Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Mar 12.
Published in final edited form as: J Aging Health. 2020 Dec 21;33(3-4):273–284. doi: 10.1177/0898264320977329

Job Strain and Late-life Cognition: Findings from the Puerto Rican Elderly Health Conditions (PREHCO) Study

Taylor F D Vigoureux 1,*, Monica E Nelson 1, Ross Andel 1,2,3, Brent J Small 1, Ana Luisa Davila-Roman 4, Michael Crowe 5
PMCID: PMC7954124  NIHMSID: NIHMS1676251  PMID: 33349101

Abstract

Objectives:

We examined associations between job strain and cognitive aging in a sample of older Puerto Ricans.

Methods:

Members of the Puerto Rican Elderly Health Conditions (PREHCO) study, aged 60-100 years at baseline, participated. Job strain indicators were quantified from O*Net (n=1632) and a matrix of Job Content Questionnaire scores (JCQ; n=1467). Global cognition was assessed twice across 4 years.

Results:

Controlling for age, sex, depressive symptoms, financial problems, hypertension, diabetes, and childhood economic hardship, low job control and high job strain were consistently associated with greater cognitive decline. Adding education attenuated these associations. High education strengthened the job control (JCQ)-cognitive change link.

Discussion:

Low job control and high job strain may accelerate cognitive aging in this population. However, it may be more difficult to disentangle the intersecting roles of education and job strain in cognitive aging among older Puerto Ricans relative to older adults from contiguous US or Europe.

Keywords: job strain, older adults, cognition, Puerto Rico

Puerto Rico is an important part of the U.S. that is often omitted from U.S. population-based studies of health and aging. Perhaps contributing to this lack of inclusion, despite more than 100 years as a U.S. territory, nearly half of Americans still don’t know that people born in Puerto Rico are U.S. citizens (Morning Consult, 2017). In the past decade, Puerto Rico has experienced rapid population aging, financial collapse, mass outward migration of younger people, Hurricane Maria, and then continuing earthquakes. The percentage of total population in Puerto Rico who are aged 65+ years has increased dramatically from 11.2% in 2000 to higher than any U.S. state at 21.3% in 2019 (U.S. Census Bureau, 2019). Since older age is the strongest demographic predictor of cognitive decline and dementia, understanding life course determinants of cognitive aging on the island is important for public health planning and policy.

Cognitive impairment has profound influence on quality of life in older adulthood and cognitive decline as Alzheimer’s disease is one of the most feared conditions of older adulthood (Alzheimer's Association, 2014; Awang et al., 2018), with almost half of surveyed older adults believing they would develop dementia (Maust et al., 2019). Even relatively minor changes in cognition can greatly affect well-being (Plassman et al., 2010), which makes understanding factors that may postpone cognitive decline and impairment that much more important. A large array of risk and protective factors determines the measurable likelihood of substantial cognitive decline (Plassman et al., 2010). Among these factors, chronic stress appears to play an important role (Lupien et al., 2009). Even before the current COVID-19 pandemic that has substantially changed the way we work and live, Americans cited money and work as the top two out of ten common sources of stress (American Psychological Association, 2011).

The fact that employed adults spend a substantial portion of their time at work (Saad, 2014) underscores the likely role of work environment in health and aging outcomes. Before Karasek’s seminal (1979) article, work-related stress was mainly operationalized by examining level of job demands. However, Karasek noted a paradox in previous work: although executives and assembly-line workers had similarly demanding jobs, their job satisfaction was very different. This paradox led Karasek to examine an important dimension previously unconsidered: job control. He theorized that job demands (i.e., expectations placed upon an employee) differentially resulted in stress as a function of job control (i.e., freedom of an employee to make decisions to meet expectations). This was insightful given Folkman’s (1984) definition of stress: “a relationship between the person and the environment that is appraised by the person as taxing or exceeding his or her resources and endangering his or her well-being” (p. 2). This definition is in line with Karasek’s job demand-control model because it implies that job demands only become problematic when a person perceives them as beyond their capabilities (which may be limited by low job control). Given the importance of work in the lives of adults, the amount of stress induced by such work characteristics can be presumed to be substantial, particularly for lower-level jobs with high workload demands and little control over work activities.

There is evidence of a relationship between job strain and an increased risk of health problems. For example, there has been consistent evidence of job strain’s association with cardiovascular disease (Nyberg et al., 2013; Schnall et al., 1994), obesity (Brunner et al., 2007), diabetes (Nyberg et al., 2013), and depression (Stansfeld et al., 2012), which are all known risk factors for cognitive decline and dementia (Anstey et al., 2019; Cooper et al., 2015). Further, studies that have directly examined job strain’s association with late-life cognition have found evidence to support an inverse association (Agbenyikey et al., 2015; Andel et al., 2011). However, previous research on job strain and cognition has been conducted primarily in countries where career opportunities and socioeconomic conditions are relatively favorable, such as the mainland U.S. (Agbenyikey et al., 2015; Andel et al., 2015), Sweden (Andel et al., 2011), Germany (Then et al., 2014), and France (Sabbath et al., 2016). Consequently, there is little research on job strain and cognition in places with relatively less favorable socioeconomic conditions, such as Puerto Rico, leaving the generalizability of job strain findings in relation to cognition largely unknown.

Also important to consider in research on cognitive aging is the role of cognitive reserve (Stern, 2002, 2009). There is evidence that indicators of higher cognitive reserve such as education are related to a lower risk of cognitive decline and dementia (Pettigrew & Soldan, 2019; Stern, 2009; Wang et al., 2019). Therefore, cognitive reserve reflected in higher education may interact with work-related stress to potentially reduce its presumably adverse effect on late-life cognition.

Job opportunities in Puerto Rico have been highly influenced by U.S. policies toward the island that residents have historically had little control over given the lack of voting representation in Congress and lack of any electoral votes for president. By 1918, all states had compulsory schooling laws for children (Katz, 1976), but there was no mandatory school attendance policy in Puerto Rico until more than three decades later in the mid-1950s. Subsequently, between 1960 and 2000, the average education of workers in Puerto Rico doubled from 6.2 to 12.2 years (Ladd & Riviera-Batiz, 2006). Puerto Rico also experienced rapid industrialization transitioning during this time period, from primarily an agricultural economy to a manufacturing/service economy, spurred by U.S. government-led initiatives such as corporate tax exemptions (i.e., Operation Bootstrap; González-Mejía & Ma, 2017; Toro, 2014). While the unemployment rate in Puerto Rico declined from 1950 to 1970, it subsequently rose sharply (Riviera-Batiz & Santiago, 1996) and remained well above the U.S. average for the next three decades. Historically, Puerto Rican women and people with lower levels of education have been much more likely to face unemployment on the island (Riviera-Batiz & Santiago, 1996).

In more recent history, the longstanding tax breaks to U.S. companies operating in Puerto Rico completed a phasing out in 2006. As a result, manufacturers, particularly those in the pharmaceutical industry, began closing plants, eliminating many jobs in Puerto Rico and exacerbating the problem of workforce migration to the mainland. Of particular concern for aging in Puerto Rico is the exodus of younger health care professionals who are needed to provide adequate care for an aging population (Perreira et al., 2017). Lower reimbursement rates for physicians in Puerto Rico compared to mainland U.S. are thought to be a major contributor to this problem (Roman, 2015). Currently, the main industries in Puerto Rico are pharmaceuticals, electronics, apparel, food products, and tourism (CIA, 2020).

