Profiles of economic and non-economic chronic stressors at midlife

Abstract

The purpose of the present study was to identify latent profiles of economic and chronic stressors during midlife. A national sample of N = 1541 adults ages 40–61 years (35.5 % male, 48.7 % white) responded to questionnaires. Latent profile analyses revealed five stressor profiles, specifically Low Stressors (46.3 % of the sample), Mixed Stressors (22.7 %), High Sicktime Stressors (14.7 %), High Unemployment Stressors (8.5 %), and High Stressors (7.9 %). Respectively, the Low and High Stressor profiles were the most and least advantaged in terms of demographic risk factors, social support, and health-related quality of life (HRQoL), while those in the three moderate profiles varied inconsistently across the covariates. The discussion considers the findings in relation to the literature and their implications for applied efforts during early and middle adulthood.

Stress and stressors have a long history of research in the social sciences, culminating in a rich literature on the effects of stress for health, with consistent evidence that high levels of stress predict adults’ own poor mental and physical health outcomes (Turner et al., 1995; Umberson et al., 2008), as well as their children’s cognitive, emotional and behavioral wellbeing (Aviles et al., 2024; Garner & Toney, 2020; Li et al., 2024). Individuals vary considerably in terms of the stressors that they experience, and it has long been speculated that specific combinations of stressors may have disproportionate effects on their health outcomes (Pearlin et al., 2005). To date, however, relatively little is known about such profiles or about their correlates. A small number of recent person-centered studies have explored such combinations of stressors in relation to covariates and health outcomes, as part of generating evidence that may be useful for targeting prevention and intervention efforts (e.g., D. Chen et al., 2024). As described below, these have generally revealed three or four profiles of stressors, corresponding to a continuum of low through high levels of exposure. Gaps remain in the literature, however, as explorations of acute and chronic stressor profiles have generally involved samples of typically-developing adolescents or young adult university students, with questionable generalizability beyond these developmental periods. This gap is particularly acute in reference to community samples of adults, who may experience varying levels of comparatively diverse stressors relative to those encountered by teens and emerging adults. Seeking to address this gap in the literature, the present study identified profiles of chronic stressors in middle-aged adults, including younger and older members of Gen X (i.e., the generational label that applies to adults born between 1965 and 1980), a cohort whose experiences with stress may be somewhat unique. We also explored associations with covariates, which permitted us to identify which individuals are most likely to have stressor profiles that may culminate in better or worse health outcomes. This information can help to inform targeted intervention efforts intended to address stress-related health consequences for at-risk middle-aged adults.

Stress and stressors at midlife

While prior inquiries into stressor profiles that have been conducted with adolescent and emerging adult samples may inform the present study’s hypotheses in a very general fashion, for several reasons, it seems unlikely that their findings will generalize to middle-aged adults. First, levels of stress vary across cohorts: Gen X adults appear to be somewhat less stressed in comparison to younger cohorts, but also more stressed than older generations (American Psychological Association, 2021, 2023). Second, the stressors that individuals encounter and their responses to them also change with age (Pearlin & Bierman, 2013): relative to teens and young adults, midlife adults report greater financial and health-related stressors (American Psychological Association, 2023), forms that tend to be chronic in nature that have generally been omitted in person-centered studies conducted with younger samples. To our knowledge, to date, financial stress has only been represented by a single item in one study of college students (Liao et al., 2018). Instead, prior investigations have included a mix of acute, chronic, and other stressors that are developmentally-relevant for younger age groups (e.g., pressure for academic achievement), but may not be broadly applicable to adults at midlife (Andretta & McKay, 2019; D. Chen et al., 2024; Liao et al., 2018; Shaw et al., 2024). Further, other stressors associated with middle adulthood have been similarly (and logically) overlooked, for example stressors associated with raising and launching children, and caring for aging family members. Finally, while the health effects of racial inequality and discrimination are visible throughout adulthood, by midlife such effects have become entrenched, widening the health gap between adults who identify as White and people of color (POC; Geronimus et al., 2006).

Thus, the current inquiry considered both economic and non-economic chronic stressors at midlife, which align with those considered in previous variable-centered investigations of stress during middle adulthood (R. Chen et al., 2022; Scott et al., 2013), and with theoretical arguments about the crucial role of economic stressors as a subgroup of chronic stressors in central life domains (Pearlin, 2010; Pearlin et al., 1981, 2005). Of the seven economic stressors represented in an updated measure, three dimensions related to employment-related economic strains (i.e., inadequate sick time, unemployment, and underemployment/problematic employment), two involved residential stressors (i.e., undesirable living situations and home ownership problems), and two concerned stressors related to inadequate health insurance and economic-related social role constraints (Richman et al., 2012). The nine focal non-economic chronic stressors included five universal forms that are broadly applicable to adults, specifically stressors related to general time pressures, isolation, social life difficulties, and residential and health problems. We also considered four role-specific forms of chronic stressors, specifically related to marital/romantic relationships, singlehood or the absence of enduring romantic ties (i.e., non-relationship stressors), divorce or separation, and parental/family stressors (Turner et al., 1995). Such stressors should be relevant for middle-aged adults, as these correspond to the central developmental tasks of midlife, which concern personal achievements in career development and in strong relationships with others, including supporting the prospects of future generations (Erikson, 1993; Hutteman et al., 2014; Vaillant, 2012).

Such chronic stressors covary in adulthood. For example, using a different measure of nine chronic stressors in a sample of German middle-aged adults, bivariate associations between the various stressor types were modest to moderate in magnitude, ranging from r = 0.21 to 0.69 (Hussenoeder et al., 2022). Similarly, in their analysis of data from the Notre Dame Study of Health and Well-Being, Scott et al. (2013) revealed modest to moderate correlations between the focal economic and non-economic chronic stressors, ranging from r = 0.13 to 0.62. Thus, there is reason to believe that the selected stressors are relevant to midlife adults and may be conducive for identifying meaningful typological profiles.

Profiles of stressors

Much of the evidence about covariates of stressors in adulthood has been derived from variable-oriented research focusing on the statistical associations between theoretically-based independent and dependent variables in aggregate (e.g., a set of stressors and a specific health outcome across a population). However, variable-centered paradigms are ill-suited for understanding how processes may operate as integrated systems within individuals (Bergman & Magnusson, 1997). As a consequence, findings of variable-focused investigations are rarely generalizable to individual people, even individuals drawn from the same population under study (Molenaar & Campbell, 2009). The person-oriented or person-centered approach addresses this limitation, and its associated methods are suited to exploring the interrelationships amongst different processes within individuals (Von Eye, 2010). A major focus in this research is describing how factors coalesce into a finite number of meaningful patterns, and subsequently, to understanding the predictors, correlates, and outcomes of those typologies (Bergman & Trost, 2006). We utilize this approach in the present study to examine the patterning of chronic stressors, and correlates of these patterns, for adults at midlife.

The stress proliferation framework provided a general basis for the present investigation, though we emphasize that this model is not developmentally-focused and does not lend itself to predictions about discrete stressor profiles or their covariates (i.e., presumptive antecedents). Per this model, serious stressors tend to beget other significant stressors, and health problems may result when such problems expand over time (Pearlin et al., 2005). Proliferation may occur through the expansion of primary stressors (e.g., loss of income resulting from unemployment can trigger new stressors related to debts) or through secondary stressors in other social roles from domains that are separate from that of the original stressor (e.g., financial strains following unemployment can lead to problems such as marital conflict; Pearlin et al., 1981). Subsequently, Wheaton and colleagues expanded upon these ideas in formulating the stress domain hypothesis, which stipulated that no individual stressor dimension could serve as a proxy representing all stressors, and that the health impacts of stressors are contingent upon the combination of stressors experienced at significant points of the lifespan (Turner et al., 1995; Wheaton, 1994; Wheaton et al., 2013). By midlife, stressors may have proliferated and accumulated to the point that individuals can experience their health effects (Pearlin et al., 2005; Price et al., 2002). While the present study is not situated to capture the temporal chain of events (e.g., in which the original stressors may “spill over” into other life domains, resulting in new stressors), we are able to determine which chronic stressors tend to occur together, and at what level, across profiles at midlife. Examining the patterning of stressors by type and level at one point in time grants insight into how certain stressor profiles may be relatively more harmful to health.

