Abstract
In the present study, we examined the prospective associations of both spousal support and spousal strain with a wide range of health and well‐being outcomes in married older adults. Applying the analytic template for outcome‐wide designs, three waves of longitudinal data from the Health and Retirement Study (n = 7788, M age = 64.2 years) were analyzed using linear regression, logistic regression, and generalized linear models. A set of models was performed for spousal support and another set of models for spousal strain (2010/2012, t1). Outcomes included 35 different aspects of physical health, health behaviors, psychological well‐being, psychological distress, and social factors (2014/2016, t2). All models adjusted for pre‐baseline levels of sociodemographic covariates and all outcomes (2006/2008, t0). Spousal support evidenced positive associations with five psychological well‐being outcomes, as well as negative associations with five psychological distress outcomes and loneliness. Conversely, spousal strain evidenced negative associations with three psychological well‐being outcomes, in addition to positive associations with three psychological distress outcomes and loneliness. The magnitude of these associations was generally small, although some effect estimates were somewhat larger. Associations of both spousal support and strain with other social and health‐related outcomes were more negligible. Both support and strain within a marital relationship have the potential to impact various aspects of psychological well‐being, psychological distress, and loneliness in the aging population.
Keywords: marital relationships, marriage, older adulthood, outcome‐wide design, social relationships, well‐being
In the last century, increases in life expectancy have led to rapid increases in the number of older adults, especially in developed countries where life expectancy is approximately 78 years for men and 83 years for women (OECD, 2021). In this rapidly growing group of aging adults, improving quality of life, and not only longevity, is a public health challenge. During the transition from adulthood to older ages, individuals face a series of physical, psychological, and social changes. For example, socioemotional selectivity theory posits that as people age, their time horizons shrink. This shift in time perception influences motivational priorities. When time is perceived as limited (often the case in older adulthood), older adults place greater value on having emotionally satisfying experiences rather than the expansion of social networks, or acquisition of new knowledge (Carstensen et al., 1999). As older adults experience important transitions that shape the extent of their social networks (e.g., children moving out and away from home, retirement), they often begin to invest more in their relationship with their romantic partner as a key source of emotional support (Fuller et al., 2020). Given the relevance of spousal relationships at later stages of life, understanding the benefits of a supportive spousal relationship in conjunction with the possible harmful consequences of a strained spousal relationship may be important for promoting the well‐being of the older adult population (Carstensen et al., 1999).
There is considerable evidence that positive and negative spousal relationships can affect well‐being in later life. Having a supportive spouse is associated with longitudinal decreases in feelings of loneliness, marital stress and tension, and depression, and with longitudinal increases of people's sense of meaning in life and healthy behaviors (Brazeau & Lewis, 2021; Brock & Lawrence, 2008; Gustavson et al., 2016; Han et al., 2019; Hudson et al., 2020; Saenz, 2021). Conversely, spousal strain has been shown to lead to marital tension, health risks, and initiating unhealthy behaviors (Chen & Feeley, 2014; Segel‐Karpas & Arbel, 2022; Uchino et al., 2014). In fact, according to the combined positivity and negativity effects model for the effects of spousal interactions, there is evidence that the effects of spousal support and strain tend to be of similar size but different directions for both positive and negative outcomes (Ingersoll‐Dayton et al., 1997). These parallel effects underscore the importance of studying both spousal support and strain together.
One construct that ought to be studied separately from spousal support is marital closeness. While spousal support can be defined as a supportive behavior performed for one spouse by the other (Goldsmith, 2004), marital closeness refers to how close the relationship between spouses is, regardless of direct spousal support or strain between them. Similar to spousal support, marital closeness has been shown to relate to better well‐being (Burke et al., 2019; Mancini & Bonanno, 2006). However, in some circumstances, high levels of marital closeness can be detrimental because they may amplify the debilitating effects of stressors, such as a partner's illness, on one's own psychological well‐being (Polenick et al., 2015).
Close relationships could be a consequence of previous spousal support, but a close relationship also increases opportunities to show commitment to the relationship through supportive acts or words. In the same sense, a relationship that is “too close” could be either the cause or the consequence of spousal strain. For this reason, it is worth studying the effect of spousal support or strain independently of the closeness level of the relationship.
Research on spousal relationships underscores the importance of the quality of the relationship for supporting individual well‐being; however, a challenge arises when we consider the behavioral theory of marriage. This theory suggests that over the life course, couples develop patterns of behavior and interacting that can either strengthen their emotional bonds or create tension across time (Gottman, 1993). For instance, couples who exhibit higher levels of positivity and less conflict not only report greater marital satisfaction but also tend to perceive more positive experiences within their marriage and more frequently disregard negative experiences (known as positive sentiment override) because the positive environment overrides momentary negative experiences (Meunier & Baker, 2012). The opposite can also happen among couples with a more antagonistic relationship, in which previous spousal strain may heighten an emphasis on negative interactions when evaluating their marriage (Birditt et al., 2010; Karney & Bradbury, 1995). These altered perceptions can then influence the dynamics of future interactions between partners, making it difficult to study the isolated effects of supportive or strain‐based interactions. Furthermore, when there is a history of accumulated strain between spouses, the marital discord model of depression suggests that a low quality marital relationship can trigger psychological distress. This onset of depression symptoms can itself exacerbate spousal strain, perpetuating the cycle (Beach et al., 1990).
Although previous research has provided an important foundation for understanding how spousal relationships can affect individual well‐being, there are several gaps in the existing empirical literature. First, because spousal relationship factors and individual well‐being are bidirectionally related, longitudinal analyses are needed to address the potential for reverse causality. The usual approach in longitudinal studies is to examine associations between one exposure at baseline and one outcome of interest assessed at a subsequent wave. However, this approach prevents us from excluding the possibility that low levels of spousal support or strain could result from previous marital dysfunction (Offer, 2021). For example, a couple's sleep problems or a spouse's phubbing habits (ignoring a spouse to pay attention to a mobile device) may result in conflict, ultimately decreasing marital satisfaction (Chen, 2018; Roberts & David, 2016). Adjusting for sleep problems, phubbing, and marital satisfaction measured before baseline spousal strain could address these concerns about reverse causation (for a more detailed explanation, see Supporting Information). Second, most prior studies on the associations of spousal support or strain with well‐being outcomes tend to focus on one or a few outcomes on some domains of well‐being and not others. If one or a few outcomes are reported in different studies, it can be challenging to compare results across studies because analytic choices are likely to vary across studies. This approach also likely hinders the publication of null results. These factors can impact the speed with which the empirical literature on a given topic advances, with a bias toward significant results that may not provide a complete picture of the linkages between spousal support or strain and well‐being. Such issues could be mitigated by adopting a wider scope in which many well‐being outcomes across various domains of functioning are examined.
The present study
This present study builds on some of the abovementioned gaps in the existing literature to examine the associations of both spousal support and spousal strain with a wide range of subsequent health and well‐being outcomes in a national sample of older U.S. adults. We generally expected that spousal support would be associated with better subsequent health and well‐being, whereas spousal strain would be associated with worse subsequent health and well‐being. However, we anticipated that there would be some variation in the strength of associations for each exposure across the outcomes.
METHOD
The current study used the analytic template for outcome‐wide longitudinal designs, which is a rigorous approach for estimating potential causal effects of a given exposure on multiple outcomes in the same sample (Vanderweele, 2017; VanderWeele et al., 2020). Not only does this analytic template enable investigators to directly compare effect sizes across various outcomes, but it can also play an important role in shaping interventions and health policies.
Study sample
Data for this study were taken from the Health Retirement Study (HRS; Sonnega et al., 2014), a prospective national cohort of U.S. adults aged 50 years or older (see https://hrs.isr.umich.edu/about). Beginning in 2006, the HRS provided mail‐in questionnaire surveys to 50% of study participants who were randomly selected to complete a range of psychosocial measures. The other 50% of HRS participants completed the same questionnaire in the next wave in 2008. The two sub‐cohorts alternate reporting so that all participants report psychosocial data every 4 years. The psychosocial questionnaire response rate was 87.7% in 2006 and 83.7% in 2008. In the present study, data from both sub‐cohorts were combined to increase sample size and power. Data from the 2006/2008 (pre‐baseline wave, t0), 2010/2012 (baseline wave, t1), and 2014/2016 (outcome wave, t2) waves were used. The starting sample for the current study was 14,664 participants who were eligible for the psychosocial questionnaire at t1. Participants who had their first interview at t1 were excluded (1666 participants excluded, n = 12,998), as well as those who were not married at t0 and t1 (5210 excluded, remaining sample = 7788).
The Health and Retirement Study has been approved by several ethics' committees, including the University of Michigan IRB. Further, informed consent was obtained from all HRS respondents. Besides, this study was exempt from additional review by the Institutional Review Board at the University of British Columbia because data were publicly available and de‐identified.
Measures
Outcomes, exposures, and covariates used in this study are summarized below (for additional details, see Supporting Information).
Exposures
The exposures were spousal support and spousal strain, which were assessed at t0 and t1 using items addressing how participants perceived their social interactions with their spouses. Spousal support was measured using three items: (1) “How much do they understand the way you feel about things?;” (2) “How much can you rely on them if you have a serious problem?;” and (3) “How much can you open up to them if you need to talk about your worries?”. The estimated internal consistency of the spousal support variable at t1 was α = 0.79. Spousal strain was assessed using four items: (1) “How often do they make too many demands on you?;” (2) “How much do they criticize you?;” (3) “How much do they let you down when you are counting on them?;” and (4) “How much do they get on your nerves?”. All seven items were rated using a four‐point response scale ranging from 1 (A lot) to 4 (Not at all). Responses were reverse‐coded and averaged to create separate indices for spousal support and spousal strain, with higher values indicating more spousal support or spousal strain. The estimated internal consistency of the spousal strain variable at t1 was also α = 0.79. For the purpose of the analysis, both spousal support and spousal strain variables were split into tertiles.