There is evidence of differences in health outcomes in older adults from Puerto Rico compared to the mainland U.S. (Perez & Ailshire, 2017), such as higher rates of hypertension and diabetes, which have been shown to be negatively related to cognitive health (Anstey et al., 2019). Older adults in Puerto Rico also experience socioeconomic disparities that can negatively contribute to poor cognitive health outcomes. For example, even before Hurricane Maria, over 40% of older adults in Puerto Rico were living below the poverty line, compared to less than 13% for all individual states (Administration on Aging, 2016). The cumulative effect of health and socioeconomic conditions on psychological well-being and, in turn, health outcomes including cognitive health, paired with the high proportion of older adults in Puerto Rico necessitates further examination of factors associated with cognitive health in this population. Moreover, job strain may be an especially important factor to consider given its potential to magnify the adverse effect of socioeconomic conditions in Puerto Rico on psychological well-being and health outcomes.

The Current Study

In order to expand knowledge on the relationship between job strain and cognition, we examined how job strain, measured with two different methods for stronger evidence, relates to change in late-life cognition in a less commonly studied, sometimes marginalized, and potentially vulnerable population: older adults from Puerto Rico who were members of the Puerto Rican Elderly Health Conditions (PREHCO) Study. To build on previous research, we raised the following research questions: (1) Does job strain (experienced in main lifetime occupation) relate to late-life cognitive change in Puerto Rican older adults? (2) Is the relationship between job strain and late-life cognitive change moderated by level of education? Based on the job demand-control model and previous findings that low job control is associated with poorer cognitive functioning (Andel et al., 2011, 2015), we expected that low job control and high job strain (i.e., high ratio of job demands to job control) would relate to greater late-life cognitive decline in Puerto Rican older adults. In line with previous work examining job strain and education as predictors of long-term health outcomes (Nilsen et al., 2014), as well as our study on work complexity and cognitive impairment in this sample (Andel et al., 2019), we also expected that education would have a major role in the association between job strain indicators and cognitive change.

Method

Participants

The PREHCO Study (McEniry & Palloni, 2010) is a longitudinal, population-based study of adults over the age of 60 residing in mainland Puerto Rico. A multistage, stratified sample design was implemented, with oversampling in regions that have a larger proportion of people of African descent and of individuals aged over 80. Data collection was approved by the institutional review boards at the University of Wisconsin and the University of Puerto Rico. Wave one of this study was conducted between 2002 and 2003, and wave two was conducted between 2006 and 2007. A third wave of data collection is currently in progress. Overall, the baseline response rate was 93.9% with no significant differences between those who did and did not respond (McEniry & Palloni, 2010). The current study was a secondary analysis of the data originally collected by the aforementioned institutions and marked as exempt from IRB approval from the University of Alabama at Birmingham (IRB Protocol #N140723002).

The study flow chart (see Figure 1) presents information about missing data for the 4,291 individuals who were initially contacted, which resulted in 2,211 participants with valid baseline cognitive data and information about main lifetime occupation with which they worked for at least 10 years, and a follow-up sample of 1,632 participants with data for all analytic variables. Included participants were younger (M=70.2, SD=7.3 vs. M=72.8, SD=9.09, t=9.64, p<.001), more educated (M=9.2, SD=4.4 vs. M=6.9, SD=4.5, t=−15.77, p<.001), and more likely to be male (49.6% vs. 33.1%, χ2(1) = 108.35, p<.001) than excluded eligible participants with baseline demographic information (n = 3,713).

Figure 1.

Figure 1.

Open in a new tab

Sample for Occupation and Cognition in PREHCO

Of the 2,190 in the baseline sample (who were not missing data for covariates), 558 were missing follow-up data. Of these, 258 dropped out and 300 had died. Those who dropped out were more likely to be older (M=75.25, SD=8.9 vs. M=70.22, SD=7.31, t=−12.02, p<.001), have fewer years of education (M=8.15, SD=4.61 vs. M=9.21, SD=4.43, t=−4.83, p<.001), be male (64.52% vs. 49.63%, χ2(1, N=2190)=37.02, p<.001), and have slightly lower baseline cognitive performance than participants (M=16.32, SD=2.49 vs. M=16.97, SD=2.30, t=5.44, p<.001).

Measures

Cognition.

Cognition was assessed using the Mini-Mental Cabán (MMC; Sanchez-Ayendez et al., 2003). The MMC is a screening measure for dementia including items similar to those of the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA). Items reflect time orientation (i.e., day and date), verbal memory (i.e., immediate and delayed word recall), visual memory (i.e., complex figure drawing), visuospatial/executive function (i.e., clock drawing, copy pentagons), abstraction (i.e., phrase meaning), and comprehension (i.e., follow a three-step command). This measure yields scores ranging from 0-20 as a measure of global cognitive function. Sanchez-Ayendez and colleagues (2003) specifically designed the MMC to be more suitable for lower education and Spanish-speaking populations than the MMSE. The MMC had better sensitivity and specificity for dementia than the Spanish translated MMSE in Puerto Rico. Despite inclusion of several different cognitive domains that are assessed briefly, the MMC has sufficient reliability, α = .68 (Sanchez-Ayendez et al., 2003).

Occupation.

Self-reports regarding longest held gainful occupation or last gainful occupation (when also identified as main occupation) were collected for 2,776, whereas the remaining 937 of the 3,713 eligible participants reported no gainful employment. Of the 937 without gainful employment, 541 had never held a job with pay for reasons including health problems, no economic need, taking care of their family, being a housewife, no job opportunities, their husband would not let them, or other reasons whereas the 394 individuals who said they held a job for pay did not report the occupation or reported occupation that could not be properly coded with categories from the 2000 US Census occupational coding scheme, which was used by PREHCO Study.

Occupation narratives were matched with categories in the 2000 Standard Occupational Classification (SOC). This coding scheme is also used in the Dictionary of Occupational Titles and easily aligns with other commonly used coding schemes. To assure sufficient exposure to work environment, only those who were employed for at least 10 years were used, as done previously (Andel et al., 2015, 2019).

Of the 3,713 eligible participants, 2,776 reported working in a gainful occupation (see Figure 1). The 937 without gainful occupation were about a year younger (M=70.6, SD=8.2 years vs. 71.6, SD=8.2 years, p=.002), less educated (M=6.8, SD=4.5 years vs. M=8.5, SD=4.5 years, p<.001), and predominantly women (85% women vs. 52% women; p<.001). Of the 2,776 participants with complete occupation, 565 worked less than 10 years in a single occupation. Compared to the 2,211 who spent at least 10 years in their main occupation, the 565 were not different in terms of age (M=71.8, SD=8.8 years vs. M=71.5, SD=8.1 years, p>.05) but were less educated (M=6.8, SD=4.3 years vs. M=8.9, SD=4.5 years, p<.001), and were mainly women (74% women vs. 46% women; p<.001).