Profiles of stressors

In person-centered analyses, the number and composition of profiles depends upon the number and quality of indicators that are included in the model (Wurpts & Geiser, 2014). Studies with teens and young adults have typically involved only five to nine stressors, and have revealed three to four profiles that correspond to low, moderate, and high levels of current or recent stressors (with one exception noted below). Our consideration of 16 developmentally-relevant domains of chronic stressors allows for the possible identification of a comparatively larger number of nuanced profiles. We review the relevant literature below, first discussing the emergent profiles, and subsequently, the differences between profiles in covariates.

In one investigation with teenagers, researchers assessed nine forms of acute and chronic stressors that respondents experienced either over the preceding six months or in their lifetime. These included chronic academic difficulties and problems in relationships with family members, peers, friends, and romantic partners, as well as prior childhood adversity (experienced at least one year prior to the interview) and acute interpersonal and non-interpersonal stressors experienced within the past year (Shaw et al., 2024). Latent profile analysis supported the retention of four profiles, the first characterized by low levels of stressors across each domain, into which the majority of the participants were classified. This was followed by a relatively small-sized second profile distinguished by high chronic peer and friend relationship stressors. The two remaining profiles were each marked by moderate chronic family relationship stressors, moderate prior childhood adversity (i.e., inclusive of familial financial difficulties), and moderate acute past year stressors. What distinguished these latter profiles were the levels of chronic academic stressors and school-related behavioral problems, which were elevated only in the smallest-sized profile (Shaw et al., 2024). While the stressors studied varied between studies, patterns demonstrated in other samples of teens and young adults are generally consistent with those in Shaw et al. (2024). In these studies, conducted with British or Chinese youth, research invariably revealed three stressor profiles, with the first and largest group representing low levels of stressors measured, a “high stress” group with generally high levels of all stressors measured, and a middle group that different in level and pattern of stressors experienced (Andretta & McKay, 2019; D. Chen et al., 2024; Liao et al., 2018).

There are limited person-centered studies with samples of adults, which reveal a bit more nuance and variability in the size of the profiles than the studies of younger samples. This may be because these studies focused on acute current stressors specifically during the early COVID-19 pandemic. Two separate investigations of current COVID-related stressors (inclusive of economic and employment-related stressors) revealed four classes, or profiles, that mirrored those in the studies of youth. In one involving 17 stressors (Luk et al., 2023), these profiles corresponded to minimal stressors (51.6 %), serious financial stressors (9.1 %), work-focused stressors (24.8 %) and family-friends COVID-related stressors (14.5 %). In the other study involving six stressors (Blais et al., 2022), the four profiles included adapting/low stressors (30 %), conflicted/moderate stressors (35 %; this profile was characterized by work-life imbalance and social isolation concerns), insecure/moderate stressors (21 %; this profile was characterized by job security and personal illness concerns), and stressed/high stressors (14 %).

In sum, studies conducted across multiple developmental periods and involving different measures of stressors tend to reveal similar numbers and types of profiles, particularly when the stressors measured are chronic in nature. Notwithstanding, it is unknown whether these previous findings will generalize to the types of chronic stressors that are developmentally-relevant to adults at midlife.

Covariates of profiles

The present study also considered a number of demographic and other covariates of profile membership, guided by the larger literatures on stress, health, and psychopathology. While individuals vary considerably in terms of their degree of exposure to stressors, adults who experience any type of social disadvantage are at elevated risk for high levels of chronic stress and for stress proliferation (Pearlin et al., 2005; Thoits, 1982; Turner et al., 1995; Turner & Avison, 2003). Selected covariates included gender, age, race, socioeconomic status (SES: i.e., income, educational attainment, and current employment), marital/cohabiting relationship status, community disadvantage, social support, and physical and mental health-related quality of life (HRQoL).

Demographic covariates

We anticipated that profiles would vary in terms of gender, though evidence concerning gender differences in stressors in adulthood is mixed. Women tend to experience higher levels of daily stressors (Stawski et al., 2023), while men tend to report higher levels of financial stressors (Richman et al., 2014). In one person-centered study of adults, women were more likely to be in the highest versus lowest COVID-related stressors profile (Blais et al., 2022). We explored age differences between profiles, but as this covariate has not been considered in prior person-centered studies, we were unable to advance a specific hypothesis regarding variations between older and younger members of Gen X. Concerning race, there is evidence that, relative to white adults, Black adults are disproportionately likely to report higher levels of economic and some chronic stressors (Brown et al., 2020). This pattern was reflected in two person-centered studies of COVID-related stressors in adults, in which individuals who identified as POC (Blais et al., 2022) or as Black (Luk et al., 2023) had elevated odds of membership in the highest stressor exposure profiles. High stressor levels may result directly from systematic marginalization and pervasive discrimination that are disproportionately experienced by POC (Dalessandro & Lovell, 2023; Wang et al., 2023). Concerning SES, adults with low income levels were disproportionately represented in high stressor classes, while those with high income levels were more likely to be classified into the lowest stressor profile (Luk et al., 2023). Such a pattern also emerged in one of the studies involving adolescents: Financial strain (indicated by receipt of free school lunch) and seeking larger numbers of school-based support services (e.g., educational psychologist supports) were linked to high probability of membership in the high stressor profile (Andretta & McKay, 2019). Similarly, low educational attainment and unstable employment have each been linked to elevated odds of membership in high versus lower stressor profiles (Blais et al., 2022; Luk et al., 2023). Person-centered studies conducted with adults have not explicitly considered the role of relationship status (i.e., whether an adult is cohabiting with or married to a partner), yet singlehood is an established predictor of elevated stress levels and economic stressors in particular (Loibl et al., 2022; Ta et al., 2017). Based on this previous research, we anticipated that membership in higher-stressor profiles would be linked to female gender, identifying as a POC, low income levels, low educational attainment, current unemployment, and singlehood.

Community disadvantage

Community disadvantage concerns the degree to which the individual’s immediate social environment is characterized by concentrated economic inequity. To date, we are unaware of any research in which it has been considered as a covariate of stressor profiles. Conceptually, this covariate represents the social context in which chronic stressors are experienced (Wheaton et al., 2013), which likely indexes the degree to which an individual’s environment (i.e., “the conditions in the environments where people are born, live, learn, work, play, worship, and age”; Healthy People 2030 et al., 2025) may produce stressors or help individuals cope with them, as posited per the social determinants of health (SDH) model (Braveman & Gottlieb, 2014). Pearlin et al. (2005) originally viewed this as a product of accumulated individual sociodemographic risk factors, conceptualizing it as an ambient stressor that could add to the burden of other stressors, ultimately contributing to eroded health. On these bases, we posited that adults classified into high-stressor profiles would live in higher-disadvantage communities, while those in relatively lower stressor profiles would reside in more advantaged settings.

Social support

Stress-related arousal may motivate individuals to use strategies for modulating that arousal to an acceptable level (i.e., to treat the physical and emotional symptoms of stress). One common coping strategy is seeking social support from others, such as romantic partners, friends, or family members, who may provide reassurance or practical assistance during challenging circumstances. A prior study with a Chinese college student sample revealed that respondents who had fulfilling close relationships were less likely to be classified into profiles characterized by relationship stresses (Liao et al., 2018). Per the buffering hypothesis (Cohen & Wills, 1985), we anticipated that individuals classified into lower stressor profiles would report higher levels of social support, while those in higher stressor profiles would report lower levels of support.