Outcomes
Thirty‐five outcome variables assessing different aspects of health and well‐being were taken from t2, including variables addressing physical health (all‐cause mortality, number of chronic conditions, diabetes, hypertension, stroke, cancer, heart disease, lung disease, arthritis, overweight/obesity, physical functioning limitations, cognitive impairment, chronic pain, sleep problems, and self‐rated health); health behaviors (heavy drinking, smoking, and frequent physical activity); psychological well‐being (positive affect, life satisfaction, optimism, purpose in life, mastery,1 health mastery, and financial mastery); psychological distress (diagnosis of depression, depressive symptoms, hopelessness, negative affect, and perceived constraints); and social factors (loneliness, living with a spouse/partner, and number of social contact with children, other family or friends). Additional details about measures that were used to assess the outcomes can be found in Supporting Information.
Covariates
The pre‐baseline (t0) levels of all outcomes and the exposures were used as covariates in all statistical models. In addition, the pre‐baseline information (t0) of the following covariates was used for our main analysis: age, gender, race, annual household income, quintile of total wealth, highest level of education, employment status, health insurance, geographic region, antecedents of childhood abuse by parents, religious service attendance, and the big five personality traits. Details about the covariates can be found in Supporting Information.
Missing data
We used the multiple imputation by chained equations (m = 5) to impute missing data on all variables (Azur et al., 2011).
Statistical analysis
Statistical analyses were performed using STATA, version 17 (StataCorp, 2021). Tests of statistical significance were two‐sided. To describe the associations between spousal support, spousal strain, and marital closeness,2 Pearson correlation coefficients were estimated at t1.
Two sets of multivariable regression models were performed to estimate the associations of spousal support and spousal strain at t1 on the outcomes at t2, one set for each exposure. To reduce the risk of reverse causation and potential mediation effects, all models of our main analysis adjusted for t0 values of all the outcomes, of the respective exposure, and of covariates (VanderWeele et al., 2016).
A separate model was estimated for each outcome. For binary outcomes with a prevalence of ≥10%, a generalized linear model with a log link and Poisson distribution was used. For less prevalent outcomes, a logistic regression model was performed. For continuous outcomes, linear regression was used. To facilitate the comparison of effect sizes, standardized beta coefficients are reported for continuous outcomes. To test the linear trend for all associations across tertiles, both spousal support and spousal strain were modeled as continuous variables instead of tertiles. In our tables, and for ease of reviewing results, we present multiple p value cutoffs (both with and without Bonferroni correction for multiple testing) because different investigators use different threshold standards for interpreting evidence based on current norms in their specific discipline.
Sensitivity analysis
E‐values were calculated for each association to evaluate the robustness of the results to potential unmeasured confounding (VanderWeele & Ding, 2017). E‐values indicate the minimum level of association (on the risk ratio scale) that an unmeasured confounder would need to have with both the exposure and the outcome to fully explain away the observed association between the two variables, conditional on the measured covariates (VanderWeele & Ding, 2017).
Additionally, we performed the following sensitivity analyses. First, the main analyses were repeated after omitting participants with health conditions at t0. In this way, we tried to look for the risk of developing these conditions among people who did not have the condition at baseline. Second, the above‐mentioned sensitivity analyses were repeated after omitting t0 values of the exposures and outcomes (i.e., only adjusting only for t0 sociodemographic variables). Third, we further replicated the main analyses using complete‐case data instead of imputing missing data. Fourth, we repeated the main analyses by adding either spousal strain at t0 (in models with spousal support as the exposure) or spousal support at t0 (in models with spousal strain as the exposure) as a covariate. Finally, the fourth sensitivity analyses were replicated by adding marital closeness at t0 as a covariate together with espousal strain and spousal support at t0.
RESULTS
Sample characteristics
Characteristics of participants at t0 by spousal support and spousal strain tertiles at t1 can be seen in Table 1. The mean age of the participants was 64.2 years (SD = 9.2). More than 50% of the sample had hypertension, arthritis, overweight, or obesity at t0. Compared with the lowest tertile of spousal support, the highest tertile had a lower proportion of women, higher annual household income, a higher proportion of Caucasians, higher religious service attendance, lower prevalence of depression diagnosis, lower depressive symptom scores, as well as higher scores in life satisfaction, health mastery, and financial mastery.
TABLE 1.
Characteristics of participants at pre‐baseline by tertiles of baseline spousal support (N = 7622) a , b , c and spousal strain (N = 7619). a , b , c
| Participant characteristics | Spousal support | Spousal strain | ||||
|---|---|---|---|---|---|---|
| Tertile 1 (n = 2890) | Tertile 2 (n = 1581) | Tertile 3 (n = 3151) | Tertile 1 (n = 2778) | Tertile 2 (n = 2190) | Tertile 3 (n = 2651) | |
| Sociodemographic factors | ||||||
| Age (year; range: 46–94), mean (SD) | 63.5 (9.0) | 64.5 (9.1) | 64.5 (9.2) | 64.7 (9.2) | 64.2 (9.0) | 63.4 (9.1) |
| Female (%) | 1734 (60.0) | 746 (47.2) | 1326 (42.1) | 1321 (47.6) | 1060 (48.4) | 1425 (53.8) |
| Race/Ethnicity (%) | ||||||
| White | 2175 (75.3) | 1260 (79.7) | 2549 (80.9) | 2287 (82.4) | 1758 (80.3) | 1938 (73.1) |
| Black | 333 (11.5) | 141 (8.9) | 240 (7.6) | 202 (7.3) | 200 (9.1) | 312 (11.8) |
| Hispanic | 307 (10.6) | 138 (8.7) | 278 (8.8) | 229 (8.3) | 186 (8.5) | 306 (11.5) |
| Other | 73 (2.5) | 42 (2.7) | 83 (2.6) | 58 (2.1) | 45 (2.1) | 95 (3.6) |
| Annual household income (%) | ||||||
| <$50,000 | 1105 (44.9) | 579 (42.4) | 1078 (40.5) | 986 (41.2) | 784 (41.8) | 990 (44.6) |
| $50,000–$74,999 | 486 (19.8) | 288 (21.1) | 504 (18.9) | 460 (19.2) | 378 (20.1) | 440 (19.8) |
| $75,000–$99,999 | 329 (13.4) | 167 (12.2) | 348 (13.1) | 309 (12.9) | 232 (12.4) | 302 (13.6) |
| ≥$100,000 | 541 (22.0) | 331 (24.3) | 733 (27.5) | 636 (26.6) | 483 (25.7) | 486 (21.9) |
| Total wealth (%) | ||||||
| 1st Quintile | 590 (24.0) | 237 (17.4) | 471 (17.7) | 402 (16.8) | 356 (19.0) | 536 (24.2) |
| 2nd Quintile | 500 (20.3) | 291 (21.3) | 507 (19.0) | 503 (21.0) | 357 (19.0) | 441 (19.9) |
| 3rd Quintile | 481 (19.5) | 296 (21.7) | 522 (19.6) | 452 (18.9) | 393 (20.9) | 453 (20.4) |
| 4th Quintile | 463 (18.8) | 263 (19.3) | 574 (21.6) | 528 (22.1) | 375 (20.0) | 396 (17.9) |
| 5th Quintile | 427 (17.4) | 278 (20.4) | 589 (22.1) | 506 (21.2) | 396 (21.1) | 392 (17.7) |
| Education (%) | ||||||
| <High School | 449 (15.6) | 217 (13.8) | 378 (12.0) | 368 (13.3) | 276 (12.7) | 398 (15.1) |
| High School | 1603 (55.6) | 836 (53.1) | 1634 (52.0) | 1471 (53.1) | 1142 (52.4) | 1460 (55.3) |
| ≥College | 829 (28.8) | 522 (33.1) | 1129 (35.9) | 934 (33.7) | 761 (34.9) | 784 (29.7) |
| Employed (%) | 1152 (46.8) | 637 (46.7) | 1193 (44.9) | 1052 (44.0) | 867 (46.2) | 1063 (47.9) |
| Health insurance (%) | 2342 (95.2) | 1313 (96.2) | 2572 (96.6) | 2306 (96.5) | 1808 (96.3) | 2110 (95.1) |
| Geographic region (%) | ||||||
| Northeast | 367 (14.6) | 182 (13.0) | 401 (14.6) | 335 (13.6) | 282 (14.6) | 332 (14.6) |
| Midwest | 726 (28.9) | 438 (31.3) | 719 (26.1) | 714 (29.0) | 552 (28.6) | 616 (27.2) |
