Abstract
Individual beliefs are known to be predictive of health. This study examines the co-construction of couple norms and links couples’ shared beliefs about aging to future individuals’ and couples’ functional limitations.
Data from the 2008 and 2014 waves of the Health and Retirement Study (1,231 couples; age range = 51–90) were analyzed using latent variables that estimated shared and individual variance in beliefs about aging in 2008 and functional limitations at follow-up in 2014. Spouses’ individual processes of physical activity and disease burden were modeled to contribute to couples’ shared beliefs about aging and subsequent functional limitations. Models progressively controlled for indicators of partner selection, couples’ shared health experiences, and similarities and differences in age.
Couples’ beliefs about aging predicted future functional limitations. The effect magnitude decreased but remained significant in all models. Physical activity predicted couples’ future functional limitations but was largely explained by shared health experiences and similarities and differences in age for wives and husbands, respectively. Disease burden contributed to couples’ shared beliefs about aging. Husbands’ contributions were explained by partner selection, but wives’ contributions remained significant in all models. The effect of couples’ shared beliefs on change in couples’ functional limitations was explained by couples’ shared health experiences.
Beliefs about aging and health occur within the context of close relationships and shared experiences. Knowledge of couples’ beliefs and health is necessary to support their individual and collective efforts to age successfully together.
Beliefs about one’s own aging are known to have significant consequences for future experiences of aging, including health, health behaviors, and longevity (Westerhof et al., 2014). Although aging is not experienced in isolation, but rather in concert with close others (Antonucci, Fiori, Birditt, & Jackey, 2010; Settersten, Richard, & Hagestad, 2015), little is known about how couples co-construct common beliefs about aging over their extended shared histories. Spouses are known to be connected in their health and well-being (Meyler, Stimpson, & Peek, 2007). Their health behaviors, perceived relationship quality, coping behaviors, and goals have been shown to influence their partners’ experiences (see Hoppmann & Gerstorf, 2009). In addition to directly influencing each other, spouses may also co-create shared norms—through shared experiences and collective action—that is greater than the sum of its parts and exerts independent influence on individual outcomes (Gonzalez & Griffin, 2002). In this study, we use older couples’ shared beliefs about aging to predict future individuals’ and couples’ functional limitations and examine how individual health and health behaviors contribute to these shared processes.
Self-perceptions of Aging and Functional Limitations
Subjective experiences of aging are multidimensional (Brothers, Miche, Wahl, & Diehl, 2015) and develop through experiences within societies (Barrett & Montepare, 2015), personal experiences of aging (Hubley & Russell, 2009), and interactions with others (Schafer & Shippee, 2010). Beliefs about aging encompass attitudes toward one’s own aging, are largely driven by age stereotypes and personal experiences of aging (Diehl et al., 2014), and consistently predict markers of successful aging such as functional ability and longevity (Westerhof et al., 2014). Beliefs about aging affect future health through cognitive and behavioral pathways (Levy, 2009). As a cognitive frame, beliefs about aging shape how individuals experience age-related change in their bodies, relationships, and roles (Stephan, Chalabaev, Kotter-Grühn, & Jaconelli, 2013; Stewart, Chipperfield, Perry, & Weiner, 2012). In addition, beliefs about aging motivate health behaviors and have been prospectively linked to physical activity and preventative care (Kim, Moored, Giasson, & Smith, 2014). Taken together, beliefs about aging form self-fulfilling prophecies for future experiences of health and well-being in old age (Levy, 2009) and are predictive of individuals’ future health (Sargent-Cox, Anstey, & Luszcz, 2014).