Health outcomes

It is well-established that economic and chronic stressors have significant implications for physical and mental health outcomes in adulthood (Graf et al., 2017; Guidi et al., 2021; Umberson et al., 2008), which is also consistent with the SDH model. Thus, we also explored profile differences in physical and mental health-related quality of life (PHRQoL and MHRQoL), which involves individuals’ self-perceptions of their physical and mental well-being (Phyo et al., 2020). These self-ratings of health status are commonly used in clinical settings and intervention evaluations, as they are consistent predictors of long-term health outcomes, including physical and mental health diagnoses and mortality (Kanesarajah et al., 2018; Wei & Mukamal, 2019). Experiencing chronic, multidimensional stressors may overwhelm adults’ coping capacities, resulting in physical symptoms (e.g., disrupted sleep and high blood pressure) that can trigger other physical health maladies and emotional distress, which ultimately undermine their self-perceptions of wellbeing (Guidi et al., 2021; Lippert et al., 2022). Simultaneously, overwhelming stressor exposure may push individuals to try to manage consequential negative affect through ineffective strategies such as rumination or brooding, which by itself may lead to depression and anxiety symptoms (Michl et al., 2013; Shaw et al., 2024). Elevated exposure to economic stressors may be of particular concern, given that chronic stressors (e.g., relationship conflict and distress) can be sequelae of economic triggers (Conger et al., 1999). In addition to these pathways, adults with high levels of economic stressors may also have limited resources for seeking routine healthcare or treatment for physical and mental health difficulties, which could further erode their HRQoL (McMaughan et al., 2020).

To date, person-centered studies with adults have considered both physical and mental health outcomes, with associations between stressor profiles and health outcomes that are generally logical per the larger literature. The studies of COVID-related stressors revealed that the lowest-stressor-exposure profiles had better physical and mental health than that reported by adults in comparatively higher-stressor-exposure profiles (Blais et al., 2022; Luk et al., 2023). Indeed, these effects were particularly pronounced for the serious financial stressor profile (i.e., the highest stressor class), even with demographic covariates modeled (Luk et al., 2023).

The current study

Owing to the gaps in the literature described above, it remains ambiguous whether the stressor profiles demonstrated in prior studies of adolescents and adults generalize specifically to middle adulthood, particularly when both economic and non-economic chronic stressors are included. Seeking to address this gap in pursuit of evidence that may inform applied efforts, the present study identified profiles of chronic stressors in a diverse national sample of middle-aged adults and connected these profiles to a broad range of covariates.

Guiding research questions were, firstly, what profiles of individual stressors are present at midlife? Firm hypotheses are not usually advanced in person-centered inquiries, and our predictions are tentative owing to the gaps in the literature. We anticipated the presence of at least one profile with high levels of multiple stressors, one with moderate stressors, and one with low stressors. Beyond anticipating that the lowest-stressor-exposure profile would be the most common, no further predictions can be advanced about relative group sizes beyond that they would vary. Secondly, how are these profiles linked to covariates? We anticipated that women would be disproportionately represented in profiles with elevated chronic stressors, while men would be overrepresented in profiles with elevated economic stressors. Grounded in the literatures on the weathering hypothesis and allostatic load (Geronimus et al., 2006; Simons et al., 2021), we predicted that individuals assigned to profiles with lower levels of stressor exposure would have relatively-advantaged demographic profiles (e.g., White race, high educational attainment, current employment, married or cohabiting, and low community disadvantage), as well as high social support and HRQoL. We hypothesized that the opposite pattern would emerge for adults classified onto profiles with high levels of stressor exposure (i.e., that they would have higher-risk demographic profiles, low social support and low HRQoL), and that any moderate profiles would be similarly moderate in terms of demographic risk factors, social support, and HRQoL.

Method

Participants and procedures

Study procedures for data collection were reviewed and approved by institutional review boards at the University of Illinois at Chicago and NORC at the University of Chicago. Participants were recruited through an address-based sampling frame. Supported by 2020 American Community Survey (ACS) data, research staff at NORC stratified all US ZIP codes into eight groups based on distributions of race/ethnicity, urbanicity, and education levels, seeking to ensure equivalent representation across these demographic dimensions. Simple random sampling of household addresses was completed within each of the eight strata, and during 2022 we mailed study recruitment packets to N = 101,200 addresses in order to identify those households that included middle-aged adult respondents. Each packet included a cover letter, a demographic response form (i.e., for determining whether the household included an adult in the qualifying age range), a modest incentive ($2), and a postage-paid return envelope. A total of n = 4925 households responded to the initial packet (i.e., a response rate of 4.9 %). Of those, n = 2500 households had one or more members who were eligible for participation, and subsequently received a second mailing with study instructions, research materials, and a $5 gift card. The instructions indicated that, in households with multiple members between the ages of 40 and 60 years, the questionnaire should be completed by the adult with the most recent birthday. All participants who returned a completed survey received an additional $10 gift card.

A total of N = 1581 adults aged 40–61 years (M = 49.81, SD = 6.49) completed surveys between February and November 2022 (74.4 % responded to the web-based survey). We removed n = 40 cases for straightlining and other forms of responding indicative of inaccuracy or inattentiveness, resulting in a final analytic sample of N = 1541 adults. Sample descriptive statistics are presented in the upper panel of Table 1.

Table 1.

Descriptive statistics.

Variable	N	M (SD) / %	Range
Male Gender	1364	35.5 %	0–1
Age	1541	49.81 (6.48)	40–61
White Race	1329	48.7 %	0–1
Income	1309	3.68 (2.61)	1–9
Educational Attainment (>high school)	1335	64.6 %	0–1
Employment Status	1337		1–5
Employed	830	62.1 %
Unemployed	253	18.9 %
Retired	90	6.7 %
Disabled	139	10.4 %
Student/Other	25	1.9 %
Relationship Status	1333		1–4
Married or Cohabiting	598	44.9 %
Single/Never Married	373	28.0%
Separated/Divorced	293	22.0%
Widowed/Other	69	5.2%
ES: Unemployment	1488	0.15 (0.26)	0–1
ES: Underemployment/Problematic Employment	1486	0.20 (0.20)	0–1
ES: Undesirable Living Situations	1501	0.22 (0.30)	0–1
ES: Home Ownership Problems	1503	0.14 (0.22)	0–1
ES: Inadequate Health Insurance	1494	0.21 (0.28)	0–1
ES: Inadequate Sick Time	1483	0.19 (0.35)	0–1
ES: Social Role Constraints	1482	0.24 (0.27)	0–1
CS: General Stressors	1473	0.54 (0.37)	0–1
CS: Isolation	1446	0.41 (0.49)	0–1
CS: Social Life	1456	0.31 (0.29)	0–1
CS: Residential	1446	0.35 (0.31)	0–1
CS: Health	1446	0.36 (0.31)	0–1
CS: Marriage or Relationship	1475	0.28 (0.34)	0–1
CS: Non-Relationship	1459	0.31 (0.41)	0–1
CS: Divorce or Separation	1456	0.16 (0.31)	0–1
CS: Parental/Family	1451	0.28 (0.33)	0–1
Community Disadvantage	1540	0.61 (0.87)	−1.17–4.78
Social Support	1410	4.60 (1.43)	1–7
PHRQoL	1289	15.04 (3.84)	6–20
MHRQoL	1304	19.41 (4.67)	6–27

Note. Descriptive statistics for the stressor scales are reported with the original variables prior to standardization. ES = Economic stressors, CS = Chronic stressors, PHRQoL = Physical health-related quality of life, MHRQoL = Mental health-related quality of life.

Measures

Unless noted as an exception, all variables were modeled as observed indicators. Scale scores were calculated only for participants who responded to a minimum of 75 % of the items. All measures were scored (with items reverse-coded as applicable) such that high values corresponded to high levels of the construct. Descriptive statistics are provided in the lower panel of Table 1.

Profile variables

Individual economy-related stressors.

Participants responded to a 43-item adaption of the 39-item Life Change Consequences of the Great Recession (LCCGR) Instrument (Richman et al., 2012). For each item, respondents indicated whether they had experienced the specific stressor during the past year (1 = yes or 0 = no/not applicable). The modified measure conceptually divided unemployment and underemployment (i.e., working in a job where one’s education or skills are at a higher level than the requirements of the job) items into separate subscales, and added four new items (i.e., difficulties with debts, received food stamps while employed, received other federal aid while employed, and concerns about having enough money for retirement) to the underemployment scale, which was renamed “underemployment and problematic employment.” The revised instrument consisted of seven subscales, specifically unemployment (4 items; sample: “Loss of job/unemployment”; α = 0.75, ω = 0.78), underemployment and problematic employment (16 items; sample: “Increased feelings of competition with fellow employees”; α = 0.84, ω = 0.84), undesirable living situations (5 items; sample: “Moving to a less expensive place to save money”; α = 0.77, ω = 0.77), home ownership problems (4 items; sample: “Problems with mortgage payments”; α = 0.54, ω = 0.54), inadequate health insurance (6 items; sample: “Lack of dental coverage”; α = 0.79, ω = 0.80), inadequate sick time (2 items; sample: “Having to work when sick because of lack of sick days”; α = 0.74), and social role constraints (6 items; sample: “Increased isolation because of concerns about money”; α = 0.75, ω = 0.78). Standardized subscale scores were used in the latent profile analysis.