| South | 944 (37.5) | 525 (37.5) | 1112 (40.4) | 969 (39.3) | 743 (38.5) | 869 (38.3) |
| West | 479 (19.0) | 255 (18.2) | 519 (18.9) | 447 (18.1) | 355 (18.4) | 450 (19.9) |
| Childhood abuse (%) | 163 (7.2) | 88 (7.0) | 120 (4.9) | 116 (5.3) | 109 (6.3) | 147 (7.3) |
| Physical health | ||||||
| Number of chronic conditions (range: 0–8), mean (SD) | 2.4 (1.4) | 2.4 (1.4) | 2.4 (1.3) | 2.3 (1.4) | 2.4 (1.4) | 2.4 (1.4) |
| Diabetes (%) | 452 (18.4) | 216 (15.8) | 404 (15.2) | 360 (15.1) | 272 (14.5) | 439 (19.8) |
| Hypertension (%) | 1270 (51.7) | 703 (51.5) | 1327 (49.9) | 1186 (49.6) | 980 (52.2) | 1132 (51.1) |
| Stroke (%) | 118 (4.8) | 94 (6.9) | 121 (4.6) | 117 (4.9) | 103 (5.5) | 113 (5.1) |
| Cancer (%) | 287 (11.7) | 176 (12.9) | 387 (14.6) | 337 (14.1) | 264 (14.1) | 251 (11.3) |
| Heart disease (%) | 466 (19.0) | 281 (20.6) | 532 (20.0) | 476 (19.9) | 365 (19.5) | 439 (19.8) |
| Lung disease (%) | 168 (6.8) | 87 (6.4) | 164 (6.2) | 152 (6.4) | 121 (6.5) | 146 (6.6) |
| Arthritis (%) | 1372 (55.8) | 752 (55.1) | 1395 (52.4) | 1224 (51.3) | 1052 (56.1) | 1245 (56.2) |
| Overweight/Obesity (%) | 1769 (72.8) | 971 (71.8) | 1943 (73.6) | 1699 (71.7) | 1358 (73.3) | 1626 (74.2) |
| Physical function limitations (%) | 445 (18.1) | 182 (13.3) | 330 (12.4) | 302 (12.6) | 265 (14.1) | 389 (17.5) |
| Cognitive impairment (%) | 275 (11.4) | 143 (10.8) | 282 (10.8) | 242 (10.3) | 194 (10.6) | 266 (12.3) |
| Chronic pain (%) | 860 (35.0) | 425 (31.1) | 762 (28.6) | 646 (27.0) | 609 (32.5) | 791 (35.7) |
| Sleep problems (%) | 576 (42.1) | 299 (39.2) | 528 (34.6) | 504 (36.5) | 386 (36.3) | 513 (42.3) |
| Self‐rated health (range: 1–5), mean (SD) | 3.3 (1.0) | 3.4 (1.0) | 3.5 (1.0) | 3.5 (1.0) | 3.4 (1.0) | 3.3 (1.0) |
| Health behaviors | ||||||
| Heavy drinking (%) | 130 (6.6) | 83 (7.7) | 182 (8.2) | 154 (7.8) | 114 (7.6) | 127 (7.2) |
| Smoking (%) | 271 (11.1) | 134 (9.9) | 236 (8.9) | 217 (9.1) | 174 (9.3) | 250 (11.4) |
| Frequent physical activity (%) | 1873 (76.1) | 1099 (80.5) | 2187 (82.2) | 1912 (80.0) | 1518 (80.9) | 1726 (77.9) |
| Religious service attendance (%) | ||||||
| Never | 580 (23.6) | 312 (22.9) | 565 (21.2) | 513 (21.5) | 431 (23.0) | 512 (23.1) |
| <1×/week | 822 (33.4) | 441 (32.3) | 832 (31.2) | 736 (30.8) | 600 (32.0) | 760 (34.3) |
| ≥1×/week | 1058 (43.0) | 611 (44.8) | 1266 (47.5) | 1142 (47.8) | 846 (45.1) | 944 (42.6) |
| Psychological well‐being | ||||||
| Positive affect (range: 1–5), mean (SD) | 3.5 (0.7) | 3.7 (0.7) | 3.9 (0.6) | 3.9 (0.6) | 3.7 (0.7) | 3.5 (0.7) |
| Life satisfaction (range: 1–7), mean (SD) | 4.9 (1.4) | 5.4 (1.3) | 5.7 (1.2) | 5.7 (1.2) | 5.4 (1.2) | 4.9 (1.4) |
| Optimism (range: 1–6), mean (SD) | 4.4 (0.9) | 4.5 (0.9) | 4.8 (0.9) | 4.8 (0.9) | 4.6 (0.9) | 4.4 (0.9) |
| Purpose in life (range: 1–6), mean (SD) | 4.5 (0.9) | 4.7 (0.9) | 5.0 (0.8) | 4.9 (0.8) | 4.7 (0.8) | 4.6 (0.9) |
| Mastery (range: 1–6), mean (SD) | 4.7 (1.1) | 4.9 (1.0) | 5.0 (1.0) | 5.0 (1.0) | 4.8 (1.1) | 4.7 (1.1) |
| Health mastery (range: 0–10), mean (SD) | 7.2 (2.2) | 7.4 (2.1) | 7.7 (2.1) | 7.6 (2.1) | 7.5 (2.0) | 7.2 (2.3) |
| Financial mastery (range: 0–10), mean (SD) | 6.9 (2.5) | 7.3 (2.3) | 7.8 (2.3) | 7.7 (2.2) | 7.5 (2.3) | 6.9 (2.6) |
| Psychological distress | ||||||
| Depression (%) | 296 (12.3) | 94 (7.1) | 127 (4.9) | 123 (5.3) | 116 (6.3) | 278 (12.9) |
| Depressive symptoms (range: 0–8), mean (SD) | 1.3 (1.8) | 0.9 (1.5) | 0.7 (1.3) | 0.7 (1.3) | 0.9 (1.4) | 1.3 (1.9) |
| Hopelessness (range: 1–6), mean (SD) | 2.4 (1.3) | 2.1 (1.1) | 1.9 (1.1) | 1.9 (1.1) | 2.1 (1.1) | 2.5 (1.3) |
| Negative affect (range: 1–5), mean (SD) | 1.7 (0.6) | 1.6 (0.5) | 1.5 (0.5) | 1.4 (0.5) | 1.6 (0.5) | 1.8 (0.6) |
| Perceived constraints (range: 1–6), mean (SD) | 2.3 (1.2) | 2.0 (1.0) | 1.8 (1.0) | 1.8 (1.0) | 2.0 (1.0) | 2.3 (1.2) |
| Social factors | ||||||
| Loneliness (range: 1–3), mean (SD) | 1.5 (0.5) | 1.3 (0.4) | 1.2 (0.4) | 1.2 (0.4) | 1.3 (0.4) | 1.6 (0.5) |
| Living with spouse (%) | 2209 (99.4) | 1253 (99.8) | 2414 (99.9) | 2170 (99.6) | 1726 (99.9) | 1976 (99.7) |
| Contact children ≥1×/week (%) | 1695 (76.6) | 931 (74.7) | 1937 (79.7) | 1730 (78.9) | 1329 (77.5) | 1503 (75.9) |
| Contact other family ≥1×/week (%) | 1154 (51.1) | 591 (46.9) | 1237 (50.8) | 1109 (50.2) | 855 (49.3) | 1019 (50.7) |
| Contact friends ≥1×/week (%) | 1345 (59.5) | 779 (61.8) | 1602 (65.6) | 1420 (64.3) | 1107 (63.7) | 1196 (59.5) |
| Personality | ||||||
| Openness (range: 1–4), mean (SD) | 2.9 (0.5) | 2.9 (0.5) | 3.0 (0.5) | 3.0 (0.5) | 3.0 (0.5) | 2.9 (0.5) |
| Conscientiousness (range: 1–4), mean (SD) | 3.3 (0.5) | 3.4 (0.4) | 3.5 (0.4) | 3.5 (0.4) | 3.4 (0.4) | 3.3 (0.5) |
| Extraversion (range: 1–4), mean (SD) | 3.1 (0.6) | 3.2 (0.5) | 3.3 (0.5) | 3.3 (0.5) | 3.2 (0.5) | 3.2 (0.6) |
| Agreeableness (range: 1–4), mean (SD) | 3.5 (0.5) | 3.5 (0.5) | 3.6 (0.4) | 3.6 (0.4) | 3.5 (0.5) | 3.5 (0.5) |
| Neuroticism (range: 1–4), mean (SD) | 2.2 (0.6) | 2.0 (0.6) | 1.9 (0.6) | 1.9 (0.6) | 2.0 (0.6) | 2.2 (0.6) |
This table was created based on non‐imputed data.
All variables in Table 1 were used as covariates and assessed in the pre‐baseline wave (t0; 2006/2008).
The percentages in some sections may not add up to 100% due to rounding.
When participant characteristics at t0 were compared as a function of spousal strain tertiles at t1, results were similar to those for spousal support but were in the opposite direction. Compared to the first tertile of spousal strain, the highest tertile had a lower proportion of Caucasians, lower annual household income, lower educational level, lower religious service attendance, higher prevalence of depression diagnosis, and higher depressive symptom scores, as well as lower scores in life satisfaction, health mastery, and financial mastery.
At t0, spousal support and spousal strain were negatively correlated (r = −0.55, p < 0.001). Marital closeness was positively correlated with spousal support (r = 0.65, p < 0.001), and it was negatively correlated with spousal strain (r = −0.50, p < 0.001). The cross‐sectional classification of tertiles of both spousal support and spousal strain at t1 is displayed in Table S1. Approximately 60% of the people who were in the highest tertile of spousal support were also in the lowest tertile of spousal strain, and vice‐versa. Around 6% of the total sample was allocated in the lowest tertiles of both spousal support and strain. Similarly, about 6% of the total sample was in the highest tertiles of both.
The percentage of individuals shifting to another tertile of spousal support or spousal strain between t0 and t1 is shown in Table S2. For each exposure, approximately one‐third of individuals in the lowest and highest tertile at t0 were classified in other tertiles at t1, while two‐thirds of participants in the second tertile at t0 shifted tertiles at t1.
Main analyses
Spousal support
Results of the outcome‐wide analysis estimating the associations of spousal support with the outcomes can be seen in Table 2. Compared to participants in the first tertile of spousal support, being in the third tertile of spousal support was associated at follow‐up with higher positive affect (β: 0.18; 95% CI: 0.12, 0.24), life satisfaction (β: 0.21; 95% CI: 0.14, 0.27), optimism (β: 0.11; 95% CI: 0.06, 0.17), purpose on life (β: 0.19; 95% CI: 0.13, 0.25), mastery (β: 0.12; 95% CI: 0.1, 0.18), and financial mastery (β: 0.15; 95% CI: 0.8, 0.22), as well as lower risk of depression diagnosis (risk ratio (RR): 0.67; 95% CI: 0.53, 0.84), depressive symptoms (β: −0.14; 95% CI: −0.19, −0.08), hopelessness (β: −0.14; 95% CI: −0.19, −0.09), negative affect (β: −0.17; 95% CI: −0.25, 0.09), perceived constraint (β: −0.11; 95% CI: −0.18, −0.05), and loneliness (β: −0.26; 95% CI: −0.33, −0.18). All these associations had p values below the Bonferroni‐correction threshold, except for mastery. Spousal support showed little evidence of associations with the other indicators of psychological well‐being or social factors, as well as the outcomes in the domains of physical health and health behaviors.
TABLE 2.
