which may influence whether some of the components of biological age are adequately controlled, we include an indicator of health insurance in 2016. Last, we include smoking status (never smoker, former smoker, or current smoker), alcohol consumption (moderate drinker, nondrinker, or heavy drinker, defined as more than one drink per occasion for women and more than two drinks per occasion for men), and body mass index (BMI) category (normal weight, underweight, overweight, and obese) in 2016, as these are lifestyle-related risk factors for accelerated aging.

Analysis

The analysis consists of a series of nested linear regression models to assess the relationship between occupational characteristics and subsequent biological aging. In preliminary analyses, we did not find significant gender differences in the associations between occupational characteristics and biological age, so we do not stratify these analyses by gender: Model 1 includes the broad occupational categorizations to characterize occupational differences in expanded biological age and controls for chronological age and gender only. Model 2 adds race/ethnicity and Model 3 adds education. Model 4 includes self-reported working conditions (high psychosocial job demands, low job control, the interaction between high demands and low control, heavy lifting, and long working hours) to assess how specific self-reported working conditions are associated with accelerated aging and whether their inclusion changes the observed differences by occupational category. Model 5 adds health insurance and household wealth quartile in 2016, and Model 6 adds health behaviors and BMI category in 2016. All models incorporate the VBS sample weights provided by the HRS. Analyses were done in R version 4.0.2.

Multiple Imputation of Missing Data

Approximately 5% of the analytic sample had missing information on at least one of the occupational characteristics or covariates. To keep individuals with missing information in our sample, we used multiple imputation to create 10 complete analytic data sets, using the Multivariate Imputation by Chained Equations (mice) package in R. In addition to the variables included in the analyses, the imputation also included the category of the longest held occupation category reported in the HRS and the overall amount of physical effort and stress involved in one’s current job, both dichotomized in the same way as heavy lifting. The percentage with missing data for each variable is displayed in Table 1, and for our regression analyses, we present the pooled results across these 10 imputed data sets.

Table 1.

Summary Statistics of Analytic Sample: Percent or Mean (SD)

	Mean (SD) or %	Percent missing
Number of respondents	1,133
Occupational category in 2010/2012		<1%
Professional/managerial	44.9
Sales/clerical	23.1
Service	13.1
Manual	19.0
Self-reported occupational characteristics in 2010/2012
High psychosocial demands	46.1	2.0%
Low job control	58.2	1.7%
High strain (high demands and low control)	29.3
Heavy lifting	8.1	<1%
Long hours (55+ hr per week)	10.2	<1%
Covariates
Chronological age in 2016	61.1 (2.7)	0%
Gender
Men	47.9
Women	52.1
Race/ethnicity		<1%
Non-Hispanic White	85.6
Non-Hispanic Black	5.5
Hispanic	5.7
Non-Hispanic Other	3.1
Educational attainment		0%
Less than high school	4.6
High school or equivalent	47.1
Some college	10.6
College or more	37.8
Household wealth quartile in 2016		0%
Lowest quartile	26.7
Second quartile	26.2
Third quartile	22.8
Top quartile	24.3
Health insurance coverage in 2016		<1%
Insured	93.0
Not insured	7.0
Smoking status in 2016		<1%
Never smoker	47.0
Former smoker	39.4
Current smoker	13.6
Alcohol consumption in 2016		<1%
Moderate drinker	49.0
Heavy drinker	25.1
Nondrinker	25.8
BMI in 2016		<1%
Underweight	1.0
Healthy weight	25.8
Overweight	40.5
Obese	32.7

Notes: Data are from the Health and Retirement Study (HRS). The sample is restricted to participants who were aged 51–60, reported working for pay in nonmilitary occupations, and participated in the leave-behind questionnaire in either the 2010 or 2012 HRS wave and who subsequently participated in the 2016 Venous Blood Study (VBS). Household wealth is limited to nonhousing assets; it is constructed by RAND and already contain imputations for missing values. Summary statistics are weighted with VBS weights. Percentages may not add to 100% due to rounding. BMI = body mass index; SD = standard deviation.

Results

Summary statistics are displayed in Table 1. The most common occupational category held by respondents in 2010 or 2012 was professional/managerial occupations, which were held by 45% of the sample, followed by sales/clerical (23%), manual (19%), and service (13%) occupations. Approximately 46% of the sample is categorized as having high psychosocial job demands, 58% as having low job control, and 29% as having high strain. Heavy lifting and long working hours were reported by 8% and 10% of respondents, respectively.