Individual non-economic chronic stressors.

Participants also responded to 38 items on their experiences with non-economic chronic stressors over the past year. Items were drawn from the 51-item chronic stress inventory detailed in Turner et al. (1995). Respondents utilized a three-point response sale ranging from 0 (not true) to 2 (very true), which for these analyses we recoded to 0 (not true) and 1 (somewhat or very true) in order to align with the response scale of economic stressors measure¹. Questions represent nine domains of chronic stress, including five universal stressors, corresponding to general (3 items; sample: “You’re trying to take on too many things at once,”; α = 0.69, ω = 0.70), isolation (1 item: “You are alone too much”), social life (4 items; sample: “You don’t have enough friends”; α = 0.57, ω = 0.59), residential (4 items; sample: “Your family lives too far away”; α = 0.53, ω = 0.59), and health stressors (5 items; sample: “You take care of an aging parent almost every day”; α = 0.67, ω = 0.67). The remaining four role-specific domains corresponded to marriage or relationship (9 items; sample: “You have a lot of conflict with your partner”; α = 0.92, ω = 0.92), non-relationship (2 items; sample: “You wonder whether you will ever get married”; α = 0.73), divorce or separation (2 items; sample: “You don’t see your children from a former marriage as much as you would like”; α = 0.58), and parental/family stressors (6 items; sample: “You feel your children don’t listen to you”; α = 0.83, ω = 0.83). Standardized subscale scores were used in the latent profile analysis. Participants also responded to a single item assessing a perceived role-specific stressor stemming from not being a parent (“You wish that you could have children but you cannot”). We omitted this relatively infrequently-endorsed item from the analysis out of concern that it may assess negative affect about childlessness at midlife (e.g., grief about infertility, regret about voluntary childlessness in hindsight), making it distinct from the other retained chronic stressors.

Covariates

Demographic characteristics.

Demographic variables included male gender (0 = cisgender or transgender women, non-binary, gender queer, or other gender identification, 1 = cisgender or transgender men), age in years, white race/ethnicity (0 = American Indian or Alaska Native, Asian, Black or African American, Hispanic/Latino, Native Hawaiian or other Pacific Islander, or Multiracial; 1 = white), prior year household income (ranging from 1 = under $15,000, to 9 = over $120,000), educational attainment (0 = high school diploma or equivalent or less, 1 = completed education beyond a high school diploma or equivalent), employment status (1 = employed, 2 = unemployed, 3 = retired, 4 = disabled, 5 = student/other; for regressions, this was recoded to 0 = not employed and 1 = currently employed), and relationship status (1 = married/cohabiting, 2 = single/never married, 3 = separated/divorced, 4 = widowed/other; for regressions, this was recoded to 0 = not married or cohabiting and 1 = married or cohabiting).

Community disadvantage.

Respondents’ census tracts were used to match five indicators of community disadvantage collected via the U.S. Census Bureau (Krieger et al., 2002) – percents of individuals who were: living below the poverty line, unemployed, receiving public assistance, under the age of 18 years, and households headed by females. We standardized these five one-year estimates from 2020 across all census tracts at the national level before matching to the participants’ census tracts and other survey datapoints. We calculated an aggregate through averaging (α = 0.82). As a check on validity, we examined the correlation between this measure and the CDC/ATSDR Social Vulnerability Index summary score (Centers for Disease Control and Prevention & Agency for Toxic Substances and Disease Registry, 2022). The correlation was high (r = 0.71, p < .001), suggestive of construct validity.

Social support.

Using a modified version of the Social Support Network Inventory (Flaherty et al., 1983), participants responded to four items assessing the degree to which they received social support from four different sources (i.e., a spouse/significant other, co-worker, a friend from outside of work, and a relative). For each item (sample: “To what extent do each of these people give you emotional support by listening, talking, or just being with you?”), response options ranged from 1 (very little) to 7 (very much). Responses to the 16 items were averaged to form a total index of social support from all sources (α = 0.92).

Physical and mental HRQoL.

Participants completed the Short-Form Health Survey (SF-12; Ware et al., 1996), a multi-dimensional measure of HRQoL. Separate six-item subscales represented PHRQoL and MHRQoL in the last four weeks. The physical health items assessed general perceptions of overall health (rated from 1 = excellent to 5 = poor), physical limitations (rated from 0 = no, does not limit at all, to 2 = yes, limits a lot), problems with work or other daily activities as a result of physical health (yes/no), and the degree to which pain interfered with normal work (rated from 1 = not at all, to 5 = extremely). The mental health items assessed problems with work or other daily activities as a result of emotional problems (yes/no), feelings in the past month (e.g., “…felt down-hearted and blue?”) and the degree to which health or emotional problems interfered with social activities (rated from 1 = all of the time, to 5 = none of the time). Raw sum scores were calculated for analyses (Hagell et al., 2017), and internal consistency was acceptable (PHRQoL α = 0.84, MHRQoL α = 0.77).

Analysis plan

Hypotheses were tested through latent profile analysis in Mplus v.8.8 (Muthén & Muthén, 1998–2017). Models were fitted using maximum likelihood with robust standard errors (MLR) estimator, specifying 100 sets of starting values, 100 optimizations, and a maximum of 100 iterations allowed in the initial stage. Following established procedures, we estimated models with one to six profiles, adding one profile at a time in order to compare its fit relative to the model with one fewer class. We then selected the best-fitting model on the basis of several indicators of fit (Ferguson et al., 2020). These included the Akaike information criterion (AIC), Bayesian information criterion (BIC), and the sample-adjusted BIC (SABIC), which are all interpreted in the same fashion (i.e., models with lower values are preferable to those with higher values). We also examined the adjusted Lo-Mendell-Rubin likelihood ratio test (aLMR; i.e., a significant p-value on this test indicates that the model provides a better fit to the data relative to a model with one fewer class) and entropy (i.e., higher values represent better fit to the data, with 0.80 considered the minimal acceptable value). Finally, we also considered the size of the smallest profile (i.e., those containing less than 5 % of the sample may be spurious and should receive extra scrutiny in light of relevant theory), and the average posterior probabilities of assignment for all groups (i.e., a mean greater than 0.90 is acceptable; B. Muthén & Muthén, 2000). Subsequently, we compared the profiles for all covariates using chi-squares and ANOVAs, in which we corrected for multiple group comparisons through Bonferroni adjustments. This was followed by a series of multinomial logistic regressions in which we predicted profile membership from all covariates.

Results

Latent profile analyses

Following the process described above, we fitted latent profile models with one to six profiles (see Table 2). While the AIC, BIC, and SABIC supported the retention of the model with six profiles, the non-significant aLMR test indicated that the addition of the sixth profile did not significantly improve the model fit relative to the five-class solution. However, this same index indicated that the five-class solution was a better fit to the data than the four-class model. Other indicators of fit suggested that the four-, five-, and six-class solutions were acceptable (i.e., entropy values and the average posterior probabilities were each above the required thresholds, and the smallest profile was larger than 5 % of the sample). Considering all indicators of fit in conjunction, we retained the five-profile model on the basis of its better fit relative to the four- and six-profile solutions (see Fig. 1).

Table 2.

Fit statistics for the latent profile models.