Spousal support and subsequent health and well‐being (health and retirement study [HRS]: N = 7788). a , b , c , d
| Spousal support | ||||
|---|---|---|---|---|
| Tertile 1 (n = 2989) (reference) | Tertile 2 (n = 1626) RR/OR/β (95% CI) | Tertile 3 (n = 3173) RR/OR/β (95% CI) | Test for trend p‐values | |
| Physical health | ||||
| All‐cause mortality | 1.00 | 0.80 (0.62, 1.03) | 0.83 (0.65, 1.05) | 0.139 |
| Number of chronic conditions | 0.00 | −0.02 (−0.07, 0.04) | −0.01 (−0.06, 0.04) | 0.798 |
| Diabetes | 1.00 | 1.03 (0.90, 1.18) | 1.04 (0.91, 1.19) | 0.540 |
| Hypertension | 1.00 | 0.98 (0.90, 1.06) | 0.99 (0.91, 1.08) | 0.878 |
| Stroke | 1.00 | 0.93 (0.77, 1.13) | 0.97 (0.79, 1.19) | 0.757 |
| Cancer | 1.00 | 0.99 (0.86, 1.15) | 1.01 (0.88, 1.16) | 0.849 |
| Heart disease | 1.00 | 0.96 (0.85, 1.09) | 0.96 (0.86, 1.09) | 0.564 |
| Lung disease | 1.00 | 1.02 (0.83, 1.24) | 0.99 (0.81, 1.20) | 0.884 |
| Arthritis | 1.00 | 1.00 (0.92, 1.08) | 1.01 (0.94, 1.09) | 0.740 |
| Overweight/obesity | 1.00 | 1.00 (0.93, 1.08) | 0.99 (0.92, 1.07) | 0.770 |
| Physical functioning limitations | 1.00 | 0.91 (0.79, 1.04) | 1.01 (0.89, 1.15) | 0.857 |
| Cognitive impairment | 1.00 | 0.97 (0.83, 1.13) | 1.04 (0.90, 1.20) | 0.548 |
| Chronic pain | 1.00 | 0.99 (0.89, 1.10) | 0.94 (0.85, 1.04) | 0.215 |
| Sleep problems | 1.00 | 0.94 (0.85, 1.04) | 0.96 (0.87, 1.07) | 0.505 |
| Self‐rated health | 0.00 | 0.00 (−0.05, 0.06) | 0.03 (−0.02, 0.08) | 0.251 |
| Health behaviors | ||||
| Heavy drinking | 1.00 | 1.35 (0.78, 2.33) | 1.26 (0.83, 1.92) | 0.278 |
| Smoking | 1.00 | 0.99 (0.58, 1.72) | 0.92 (0.59, 1.45) | 0.706 |
| Frequent physical activity | 1.00 | 1.06 (0.98, 1.15) | 1.05 (0.98, 1.14) | 0.210 |
| Psychological well‐being | ||||
| Positive affect | 0.00 | 0.11 (0.05, 0.17)*** | 0.18 (0.12, 0.24)*** | <0.001 |
| Life satisfaction | 0.00 | 0.17 (0.11, 0.23)*** | 0.21 (0.14, 0.27)*** | <0.001 |
| Optimism | 0.00 | 0.07 (0.02, 0.12)** | 0.11 (0.06, 0.17)*** | <0.001 |
| Purpose in life | 0.00 | 0.12 (0.07, 0.18)*** | 0.19 (0.13, 0.25)*** | <0.001 |
| Mastery | 0.00 | 0.09 (0.03, 0.16)** | 0.12 (0.05, 0.18)** | 0.002 |
| Health mastery | 0.00 | 0.04 (−0.02, 0.09) | 0.06 (0.00, 0.13) | 0.055 |
| Financial mastery | 0.00 | 0.09 (0.03, 0.16)** | 0.15 (0.08, 0.22)*** | <0.001 |
| Psychological distress | ||||
| Depression | 1.00 | 0.68 (0.55, 0.85)*** | 0.67 (0.53, 0.84)*** | <0.001 |
| Depressive symptoms | 0.00 | −0.13 (−0.19, −0.08)*** | −0.14 (−0.19, −0.08)*** | <0.001 |
| Hopelessness | 0.00 | −0.08 (−0.14, −0.02)* | −0.14 (−0.19, −0.09)*** | <0.001 |
| Negative affect | 0.00 | −0.12 (−0.19, −0.05)*** | −0.17 (−0.25, −0.09)*** | <0.001 |
| Perceived constraints | 0.00 | −0.09 (−0.16, −0.02)* | −0.11 (−0.18, −0.05)*** | <0.001 |
| Social factors | ||||
| Loneliness | 0.00 | −0.19 (−0.25, −0.13)*** | −0.26 (−0.33, −0.18)*** | <0.001 |
| Living with a spouse | 1.00 | 1.05 (0.98, 1.12) | 1.06 (0.99, 1.13) | 0.086 |
| Contact children ≥1×/week | 1.00 | 1.03 (0.95, 1.12) | 1.04 (0.96, 1.12) | 0.319 |
| Contact other family ≥1×/week | 1.00 | 1.04 (0.94, 1.16) | 1.07 (0.98, 1.18) | 0.111 |
| Contact friends ≥1×/week | 1.00 | 1.04 (0.95, 1.14) | 1.00 (0.91, 1.09) | 0.918 |
Abbreviations: CI, confidence interval; OR, odds ratio; RR, risk ratio.
If the reference value is “1,” the effect estimate is OR or RR; if the reference value is “0,” the effect estimate is β.
The analytic sample was restricted to those who had participated in the baseline wave (t1; 2010 or 2012). Multiple imputation was performed to impute missing data on the exposure, covariates, and outcomes. All models were adjusted for the following: sociodemographic characteristics (age, sex, race/ethnicity, annual household income, total wealth, level of education, employment status, health insurance, geographic region), childhood abuse, religious service attendance, all outcome variables (diabetes, hypertension, stroke, cancer, heart disease, lung disease, arthritis, overweight/obesity, physical functioning limitations, cognitive impairment, chronic pain, self‐rated health, heavy drinking, current smoking status, physical activity, sleep problems, positive affect, life satisfaction, optimism, purpose in life, mastery, health mastery, financial mastery, depressive symptoms, hopelessness, negative affect, perceived constraints, loneliness, living with spouse, contact children ≥1×/week, contact other family ≥1×/week, contact friends ≥1×/week), personality factors (openness, conscientiousness, extraversion, agreeableness, neuroticism), and the pre‐baseline value of the exposure (receiving social support from a spouse). These variables were adjusted for in the wave pre‐baseline to the exposure assessment (t0; 2006 or 2008).
An outcome‐wide analytic approach was used, and a separate model for each outcome was run. A different type of model was run depending on the nature of the outcome: (1) for each binary outcome with a prevalence of ≥10%, a generalized linear model (with a log link and Poisson distribution) was used to estimate a RR; (2) for each binary outcome with a prevalence of <10%, a logistic regression model was used to estimate an OR; and (3) for each continuous outcome, a linear regression model was used to estimate a β.
All continuous outcomes were standardized (mean = 0; standard deviation = 1), and β was the standardized effect size.
*p < 0.05 before Bonferroni correction. **p < 0.01 before Bonferroni correction. ***p < 0.05 after Bonferroni correction (the p value cutoff for Bonferroni correction is p = 0.05/35 outcomes = p < 0.001).
Spousal strain
The outcome‐wide analysis that assessed the effect of spousal strain can be seen in Table 3. Compared to the first tertile, being in the third tertile of spousal strain was associated at follow‐up with lower scores of positive affect (β: −0.16; 95% CI: −0.23, −0.09), life satisfaction (β: −0.26; 95% CI: −0.32, −0.20), and optimism (β: −0.12; 95% CI: −0.17, −0.06), and higher scores of depressive symptoms (β: 0.18; 95% CI: 0.10, 0.25), hopelessness (β: 0.16; 95% CI: 0.10, 0.22), negative affect (β: 0.23; 95% CI: 0.14, 0.31), and loneliness (β: 0.27; 95% CI: 0.20, 0.33). There was also some evidence that spousal strain was positively associated with higher risk of depression diagnosis (RR: 1.46; 95% CI: 1.16, 1.84) and perceived constraints (β: 0.13; 95% CI: 0.05, 0.22), while being negatively associated with self‐related health (β: −0.08; 95% CI: −0.14, −0.02), purpose in life (β: −0.11; 95% CI: −0.18, −0.04), mastery (β: −0.09; 95% CI: −0.17, −0.01), health mastery (β: −0.11; 95% CI: −0.19, −0.03), and financial mastery (β: −0.11; 95% CI: −0.18, −0.03), but these latter associations did not pass the p < 0.05 cutoff after Bonferroni correction. Spousal strain showed little evidence of associations with the other indicators of psychological well‐being or social factors. Except for self‐rated health, there was little evidence of associations between spousal strain and the outcomes on the domains of physical health and health behaviors. Even so, the association between spousal strain and self‐rated health did not pass the Bonferroni‐corrected threshold (β: −0.08; 95% CI: −0.14, −0.03).
TABLE 3.