The most common occupational categories in the PREHCO sample were production occupations (most common was sewing machine operators at 9.1%), housekeeping and janitorial workers (building cleaners 3.7%), service occupations (cooks 3.4%), administrative workers (secretaries and administrative assistants 3.4%), managers (general/operations managers 2.6%), agricultural workers (miscellaneous agricultural workers 2.3%), and sales persons (2.2%). Most (87%) participants were fully retired at baseline, 10% worked at various part-time capacities, and 3% were still in full-time employment.

Job strain.

Job demands and job control (decision latitude) were used to measure job strain (Karasek, 1979; Karasek & Theorell, 1990). Job demands was designed to measure psychological stress (e.g., intense, hectic work schedule, extreme workload). Job control is a measure of the extent to which one can use personal judgment or decision authority to assert control in the workplace.

We used two different occupation-based measures of job strain to assign job demands and job control to participants. The first utilized data from the Occupational Information Network (https://www.onetcenter.org) Version 3.1 (subsequently referred to as O*Net). The O*Net database, an industry-based coding scheme developed by the Department of Labor to replace the Dictionary of Occupational Titles, is the main source of information about job characteristics in the United States (Schwartz, 2000). Job ratings were based on national surveys of job incumbents and the work of trained job analysts who rate jobs using a priori defined criteria. Census-based codes used by PREHCO were linked to occupation codes from O*Net to determine job demands and job control. Job demands was measured with the following eight O*Net job ability items: Required level and importance of (a) concentrating on a single task without interruption, (b) making decisions quickly, (c) shifting back and forth between two or more tasks, and (d) focusing on a single sound in the presence of multiple sounds; Cronbach’s α was .92. Job control was measured with the following four work value items: (a) freedom to try out new ideas, (b) ability to make own decisions, (c) ability to work without supervision, and (d) freedom to use own abilities; Cronbach’s α was .96. The variables were transformed into z-scores, then the relevant items were averaged to create the job demands and job control variables.

The second occupation-based measure of job strain utilized Karasek and colleagues’ (1998) job strain standardized demands and control scores for reported occupations (subsequently referred to as JCQ). These were developed through compiling data from 10,288 men and 6,313 women from six studies conducted in the U.S., Sweden, France, and Japan. All participants responded to Karasek’s Job Content Questionnaire (1979) and reported their occupations. For each occupation, Karasek and colleagues averaged reported demands and control for men and women, subsequently creating standardized scores and matched them to the 1970 US Census. We requested these data from Karasek and colleagues as a basis for coding demands and control for our participants and matched them to 2000 Census codes.

There were 166 job titles from 20 different job categories that were not sufficiently represented in the six samples that Karasek used to create standardized job strain scores. The job titles are variable and do not show a consistent pattern. The five categories with the highest frequency (n > 50) were Management Occupations, Food Preparation and Serving-Related, Office and Administrative Support Occupations, Production Occupations, and Transportation and Material Moving. This resulted in missing job strain scores for 695 (42.5%) of the 1,632 participants. We matched the categories from the 1970 US Census that contained job strain scores to the 2000 US Census categories represented by the 695 participants with missing job strain scores. The first three authors separately matched the pertinent categories across the two versions of the US Census, then they met to come up with consensus for any discrepancies in codes. There was 80% initial agreement; discrepancies mostly concerned overlapping categories. The procedure resulted in job strain scores for additional 530 participants, resulting in a sample of 1,467 PREHCO participants with JCQ-based data. Raw scores for job demands and job control range from 0-48 and 0-96, respectively, with higher scores indicating higher demands or control.

For both job strain measurement approaches, we used job demands and control as individual independent variables. Additionally, as suggested by Karasek and Theorell (1990), we used a quotient of job demands/job control to measure job strain. Given the difficulty of using z-scores in quotient calculations, we used standard scores with a mean of 100 and standard deviation of 15 for job demands and control to generate a quotient. Additionally, to compare results from our sample to that of the Health and Retirement Study (HRS; mainland U.S.), we created a score with a range of 0 to 100 as done in previous studies (Andel et al., 2015; Cifuentes et al., 2007). For the O*Net measure, the mean (SD) was 39.5 (15.4) for job demands and 42.3 (21.3) for job control. In comparison, the same variables yielded means (SD) of 54.6 (4.9) for job demands and 77.8 (7.7) for job control in the HRS data (Andel et al., 2015).

Covariates.

At baseline, participants responded to questionnaires assessing mental health and demographic information. Given the relationship between depression and cognition, participants responded to a Spanish translation of the Geriatric Depression Scale (GDS; Yesavage et al., 1983). The GDS consists of 15 yes or no items asking whether participants experienced depressive symptoms. Of the 15 items, five were positively worded (e.g., “During the last two weeks… you felt basically satisfied with your life.”) and 10 were negatively worded (e.g., “During the last two weeks… did you feel hopeless in your current situation?”). For positively worded items, an answer of no was scored as ‘1’; for negatively worded items, an answer of yes was scored as ‘1’. The sum score ranged from 0 to 15 depressive symptoms. Both the English and Spanish version of the GDS have acceptable reliability (Cronbach’s α = .77; Friedman et al., 2005) and convergent validity with the Cornell Scale of Depression in Dementia and the Hamilton Self-Rating Depression Scale (both rs = .77; Kørner et al., 2006) The GDS also has sufficient sensitivity and specificity for detection of depression (Aguilar-Navarro et al., 2007; Pocklington et al., 2016; Yesavage et al., 1983).

For self-rated health, participants indicated whether they would describe their current health status as bad (coded as ‘0’), average (coded as ‘1’), good (coded as ‘2’), very good (coded as ‘3’), or excellent (coded as ‘4’). Two items relating to difficulty paying for goods or health care services were used to assess financial problems at baseline. After summing the responses of these two items, the resulting financial problems variable ranged from 0 to 3 with ‘0’ representing no financial problems, ‘1’ representing low financial problems, ‘2’ representing moderate financial problems, and ‘3’ representing high financial problems. Childhood economic hardship was measured using two items. Participants responded to the following question, with higher scores reflecting greater hardship: “In general, would you say that the economic conditions in the household in which you grew up were good (coded as ‘0’), average (coded as ‘1’), or bad (coded as ‘2’)?”. Participants also responded yes (coded as ‘1’) or no (coded as ‘0’) to “Did you suffer economic hardship that prevented you from eating regularly, adequately clothing yourself or receiving the necessary medical attention?”. A sum score from these two questions was created, with a range of 0 to 3 with ‘0’ representing no childhood economic hardship, ‘1’ representing low hardship, ‘2’ representing moderate hardship, and ‘3’ representing high hardship. Participants self-reported whether they had ever been diagnosed with hypertension or diabetes. Participants also indicated how many years of education they obtained.