Table 2 presents results from regressions predicting expanded biological age from characteristics of occupations. In the minimally adjusted Model 1, those employed in service occupations appear 3.02 years (95% confidence interval [CI] = 1.87–4.17) older biologically in 2016 than those employed in professional/managerial occupations, but there are no significant differences in biological age detected for those in sales/clerical or manual occupations. Controlling for race/ethnicity in Model 2 does not substantially change these estimates, and while adjustment for educational attainment in Model 3 slightly attenuates this association, service occupations remain significantly associated with a 2.49-year (95% CI = 1.24–3.74) older biological age. In Model 4, which includes self-reported working conditions, service work is associated with 2.20 years (95% CI = 0.95–3.44) older biological age, while low job control is associated with 1.65 years (95% CI = 0.66–2.65), heavy lifting with 2.04 years (95% CI = 0.67–3.40), and long working hours with 1.58 years (95% CI = 0.34–2.82) older biological age. Neither psychosocial job demands nor job strain (the interaction between high demands and low control) were significantly associated with biological age. The inclusion of household wealth and health insurance in Model 5 and the addition of health behaviors and BMI category in Model 6 both further reduce the coefficient on service work, but even in Model 6, service occupations remain significantly associated with a 1.65-year (95% CI = 0.41–2.88) increase in biological age. Low job control remains associated with 1.40 years (95% CI = 0.44–2.36), heavy lifting with 2.08 years (95% CI = 0.75–3.40), and long working hours with 1.87 years (95% CI = 0.68–3.05) older biological age, even after adjusting for financial resources and lifestyle-related risk factors for accelerated aging.

Table 2.

Associations Between Occupational Characteristics and Expanded Biological Age (Coefficients and 95% Confidence Intervals)

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Occupational category in 2010/2012 (ref = professional/managerial)
Sales/clerical	0.29 (−0.65, 1.23)	0.28 (−0.66, 1.23)	−0.23 (−1.24, 0.79)	−0.18 (−1.18, 0.82)	−0.37 (−1.37, 0.63)	−0.34 (−1.32, 0.64)
Service	3.02*** (1.87, 4.17)	3.11*** (1.94, 4.28)	2.49*** (1.24, 3.74)	2.20** (0.95, 3.44)	1.80** (0.53, 3.06)	1.65** (0.41, 2.88)
Manual	0.52 (−0.53, 1.57)	0.54 (−0.51, 1.60)	−0.30 (−1.51, 0.91)	−0.66 (−1.88, 0.55)	−1.00 (−2.22, 0.22)	−1.12 (−2.32, 0.09)
Self-reported working conditions in 2010/2012
High psychosocial job demands				0.00 (−1.16, 1.15)	−0.04 (−1.20, 1.11)	−0.35 (−1.48, 0.77)
Low job control				1.65** (0.66, 2.65)	1.72** (0.72, 2.71)	1.40** (0.44, 2.36)
High demands × low job control				0.81 (−0.68, 2.30)	0.61 (−0.88, 2.11)	0.96 (−0.49, 2.41)
Heavy lifting				2.04** (0.67, 3.40)	2.10** (0.74, 3.46)	2.08** (0.75, 3.40)
Long working hours				1.58* (0.34, 2.82)	1.70** (0.47, 2.93)	1.87** (0.68, 3.05)
Chronological age (centered on 56)	0.88*** (0.75, 1.02)	0.88*** (0.75, 1.02)	0.89*** (0.75, 1.02)	0.94*** (0.80, 1.07)	0.95*** (0.82, 1.08)	0.93*** (0.80, 1.06)
Woman	−0.05 (−0.84, 0.75)	−0.02 (−0.81, 0.77)	−0.09 (−0.88, 0.71)	−0.03 (−0.81, 0.76)	−0.06 (−0.84, 0.72)	−0.07 (−0.87, 0.72)
Race/ethnicity (ref = non-Hispanic White)
Non-Hispanic Black		−0.43 (−2.04, 1.18)	−0.67 (−2.30, 0.95)	−0.73 (−2.33, 0.87)	−1.34 (−2.95, 0.28)	−1.84* (−3.42, −0.27)
Hispanic		−0.16 (−1.76, 1.43)	−0.70 (−2.46, 1.07)	−0.74(−2.48, 0.99)	−1.26 (−3.04, 0.52)	−0.83 (−2.56, 0.91)
Non-Hispanic Other		−1.38 (−3.49, 0.73)	−1.48 (−3.59, 0.63)	−1.04 (−3.12, 1.03)	−1.33 (−3.40, 0.74)	−1.80 (−3.82, 0.22)
Education (ref = college or more)
Some college			0.59 (−0.74, 1.91)	0.19 (−1.11, 1.50)	−0.31 (−1.63, 1.01)	−0.41 (−1.70, 0.88)
High school			1.24* (0.30, 2.17)	1.14* (0.22, 2.06)	0.83 (−0.11, 1.77)	0.60 (−0.32, 1.52)
Less than high school			2.16 (−0.02, 4.34)	2.33* (0.19, 4.47)	1.99 (−0.16, 4.13)	1.35 (−0.74, 3.44)
Household wealth quartile (ref = top quartile) in 2016
Quartile 3					1.09* (0.04, 2.13)	0.80 (−0.23, 1.82)
Quartile 2					1.50** (0.45, 2.55)	0.80 (−0.24, 1.83)
Quartile 1					2.43*** (1.31, 3.55)	1.25* (0.13, 2.36)
Uninsured in 2016					−0.56 (−2.07, 0.94)	−0.19 (−1.66, 1.27)
Smoking status in 2016 (ref = never smoker)
Former smoker						0.08 (−0.68, 0.84)
Current smoker						1.86** (0.71, 3.01)
Alcohol consumption in 2016 (ref = moderate drinker)
Heavy drinker						0.45 (−0.42, 1.32)
Nondrinker						2.25*** (1.35, 3.14)
BMI category in 2016 (ref = healthy weight)
Underweight						4.89** (1.23, 8.54)
Overweight						1.31** (0.40, 2.21)
Obese						2.81*** (1.87, 3.75)
Intercept	54.73*** (53.75, 55.70)	54.77*** (53.79, 55.76)	54.45*** (53.43, 55.48)	52.80*** (51.57, 54.02)	51.98*** (50.65, 53.31)	50.55*** (49.06, 52.03)
Number of observations	1,133	1,133	1,133	1,133	1,133	1,133