Spousal strain and subsequent health and well‐being (health and retirement study [HRS]: N = 7788). a , b , c , d
| Spousal strain | ||||
|---|---|---|---|---|
| Tertile 1 (n = 2814) (reference) | Tertile 2 (n = 2229) RR/OR/β (95% CI) | Tertile 3 (n = 2745) RR/OR/β (95% CI) | Test for trend p‐values | |
| Physical health | ||||
| All‐cause mortality | 1.00 | 0.96 (0.76, 1.20) | 1.24 (0.97, 1.59) | 0.090 |
| Number of chronic conditions | 0.00 | −0.01 (−0.05, 0.02) | 0.03 (−0.01, 0.07) | 0.137 |
| Diabetes | 1.00 | 0.96 (0.85, 1.08) | 1.04 (0.92, 1.19) | 0.536 |
| Hypertension | 1.00 | 0.99 (0.92, 1.07) | 1.00 (0.92, 1.09) | 0.941 |
| Stroke | 1.00 | 1.00 (0.82, 1.22) | 1.14 (0.91, 1.42) | 0.247 |
| Cancer | 1.00 | 0.97 (0.85, 1.10) | 0.94 (0.82, 1.09) | 0.419 |
| Heart disease | 1.00 | 1.03 (0.92, 1.14) | 1.06 (0.94, 1.19) | 0.347 |
| Lung disease | 1.00 | 0.88 (0.73, 1.05) | 0.94 (0.78, 1.14) | 0.516 |
| Arthritis | 1.00 | 1.01 (0.94, 1.08) | 1.03 (0.95, 1.12) | 0.484 |
| Overweight/obesity | 1.00 | 1.00 (0.93, 1.07) | 1.01 (0.93, 1.09) | 0.843 |
| Physical functioning limitations | 1.00 | 1.00 (0.88, 1.14) | 1.09 (0.95, 1.26) | 0.203 |
| Cognitive impairment | 1.00 | 0.92 (0.80, 1.06) | 0.91 (0.79, 1.06) | 0.226 |
| Chronic pain | 1.00 | 0.99 (0.90, 1.10) | 1.02 (0.92, 1.13) | 0.690 |
| Sleep problems | 1.00 | 1.04 (0.94, 1.15) | 1.09 (0.97, 1.21) | 0.151 |
| Self‐rated health | 0.00 | −0.02 (−0.07, 0.03) | −0.08 (−0.14, −0.02)** | 0.007 |
| Health behaviors | ||||
| Heavy drinking | 1.00 | 1.04 (0.75, 1.44) | 0.90 (0.59, 1.37) | 0.619 |
| Smoking | 1.00 | 1.14 (0.79, 1.62) | 0.88 (0.47, 1.65) | 0.670 |
| Frequent physical activity | 1.00 | 1.04 (0.97, 1.12) | 0.99 (0.90, 1.09) | 0.938 |
| Psychological well‐being | ||||
| Positive affect | 0.00 | −0.09 (−0.14, −0.03)** | −0.16 (−0.23, −0.09)*** | <0.001 |
| Life satisfaction | 0.00 | −0.09 (−0.14, −0.03)** | −0.26 (−0.32, −0.20)*** | <0.001 |
| Optimism | 0.00 | −0.06 (−0.11, −0.01)* | −0.12 (−0.17, −0.06)*** | <0.001 |
| Purpose in life | 0.00 | −0.05 (−0.1, 0.01) | −0.11 (−0.18, −0.04)** | 0.003 |
| Mastery | 0.00 | −0.07 (−0.13, −0.01)* | −0.09 (−0.17, −0.01)* | 0.031 |
| Health mastery | 0.00 | −0.03 (−0.09, 0.04) | −0.11 (−0.19, −0.03)* | 0.013 |
| Financial mastery | 0.00 | −0.06 (−0.12, −0.01)* | −0.11 (−0.18, −0.03)** | 0.006 |
| Psychological distress | ||||
| Depression | 1.00 | 1.25 (1.00, 1.55)* | 1.46 (1.16, 1.84)** | 0.002 |
| Depressive symptoms | 0.00 | 0.07 (0.01, 0.13)* | 0.18 (0.10, 0.25)*** | <0.001 |
| Hopelessness | 0.00 | 0.07 (0.02, 0.12)* | 0.16 (0.10, 0.22)*** | <0.001 |
| Negative affect | 0.00 | 0.13 (0.06, 0.19)*** | 0.23 (0.14, 0.31)*** | <0.001 |
| Perceived constraints | 0.00 | 0.06 (0.00, 0.12) | 0.13 (0.05, 0.22)** | 0.004 |
| Social factors | ||||
| Loneliness | 0.00 | 0.10 (0.04, 0.17)** | 0.27 (0.20, 0.33)*** | <0.001 |
| Living with a spouse | 1.00 | 0.99 (0.93, 1.05) | 0.96 (0.90, 1.03) | 0.264 |
| Contact children ≥1×/week | 1.00 | 1.00 (0.93, 1.08) | 0.99 (0.91, 1.07) | 0.798 |
| Contact other family ≥1×/week | 1.00 | 0.96 (0.88, 1.05) | 0.93 (0.85, 1.02) | 0.137 |
| Contact friends ≥1×/week | 1.00 | 1.02 (0.94, 1.11) | 0.99 (0.89, 1.09) | 0.809 |
Abbreviations: CI, confidence interval; OR, odds ratio; RR, risk ratio.
If the reference value is “1,” the effect estimate is OR or RR; if the reference value is “0,” the effect estimate is β.
The analytic sample was restricted to those who had participated in the baseline wave (t1; 2010 or 2012). Multiple imputation was performed to impute missing data on the exposure, covariates, and outcomes. All models were adjusted for the following variables: sociodemographic characteristics (age, sex, race/ethnicity, annual household income, total wealth, level of education, employment status, health insurance, geographic region), childhood abuse, religious service attendance, all outcome variables (diabetes, hypertension, stroke, cancer, heart disease, lung disease, arthritis, overweight/obesity, physical functioning limitations, cognitive impairment, chronic pain, self‐rated health, heavy drinking, current smoking status, physical activity, sleep problems, positive affect, life satisfaction, optimism, purpose in life, mastery, health mastery, financial mastery, depressive symptoms, hopelessness, negative affect, perceived constraints, loneliness, living with spouse, contact children ≥1×/week, contact other family ≥1×/week, contact friends ≥1×/week), personality factors (openness, conscientiousness, extraversion, agreeableness, neuroticism), and the pre‐baseline value of the exposure (receiving social strain from a spouse). These variables were adjusted for in the wave pre‐baseline to the exposure assessment (t0; 2006 or 2008).
An outcome‐wide analytic approach was used, and a separate model for each outcome was run. A different type of model was run depending on the nature of the outcome: (1) for each binary outcome with a prevalence of ≥10%, a generalized linear model (with a log link and Poisson distribution) was used to estimate a RR; (2) for each binary outcome with a prevalence of <10%, a logistic regression model was used to estimate an OR; and (3) for each continuous outcome, a linear regression model was used to estimate a β.
All continuous outcomes were standardized (mean = 0; standard deviation = 1), and β was the standardized effect size.
*p < 0.05 before Bonferroni correction. **p < 0.01 before Bonferroni correction. ***p < 0.05 after Bonferroni correction (the p value cutoff for Bonferroni correction is p = 0.05/35 outcomes = p < 0.001).
Test for linear trends
For each of the significant associations in the main analyses, there was evidence of a linear trend when we modeled the exposures as continuous variables. There was little evidence supporting additional associations that were not observed in the main analyses.
Sensitivity analyses
Tables 4 and 5 show the E‐values for the robustness of the associations found in Tables 3 and 4, respectively. Most of the E‐values for the point estimates that showed evidence of associations in our main analyses were above 1.30. For some of the outcomes, the E‐value for the confidence interval was above 1.50, suggesting that some of the associations between spousal support (or strain) and the respective outcomes were moderately robust to unmeasured confounding, conditional on the measured covariates. For example, an unmeasured confounder would need to be associated with both spousal support and loneliness by risk ratios of 1.84 each to explain away the associations between them, conditional on the measured covariates; weaker joint confounder associations could not. To shift the confidence interval for the association between spousal support and loneliness to include the null, unmeasured confounder risk ratio associations of 1.66 each could suffice, but weaker joint confounder associations could not.
TABLE 4.
Robustness to unmeasured confounding (E‐values) for the associations between spousal support (3rd Tertile vs. 1st Tertile) and subsequent health and well‐being (N = 7788). a
| Effect estimate b | Confidence interval limit c | |
|---|---|---|
| Physical health | ||
| All‐cause mortality | 1.72 | 1.00 |
| Number of chronic conditions | 1.09 | 1.00 |
| Diabetes | 1.25 | 1.00 |
| Hypertension | 1.10 | 1.00 |
| Stroke | 1.23 | 1.00 |
| Cancer | 1.12 | 1.00 |
| Heart disease | 1.23 | 1.00 |
| Lung disease | 1.13 | 1.00 |
| Arthritis | 1.12 | 1.00 |
| Overweight/obesity | 1.11 | 1.00 |
| Physical functioning limitations | 1.11 | 1.00 |
| Cognitive impairment | 1.25 | 1.00 |
| Chronic pain | 1.32 | 1.00 |
| Sleep problems | 1.23 | 1.00 |
| Self‐rated health | 1.19 | 1.00 |
| Health behaviors | ||
| Binge drinking | 1.83 | 1.00 |
| Smoking | 1.39 | 1.00 |
| Frequent physical activity | 1.29 | 1.00 |
| Psychological well‐being | ||
| Positive affect | 1.64 | 1.49 |
| Life satisfaction | 1.71 | 1.54 |
| Optimism | 1.45 | 1.29 |
| Purpose in life | 1.67 | 1.51 |
| Mastery | 1.46 | 1.27 |
| Health mastery | 1.31 | 1.03 |
| Financial mastery | 1.55 | 1.36 |
| Psychological distress | ||
| Depression | 2.36 | 1.68 |
| Depressive symptoms | 1.52 | 1.37 |
| Hopelessness | 1.52 | 1.38 |
| Negative affect | 1.62 | 1.42 |
| Constraints | 1.46 | 1.28 |
| Social factors | ||
| Loneliness | 1.84 | 1.66 |
| Living with spouse | 1.32 | 1.00 |
| Contact children ≥1×/week | 1.24 | 1.00 |
| Contact other family ≥1×/week | 1.36 | 1.00 |
| Contact friends ≥1×/week | 1.03 | 1.00 |
See VanderWeele and Ding (2017) for the formula for calculating E‐values.
The E‐values for effect estimates are the minimum strength of association on the risk ratio scale that an unmeasured confounder would need to have with both the exposure and the outcome to fully explain away the observed association between the exposure and outcome, conditional on the measured covariates.
The E‐values for the limit of the 95% confidence interval (CI) closest to the null denote the minimum strength of association on the risk ratio scale that an unmeasured confounder would need to have with both the exposure and the outcome to shift the confidence interval to include the null value, conditional on the measured covariates.
TABLE 5.