How individuals understand their own aging is intersubjective—the meaning of age and the experience of aging is constructed through interactions in society and with others (Settersten et al., 2015). Married couples are intimate family units of unrelated individuals who elected each other to navigate a shared life course. Many couples in older adulthood share extended life histories and have traveled together through age-graded roles and responsibilities. Within their tightly interwoven lives, a spouse’s beliefs have implications for the partner’s experience of aging (Drewelies, Chopik, Hoppmann, Smith, & Gerstorf, 2016). Similarly, one spouse’s experiences of health and illness are known to shape relationship processes (Martire et al., 2006). This study examines how individual beliefs about aging, health behaviors, and health experiences contribute to the co-construction of norms that are unique to the couple and exert an independent and unique influence on health outcomes (see Figure 1).
Common fate model of couples’ self-perceptions of aging on future functional limitations with contribution of individual processes on shared processes. Note: Loadings within constructs are constrained to 1. Variances within constructs across gender are constrained to equal. Individual variances (i) were allowed to covary within gender and were tested for equality/inequality across gender.
Common fate model of couples’ self-perceptions of aging on future functional limitations with contribution of individual processes on shared processes. Note: Loadings within constructs are constrained to 1. Variances within constructs across gender are constrained to equal. Individual variances (i) were allowed to covary within gender and were tested for equality/inequality across gender.
Couples’ Co-construction of Shared Beliefs
Individual beliefs about aging are constructed in part through experiences of health. In this study, we examine how physical activity and disease burden—two individual processes—shape couples’ shared beliefs about aging and predict future functional limitations. We selected physical activity and disease burden because of their relevance for subjective experiences of aging (Hubley & Russell, 2009), relationship processes (Rook, 2015), and health policy and interventions (Martire, Schulz, Helgeson, Small, & Saghafi, 2010). Spouses establish norms that guide how they support one another in managing chronic illness, as well as encourage each other to engage in positive health behaviors (Tucker & Mueller, 2000). Evidence also suggests that one spouse’s physical activity and disease burden has implications for both spouses’ health and well-being. For example, one spouse’s engagement in physical activity could influence the partner’s activity engagement (Franks, Wendorf, Gonzalez, & Ketterer, 2004; Li, Cardinal, & Acock, 2013) and potentially shape beliefs about aging (Levy, 2009). Similarly, partners are intimately involved in the disease process (Franks, Lucas, Stephens, Rook, & Gonzalez, 2010; Monin et al., 2010). Managing chronic conditions can increase caregiving burden and limit opportunities for both partners (Polenick, Martire, Hemphill, & Stephens, 2015), to the detriment of couples’ beliefs about aging and functional limitations. Experiences of aging are also gendered (Calasanti, 2010; Kornadt, Voss, & Rothermund, 2013). Husbands and wives may therefore differ in how their individual experiences contribute to couples’ shared norms of health and shared beliefs about aging.
Couple-level processes stem from partner selection and shared experiences, which have implications for how interventions and policy can support couples in constructing positive shared beliefs about aging. Partners select one another across the axes of age, education, and race (Schwartz, 2013), which distribute opportunities and exert constrains on the couple unit and guide how couples move together through their shared life course. Through their shared experiences, couples co-construct norms of support, companionship, and social control (Rook, 2015), which shape couple processes and amplify individual experiences (Gonzalez & Griffin, 2002). In this study, we focus on couples’ shared experiences of health and health behaviors because health involves both spouses and health behaviors are formed in response to experiences within the couples’ shared environment (Franks et al., 2004; Li et al., 2013; Tucker & Mueller, 2000). In addition to guiding partner selection, age indicates the experience of historical events, biological and psychological aging processes, and transitions through age-graded roles and responsibilities (Settersten et al., 2015). Similarities and differences in age within couples therefore have implications for how couples’ shared beliefs are shaped by individual processes and predict future health.
Present Study
Together, extant research on the interdependence of health and well-being within couples offers a compelling foundation to examine older couples’ shared beliefs about aging. Our study has three aims. Our first aim is to quantify the extent to which beliefs about aging are shared within couples and linked to individuals’ and couples’ future functional limitations. Our second aim is to examine how individual processes contribute to shared processes of couples’ beliefs about aging and functional limitations. We test the contributions of spouses’ individual physical activity and disease burden to couples’ beliefs about aging and subsequent functional limitations. Our third aim is to examine the extent to which indicators of partner selection, shared health experiences, and similarities and differences in age within couples account for these associations.