Statistical Analyses

We tested differences between participants using t-tests and chi-squares as applicable. Next, we tested our hypotheses using residualized change linear regressions (Dalecki & Willits, 1991) to examine cognitive change between wave 1 and wave 2. Each research question was tested once with job demands and job control as independent variables and once with the quotient of job demands to job control (i.e., job strain) as the independent variable. All analyses first controlled for age, sex, depressive symptoms, baseline financial problems, childhood economic hardship, hypertension, diabetes, and baseline cognition in model 1, and then additionally controlled for education in model 2 as done in previous work in this area (Nilsen et al., 2014). To assess our second research question, an interaction term between job strain indicators was included in a fully adjusted model. Due to known gender differences in occupation in Puerto Rico, we also assessed whether gender moderated the association between job strain indicators and cognitive decline as supplemental analyses. Significant interactions were probed and represented with Johnson-Neyman plots. All analyses were conducted in SAS Version 9.4 of the SAS System for Windows. Significant moderations were probed and plotted using output from the PROCESS Macro for SAS Version 3.4.1 (Hayes, 2017).

Results

Sample Characteristics at Baseline

Table 1 contains descriptive statistics for all study variables based on the overall sample after exclusion criteria (n=1,632). Table 1 also includes descriptive statistics for this sample stratified by education (i.e., no high school, some high school, more than high school). Participants were 70 years old on average (SD=7.31), with an approximately equal ratio of men to women, and an average of 9.21 years of education (SD=4.43). Only approximately one fifth of the sample did not report any kind of childhood economic hardship. On average, participants spent 25 years in their main occupation (SD=10.34). For the O*Net sample, job strain was negatively correlated with education (r=−.08, p=.001), but not cognition. Job control was positively correlated with education (r=.39, p<.001) and follow-up cognition (r=.14, p<.001). Job demands were positively correlated with education (r=.33, p<.001) and follow-up cognition (r=.11, p<.001). For the JCQ sample, job strain was negatively correlated with education (r=−.17, p<.001) and follow-up cognition (r=−.13, p<.001). Job control was positively correlated with education (r=.27, p<.001) and follow-up cognition (r=.15, p<.001). Job demands were not correlated with education or cognition for this sub-sample.

Table 1.

Descriptive statistics for study variables for full sample and stratified by education.

Variable Name Full Sample
(n = 1632)		 No HS
(n = 799)		 Some HS
(n = 406)		 Beyond HS
(n = 427)
M/N SD/% M/N SD/% M/N SD/% M/N SD/%
Age 70.24 7.31 71.50 7.73 68.41 6.54 69.56 6.75
Female 822 50.37 375 46.58 200 49.26 247 57.71
Education 9.21 4.43 5.30 2.72 11.61 0.69 14.26 1.06
Baseline Depression 3.03 3.29 3.31 3.38 2.88 3.24 2.67 3.13
Baseline Financial Problems 1.01 1.12 1.14 1.13 1.04 1.12 0.73 1.04
Childhood Economic Hardship 1.54 1.11 1.73 1.09 1.58 1.11 1.15 1.05
Hypertension 931 57.05 466 58.18 221 54.43 246 57.61
Diabetes 420 25.74 209 26.03 100 24.63 113 26.4
Baseline Cognition 16.97 2.30 16.37 2.41 17.28 2.18 17.79 1.83
Follow-Up Cognition 16.16 2.79 15.29 2.96 16.58 2.35 17.39 2.23
O*Net Job Demands 100.03 12.12 96.29 11.22 101.69 12.24 105.44 11.21
O*Net Job Control 100.05 14.24 95.24 12.15 99.07 12.90 109.96 14.11
O*Net Job Strain 1.01 0.15 1.02 0.13 1.04 0.15 0.98 0.17
Full Sample
(n = 1467)		 No HS
(n = 697)		 Some HS
(n = 372)		 Beyond HS
(n = 398)
JCQ Job Demands 30.49 5.62 30.53 5.89 29.95 5.78 30.94 4.92
JCQ Job Control 70.05 14.05 66.66 13.97 69.96 13.48 76.09 12.64
JCQ Job Strain 1.03 0.26 1.07 0.27 1.01 0.26 0.97 0.20
Open in a new tab

Note. JCQ job control is on a scale of 0-96 (original scale of variables from JCQ crosswalk), and JCQ job demands are on a scale of 0-48. Higher scores indicate higher job control and job demands, respectively. Job strain is a quotient of job demands to job control (each on a standard scale of M of 100 and SD of 15), with higher scores indicating higher strain.

Main Results

Job demands and job control.

Table 2 contains results from regression analyses using job demands and control as independent variables. In line with our hypothesis, after controlling for all covariates except education, the positive regression coefficient b indicates that higher job control was associated with less cognitive decline (increasingly better cognitive scores compared to those with lower job control) whether measured with the O*Net measure (b=0.31, SE=0.07, p<.001) or the JCQ measure (b=0.28, SE=0.07, p=.015. When education was also controlled, the job control—cognition relationship was no longer significant for the O*Net measure (p>.05), but it was retained for the JCQ measure (b=0.16, SE=.07, p=.015), despite the magnitude of the effect being reduced by over 40% (from b=0.28 to b=0.16).

Table 2.

Results from regression analyses for the association between job control or demands and cognition.

Model 1: Before Controlling for Education Model 2: Controlling for Education
Variable Name O*Net (n = 1632a) JCQ (n = 1467b) O*Net (n = 1632a) JCQ (n = 1467b)
B SE p B SE p B SE p B SE p
Intercept 10.72 0.51 10.78 0.55 11.45 0.51 11.43 0.55
Baseline Cognition 0.38 0.03 <.001 0.38 0.03 <.001 0.32 0.03 <.001 0.33 0.03 <.001
Age −0.11 0.01 <.001 −0.11 0.01 <.001 −0.10 0.01 <.001 −0.09 0.01 <.001
Sex (Female) 0.31 0.12 .013 0.30 0.13 .024 0.20 0.12 .111 0.25 0.13 .051
Baseline Depression −0.03 0.02 .177 −0.04 0.02 .054 −0.03 0.02 .238 −0.03 0.02 .085
Baseline Financial Problems −0.02 0.06 .765 −0.01 0.06 .811 0.04 0.06 .526 0.04 0.06 .530
Childhood Economic Hardship 0.03 0.06 .631 −0.003 0.06 .957 0.08 0.06 .153 0.07 0.06 .242
Hypertension −0.01 0.13 .943 −0.03 0.13 .841 −0.03 0.12 .804 −0.05 0.13 .729
Diabetes −0.21 0.14 .137 −0.27 0.15 .076 −0.18 0.14 .197 −0.24 0.15 .097
Job Demands 0.11 0.08 .186 −0.10 0.06 .137 −0.02 0.08 .839 −0.10 0.06 .122
Job Control 0.31 0.07 <.001 0.28 0.07 <.001 0.13 0.07 .064 0.16 0.07 .122
Education --- --- --- --- --- --- 0.13 0.02 <.001 0.13 0.12 <.001
Open in a new tab

Note. Age is centered at the minimum (60), depression and education are mean-centered; job demands and job control are in z-scores. B = unstandardized regression coefficient; SE = standard error.

a

There were missing values for money problems (n=5) or information about high blood pressure (n=5) or diabetes (n=2)

b

There were missing values for money problems (n=1) or information about high blood pressure (n=5) or diabetes (n=2)

Next, we examined the education-by-control moderation and found that the job control was more strongly related to change in cognition with more years of education, b=0.03, SE=0.01, p=.026. Figure 2 shows the Johnson-Neyman plot which indicates that low job control is only significantly associated with greater cognitive decline for individuals with 8+ years of education, whereas the association does not reach statistical significance when education is lower. Job demands were not significantly associated with late-life cognitive change (ps=.186 and .137, respectively).