Robustness to unmeasured confounding (E‐values) for the associations between spousal strain (3rd Tertile vs. 1st Tertile) and subsequent health and well‐being (N = 7788). a
| Effect estimate b | Confidence interval limit c | |
|---|---|---|
| Physical health | ||
| All‐cause mortality | 1.79 | 1.00 |
| Number of chronic conditions | 1.14 | 1.00 |
| Diabetes | 1.25 | 1.00 |
| Hypertension | 1.06 | 1.00 |
| Stroke | 1.54 | 1.00 |
| Cancer | 1.31 | 1.00 |
| Heart disease | 1.30 | 1.00 |
| Lung disease | 1.32 | 1.00 |
| Arthritis | 1.20 | 1.00 |
| Overweight/obesity | 1.10 | 1.00 |
| Physical functioning limitations | 1.41 | 1.00 |
| Cognitive impairment | 1.42 | 1.00 |
| Chronic pain | 1.17 | 1.00 |
| Sleep problems | 1.39 | 1.00 |
| Self‐rated health | 1.36 | 1.18 |
| Health behaviors | ||
| Binge drinking | 1.46 | 1.00 |
| Smoking | 1.53 | 1.00 |
| Frequent physical activity | 1.08 | 1.00 |
| Psychological well‐being | ||
| Positive affect | 1.58 | 1.41 |
| Life satisfaction | 1.86 | 1.70 |
| Optimism | 1.46 | 1.30 |
| Purpose in life | 1.46 | 1.26 |
| Mastery | 1.38 | 1.12 |
| Health mastery | 1.44 | 1.21 |
| Financial mastery | 1.44 | 1.22 |
| Psychological distress | ||
| Depression | 2.28 | 1.59 |
| Depressive symptoms | 1.63 | 1.44 |
| Hopelessness | 1.58 | 1.41 |
| Negative affect | 1.76 | 1.55 |
| Constraints | 1.51 | 1.29 |
| Social factors | ||
| Loneliness | 1.87 | 1.70 |
| Living with spouse | 1.25 | 1.00 |
| Contact children ≥1×/week | 1.12 | 1.00 |
| Contact other family ≥1×/week | 1.35 | 1.00 |
| Contact friends ≥1×/week | 1.13 | 1.00 |
See VanderWeele and Ding (2017) for the formula for calculating E‐values.
The E‐values for effect estimates are the minimum strength of association on the risk ratio scale that an unmeasured confounder would need to have with both the exposure and the outcome to fully explain away the observed association between the exposure and outcome, conditional on the measured covariates.
The E‐values for the limit of the 95% confidence interval (CI) closest to the null denote the minimum strength of association on the risk ratio scale that an unmeasured confounder would need to have with both the exposure and the outcome to shift the confidence interval to include the null value, conditional on the measured covariate.
The direction and magnitude of the associations were mostly similar when we replicated the main analyses after excluding participants with pre‐baseline health conditions (see fully‐adjusted models in Tables S3 and S4), after excluding participants with pre‐baseline health conditions and omitting pre‐baseline values of the exposures and outcomes (see conventionally‐adjusted models in Tables S3 and S4), and after complete‐case analysis (Tables S5 and S6). However, when adjusting exclusively for sociodemographic variables among those without pre‐baseline health conditions (conventionally‐adjusted models in Tables S3 and S4), effect sizes for associations generally strengthened, and additional associations arose for both exposures. On the one hand, additional associations were found for spousal support with a lower risk of sleep problems, higher self‐rated health (both passed the Bonferroni‐corrected threshold), a lower risk of all‐cause mortality and smoking, a higher risk of frequent physical activity, lower number of chronic conditions, a lower risk of chronic pain, and a higher likelihood of living with one's spouse, having contact with children >1×/week, and having contact with other family and friends >1×/week (none of which passed the Bonferroni‐corrected threshold). On the other hand, additional associations were found for spousal strain with higher number of chronic conditions, a higher risk of sleep problems (both passed the Bonferroni‐corrected threshold), arthritis, and physical functioning limitations, a higher risk of chronic pain, and a lower likelihood of living with one's spouse/partner, having contact with other family than children >1×/week, and having contact friends >1×/week (none of which passed the Bonferroni‐corrected threshold).
After adding spousal strain at t0 to the models assessing the effects of spousal support at t1 (Table S7), spousal support showed evidence of associations with higher positive affect, life satisfaction, purpose in life, and financial mastery, together with less depressive symptoms, hopelessness, and loneliness (even after Bonferroni correction). After further adding marital closeness at t0 as a covariate (Table S8), the strength of the associations for life satisfaction, depressive symptoms, hopelessness, and loneliness attenuated (and no longer passed the Bonferroni‐corrected threshold). Similarly, there was evidence supporting associations of spousal strain at t1 with life satisfaction, optimism, hopelessness, negative affect, and loneliness after Bonferroni correction when spousal support at t0 was added as a covariate (Table S9), while the associations with depression symptoms no longer passed the Bonferroni‐corrected threshold, and the associations with perceived constraints and purpose in life attenuated to include the null. After additionally including marital closeness at t0 as a covariate, the magnitude of the association between spousal strain and higher depression symptoms remained similar but passed the Bonferroni‐corrected threshold (Table S10).
DISCUSSION
In this study, we assessed the associations between spousal support and spousal strain with a wide range of health and well‐being outcomes in a sample of married older adults. After 4 years of follow‐up, spousal support evidenced robust positive associations with five psychological well‐being outcomes, as well as negative associations with five psychological distress outcomes and loneliness. Conversely, spousal strain evidenced robust negative associations with three psychological well‐being outcomes, in addition to positive associations with three psychological distress outcomes and loneliness. There was little evidence of associations found for spousal support or strain with outcomes related to physical health, health behaviors, or other social factors.
For most outcomes for which there was evidence of associations, support and strain generally had similar effect sizes. These results are consistent with the combined positivity and negativity effects model (Ingersoll‐Dayton et al., 1997), which states that spousal support and strain have parallel effects on negative and positive outcomes. Therefore, the implications of higher perceived spousal strain for well‐being are likely to be similar to having a lower perceived spousal support. However, despite the comparable results across most outcomes that were assessed, there were some subtle differences between the associations found for spousal support and strain. For example, the strongest associations for spousal support were found for purpose in life and financial mastery, whereas the strongest associations for spousal strain emerged for life satisfaction, negative affect, self‐rated health, and health mastery. These differences were more evident in the sensitivity analysis that controlled for pre‐baseline levels of spousal support, spousal strain, and marital closeness. For example, after controlling for pre‐baseline values of both exposures and marital closeness, spousal support had stronger associations with purpose in life and financial mastery than spousal strain, while spousal strain had stronger associations with loneliness and negative affect. Most of these subtle differences might be explained by the fact that experiences of spousal support involve feeling understood and a sense of trust in one's spouse, which makes it easier to nurture a positive mindset that attenuates reactivity to stressors (Selcuk et al., 2016). In contrast, spousal strain may arise from demanding, critical, and unreliable spousal behavior that can have negative implications for an individual's well‐being (Stanton et al., 2019).
Our results align with previous literature about the psychosocial effects of spousal support or spousal strain in older people. Spousal support can lead to greater life satisfaction and psychological well‐being directly or by reducing feelings of loneliness (Chen & Feeley, 2014; Saenz, 2021) or stress (Shrout, 2021). Because older adults tend to prioritize socioemotional well‐being, those who are married should be encouraged to nurture their spousal relationships (Carstensen et al., 1999).
Our results also align with the marital discord model of depression (Beach et al., 1990), which states that spousal strain may precipitate depression. In contrast with previous literature, there was little evidence of associations between spousal support (or strain) with physical health or health behavior outcomes in the main analyses. However, we found some evidence of associations in the sensitivity analysis that adjusted exclusively for sociodemographic characteristics. For example, spousal support was associated with an increase in self‐related health and health mastery, as well as a decrease in sleep problems. On the other hand, spousal strain was associated with an increase in number of chronic conditions and sleep problems, as well as with a decrease in self‐related health and health mastery. One possible explanation for these findings might be reverse causation, which can be considered in light of the behavioral theory of marriage. According to this theory, couples with a history of supportive relationship tend to adapt better to stressors, such as changes in health status, whereas couples with a history of strained relationships are more likely to have strain‐based responses. Thus, in our data, changes in health status from pre‐baseline to baseline wave could be associated with higher support (or higher strain) from the couple at the baseline wave (Fu & Noguchi, 2018). At follow‐up, this spousal support (or strain) could buffer (or increase) the individual's perceptions about their health (Slatcher & Selcuk, 2017), especially among individuals with functional limitations (Saenz, 2021). However, when adjusting from both pre‐baseline health conditions and the exposure, our study evaluates the effects of incident, not prevalent, spousal support and strain (thus, only evaluating the cumulative effect of up to 4 years of spousal support or strain), as it is seen in Supporting Information. Although this approach makes our findings appealing to any married older couple because the results provide an indication of how a change in spousal support or strain from pre‐baseline (t0) to baseline (t1) might affect subsequent health and well‐being (VanderWeele et al., 2020), 4 years of follow‐up might be insufficient for estimating potential causal effects on certain outcomes (e.g., chronic health conditions). In further studies on marital relationships, we recommend using more waves of data with longer follow‐up periods to explore whether our findings replicate over a longer follow‐up period.
The impact of spousal strain on self‐rated health has been discussed previously in the literature, and it is especially important in the absence of spousal support (Chen & Feeley, 2014). One possible explanation of our findings is that continuous spousal strain may burden the individual's well‐being over time, causing marital burnout, which in turn might cause somatic manifestations of stress and worse perceptions of health (Nejatian et al., 2021). An alternative explanation is that spousal strain could trigger depressive symptoms (Beach et al., 1990), which could lower a person's perception of their physical health. Another possible mechanism for this association might be that perceived spousal strain reflects the spousal strain exerted on the partner. Spousal strain can be reciprocal because negative comments from one partner may incite negative responses from the other partner, which may be verbal or behavioral. This spousal strain perceived by a spouse could lead them to withdraw away from their demanding partner (Holley et al., 2013), which may decrease the availability of spousal support in service of maintaining their own health status (Ryan et al., 2014). Dyadic data will likely contribute to exploring these speculative possibilities more comprehensively in the future.
Both spousal support and strain showed little evidence of association with the social factors, except for loneliness, which could also be considered a form of psychosocial distress because it is a psychological manifestation of social isolation. These findings resonate with insights from socioemotional selectivity theory, which posits that older adults often turn to emotionally fulfilling relationships in the later stages of life. When opportunities for social relationships with friends decrease due to aging, familiar social partners become the preferred source for social relationships (Fung et al., 1999). However, even with the presence of a spouse, it is possible to feel lonely when relationship quality is low. Indeed, the outcome with one of the strongest associations with both exposures was loneliness. These findings may be because loneliness is a key outcome of a low‐quality marital relationship at advanced ages, and an increase in loneliness could partially mediate associations of spousal support and strain with other well‐being outcomes (Cacioppo et al., 2010). Both exposures' effects on loneliness could vary depending on the marital context. For instance, marital strain has little effect on a highly supportive marital relationship (Chen & Feeley, 2014), while social strain's deleterious effects could be stronger when the individual has physical limitations (Saenz, 2021). Given our application of the outcome‐wide analytic design in this study, which estimated associations between the exposures and all subsequent outcomes taken from a single point in time, this theorizing would require examination in future research.