Notes: Results are pooled across 10 imputations. BMI = body mass index.

*p < .05. **p < .01. ***p < .001.

Supplementary Analyses

Because individuals may switch into less demanding jobs at midlife than they held at earlier points in the life course, we consider the category of the longest held occupation of the occupations reported in the HRS instead of the current occupation in 2010 or 2012. This is a variable constructed by RAND that is the job with the longest tenure out of the respondent’s current job, last job, or the most recent job before the last job that was held for five or more years, so it is not necessarily the longest held job. These results, displayed in Supplementary Table A1, are consistent with those of the main analysis in that service work is associated with older biological age after controlling for race/ethnicity and education, while sales/clerical and manual work are not. In another set of supplementary analyses (Supplementary Table A2), in lieu of using psychosocial job demands, job control, and job strain as measures of psychosocial stressors and heavy lifting as a marker of physical demands, we used the more general self-reports of whether one’s current job involves a high degree of stress and whether one’s current job requires high physical effort. High stress is associated with significantly higher biological age before including lifestyle-related risk factors, but is no longer significant once these are included. High physical effort is not significantly associated with biological age in any of these models.

Discussion

Extensive research has examined the effects of adverse working conditions on health at older ages, finding that individuals with difficult working conditions have worse health on numerous dimensions, ranging from worse physical functioning to higher rates of cardiovascular disease and mortality (Andrasfay et al., 2021; Brussig & Drescher, 2022; Landsbergis, 2010; Townsend & Mehta, 2021). Studies have also found associations between occupational characteristics and biomarkers indicating early signs of dysregulation in the cardiovascular, metabolic, or inflammatory systems (Holtermann et al., 2020; Landsbergis et al., 2013; Virtanen et al., 2012). This study extends this prior work and is the first to examine occupational characteristics as risk factors for expanded biological age, a novel measure that summarizes multisystem dysregulation and can detect differences in underlying health status between individuals that may otherwise appear healthy (Crimmins et al., 2021). Using a nationally representative sample that includes individuals employed in a diverse set of occupations at midlife, we show that service work and several self-reported working conditions are longitudinally associated with one’s biological age.

In our analyses, we demonstrate that individuals employed in service work exhibit accelerated biological aging compared to those employed in professional/managerial occupations. This association between service work and older biological age remains even after adjustment for household wealth, health insurance, health behaviors, and BMI. We also find that low job control, heavy lifting, and long working hours are associated with accelerated biological aging independent of their associations with background characteristics, occupation category, financial resources, and lifestyle-related risk factors for accelerated aging. Our findings are consistent with prior research that has found low job control, physically demanding work, and long hours at work to be risk factors for biomarkers related to the individual components of expanded biological age (Clays et al., 2005; Davila et al., 2011; Holtermann et al., 2020; Magnusson Hanson et al., 2019; Song et al., 2021; Steptoe & Willemsen, 2004; Virtanen et al., 2012, 2019). Although prior research has found psychosocial job demands and job strain to be associated with individual biomarkers, we do not find these psychosocial job stressors to be significantly associated with expanded biological age in this study (Landsbergis et al., 2013; Magnusson Hanson et al., 2019; Netterstrøm & Sjøl, 1991). While ours is the first study to examine expanded biological age, prior research has found associations between psychosocial stressors at work and increased allostatic load—a related summary measure of the cumulative biological burden of stress that typically includes other biomarkers directly measuring activation of the stress response—and between general stress, whether at work or in other domains, and a summary measure of physiologic dysregulation (Cuitún Coronado et al., 2018; Goldman et al., 2005; Guidi et al., 2021). Taken together, our findings and those of previous work highlight the potential role of psychosocial stress at work, physically demanding work, and long working hours in experiencing more rapid age-related dysregulation in multiple body systems.