Number of Profiles	LL	AIC	BIC	SABIC	aLMR	Entropy	Smallest Profile %	Average PP
1	−33,408.91	66,881.82	67,052.12	66,950.46
2	−30,841.88	61,781.77	62,042.54	61,896.87	<0.001	0.90	32.1 %	0.97
3	−30,100.08	60,332.16	60,683.41	60,473.74	<0.001	0.87	13.2%	0.94
4	−29,546.08	59,258.16	59,699.88	59,436.21	<0.001	0.92	9.3 %	0.95
5	−29,235.31	58,670.60	59,202.79	58,885.11	0.019	0.92	7.9 %	0.95
6	−28,970.99	58,175.98	58,798.64	58,426.96	0.349	0.89	6.7 %	0.93

Note. The retained model is denoted in bold font. LL = Loglikelihood, FP = Free parameters, AIC = Akaike information criterion, BIC = Bayesian information criterion, SABIC = Sample-adjusted BIC, CAIC = Consistent AIC, aLMR = Adjusted Lo-Mendell-Rubin likelihood ratio test, PP = Posterior probability.

Fig. 1.

Standardized means of the five profiles.

Note. P = Profile.

The largest profile (46.3 % of the sample) was characterized by low levels of all stressors (i.e., Profile 2, Low Stressors). The second largest profile (22.7 %) was characterized by low to average levels of economic stressors and elevated levels of chronic stressors (i.e., Profile 4, Mixed Stressors). The third largest profile (14.7 %) reported high levels of sicktime-related economic stressors, below average levels of unemployment stressors, and average to slightly above average levels of all other stressors (i.e., Profile 3, High Sicktime Stressors). The fourth largest profile (8.5 %) resembled the preceding profile on the economic and chronic stressors, except that they reported high levels of unemployment-related stressors and below-average levels of sicktime-related stressors (i.e., Profile 1, High Unemployment Stressors). Finally, the fifth profile was the smallest in size (7.9 %), and was characterized by high levels of all economic stressors and elevated levels of all chronic stressors (i.e., Profile 5, High Stressors).

Detailed results of mean comparisons are reported in Supplemental Table S1. With two exceptions, the Low Stressors profile reported lower levels of all stressors and the High Stressors reported higher levels of all stressors relative to all other classes. Exceptions involved sicktime stressors (i.e., profile means were equal for the Low and Mixed Stressors profiles) and isolation (i.e., profile means were equal for the High, High Unemployment, and the Mixed Stressors profiles). For the High Unemployment, High Sicktime, and the Mixed Stressors profiles, there were no significant differences for eight of the 16 variables. For the stressors with mean differences, the patterns were variable, consisting of instances in which 1) all three profiles differed significantly from each other (two variables: Unemployment and Inadequate Sicktime, defining the High Unemployment and High Sicktime Stressor profiles), 2) the High Unemployment and High Sicktime Stressors profiles were equal to each other and both were different from the Mixed Stressors profile (three variables: Underemployment/Problematic Employment, Inadequate Health Insurance, and Social Life), and 3) two profiles were equal to each other but only one differed from the third profile (three variables: Isolation, Non-relationship, and Parental/Family).

Covariates of profile membership

We compared the profiles for each covariate via chi-square tests and ANOVAs (see Table 3 and Figs. 2a-2f). These findings were comparable to those from multinomial logistic regressions predicting profile membership (see Supplemental Table S2). As such, below we interpret the results of the first set of comparisons.

Table 3.

Comparison of latent profiles: distributional and mean differences in covariates.

	M (SD) / %
Profile/Variable	P1: High Unemp. Stressors	P2: Low Stressors	P3: High Sicktime Stressors	P4: Mixed Stressors	P5: High Stressors	Significant Differences
Male Gender	33.3 %	38.0%	34.3 %	30.4 %	41.4 %
White Race Versus All Others	40.0%	53.6 %	43.4 %	50.3 %	33.7 %	2,4 > 5
Educational Attainment (>HS)	67.3 %	64.2%	62.9 %	70.1 %	51.6 %	4 >5
Employment Status
Employed	39.1 %	68.0%	85.7 %	48.1 %	46.8 %	3 > 2 > 1,4,5
Unemployed	55.5 %	13.4 %	5.6 %	18.4 %	41.5 %	1,5 > 2,4 > 3
Retired	0.0%	9.5 %	2.0%	7.9 %	1.1 %	2 > 1,3; 4 > 1
Disabled	2.7 %	7.4 %	5.1 %	23.1 %	7.4 %	4 > 1,2,3,5
Student/Other	2.7 %	1.6 %	1.5 %	2.5 %	3.2%
Relationship Status
Married or Cohabiting	30.9 %	56.0%	48.0%	32.4 %	25.0%	2,3 > 1,4,5
Single/Never Married	36.4 %	26.2%	23.0%	27.3 %	41.7 %	5 > 2,3
Separated/Divorced	23.6 %	13.7 %	24.5 %	34.3 %	27.1 %	3,4,5 > 2
Widowed/Other	9.1 %	4.1 %	4.6 %	6.0%	6.3 %

Note. P = Profile, Unemp. = Unemployment, HS=High School. Bonferroni adjustments were made for all pairwise comparisons.

Figs. 2.

a-2f. Profile mean differences in covariates.

Note. P = Profile, UN = Unemployment Stressors, PHRQoL = Physical health-related quality of life, MHRQoL = Mental health-related quality of life. Error bars denote SDs, and the values above each bar denote significant pairwise differences. Bonferroni adjustments were made for all pairwise comparisons.

The profiles did not differ in terms of gender treated as a dichotomous variable, χ²(4) = 7.18, p = .13, Cramér’s V = 0.07. As a supplemental analysis, we explored profile differences in terms of gender as a categorical variable, and results are provided in the upper panel of Supplemental Table S3. There were subtle differences in average age, such that the Low and Mixed Stressors profiles were older than those in the High Sicktime and High Stressors profiles, F(4, 1508) = 7.66, p < .001, η² = 0.02 (see Fig. 2a). There were similarly modest differences in race treated as a dichotomous variable, such that the Low Stressors profile was disproportionately White relative to the High Stressors profile, χ²(4) = 20.34, p < .001, Cramér’s V = 0.12. As a second supplemental analysis, we also explored profile differences with a categorical race variable, and these results are provided in the lower panel of Supplemental Table S3. In terms of annual household income, the Low Stressors profile had the highest, followed by the High Sicktime and Mixed Stressors profiles (who did not differ from each other), and the High Unemployment and the High Stressors profiles (who also did not differ), F(4, 1302) = 30.79, p < .001, η² = 0.09 (see Fig. 2b). Educational attainment differences were minimal, such that the Mixed Stressors profile was more likely to have completed education beyond high school relative to the High Stressors profile, χ²(4) = 11.78, p = .019, Cramér’s V = 0.09.

The profiles differed substantially in terms of employment status, χ²(16) = 270.53, p < .001, Cramér’s V = 0.23. The High Sicktime Stressors profile was more likely to report current employment relative to the Low Stressors profile, and individuals in both profiles were more likely to be employed than those in the three other classes, which did not differ from each other. There were parallel differences in unemployment, such that the High Unemployment and the High Stressors profiles were most likely to be unemployed, followed by the Low and the Mixed Stressors profiles, and finally, the High Sicktime Stressors profile. Concerning retirement, those in the Low and Mixed Stressors profiles did not differ, and each group included a larger percentage of retired respondents relative to the High Unemployment profile. For the Low Stressors profile only, this proportion was greater than in the High Sicktime Stressors class. Additionally, the Mixed Stressors profile had a higher percentage of disabled participants relative to all other profiles. The profiles did not differ in terms of other employment status.

The profiles also differed in terms of relationship status, χ²(12) = 101.61, p < .001, Cramér’s V = 0.16. The Low and the High Sicktime Stressors profiles were more likely to report being married or cohabiting than all three other classes. The High Stressors profile was more likely to report being single/never married relative to the Low and the High Sicktime Stressors profiles. The Low Stressors profile was also less likely to report being separated or divorced relative to the High Sicktime, Mixed and High Stressors classes. There were no group differences in widowhood/other relationship status.

There were subtle profile differences in community disadvantage, F(4, 1507) = 10.45, p < .001, η² = 0.03 (see Fig. 2c). The High Unemployment and High Stressors profiles were statistically equivalent, and each profile reported higher disadvantage relative to all other classes, which did not differ from each other.