Additionally, these effects could depend on both actor and partner's perceived spousal support or strain, setting up a feedback system within the marital relationship where increases in the well‐being of one spouse could help to reduce stress and increase the well‐being of the other spouse (Shrout, 2021). Although our three‐wave analytic approach enabled us to adjust for many salient contextual variables, additional research using dyadic data analysis methods (e.g., actor–partner interdependence model) is needed to examine this possibility further.
Implications for research and public health
The findings of this study suggest that spousal support and strain have parallel effects on some domains of well‐being among married older adults (most notably for indicators of psychological well‐being, indicators of psychological distress, and loneliness). Given that our data are from an older sample of non‐clinical adults, our results could be useful for public health interventions in the form of community, clinic‐based, or public health psychoeducation programs aimed at supporting the marital relationships of older adults. Addressing unresolved hurts from a history of strain within the spousal relationship is crucial, as it has been shown to degenerate the quality of a marriage, leading to resentment, distrust, and withdrawal from the relationship that can have negative downstream consequences for the well‐being of both spouses (Manalel et al., 2019). Forgiveness is one potential avenue through which married couples may be able to reconcile and restore the quality of their relationship (Eyring et al., 2021).
Although most of the existing interventions to promote forgiveness require a mental health professional (Wade et al., 2014), a growing number of studies have shown that do‐it‐yourself psychoeducation forgiveness workbooks may be effective in promoting forgiveness (e.g., Greer et al., 2014). For example, a recent multinational randomized controlled trial with more than 4500 adults found evidence suggesting that a nominal (2–3 h) do‐it‐yourself forgiveness workbook promoted forgiveness, mental health, and well‐being (Ho et al., in press). Because the workbook is freely accessible to the public and can be downloaded for print or electronic use (https://reach.discoverforgiveness.org/), it is particularly well suited to non‐treatment‐seeking individuals. Such resources may support older married adults with reducing unresolved resentment and bitterness toward their spouse as a potential barrier to a higher quality marital relationship, with possible downstream implications for improving their (and perhaps their spouse's) well‐being. In a similar vein, the findings of this study suggest that self‐directed interventions to promote supportive spousal interactions may have some utility to improve well‐being, as has been explored previously through a self‐directed gratitude activity between younger couples (Parnell et al., 2020). Although interventions focusing on forgiveness and gratitude offer promising avenues for addressing spousal strain and enhancing support, there could be other pathways to fostering marital well‐being (e.g., relational reflexivity) that could be considered in future work (Moscatelli et al., 2022).
Apart from public health interventions, the findings of this study could also be relevant to the work of mental health professionals who provide family and couples therapy with older adults, such as when early signs of loneliness or psychological distress are observed in one or both spouses. Even among older couples with a long history of high spousal strain and low spousal support, improvements in these areas could be beneficial to well‐being because of how important close relationships are at this stage of life (Fung et al., 1999). In this regard, couple interventions that foster empathy and mutual support, or strengthen communication skills, can be implemented with the aim of reducing loneliness and improving psychological distress (Cohen et al., 2010).
Future research might consider investigating the mechanisms underlying the observed associations, especially those pertaining to outcomes that showed different effect sizes between spousal support and strain. Another potential line of research might involve evaluating the effects of perceived spousal support and spousal strain on the other spouse. Although prior research has explored this question with some outcomes included in this study, such as loneliness (Chen & Feeley, 2014; Saenz, 2021), other important outcomes (e.g., hopelessness) have yet to be rigorously examined.
Limitations and strengths
This study has several limitations. First, we cannot completely rule out the possibility of unmeasured confounding variables. However, as part of the three‐wave analytic strategy used, our main analyses adjusted for a wide set of covariates, including prior levels of all outcomes and the respective exposure. Second, self‐report bias could be present for all variables, as the data we used was entirely self‐reported. Third, the follow‐up period of 4 years may not be sufficient to capture the onset of chronic conditions, such as cardiovascular diseases, so we encourage conducting similar research with longer follow‐up windows. Fourth, we did not have information about the spousal relationship duration; therefore, it is possible that some of the associations observed for spousal support or strain might vary based on the length of time individuals have been married.
Despite these limitations, the prospective nature of the data and the robust control for pre‐baseline covariates and outcomes help to mitigate some of these possible concerns. Additionally, E‐values suggested that some of the results from the main analyses were moderately robust to potential unmeasured confounding. Other strengths of our study are related to the HRS database, including the large sample of older U.S. adults, and the longitudinal design with three waves of data.
CONCLUSION
This study offered evidence linking both spousal support and strain with various well‐being outcomes among older U.S. adults, including many indicators of psychological well‐being, several indicators of psychological distress, and loneliness. Mental health professionals involved in supporting the well‐being of married older adults should consider the quality of their spousal relationships.
FUNDING INFORMATION
We would like to acknowledge and thank the Health and Retirement Study conducted by the Institute for Social Research at the University of Michigan, with grants from the National Institute on Aging (U01AG09740) and the Social Security Administration. Besides, E. Kim was funded by Michael Smith Health Research BC, J Nakamura was funded by Vanier Canada Graduate Scholarships (Vanier CGS) program, and PA de la Rosa was funded by Funciva and Proeduca Summa, and Colegio de Médicos de Navarra (Beca Senior 2020). T. VanderWeele wishes to acknowledge John Templeton Foundation (Grant 61665) as it supported general well‐being research within the team. The funders had no role in preparing the manuscript or the decision to publish it.
Supporting information
Tables S1‐S10
Data S1
ACKNOWLEDGMENTS
This study was exempt from additional review by the Institutional Review Board at the University of British Columbia because data were publicly available and de‐identified. We would also like to thank Martiño Rodríguez‐González for his guidance during the preparation of the manuscript [Correction added on 04 October 2024, after first online publication: The previous sentence has been included in the Acknowledgment section.]
de la Rosa, P. A. , Nakamura, J. , Cowden, R. G. , Kim, E. , Osorio, A. , & VanderWeele, T. J. (2025). Longitudinal associations of spousal support and strain with health and well‐being: An outcome‐wide study of married older U.S. Adults. Family Process, 64(1), e13057. 10.1111/famp.13057
Footnotes
Derived from the Personal Mastery subscale from the Sense of Control Scale (Lachman & Weaver, 1998), which assesses whether a person has a sense of control over their life (e.g., “I can do just about anything I really set my mind to”).
Assessed using the item “How close is your relationship with your spouse or partner?” (1 = Very close; 4 = Not at all close). Responses were reverse‐coded for the analysis.
DATA AVAILABILITY STATEMENT
Data are available upon reasonable request.
REFERENCES
- Azur, M. J. , Stuart, E. A. , Frangakis, C. , & Leaf, P. J. (2011). Multiple imputation by chained equations: What is it and how does it work? International Journal of Methods in Psychiatric Research, 20(1), 40–49. 10.1002/mpr.329 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Beach, S. R. , Sandeen, E. , & O'Leary, K. (1990). Depression in marriage: A model for etiology and treatment. Guilford Press. [Google Scholar]
- Birditt, K. S. , Brown, E. , Orbuch, T. L. , & McIlvane, J. M. (2010). Marital conflict behaviors and implications for divorce over 16 years. Journal of Marriage and Family, 72(5), 1188–1204. 10.1111/j.1741-3737.2010.00758.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- Brazeau, H. , & Lewis, N. A. (2021). Within‐couple health behavior trajectories: The role of spousal support and strain. Health Psychology, 40(2), 125–134. 10.1037/hea0001050 [DOI] [PubMed] [Google Scholar]
- Brock, R. L. , & Lawrence, E. (2008). A longitudinal investigation of stress spillover in marriage: Does spousal support adequacy buffer the effects? Journal of Family Psychology, 22(1), 11–20. 10.1037/0893-3200.22.1.11 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Burke, L. K. , Rauer, A. , & Sabey, A. K. (2019). A developmental perspective on marital closeness and health in older adulthood. Journal of Social and Personal Relationships, 36(8), 2397–2415. 10.1177/0265407518788696 [DOI] [Google Scholar]
- Cacioppo, J. T. , Hawkley, L. C. , & Thisted, R. A. (2010). Perceived social isolation makes me sad: 5‐year cross‐lagged analyses of loneliness and depressive symptomatology in the Chicago health, aging, and social relations study. Psychology and Aging, 25(2), 453–463. 10.1037/a0017216 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Carstensen, L. L. , Isaacowitz, D. M. , & Charles, S. T. (1999). Taking time seriously: A theory of socioemotional selectivity. American Psychologist, 54(3), 165–181. 10.1037/0003-066X.54.3.165 [DOI] [PubMed] [Google Scholar]
- Chen, J. H. (2018). Couples' sleep and psychological distress: A dyadic perspective. The Journals of Gerontology. Series B, Psychological Sciences and Social Sciences, 73(1), 30–39. 10.1093/geronb/gbx001 [DOI] [PubMed] [Google Scholar]