Methods
The current study examines couples’ shared beliefs about aging and future functional limitations using data from the 2008 and 2014 waves of the Health and Retirement Study (HRS). The HRS is a National Institute of Aging (NIA U01AG009740) sponsored biennial panel study that is conducted by the University of Michigan and nationally representative of Americans aged 51 and older (Sonnega et al., 2014). Sampling is conducted using a multistage probability design at the household level. If the primary respondent is coupled, both partners are interviewed and brought into the HRS panel.
Beginning in 2006, 50% of households were randomly assigned to an in-person interview at their place of residence and the other 50% was interviewed by telephone. Interviews included questions about functional limitations, health, and demographics. Following the in-person interview, respondents were left a paper self-administered psychosocial questionnaire (SAQ) to complete and return by mail. Starting in 2008, the baseline for this study, the SAQ included questions on beliefs about aging.
Participants
In the 2008 wave, 1,992 married couples (3,984 individuals), where both spouses were age eligible (M = 68, SD = 9.01, range = 51–95), community-residing, and non-proxy respondents, were assigned to the in-person interview. Among eligible couples, 79% identified as White/Caucasian, 9% identified as African American, 12% identified as Hispanic/other, and 29% had completed at least some college. From this sample of eligible couples, there were 1,883 couples (86%) where both (n = 1,723) or at least one (n = 160) spouse completed and returned the SAQ. Logistic regression showed those who completed the SAQ tended to be older and identify as White.
Measures
The measures of individual and shared processes follow. Descriptive statistics are provided for the analytic sample (1,231 couples).
Shared Processes
Beliefs about aging and functional limitations were modeled as the couples’ shared processes and were adjusted for variance at the individual level.
Beliefs About Aging.
Respondents’ beliefs about aging were measured at the 2008 wave using the Philadelphia Geriatric Center Morale subscale on attitudes toward on aging (Lawton, 1975). The subscale included three positive (e.g., I have as much as pep as I did last year) and two negative (e.g., Things keep getting worse as I get older) statements about one’s own aging. Respondents rated each statement from (1) strongly disagree to (6) strongly agree. The beliefs score was constructed for each respondent by reverse coding the negative items and then calculating the mean (M = 4.12, SD = 1.08). The scale showed good internal consistency (Cronbach’s α = .72).
Functional Limitations.
Respondents’ reports of difficulty completing basic physical activities were measured in the 2008 and 2014 waves (Fonda & Herzog, 2004). Activities (walking several blocks, jogging 1 mile, walking 1 block, sitting for 2 hr, getting up from a chair, climbing stairs, climbing 1 flight of stairs, stooping, reaching arms above shoulders, pushing and pulling large objects, lifting weights, and picking up a dime) that respondents reported having difficulty completing due to health problems were summed (M = 2.59, SD = 2.71, range = 0–10).
Individual Processes
Physical activity and disease burden were adjusted for interdependence within couples and examined separately as individual processes that contribute to couples’ shared beliefs about aging and future functional limitations.
Physical Activity.
The frequency of engagement in moderate activity was measured in the 2008 wave. Respondents reported their frequency of doing moderately energetic activities such as gardening, cleaning the car, or walking. Responses on the single-item scale ranged from (4) “hardly ever or never” to (1) “more than once a week” and were reverse coded so that a higher score indicated greater frequency of moderate activity (M = 3.11, SD = 1.16). We acknowledge the conceptual overlap in the items that comprise functional ability and physical activity. We assume the future ability to do an activity (functional ability) to be distinct from the frequency of actually doing the activity (physical activity).
Disease Burden.