Figure 2. Johnson-Neyman plot of education moderating job control-cognition relationship.

Figure 2.

Open in a new tab

Note. Johnson-Neyman plot depicting the moderation effect of education (centered at the mean of the JCQ sub-sample; M = 9.3) on the job control (JCQ) and late-life cognition relationship. The job control—cognition relationship is significant only for individuals with at least 8 years of education.

Job strain.

Table 3 contains results from regression analyses using job strain (quotient of job demands to job control) as the independent variable. As with the job control measure, after controlling for all covariates except education, higher job strain was significantly associated with greater cognitive decline whether measured with the O*Net measure (b=−0.88, SE=0.42, p=.035) or the JCQ measure (b=−1.09, SE=0.25, p<.001), and this association was explained away by education for the O*Net sample, but not for the JCQ sample (b=−0.78, SE=0.25, p=.002) despite over 30% reduction in the effect. We next examined whether education was a moderator for the association between job strain and cognition and found no significant interaction between job strain and education for either operationalization of strain (b=−0.08, SE=0.09, p=.403; b=−0.08, SE=.06, p=.159, respectively).

Table 3.

Results from regression analyses for the association between job strain and cognition for the JCQ and O*Net measures.

Model 1: Before Controlling for Education Model 2: Controlling for Education
Variable Name O*Net (n = 1632a) JCQ (n = 1467b) O*Net (n = 1632a) JCQ (n = 1467b)
B SE p B SE p B SE p B SE P
Intercept 11.47 0.67 11.34 0.62 11.99 0.65 12.23 0.61
Baseline Cognition 0.39 0.03 <.001 0.38 0.03 <.001 0.32 0.03 <.001 0.33 0.03 <.001
Age −0.11 0.01 <.001 −0.11 0.01 <.001 −0.09 0.01 <.001 −0.09 0.01 <.001
Sex (Female) 0.27 0.13 .034 0.31 0.13 .019 0.18 0.12 .145 0.25 0.13 .047
Baseline Depression −0.03 0.02 .129 −0.04 0.02 .051 −0.02 0.02 .226 −0.04 0.02 .082
Baseline Financial Problems −0.04 0.06 .521 −0.02 0.06 .699 0.03 0.06 .540 0.04 0.06 .547
Childhood Economic Hardship −0.01 0.06 .901 −0.01 0.06 .821 0.08 0.06 .175 0.07 0.06 .257
Hypertension −0.01 0.13 .943 −0.03 0.13 .795 −0.03 0.12 .794 −0.03 0.13 .795
Diabetes −0.19 0.14 .179 −0.26 0.15 .081 −0.17 0.14 .213 −0.26 0.15 .081
Job Strain −0.88 0.42 .035 −1.09 0.25 <.001 −0.53 0.41 .197 −0.78 0.25 .002
Education - - - - - - 0.14 0.01 <.001 0.13 0.02 <.001
Open in a new tab

Note. Age is centered at the minimum (60), depression and education are mean-centered; job strain is a quotient of standard scores (M = 100, SD = 15) of job demands and job control. B = unstandardized regression coefficient; SE = standard error.

a

There were missing values for money problems (n=5) or information about high blood pressure (n=5) or diabetes (n=2)

b

There were missing values for money problems (n=1) or information about high blood pressure (n=5) or diabetes (n=2)

Supplemental Results

As supplemental analyses, given the known associations among gender, occupation, and cognition, we tested whether gender moderated significant job strain—cognition relationships. Gender did not moderate the association between job control or job strain and late-life cognitive change (results not included but available upon request).

As post-hoc analyses, we also tested whether we could replicate the results reported above when assessing job strain indicators in relation to cognitive performance at follow-up (rather than change in cognitive performance between baseline and follow-up). Job control (both O*Net and JCQ operationalizations) was associated with a higher follow-up cognitive score before controlling for education (b=0.39, SE=.07, p<.001; b=0.38, SE=0.07, p=.002, respectively), but only remained significant after controlling for education in the JCQ sample (b=0.21, SE=0.07, p=.002). Education did not moderate these relationships (ps=.076 and .054, respectively). Higher JCQ job strain was associated with lower follow-up cognitive score before (b=−1.36, SE=0.26, p<.001) and after (b=−0.91, SE=0.26, p<.001) controlling for education. However, higher O*Net job strain was associated with poorer follow-up cognitive score before controlling for education (b=−1.05, SE=0.44, p=.018), but not after (b=−0.54, SE=0.42, p=.199). Education did not moderate these relationships either (ps=.116 and .473, respectively).

Discussion

In the current study, we sought to examine the relationship between job strain and late-life cognition in a less commonly studied, sometimes marginalized, and potentially vulnerable population: older adults from Puerto Rico. Consistent with our hypotheses, using residualized change regression analyses that adjust for baseline cognitive performance, we found that low job control and high job strain, whether assessed with the O*Net scheme or the JCQ-based job exposure matrix, were associated with poorer cognitive performance at follow-up. These results were independent of age, sex, depressive symptoms, baseline financial problems, childhood economic hardship, hypertension, and diabetes. Also in line with our hypotheses was the finding that education played a major role in the association between job strain indicators and cognitive decline. First, education explained away the negative association between greater job control and less cognitive decline, and the positive association between more job strain and more cognitive decline, when job strain was measured with the O*Net based scoring. These same associations were retained after education was added to the models with JCQ-based job control and job strain, but the magnitude of these associations was reduced by 40% and 30%, respectively. Second, we found an interaction of education by JCQ-based job strain, whereby there was a significant relationship between low job control and cognitive decline only for those with at least 8 years of education, suggesting that job strain had greater influence on cognitive performance among the more educated participants. This suggests that, among our sample of Puerto Ricans, the role of work environment in cognitive aging may grow with more years of education. Conversely, it may be that, for those with low education, influences on cognitive aging may come more predominantly from outside of the workplace.

Previous research has highlighted the importance of job control for late-life cognitive decline (Andel et al., 2011, 2015). Similarly, we found that low job control alone was related to late-life cognitive decline in our sample. We also found that a higher ratio of demands to control (job strain) was associated with greater late-life cognitive decline. These results fit with the theoretical model proposed by Karasek (1979) and indicate that a lack of job control, or a lack of job control in combination with high job demands, may later reflect in poorer cognitive outcomes. Notably, this finding applied even in our Puerto Rico-based sample, where employment opportunities may be more limited than elsewhere in the U.S. and job control scores where substantially lower than in the mainland U.S.

Another theoretical explanation for the adverse role of job strain indicators in poorer late-life cognition is Kahn and colleagues’ (1964) role conflict theory. In the case of job strain, an individual has certain responsibilities (i.e., demands) to meet, but role restrictions resulting in a lack of freedom to make decisions (i.e., control) may impede that individual’s ability to meet those responsibilities (Jackson, 1989). This internal conflict can result in stress that is detrimental to health. Unfortunately, this study did not include direct measures of stress (e.g., cortisol, HPA axis functionality). However, our use of job strain measures as a proxy for stress is strongly grounded in theory. Including physiological measures of stress in future studies would allow researchers to provide support for the glucocorticoid-cascade hypothesis (Sapolsky et al., 1986), which would indicate that prolonged exposure to stress (e.g., job strain) is related to accelerated aging (e.g., cognitive decline).