- Chen, Y. , & Feeley, T. H. (2014). Social support, social strain, loneliness, and well‐being among older adults: An analysis of the health and retirement study. Journal of Social and Personal Relationships, 31(2), 141–161. 10.1177/0265407513488728 [DOI] [Google Scholar]
- Cohen, S. , O'Leary, K. D. , & Foran, H. (2010). A randomized clinical trial of a brief, problem‐focused couple therapy for depression. Behavior Therapy, 41(4), 433–446. 10.1016/j.beth.2009.11.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Eyring, J. B. , Leavitt, C. E. , Allsop, D. B. , & Clancy, T. J. (2021). Forgiveness and gratitude: Links between couples' mindfulness and sexual and relational satisfaction in new cisgender heterosexual marriages. Journal of Sex and Marital Therapy, 47(2), 147–161. 10.1080/0092623X.2020.1842571 [DOI] [PubMed] [Google Scholar]
- Fu, R. , & Noguchi, H. (2018). Does the positive relationship between health and marriage reflect protection or selection? Evidence from middle‐aged and elderly Japanese. Review of Economics of the Household, 16(4), 1003–1016. 10.1007/s11150-018-9406-4 [DOI] [Google Scholar]
- Fuller, H. R. , Ajrouch, K. J. , & Antonucci, T. C. (2020). The convoy model and later‐life family relationships. Journal of Family Theory and Review, 12(2), 126–146. 10.1111/jftr.12376 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Fung, H. H. , Carstensen, L. L. , & Lutz, A. M. (1999). Influence of time on social preferences: Implications for life‐span development. Psychology and Aging, 14(4), 595–604. 10.1037//0882-7974.14.4.595 [DOI] [PubMed] [Google Scholar]
- Goldsmith, D. J. (2004). Communicating social support. Cambridge University Press. 10.1017/CBO9780511606984 [DOI] [Google Scholar]
- Gottman, J. M. (1993). A theory of marital dissolution and stability. Journal of Family Psychology, 7(1), 57–75. 10.1037/0893-3200.7.1.57 [DOI] [Google Scholar]
- Greer, C. L. , Worthington, E. L. , Lin, Y. , Lavelock, C. R. , & Griffin, B. J. (2014). Efficacy of a self‐directed forgiveness workbook for Christian victims of within‐congregation offenders. Spirituality in Clinical Practice, 1(3), 218–230. 10.1037/scp0000012 [DOI] [Google Scholar]
- Gustavson, K. , Røysamb, E. , Borren, I. , Torvik, F. A. , & Karevold, E. (2016). Life satisfaction in close relationships: Findings from a longitudinal study. Journal of Happiness Studies, 17(3), 1293–1311. 10.1007/s10902-015-9643-7 [DOI] [Google Scholar]
- Han, S. H. , Kim, K. , Burr, J. A. , & Pruchno, R. (2019). Social support and preventive healthcare behaviors among couples in later life. Gerontologist, 59(6), 1162–1170. 10.1093/geront/gny144 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ho, M. Y. , Worthington, E. L., Jr. , Cowden, R. G. , Bechara, A. O. , Chen, Z. J. , Gunatirin, E. Y. , Joynt, S. , Khalanskyi, V. V. , Korzhov, H. , Kurniati, N. M. T. , Rodríguez, N. , Salnykova, A. , Shtanko, L. , Tymchenko, S. , Voytenko, V. L. , Zulkaida, A. , Mathur, M. B. , & VanderWeele, T. J. (2024). International REACH forgiveness intervention: A multi‐site randomized controlled trial. BMJ Public Health, 2(1), e000072. 10.1136/bmjph-2023-000072 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Holley, S. R. , Haase, C. M. , & Levenson, R. W. (2013). Age‐related changes in demand‐withdraw communication behaviors. Journal of Marriage and Family, 75(4), 822–836. 10.1111/jomf.12051 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hudson, N. W. , Lucas, R. E. , & Donnellan, M. B. (2020). Are we happier with others? An investigation of the links between spending time with others and subjective well‐being. Journal of Personality and Social Psychology, 119(3), 672–694. 10.1037/pspp0000290 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ingersoll‐Dayton, B. , Morgan, D. , & Antonucci, T. (1997). The effects of positive and negative social exchanges on aging adults. Journals of Gerontology ‐ Series B Psychological Sciences and Social Sciences, 52(4). 10.1093/geronb/52B.4.S190 [DOI] [PubMed] [Google Scholar]
- Karney, B. R. , & Bradbury, T. N. (1995). The longitudinal course of marital quality and stability: A review of theory, method, and research. Psychological Bulletin, 118(1), 3–34. 10.1037/0033-2909.118.1.3 [DOI] [PubMed] [Google Scholar]
- Lachman, M. E. , & Weaver, S. L. (1998). The sense of control as a moderator of social class differences in health and well‐being. Journal of Personality and Social Psychology, 74(3), 763–773. 10.1037/0022-3514.74.3.763 [DOI] [PubMed] [Google Scholar]
- Manalel, J. A. , Birditt, K. S. , Orbuch, T. L. , & Antonucci, T. C. (2019). Beyond destructive conflict: Implications of marital tension for marital well‐being. Journal of Family Psychology, 33(5), 597–606. 10.1037/fam0000512 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mancini, A. D. , & Bonanno, G. A. (2006). Marital closeness, functional disability, and adjustment in late life. Psychology and Aging, 21(3), 600–610. 10.1037/0882-7974.21.3.600 [DOI] [PubMed] [Google Scholar]
- Meunier, V. , & Baker, W. (2012). Positive couple relationships: The evidence for long‐lasting relationship satisfaction and happiness. In Roffey S. (Ed.), Positive relationships. Springer. 10.1007/978-94-007-2147-0_5 [DOI] [Google Scholar]
- Moscatelli, M. , Ferrari, C. , Parise, M. , Serrano, C. , & Carrà, E. (2022). “Constructing the we”: Relational reflexivity of couples with children in Italy. A mixed‐method study. Marriage & Family Review, 58(5), 383–412. 10.1080/01494929.2021.1997873 [DOI] [Google Scholar]
- Nejatian, M. , Alami, A. , Momeniyan, V. , Delshad Noghabi, A. , & Jafari, A. (2021). Investigating the status of marital burnout and related factors in married women referred to health centers. BMC Women's Health, 21(1), 1–9. 10.1186/s12905-021-01172-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- OECD . (2021). Health at a glance 2021: OECD indicators. OECD Publishing. [Google Scholar]
- Offer, S. (2021). Negative social ties: Prevalence and consequences. Annual Review of Sociology, 47(1), 177–196. 10.1146/annurev-soc-090820-025827 [DOI] [Google Scholar]
- Parnell, K. J. , Wood, N. D. , & Scheel, M. J. (2020). A gratitude exercise for couples. Journal of Couple & Relationship Therapy, 19(3), 212–229. 10.1080/15332691.2019.1687385 [DOI] [Google Scholar]
- Polenick, C. A. , Martire, L. M. , Hemphill, R. C. , & Stephens, M. A. P. (2015). Effects of change in arthritis severity on spouse well‐being: The moderating role of relationship closeness. Journal of Family Psychology, 29(3), 331–338. 10.1037/fam0000093 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Roberts, J. A. , & David, M. E. (2016). My life has become a major distraction from my cell phone: Partner phubbing and relationship satisfaction among romantic partners. Computers in Human Behavior, 54, 134–141. 10.1016/j.chb.2015.07.058 [DOI] [Google Scholar]
- Ryan, L. H. , Wan, W. H. , & Smith, J. (2014). Spousal social support and strain: Impacts on health in older couples. Journal of Behavioral Medicine, 37(6), 1108–1117. 10.1007/s10865-014-9561-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- Saenz, J. L. (2021). Spousal support, spousal strain, and loneliness in older Mexican couples. The Journals of Gerontology. Series B, Psychological Sciences and Social Sciences, 76(4), 176–186. 10.1093/geronb/gbaa194 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Segel‐Karpas, D. , & Arbel, R. (2022). Optimism, pessimism and support in older couples: A longitudinal study. Journal of Personality, 90(4), 645–657. 10.1111/jopy.12688 [DOI] [PubMed] [Google Scholar]
- Selcuk, E. , Gunaydin, G. , Ong, A. D. , & Almeida, D. M. (2016). Does partner responsiveness predict hedonic and eudaimonic well‐being? A 10‐year longitudinal study. Journal of Marriage and the Family, 78(2), 311–325. 10.1111/JOMF.12272 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Shrout, M. R. (2021). The health consequences of stress in couples: A review and new integrated dyadic biobehavioral stress model. Brain, Behavior, and Immunity Health, 16, 100328. 10.1016/j.bbih.2021.100328 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Slatcher, R. B. , & Selcuk, E. (2017). A social psychological perspective on the links between close relationships and health. Current Directions in Psychological Science, 26(1), 16–21. 10.1177/0963721416667444 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sonnega, A. , Faul, J. D. , Ofstedal, M. B. , Langa, K. M. , Phillips, J. W. R. , & Weir, D. R. (2014). Cohort profile: The health and retirement study (HRS). International Journal of Epidemiology, 43(2), 576–585. 10.1093/ije/dyu067 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Stanton, S. C. E. , Selcuk, E. , Farrell, A. K. , Slatcher, R. B. , & Ong, A. D. (2019). Perceived partner responsiveness, daily negative affect reactivity, and all‐cause mortality: A 20‐year longitudinal study. Psychosomatic Medicine, 81(1), 7–15. 10.1097/PSY.0000000000000618 [DOI] [PMC free article] [PubMed] [Google Scholar]
- StataCorp . (2021). Stata statistical software: Release 17. StataCorp LLC. [Google Scholar]
- Uchino, B. N. , Smith, T. W. , & Berg, C. A. (2014). Spousal relationship quality and cardiovascular risk: Dyadic perceptions of relationship ambivalence are associated with coronary‐artery calcification. Psychological Science, 25(4), 1037–1042. 10.1177/0956797613520015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Vanderweele, T. J. (2017). Outcome‐wide epidemiology. Epidemiology, 28(3), 399–402. 10.1097/EDE.0000000000000641 [DOI] [PMC free article] [PubMed] [Google Scholar]
- VanderWeele, T. J. , & Ding, P. (2017). Sensitivity analysis in observational research: Introducing the E‐value. Annals of Internal Medicine, 167(4), 268–274. 10.7326/M16-2607 [DOI] [PubMed] [Google Scholar]
- VanderWeele, T. J. , Jackson, J. W. , & Li, S. (2016). Causal inference and longitudinal data: A case study of religion and mental health. Social Psychiatry and Psychiatric Epidemiology, 51(11), 1457–1466. 10.1007/s00127-016-1281-9 [DOI] [PubMed] [Google Scholar]
- VanderWeele, T. J. , Mathur, M. B. , & Chen, Y. (2020). Outcome‐wide longitudinal designs for causal inference: A new template for empirical studies. Statistical Science, 35(3), 437–466. 10.1214/19-STS728 [DOI] [Google Scholar]
- Wade, N. G. , Hoyt, W. T. , Kidwell, J. E. M. , & Worthington, E. L. (2014). Efficacy of psychotherapeutic interventions to promote forgiveness: A meta‐analysis. Journal of Consulting and Clinical Psychology, 82(1), 154–170. 10.1037/a0035268 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Tables S1‐S10
Data S1
Data Availability Statement
Data are available upon reasonable request.