There were significant profile differences in social support, F(4, 1403) = 37.40, p < .001, η² = 0.10 (see Fig. 2d). Respectively, the Low and High Stressors profiles reported higher and lower levels than all other classes. The three moderate profiles did not differ from each other.

For PHRQoL, the Low Stressors profile reported higher levels than all other profiles, who did not differ from each other, F(4, 1282) = 50.59, p < .001, η² = 0.14 (see Fig. 2e). For MHRQoL, the Low Stressors profile reported higher levels than all other profiles, F(4, 1297) = 85.43, p < .001, η² = 0.21 (see Fig. 2f). Additionally, the High Sicktime Stressors profile reported higher levels than the Mixed and High Stressors profiles, while the High Unemployment Stressors profile reported better MHRQoL than the High Stressors class.

Sensitivity analyses

Owing to the emergence of the High Sicktime Stressors profile, we conducted sensitivity analyses in order to ascertain whether this profile was disproportionately at-risk relative to the other profiles in terms of two additional covariates. We considered participants’ residence in Medicaid expansion states (coded 0 = residence in a non-expansion state, 1 = residence in an expansion state; 51.7 %) and whether they lacked health insurance coverage (coded 0 = has some form of public and/or private health insurance, 1 = uninsured; 16.3 %). The profiles differed in terms of residence in states with the Medicaid expansion, χ²(4) = 13.14, p = .01, Cramér’s V = 0.09. Those in the High Sicktime Stressors profile (42.8 %) were less likely to report residence in an expansion state relative to the Low (53.3 %) and Mixed Stressors profiles (55.1 %), who did not differ from each other. The High Unemployment (56.3 %) and High Stressors (44.2 %) profiles did not differ from any other classes. The profiles also varied in terms of the likelihood of being uninsured, χ²(4) = 46.57, p < .001, Cramér’s V = 0.18. Respondents in the High Unemployment (25.6 %), High Sicktime (24.5 %), and High Stressors (29.5 %) profiles were more likely to be uninsured than those in the Low (13.3 %) and Mixed Stressors profiles (9.6 %), who did not differ from each other.

We also explored whether the profiles differed in terms of reporting any debt (coded as 0 = no debt, 1 = any debt). The profiles differed in terms of reporting any debt, χ²(4) = 274.14, p < .001, Cramér’s V = 0.43. The group with the largest proportion of members reporting any debt was the High Stressors profile (73.5 %), followed closely by the High Sicktime (63.9 %) and High Unemployment (59.5 %) profiles, though these three groups did not differ significantly from each other. However, the Mixed Stressors profile (55.4 %) was significantly different from the High and Low Stressors (19.4 %) profiles, the latter of which included the smallest proportion of members reporting any debt. For descriptive purposes only, we explored profile differences in the amount of debt in US dollars (Raw M = $17,222.08, SD = $61,473.29, range = 0–1,158,000; after winsorizing outliers, M = $11,625.86, SD = $25,553.55, range = 0–100,000). Results of the analysis for amount of winsorized debt were comparable to those for any debt, F (4, 1382) = 23.47, p < .001, η² = 0.06 (see Supplemental Fig. S1). On average, the High Stressors profile reported higher levels of debt than all other profiles, followed by the Sicktime, Mixed, and High Unemployment Stressors profiles, who each reported higher levels of debt than the Low Stressors profile. We repeated the logistic regressions predicting profile membership with these three additional bivariate controls included, and the results of these models were generally comparable to those reported in Supplemental Table S2 and above (see Supplemental Table S4).

Discussion

Contemporary middle-aged adults are more stressed relative to older and younger adults, respectively (American Psychological Association, 2021, 2023), and demonstrate unique elevations in levels of financial and health-related stressors in accordance with the characteristic developmental tasks of midlife (Arnett, 2018). Seeking to address gaps in the literature about stressor exposure during this period of adulthood, the present study identified profiles of chronic stressors in a sample of GenX adults. Findings extended those of prior studies of stressor profiles conducted at earlier developmental periods, which to date have received greater emphasis in variable- and person-centered research on stress. While previous investigations have generally revealed three or four profiles, the present study identified five stressor classes, including three relatively distinct moderate stressor exposure profiles in addition to low and high stressor profiles. Notably, there was greater variability across the profiles in terms of economic stressors in comparison to the non-economic chronic stressors, and noteworthy variability in the levels of economic stressors within each profile except for the low stressor exposure class. Economic stressors have been neglected in previous studies, and the present findings illustrate the importance of their inclusion, particularly in light of theoretical assertions about the impacts of financial stressors for stress proliferation processes (Pearlin, 2010; Pearlin et al., 1981, 2005). Additionally, those with the lowest and highest stressor profiles were respectively the most and least demographically advantaged in terms of demographic risk factors, social support, and HRQoL, while those in the three moderate profiles varied inconsistently across the covariates. We interpret these findings below.

Profiles of stressors at midlife

With the caveat about the study’s cross-sectional design precluding the possibility of disentangling the causal chain of events, these findings reflect the core notions of the stress proliferation framework (Pearlin et al., 2005), in that all profiles with elevated stressors demonstrated high levels of multiple stressors, suggestive of possible “spillover” from an original stressor to others. Proliferation may be assumed for the High Stressors and for the three moderate stressor profiles (and perhaps to have not occurred for the Low Stressors profile), though there is certain to be variability in the chains of stressors experienced by individuals within each profile. Associations with covariates provide possible hints about antecedents. Below, we discuss each profile in turn.

Low stressors profile

Notably, the largest-sized profile was the Low Stressors profile, which was characterized by low levels of all stressors, echoing prior person-centered investigations in other samples with diverse measures of stressors (Andretta & McKay, 2019; D. Chen et al., 2024; Liao et al., 2018; Luk et al., 2023; Shaw et al., 2024). Consistent with expectations and prior studies, this profile demonstrated considerable advantage relative to the others in terms of income, debt, insurance coverage, educational attainment, employment, relationship status, and social support (Blais et al., 2022; Liao et al., 2018; Luk et al., 2023; Ta et al., 2017; Wolfe et al., 2022). Such privilege may directly translate into ready navigation of the developmental tasks of midlife, culminating in low exposure and/or susceptibility to stressors, low risk for stressor proliferation, advantageous positioning for coping with any stressors that arise, and consequently, better physical and mental health at midlife (Drentea & Reynolds, 2015; Thoits, 1982). For example, marriage is linked to lower stressor levels and greater social support (Ta et al., 2017; Vaingankar et al., 2020). Further, the majority of the sample’s retired participants were in this profile, and the economic privilege reflected in being able to retire is by itself an indicator of lower stressor exposure levels (Fonseca et al., 2024; Kuhn et al., 2021).

High stressors profile

In contrast, the smallest-sized profile was the High Stressors profile, which in line with some prior studies, demonstrated high levels of essentially all of the stressors (Blais et al., 2022; D. Chen et al., 2024; Luk et al., 2023). This profile was disproportionately disadvantaged relative to the others, suggesting distinct vulnerability for exposure to a broad range of chronic stressors (Thoits, 1982). Consistent with other investigations, relative to the comparatively less-stressed profiles, this profile was disproportionately comprised of individuals who identified as POC, and reported lower incomes, educational attainment, lower likelihood of current employment (though at levels equivalent to the High Unemployment Stressors profile) and marriage or cohabiting (though at levels equivalent to the High Unemployment and the Mixed Stressors profiles), and the worst debt (Andretta & McKay, 2019; Blais et al., 2022; Brown et al., 2020; Luk et al., 2023; Ta et al., 2017). Additionally, this profile was somewhat younger in age, a finding that is unique to this investigation but is generally consistent with evidence on financial stress (Loibl et al., 2022). Consistent with theoretical assertions, this profile was also more likely to reside in comparatively highly disadvantaged communities (Pearlin et al., 2005), which likely contributes to this subgroup’s demographic profile, particularly experiences with unemployment and being unpartnered. This accumulation of risk factors likely results in this group being the most vulnerable to the establishment and proliferation of economic stressors to non-economic stressor domains (Conger et al., 1999). Such expansion may flourish in the face of low levels of social support reported by this group, a finding that was consistent with a previous study (Liao et al., 2018): at midlife, those who are lacking romantic partners may be particularly at risk for low total support levels, as spouses/partners tend to provide higher levels of support than do other family members and friends, especially as social networks shrink with age (Gurung et al., 2003; Kalmijn, 2012). In the face of such high levels of stressors, demographic risk factors, and low social support, it is perhaps unsurprising that the High Stressors profile is worse off than the Low Stressors profile in terms of PHRQoL, and that this profile is particularly disadvantaged relative to the others in MHRQoL (Blais et al., 2022; D. Chen et al., 2024; Liao et al., 2018; Luk et al., 2023; Shaw et al., 2024). That this profile consists of disproportionately younger respondents who identify as POC while reporting comparatively poorer health lends additional support for the weathering hypothesis (Geronimus et al., 2006; Simons et al., 2021): while not measured in the present investigation, discrimination and systematic marginalization and inequality are plausible mechanisms linking demographic covariates, accumulated experiences with stressors, and health outcomes.