We also found that education was more strongly related to cognitive decline than job strain indicators, and that education explained away associations between some indicators of job strain and cognitive aging. Higher education is associated with better cognitive functioning, which is thought to be at least partially due to increased cognitive reserve (Stern, 2002, 2012). It appears that, in this specific cohort of Puerto Ricans born before 1943, education may be a more relevant factor for the rate of cognitive aging than the characteristics of work such as job strain. These results go along with the findings for work complexity and cognition in the PREHCO sample reported previously (Andel et al., 2019).

Since education can be a strong determinant of job placement (Raffe, 2003), the effect of education on working conditions may have also limited the range of job strain indicators present in the current sample. Specifically, lower average education (9.2 vs. 12.6 years in the HRS) in this older sample from Puerto Rico may have led to a higher rate of lower-level jobs categorized as having both lower demands and less control (see comparison to job control and demands in the HRS from the Methods section). On average, jobs appeared to have substantially less control than those in the mainland U.S. (M = 42.3 vs. M = 77.8 on a 0-100 scale, or 45% lower, in the PREHCO vs. HRS). Therefore, less opportunity for jobs with high control may have led to differences in the current findings compared to previous studies. Examining subjective job strain and important buffers of strain such as social support at work in Puerto Rico in future studies may help elucidate these findings and better disentangle education—work—cognition associations.

Job control as measured with the JCQ was associated with late-life cognitive decline only for those who were highly educated. Given that education frequently determines a person’s occupation, this finding seems plausible. It may be that those who are highly educated end up in an occupation where they not only have high job control, but they are frequently utilizing this control because the job demands are also high. In Karasek’s (1979) job demand-control model, this example is called an “active job”. Previous studies found that active jobs have a protective effect (Andel et al., 2011) or that compared to high strain jobs (i.e., high job demands, low job control) and passive jobs (i.e., low job demands and low job control); people in active jobs were less likely to develop dementia (Wang et al., 2012).

Results may have diverged from traditional patterns of job strain—cognition relationships due to the stage of industrialization of countries under investigation. This may be tied to the economic contexts. Specifically, Puerto Rico was experiencing significant changes in the economic climate due to the rapid industrialization during the time the adults in the sample would have been working (Toro, 2014, 2017), whereas previous samples from European countries and the mainland U.S. would have already had established and highly industrialized economies. Acknowledging that the economic context may influence these relationships is important for understanding the generalizability of previous job strain findings.

Limitations and Future Directions

We measured occupation-based rather than self-reported job strain. Self-reported job strain would capture between-person differences within one occupation and allow for examination of the effect of unique job factors (e.g., managerial styles, corporate vs. family-owned work environments) and individuals’ perception on job strain. Although occupation-based measures do not account for these individual-level factors on job strain since they assess strain on a job-category-level (e.g., assuming all accountants have the same level of demands and control), occupation-based measures do not suffer from many of the limitations of self-reported indicators, such as bias due to recall, self-perception, social desirability, and current psychological state. Given the objective nature of occupation-based job strain measures and the fact that this sample consisted of older adults who are more likely to have memory difficulties, the occupation-based measure may be more appropriate. Ideally, future studies would use both types of measurement for job strain.

It is important to note that our study looked at the occupation that individuals spent the greatest amount of time in. We did not account for career changes, resulting in the examination of a portion rather than the totality of participants’ occupational history. Additionally, we did not have information on how much time had passed since participants were in their primary occupations. On average, our sample reached adulthood in the 1950s, when average education level was beginning to increase rapidly on the island. Industrialization in the 1960s and 1970s along with the economic crisis that occurred in the 1970s through the early 1980s resulted in great workforce fluctuations in Puerto Rico (e.g., a decline in agricultural jobs corresponding with increased manufacturing jobs, followed by a sharp rise in unemployment) (Riviera-Batiz & Santiago, 1996; Toro, 2014). Although traditionally people tend to stay in one line of occupation, future studies should obtain a more detailed occupational history. Finally, the current analytic sample only included individuals with long-term employment, and thus cannot be generalized to the portion of the Puerto Rican population that was not part of the labor force. The male labor force participation rate in Puerto Rico for those age 16+ years declined from 70.7% in 1950 to 54.5% in 1980. For women, there was an increase in labor force participation during this same period of time (21.3% in 1950 and 29.1% in 1980; Riviera-Batiz & Santiago, 1996).

We found that participants who were included in our analysis were younger, more likely to be male, and had higher educational attainment than participants who were excluded. With the higher educational attainment and younger age, our sample may have had less cognitive decline than the full PREHCO sample. Due to the known historical differences in occupations found between men and women in Puerto Rico (Toro, 2017) as well as the previously described cohort changes in education and occupation, results may not be generalizable to all older Puerto Ricans or to future cohorts of older Puerto Ricans.

Conclusion

We provide initial evidence suggesting that low job control and greater job strain may contribute to cognitive decline among Puerto Rican older adults, which is in line with results from other populations. However, we note that this Puerto Rican sample had occupations with lower job control and job demands scores than what has been reported in previous work from the mainland U.S. Although both more job control was found to reduce cognitive decline and more job strain was found to accelerate it, the associations were substantially influenced by education. Considering that Puerto Rico has a higher proportion of older adults than the mainland U.S., it is important to recognize how factors contributing to cognitive decline may differ from other populations to best understand the unique experience of this population. Education is a modifiable factor with clear disparities across race and ethnicity in the U.S. but without consensus on how best to provide equal access to educational opportunities. Further, unequal educational opportunities could also result in differential job placement, limiting the amount of control adults would have in their occupation. For adults already in their working years, it is possible that we could still improve job control with relatively small changes such as more flexibility with work schedule or more involvement in decision-making including order of day-to-day work tasks. Together, these findings provide groundwork for better understanding the role of life course factors such as education and occupation in relation to cognitive aging across diverse populations of older adults.

Acknowledgments

Funding: This work was supported in part by National Institute on Aging (NIA) (grant numbers R21 AG045722, R01 AG064769). The content is solely the responsibility of the authors and does not necessarily represent the official views of NIA or the National Institutes of Health.

Footnotes

Declaration of Conflicting Interests: The authors declare that there is no conflict of interest.