The persistent association between service work and accelerated biological aging, even after controlling for working conditions, is consistent with the findings of Schmitz (2016), who found that service workers had significantly worse self-rated health than white-collar workers, even after adjustment for SES and working conditions (Schmitz, 2016). Although the outcome of that study was a self-reported global measure of health, it suggests that there is something about service work that is detrimental to health that is not captured in commonly collected occupational measures. The particularly strong and persistent association between service work and accelerated aging is consistent with a growing body of work emphasizing the service sector’s unique confluence of adverse working conditions, including schedule unpredictability, job precarity, lack of paid sick leave, and emotional demands (Abrams et al., 2021; Dollard et al., 2003; Harknett & Schneider, 2020; Schneider & Harknett, 2020). While we consider psychosocial job demands, low job control, and long working hours, these measures do not necessarily capture the emotional stress related to working with customers or patients or the specific shifts individuals work, which likely contribute to the accelerated aging observed among service workers. Several of the fastest growing occupations in the next decade are projected to be in the service sector, underscoring the need for more research to understand the aspects of service jobs that are particularly detrimental to these physiologic systems and that may be amenable to interventions (Bureau of Labor Statistics, 2022; Ghilarducci & Schuster, 2021).

Limitations

This study has several limitations. Although we chose to limit the analysis to individuals working at ages 51–60 to capture working conditions before widespread retirement, there are still many individuals who have exited the labor force before these ages. Because early exit from the labor force is a common response to poor health, the range of biological ages we observe in this sample may be limited by the healthy worker effect and our estimates of the associations between working conditions and biological age may be biased downward. A related concern is health-related selection out of manual occupations by midlife, which may explain why we do not find a significant association between manual work and biological age, in contrast to prior work finding associations between manual work and worse levels of individual biomarkers of aging (Engström et al., 2006; Fraga et al., 2015). Future research should examine these associations in a younger cohort to address concerns about selective attrition from the labor force or certain occupations at older ages. The analysis is also limited by the available measures of occupational characteristics. The HRS asks respondents to report on a limited set of working conditions that do not capture other important occupational risk factors for biological aging, particularly those related to environmental exposures.

Conclusion

In response to population aging, countries around the world, including the United States, have implemented increases in retirement eligibility ages to encourage working longer and there is ongoing discussion about further increasing these eligibility ages. At the same time, many of the jobs with the highest projected growth in the next decade are low-wage jobs in the service sector (Bureau of Labor Statistics, 2022; Ghilarducci & Schuster, 2021). Our findings challenge the feasibility of increases in retirement age for all workers as they suggest that workers with certain occupational characteristics at midlife, particularly service work, low job control, heavy lifting, and long work hours, are systematically approaching retirement ages with accelerated biological aging. These workers are likely to face a greater burden of disability and chronic diseases during the ages at which they are expected to work longer, and if they manage to continue working in these jobs it may negatively affect their health. Although more research is needed to understand the mechanisms through which these characteristics affect the aging process, improving the work environment may be a way to help older adults who wish to continue working while preventing accelerated aging and reducing the risk of several age-related diseases.

Supplementary Material

Supplementary data are available at The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences online.

Author Notes

1. The majority of these biomarkers were assayed from a venous blood sample, which was collected in respondents’ homes by a phlebotomist and shipped to the Advanced Research and Diagnostic Laboratory at the University of Minnesota. Systolic blood pressure and peak flow were measured by HRS interviewers during enhanced face-to-face interviews that took place in 2014 for half the sample and 2016 for the other half. HbA1c was assayed from dried blood spots provided by respondents in 2014 for half the sample and 2016 for the other half.

2. Professional/managerial comprises management; business and financial operations; computer and mathematical; architecture and engineering; life, physical, and social science; community and social service; education, training, and library; arts, design, entertainment, sports, and media; health care practitioners and technical occupations. Sales/clerical comprises sales and related occupations and office and administrative support occupations. Service comprises health care support; protective service; food preparation and serving; building and grounds cleaning and maintenance; personal care and service occupations. Manual comprises farming, fishing, and forestry; construction and extraction; installation, maintenance, and repair; production; and transportation and material moving. We exclude individuals in military occupations because military occupations do not neatly fall into these categories and there were too few individuals to constitute a separate category.

Funding

This study was supported by the National Institute on Aging (grants T32-AG000037 and P30-AG017265). The Health and Retirement Study is sponsored by the National Institute on Aging (grant number NIA U01-AG009740) and the Social Security Administration and is conducted by the University of Michigan.

Conflict of Interest

None declared.