Moderate/mixed profiles

The three moderate/mixed economic stressors profiles were collectively better off than the High Stressors profile and worse off than the Low Stressors profile for both stressors and covariates, but variations between these three profiles were comparatively subtle and less consistent in comparison. These three profiles did not differ statistically for three economic (i.e., undesirable living situations, home ownership, and social role constraints) and four non-economic chronic stressors (i.e., general, residential, health and divorce/separation stressors), but diverged fully for two stressors (i.e., unemployment and sicktime – the defining stressors of the High Unemployment and High Sicktime Stressors profiles, and the only two indicators for which these two groups significantly differed). For the remaining seven stressors, two profiles were equivalent while the third was distinct. In particular, for underemployment and inadequate health insurance, the High Unemployment and High Sicktime Stressors profiles were equal to each other while both were higher than the Mixed Stressors profile, suggesting that stressors related to job quality and inadequate benefits are particularly salient for those in these two profiles. For the remaining five chronic stressors, the Mixed Stressors profile was higher than both the High Unemployment and High Sicktime Stressors profiles (i.e., isolation and non-relationship), higher than the High Sicktime Stressors profile only (i.e., social life), or higher than the High Unemployment Stressors profile only (i.e., marriage or relationships, and parental/family). This suggests that chronic stressors associated with interpersonal relationships best distinguished members of the Mixed Stressors profile from those in other profiles. Regarding covariates, differences between these three moderate profiles were relatively minimal, and did not follow a consistent pattern. Ultimately, while these profiles were comparable on social support (i.e., a key protective factor), debt, and PHRQoL, their divergent economic and interpersonal relationship-related vulnerabilities may trigger distinct stressor profiles, hinting at separate proliferation pathways (Pearlin et al., 1997).

High unemployment stressors profile.

The High Unemployment Stressors profile (i.e., the second smallest profile) was unique to the present investigation, and was defined in terms of high unemployment stressor exposure within the past year. Like the High Sicktime Stressors profile, this profile had elevated stressor exposure related to underemployment or problematic employment, as well as inadequate insurance, which may itself be a symptom of unstable employment and low income levels (Tolbert et al., 2023). Simultaneously, the High Unemployment Stressors group was more disadvantaged relative to the two other moderate profiles. In particular, this profile was associated with lower incomes and higher community disadvantage relative to both other moderate profiles, and individuals in this profile were less likely to be married/cohabiting in comparison to those in the High Sicktime Stressors profile. These variations in covariates could independently or collectively explain the observed elevations in discrete stressors that are manifest in the High Unemployment Stressors profile. For example, living in highly disadvantaged settings heightens risk for unemployment and low income, which can each lead to economic stressors that impact family relationships (Pearlin et al., 1981). Community disadvantage is also a risk factor for singlehood (South & Crowder, 1999), which could explain elevations in loneliness and non-relationship stressors reported by this group. Romantic partners may also act as an additional safety net in the context of economic stressor exposure (e.g., job loss may not be as devastating if another adult in the household is working for pay). These stressors notwithstanding, that this profile did not differ from either of the other moderate stressor profiles in terms of HRQoL suggests equifinality despite variability across individual stressors and risk factors.

While those in the High Unemployment stressors profile reported fewer stressors than the High Stressors group, these variations in exposure do not appear to be the result of differential demographic risk or physical health, as these two groups were comparable in terms of annual household incomes, educational attainment, current employment and marriage/cohabitation, community disadvantage, and PHRQoL. Notwithstanding, the High Unemployment Stressors profile was advantaged in terms of social support and MHRQoL: these adults may have access to resources for coping with stressors, and modulating stressor levels may slow the proliferation process, ultimately helping to support resiliency (Wheaton, 2009). Individuals in the High Unemployment Stressors profile may be at an earlier stage of the stress proliferation process relative to those in the High Stressors group, and it is possible that over time, the stressor exposure levels of those in the former profile will catch up to the latter (i.e., an explanation that cannot be explored in these cross-sectional data). Additionally, these two profiles may differ meaningfully on crucial antecedents that were not measured in this study, such as experiences of adversity in early childhood (Arpawong et al., 2022). These possibilities warrant consideration in suitable datasets.

High sicktime stressors profile.

The High Sicktime Stressors profile was the third largest profile, and was also unique to the present investigation. This group reported exceptionally high levels of stressor exposure related to inadequate sicktime at work, elevated levels of underemployment and problematic employment, and in insurance stressors that were equivalent to those in the High Unemployment Stressors class, and elevated levels of parental-family stressors that were equivalent to the Mixed Stressors class. Despite appearing to be under some strain (i.e., manifest as lower social support and poorer MHRQoL) relative to the Low Stressors profile, economic disadvantage did not seem to be the reason for that distress, as in some ways, this profile appeared to be relatively advantaged. In particular, this profile reported the highest rates of current employment of all classes, as well as moderately high rates of marriage/cohabiting and elevated income levels relative to comparatively higher-stressor-exposure classes. Simultaneously, this profile was somewhat younger than the Low Stressors and the Mixed Stressors profiles. Interestingly, the findings for HRQoL do not clearly suggest that the individuals in this profile have health difficulties that would result in greater need for sicktime, which may thusly increase exposure to that stressor (Boot et al., 2011). On average, adults in this profile were relatively healthy, reporting lower PHRQoL and MHRQoL relative only to the Low Stressors class. This profile’s elevated levels of parental/family stressors suggest that work-family conflict (e.g., having to use sicktime or unpaid days off in order to meet children’s needs or to care for aging and/or ill relatives) may contribute to individuals’ perceptions of sicktime as being inadequate (Byron, 2005), a possibility that cannot be tested using these data. Further seeking to understand this unexpected class, we conducted additional exploratory analyses of profile-level differences in the underemployment and problematic employment items, which revealed that this profile was more likely to report experiencing lost income (e.g., pay cuts, lost hours, unpaid furloughs, and no raises or bonuses) and increasing workloads due to attrition amongst coworkers. It may be that insufficient sicktime is itself a symptom of such problematic or precarious employment circumstances, which could also be intertwined with work-family conflict.

Mixed stressors profile.