References

  1. Agbenyikey W, Karasek R, Cifuentes M, Wolf P, Seshadri S, Taylor J, Beiser A, & Au R (2015). Job strain and cognitive decline: A prospective study of the framingham offspring cohort. The International Journal of Occupational and Environmental Medicine, 6(2), 79–94. 10.15171/ijoem.2015.534 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Aging, A. on. (2016). A Profile of Older Americans: 2016. https://www.aarp.org/content/dam/aarp/livable-communities/learn/demographics/a-profile-of-older-americans-2011-aarp.pdf.
  3. Aguilar-Navarro SG, Fuentes-Cantú A, Ávila-Funes JA, & García-Mayo EJ (2007). Validez y confiabilidad del cuestionario del ENASEM para la depresión en adultos mayores. Salud Publica de Mexico, 49(4), 256–262. 10.1590/S0036-36342007000400005 [DOI] [PubMed] [Google Scholar]
  4. American Psychological Association. (2011). Stressed in America. Monitor on Psychology. [Google Scholar]
  5. Andel R, Crowe M, Kareholt I, Wastesson J, & Parker MG (2011). Indicators of job strain at midlife and cognitive functioning in advanced old age. Journals of Gerontology - Series B Psychological Sciences and Social Sciences, 66 B(3), 287–291. 10.1093/geronb/gbq105 [DOI] [PubMed] [Google Scholar]
  6. Andel R, Dávila-Roman AL, Grotz C, Small BJ, Markides KS, & Crowe M (2019). Complexity of work and incident cognitive impairment in Puerto Rican older adults. Journals of Gerontology - Series B Psychological Sciences and Social Sciences, 74(5), 785–795. 10.1093/geronb/gbx127 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Andel R, Infurna FJ, Rickenbach EAH, Crowe M, Marchiondo L, & Fisher GG (2015). Job strain and trajectories of change in episodic memory before and after retirement: results from the Health and Retirement Study. Journal of Epidemiology and Community Health, 69, 442–446. 10.1136/jech-2014-204754 [DOI] [PubMed] [Google Scholar]
  8. Anstey KJ, Ee N, Eramudugolla R, Jagger C, & Peters R (2019). A systematic review of meta-analyses that evaluate risk factors for dementia to evaluate the quantity, quality, and global representativeness of evidence. Journal of Alzheimer’s Disease, 70, S165–S186. 10.3233/JAD-190181 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Association, A. (2014). Alzheimer’s Association Report: 2014 Alzheimers disease facts and figures. Alzheimer’s and Dementia, 10, e47–e92. 10.1016/j.jalz.2014.02.001 [DOI] [PubMed] [Google Scholar]
  10. Awang H, Mansor N, Peng TN, & Osman NAN (2018). Understanding ageing: Fear of chronic diseases later in life. Journal of International Medical Research, 46(1), 175–184. 10.1177/0300060517710857 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Brunner EJ, Chandola T, & Marmot MG (2007). Prospective effect of job strain on general and central obesity in the Whitehall II Study. American Journal of Epidemiology, 165(7), 828–837. 10.1093/aje/kwk058 [DOI] [PubMed] [Google Scholar]
  12. Bureau, U. S. C. (2019). QuickFacts: Puerto Rico. https://www.census.gov/quickfacts/PR
  13. CIA. (2020). Central America: Puerto Rico. The World Factbook. https://www.cia.gov/library/publications/the-world-factbook/geos/rq.html [Google Scholar]
  14. Cifuentes M, Boyer J, Gore R, D’Errico A, Tessler J, Scollin P, Lerner D, Kriebel D, Punnett L, & Slatin C (2007). Inter-method agreement between O*NET and survey measures of psychosocial exposure among healthcare industry employees. American Journal of Industrial Medicine, 50(7), 545–553. 10.1002/ajim.20480 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Consult, M. (2017). Morning Consult. National Tracking Poll; #170916. [Google Scholar]
  16. Cooper C, Sommerlad A, Lyketsos CG, & Livingston G (2015). Modifiable predictors of dementia in mild cognitive impairment: A systematic review and meta-analysis. American Journal of Psychiatry, 172(4), 323–334. 10.1176/appi.ajp.2014.14070878 [DOI] [PubMed] [Google Scholar]
  17. Dalecki M, & Willits FK (1991). Examining change using regression analysis: Three approaches compared. Sociological Spectrum, 11(2), 127–145. 10.1080/02732173.1991.9981960 [DOI] [Google Scholar]
  18. Folkman S (1984). Personal control and stress and coping processes: A theoretical analysis. Journal of Personality and Social Psychology, 46(4), 839–852. 10.1037/0022-3514.46.4.839 [DOI] [PubMed] [Google Scholar]
  19. Friedman B, Heisel MJ, & Delavan RL (2005). Psychometric properties of the 15-item geriatric depression scale in functionally impaired, cognitively intact, community-dwelling elderly primary care patients. Journal of the American Geriatrics Society, 53(9), 1570–1576. 10.1111/j.1532-5415.2005.53461.x [DOI] [PubMed] [Google Scholar]
  20. González-Mejía AM, & Ma X (Cissy). (2017). The Emergy Perspective of Sustainable Trends in Puerto Rico From 1960 to 2013. Ecological Economics, 133, 11–22. 10.1016/j.ecolecon.2016.11.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Hayes AF (2017). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach (Second). Guilford Press. [Google Scholar]
  22. Jackson SE (1989). Does job control control job stress? Job Control and Worker Health, 25–53. [Google Scholar]
  23. Kahn RL, Wolfe DM, Quinn RP, Snoek JD, & Rosenthal RA (1964). Organizational stress: Studies in role conflict and ambiguity. Wiley. [Google Scholar]
  24. Karasek RA, Kawakami N, Brisson C, Houtman I, Bongers P, & Amick B (1998). The Job Content Questionnaire (JCQ): An instrument for internationally comparative assessments of psychosocial job characteristics. Journal of Occupational Health Psychology, 3(4), 322–355. [DOI] [PubMed] [Google Scholar]
  25. Karasek R, & Theorell T (1990). Healthy work: Stress, productivity, and the reconstruction of working life. Basic Books. [Google Scholar]
  26. Karasek Robert A. (1979). Job demands, Job decision latitude, and mental strain: Implications for job redesign. Administrative Science Quarterly, 24(2), 285. 10.2307/2392498 [DOI] [Google Scholar]
  27. Katz MS (1976). A history of compulsory education laws. In Fastback Series, No. 75. Bicentennial Series. [Google Scholar]
  28. Kørner A, Lauritzen L, Abelskov K, Gulmann N, Marie Brodersen A, Wedervang-Jensen T, & Marie Kjeldgaard K (2006). The Geriatric Depression Scale and the Cornell Scale for Depression in Dementia. A validity study. Nordic Journal of Psychiatry, 60(5), 360–364. 10.1080/08039480600937066 [DOI] [PubMed] [Google Scholar]
  29. Ladd HF, & Riviera-Batiz FL (2006). Education and economic development. In The Economy of Puerto Rico: Restoring Growth (pp. 189–238). [Google Scholar]
  30. Lupien SJ, McEwen BS, Gunnar MR, & Heim C (2009). Effects of stress throughout the lifespan on the brain, behaviour and cognition. Nature Reviews Neuroscience, 10(6), 434–445. 10.1038/nrn2639 [DOI] [PubMed] [Google Scholar]
  31. Maust DT, Solway E, Langa KM, Kullgren JT, Kirch M, Singer DC, & Malani P (2019). Perception of dementia risk and preventative actions among US adults aged 50 to 64 years. JAMA Neurology, 77(2), 259–261. 10.1001/jamaneurol.2019.3946 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. McEniry M, & Palloni A (2010). Early life exposures and the occurence and timing of heart disease among the older adult Puerto Rican population. Demography, 47(1), 23–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Nilsen C, Andel R, Fors S, Meinow B, Darin Mattsson A, & Kåreholt I (2014). Associations between work-related stress in late midlife, educational attainment, and serious health problems in old age: A longitudinal study with over 20 years of follow-up. BMC Public Health, 14(1), 1–12. 10.1186/1471-2458-14-878 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Nyberg ST, Fransson EI, Heikkilä K, Alfredsson L, Casini A, Clays E, De Bacquer D, Dragano N, Erbel R, Ferrie JE, Hamer M, Jöckel KH, Kittel F, Knutsson A, Ladwig KH, Lunau T, Marmot MG, Nordin M, Rugulies R, … Kivimäki M (2013). Job strain and cardiovascular disease risk factors: Meta-analysis of individual-participant data from 47,000 men and women. PLoS ONE, 8(6), e67323. 10.1371/journal.pone.0067323 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Perez C, & Ailshire JA (2017). Aging in Puerto Rico: A comparison of health status among island Puerto Rican and mainland U.S. older adults. Journal of Aging and Health, 29(6), 1056–1078. 10.1177/0898264317714144 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Perreira K, Peters R, Lallemand N, & Zuckerman S (2017). Puerto Rico health care infrastructure assessment. [Google Scholar]
  37. Pettigrew C, & Soldan A (2019). Defining cognitive reserve and implications for cognitive aging. Current Neurology and Neuroscience Reports, 19(1). 10.1007/s11910-019-0917-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Plassman BL, Williams JW, Burke JR, Holsinger T, & Benjamin S (2010). Systematic review: Factors associated with risk for and possible prevention of cognitive decline in later life. Annals of Internal Medicine, 153(3), 182–193. 10.7326/0003-4819-153-3-201008030-00258 [DOI] [PubMed] [Google Scholar]
  39. Pocklington C, Gilbody S, Manea L, & McMillan D (2016). The diagnostic accuracy of brief versions of the Geriatric Depression Scale: a systematic review and meta-analysis. International Journal of Geriatric Psychiatry, 31(8), 837–857. 10.1002/gps.4407 [DOI] [PubMed] [Google Scholar]
  40. Raffe D (2003). Pathways linking education and work: A review of concepts, research, and policy debates. Journal of Youth Studies, 6(1), 3–19. 10.1080/1367626032000068136 [DOI] [Google Scholar]
  41. Riviera-Batiz FL, & Santiago CE (1996). Puerto Rico in the 1990s. Russell Sage Foundation. [Google Scholar]
  42. Roman J (2015). The Puerto Rico healthcare crisis. Annals of the American Thoracic Society, 12(12), 1760–1763. [DOI] [PubMed] [Google Scholar]
  43. Saad L (2014). The “40 hour” workweek is actually longer - by seven hours. http://news.gallup.com/poll/175286/hour-workweek-actually-longer-seven-hours.aspx [Google Scholar]
  44. Sabbath EL, Andel R, Zins M, Goldberg M, & Berr C (2016). Domains of cognitive function in early old age: which ones are predicted by pre-retirement psychosocial work characteristics? Occupational and Environmental Medicine, 73(10), 640–647. 10.1136/oemed-2015-103352 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Sanchez-Ayendez M, Caban C, Fernandez L, Rosich W, Davila AL, Larriuz M, Hernandez J, Guruxarri A, & Palloni A (2003). Una escala psicometrica breve para evaluar el estado cognitivo de hispanoparlantes de edad mayor. Puerto Rican Health Sciences Journal, 22(4), 377–383. [PubMed] [Google Scholar]
  46. Sapolsky RM, Krey LC, & McEwen BS (1986). The neuroendocrinology of stress and aging: The glucocorticoid cascade hypothesis. Endocrine Reviews1, 7(3), 284–301. [DOI] [PubMed] [Google Scholar]
  47. Schnall PL, Landsbergis PA, & Baker D (1994). Job strain and cardiovascular disease. Annual Review of Public Health, 15, 381–411. https://doi.org/0163-7525194/0510-0381$05.00 [DOI] [PubMed] [Google Scholar]
  48. Schwartz JE (2000). Imputation of job characteristics scores at “measurement of psychosocial workplace exposure variables”. In Schnall PL, Belkic K, Landbergis PA, & Baker D (Eds.), State of the art reviews; occupational medicine (pp. 172–175). Hanley & Belfus Inc. [PubMed] [Google Scholar]
  49. Stansfeld SA, Shipley MJ, Head J, & Fuhrer R (2012). Repeated job strain and the risk of depression: Longitudinal analyses from the Whitehall II study. American Journal of Public Health, 102(12), 2360–2366. 10.2105/AJPH.2011.300589 [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Stern Y (2002). What is cognitive reserve? Theory and research application of the reserve concept. Journal of the International Neuropsychological Society, 8(3), 448–460. 10.1017/S1355617702813248 [DOI] [PubMed] [Google Scholar]
  51. Stern Y (2009). Cognitive reserve. Neuropsychologia, 47(10), 2015–2028. 10.1016/j.neuropsychologia.2009.03.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Stern Y (2012). Cognitive reserve in ageing and Alzheimer’s disease. The Lancet Neurology, 11(11), 1006–1012. 10.1016/S1474-4422(12)70191-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Then FS, Luck T, Luppa M, Arélin K, Schroeter ML, Engel C, Löffler M, Thiery J, Villringer A, & Riedel-Heller SG (2014). Association between mental demands at work and cognitive functioning in the general population - Results of the health study of the Leipzig research center for civilization diseases (LIFE). Journal of Occupational Medicine and Toxicology, 9(1). 10.1186/1745-6673-9-23 [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Toro HJ (2014). Economic change and occupational stasis: Puerto Rico as a case study of stratification and development. Research in Social Stratification and Mobility, 36, 101–119. 10.1016/j.rssm.2014.02.003 [DOI] [Google Scholar]
  55. Toro HJ (2017). Legacies of opportunity and economic integration: Path dependent labor force participation in Puerto Rico’s development. Research in Social Stratification and Mobility, 49(March), 47–60. 10.1016/j.rssm.2017.03.006 [DOI] [Google Scholar]
  56. Wang HX, Wahlberg M, Karp A, Winblad B, & Fratiglioni L (2012). Psychosocial stress at work is associated with increased dementia risk in late life. Alzheimer’s and Dementia, 8(2), 114–120. 10.1016/j.jalz.2011.03.001 [DOI] [PubMed] [Google Scholar]
  57. Wang Y, Du Y, Li J, & Qiu C (2019). Lifespan intellectual factors, genetic susceptibility, and cognitive phenotypes in aging: Implications for interventions. Frontiers in Aging Neuroscience, 11(MAY), 1–11. 10.3389/fnagi.2019.00129 [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Yesavage JA, Brink TL, Rose TL, Lum O, Huang V, Adey M, & Leirer VO (1983). Development and validation of a geriatric depression screening scale: A preliminary report. Journal of Psychiatric Research, 17(1), 37–49. 10.1016/0022-3956(82)90033-4 [DOI] [PubMed] [Google Scholar]

RESOURCES