Finally, the mid-sized Mixed Stressors profile was also unique to the present investigation. As with the other moderate stressor exposure profiles, this profile had more frequent stressor exposure than the Low Stressors profile on most of the economic stressors (i.e., exceptions were the unemployment and sicktime dimensions) and on all non-economic chronic stressors. In turn, this profile reported less frequent exposure than High Unemployment and High Sicktime Stressors profiles on those key indicators, as well as on inadequate health insurance, and underemployment and problematic employment stressors. This profile reported comparatively more stressors than either or both of those profiles for five of the chronic stressors involving close social relationships (i.e., being socially isolated, being unable to find a life partner, lacking fulfillment in their romantic relationships, and experiencing strain due to separation or divorce). Further, this profile had lower average incomes, moderate debt, low rates of marriage/cohabiting and current employment, and comparatively higher likelihood of disability, all of which represent sources of vulnerability to stressors (Avison & Turner, 1988; Blais et al., 2022; Luk et al., 2023; Okoro et al., 2021; Ta et al., 2017). For this profile, it is particularly difficult to speculate about pathways of stress proliferation in the absence of longitudinal data, though two possibilities seem feasible. First, experiencing specific chronic social stressors may have triggered a subset of economic stressors rather than the converse. That approximately two-thirds of this group were unpartnered may explain why this profile reported low average incomes (i.e., a larger proportion of this profile were likely in single-earner households), and elevated levels of relationship-related non-economic chronic stressors. Thus, it follows that this profile reported lower levels of social support (Stronge et al., 2019), which may contribute to their relatively poorer HRQoL: individuals in this profile reported poorer mental and physical HRQoL than the Low Stressors profile, and worse MHRQoL than the High Sicktime Stressors profile. These differences suggest the second possible explanation, in that health problems could antecede both sets of stressors (Houle & Keene, 2015). In other words, rather than certain stressors directly causing poor health, suboptimal health may trigger the proliferation of other stressors by giving rise to disability, consequential unemployment, and low income paired with high medical bills and lost health insurance. These in turn lead to further economic stressors, such as mortgage foreclosure, and to their sequelae, such as strained social relationships (Houle & Keene, 2015; Smith, 2013; Terrill & Molton, 2019). Notably, 51.1 % of all individuals in the sample who identified as disabled were classified into this profile. This alone may explain this profile’s relatively poor MHRQoL, in that the impact of chronic stressors on mental health tends to be more detrimental for individuals with disabilities relative to their peers who are not disabled (Avison & Turner, 1988). Individual variability seems likely (i.e., some individuals may experience one proliferation pathway while others follow another), as do the possibility of bidirectional associations between risk factors, chronic stressors, and health: for example, divorce is a risk factor for disability and for growth in economic stressors in women (Lin & Brown, 2021; Tamborini et al., 2016). Longitudinal data are required in order to explore the various experiences that culminate in this profile.

Contributions, limitations, and future directions

As is true of all investigations, ours was characterized by several limitations. Our study’s cross-sectional, single-informant design was consistent with other person-centered investigations into stressor profiles, and thus this investigation was similarly constrained as are prior works in this area. Self-report surveys are the norm for research on individuals’ self-perceptions of stressors, which cannot be readily assessed through other approaches (e.g., observation). As discussed above, longitudinal data may grant new insights into proliferation pathways, transitions between stressor profiles over time, and into the long-term physical, mental, and behavioral health impacts of exposure to profiles of economic and non-economic chronic stressors, as has already been evidenced in the debt literature (Sun & Houle, 2020).

Other limitations stem from unique practical barriers imposed by the study’s design. In particular, our data derive from a national sample, but not a nationally representative sample, of middle-aged adults in the US. The sampling strategies reflect the investigators’ intentions of comparing Black and White adults in rural and urban/suburban contexts. Funding limitations and practical challenges imposed by the COVID-19 pandemic created complications that resulted in the eventual recruitment of a sample that included relatively small groups of adults of other ancestries. The study’s low baseline response rate is also suboptimal, though this investigation is far from unique in encountering this limitation around the time of data collection (Rothbaum & Bee, 2021; Stedman et al., 2019). The potential for nonresponse bias cannot be ruled out, particularly for adults who are socioeconomically disadvantaged and have the highest levels of chronic stress (Barr et al., 2008; Enzenbach et al., 2019), which may have implications for the size of the High Stressors profile (i.e., it may be an underestimation). Further, while our use of a binary response scale was consistent with previous studies involving the focal measure of economic stressors (Richman et al., 2012), its usage precludes a finer-grained analysis of stressor frequency or intensity, which may consequently have obscured meaningful variations in individuals’ experiences with economic stressors over the prior year. As noted above, we initially employed the recommended three-point response scale for the items assessing non-economic chronic stressors (Turner et al., 1995), and additional analyses revealed that there was no meaningful additional value conferred through the retention of the original metric of the non-economic chronic stressors subscales, supporting our decision to present findings involving subscales computed using dichotomous indicators. Relatedly, survey length limitations precluded our consideration of other individual-level covariates of stressor profiles. For future studies, it would be prudent to consider adverse childhood experiences (i.e., ACES) and personality-level factors, as these can influence all stages of the stress process, which may be particularly informative for understanding antecedents of the three moderate stressor profiles and their health outcomes (Mosley-Johnson et al., 2021; Williams et al., 2011). Additionally, the survey’s employment status item did not include “disabled” as a response option, and we derived this from open-ended responses clarifying the “other employment” option. Thus, a larger percentage of the sample may be disabled but categorized as unemployed. Providing this response option is an important modification for future studies.

In sum, the present study provides new evidence of the existence of distinct stressor profiles in middle adulthood, some of which replicate prior studies conducted with other age groups and types of stressors. Consistent with the SDH model (e.g., Braveman & Gottlieb, 2014), our data suggest that sociodemographic advantage appears to translate into lower levels of chronic stressors, which in turn may convey further benefits for physical and mental HRQoL. In contrast, socioeconomic disadvantage has a parallel association with greater exposure to chronic stressors, culminating in relatively poor HRQoL. While different configurations of moderate stressors emerged, the degree to which these profiles varied systematically relative to each other was comparatively limited, and in some instances, membership in a moderate profile was not disadvantageous relative to the Low Stressors profile, nor advantageous in comparison to the High Stressors class. Overall, the presence of factors that may protect against elevated stressor levels (e.g., being employed or partnered) may produce circumstances that give rise to high stress for some, but not all middle-aged adults. Notwithstanding, and in alignment with the SDH model, the present study’s findings illustrate the potential value of diverse social policies intended to reduce socioeconomic inequality (e.g., funding for higher education, universal basic income, etc.), in addition to those supporting health and wellbeing (e.g., universal healthcare).

These findings further illustrate the vital importance of interventions that target population health and systematic socioeconomic inequality by helping individuals and families cope with chronic stressors. There is a persistent need for comprehensive social and economic policies that may directly improve the well-being of adults at midlife, which may additionally convey distal benefits to their offspring (Burak, 2019; Waldfogel et al., 2019). As noted by Braveman and Gottlieb (2014), socioeconomic status is a key factor that shapes population health. Our data show that POC were more likely to be classified into the High Stress profile, where each of the chronic economic stressors (along with several of the non-economic chronic stressors) were significantly more common than in the other profiles. Such persistent inequities in stressor exposures that impact the health of POC illustrate the vital importance of support for social movements that address systemic racism and racial discrimination. Other potential changes to reduce inequities might occur at the political level, by improving the social safety net in order to diminish the deleterious effects of social stressors linked to job loss, precarious employment, and disability (Case & Deaton, 2020). One important federal change would be implementing universal health care coverage to fully eliminate the need for employer-sponsored health insurance, and thus mitigate health-related costs of unemployment or of precarious forms of employment that lack health care benefits. Likewise, strengthening collective bargaining rights for workers could help raise job standards, including better pay and benefits, thus reducing some of the main drivers of economic stress for workers. At the community, state, or federal level, job training programs for adolescents and adults could provide needed skills for workers to improve employability, and financial literacy education or interventions and other protections may prevent youth from accumulating excessive debt long before midlife (Gerrans, 2021; Martinchek et al., 2022). More generous paid leave policies, at the workplace, state, or federal level, could mitigate a big source of chronic stress for caregivers by allowing the protected time needed to devote to caregiving activities without the fear of losing one’s income or job. In particular, government programs which provide unpaid rather than paid leave do not help the most economically disadvantaged groups who lack an employed spouse or other financial resources to facilitate taking unpaid time off from work (Umberson & Karas Montez, 2010). Community programs to improve the availability of affordable housing could also reduce a major source of economic stress. Clearly, addressing socioeconomic inequality is a challenging task, but one that also holds many opportunities for intervention within individual workplaces, in communities, and at the state and federal levels. In 2023, the White House published a report which outlined key federal actions that can be taken to address social and environmental challenges that Americans face (Domestic Policy Council, Office of Science and Technology Policy, 2023). We urge policy makers and politicians to continue to pursue these efforts, to reduce chronic stressors and improve U.S. population health.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.appdev.2025.101887.

Data availability

Data for participants who agreed to data sharing will be made available through the National Institute on Alcohol Abuse and Alcoholism Data Archive.