Respondents’ disease burden was measured in 2008 as the count of reported diagnoses of: heart disease, lung disease, diabetes, cancer, stroke, and arthritis. Diagnoses of cancer and stroke were counted if the respondent reported receiving treatment for the condition within the last 2 years. Signs of depression were counted toward disease burden if respondents reported more than four symptoms in the 8-item Center for Epidemiological Studies Depression scale (CES-D; Steffick, 2000). Disease burden ranged from 0 to 7 (M = 1.53, SD = 1.19).
Covariates
Covariates were added in blocks to explain how couples’ individual and shared processes predicted future functional limitations. The first block addressed partner selection and included sociodemographic characteristics of educational attainment (25% with at least some college), race, and sample quintiles of household wealth. The second covariate block added indicators of couples’ shared health experiences and adjusted for: body mass index (BMI categorized as <18.5–24.9 [27%]; 25–29.9 [40%]; 30+ [33%]), alcohol use (59% currently drink alcohol), and whether or not the respondent had ever smoked (57% have smoked). The health and health behavior block adjusted for disease burden in physical activity models, and physical activity in disease burden models. The final model added age to adjust for similarities and differences in age within couples. All covariates, except for BMI (missing = 19) and smoking (missing = 18) were fully present within the analytic sample. Missing values were accommodated using full information maximum likelihood.
Loss to Follow-up
In total, 562 (14%) respondents died before the 2014 follow-up. Logistic regression found couples’ and individuals’ functional limitations, age, disease burden, frequency of physical activity, and smoking to predict vital status at follow-up. In addition, 53 couples (106 individuals) were alive but formed separate households between 2008 and 2014, and 239 individuals were alive but missed the 2014 interview. The likelihood of exclusion from analysis due to the couples’ separation or missed interview did not systematically vary across study covariates. Across all causes of loss to follow-up, those who attritted tended to have more negative couples’ (but not individual) beliefs about aging as well as greater functional limitations at the couple level and fewer functional limitations relative to their couple norm (see Supplementary Table 1). The final analytic sample included 1,231 couples where at least one spouse completed the SAQ in 2008 and both spouses were alive and completed the follow-up interview in 2014 (see Supplementary Figure 1).
Analytic Plan
Common fate models (Ledermann & Kenny, 2012), also known as latent dyadic models (Gonzalez & Griffin, 2002), were used to link couples’ shared variance in beliefs about aging to their shared variance in functional limitations at follow-up 6 years later. The common fate model is a form of structural equation model that parses observed variables into variation that is attributed to the couple and individual. A latent variable represents couples’ shared variance (similar to a couple mean). The residual represents the spouses’ remaining individual (or unique) variance to depict individual variation above and below the couple mean. The common fate model allows couples to vary from other couples (some couples may have more positive beliefs about aging than others) and for spouses within the dyad to vary within couples (spouses may differ from each other in their beliefs about aging). We used the common fate model to link couple variances over and above effects at the individual level. This approach is distinct from the commonly used actor-partner interdependence model, which adjusts for interdependence within couples and then links the remaining individual variance to spouses’ individual outcomes.
In preparation for analysis, separate measurement models were constructed to parse couple from individual variance. Factor loadings were constrained to 1 and residuals were constrained to be equal across gender. This model is mathematically equivalent to an unconditional model in the multilevel context.
To accomplish Aim 1, we first used the unconditional models to estimate the intraclass correlations, which represent both the proportion of total variation attributed to couple membership and the extent of similarity within couples. We then regressed couples’ shared processes of functional limitations at follow-up on couples’ beliefs about aging, adjusting for individual covariances. To address Aim 2, we modeled physical activity and disease burden as individual processes—adjusted for interdependence within couples—to contribute to couples’ beliefs about aging and predict couples’ functional limitations. Individual variances were constrained to be equal within each construct to scale the betas and then covaried within gender to simultaneously estimate couple and individual processes. Functional limitations and disease burden were measured as continuous rather than count variables because an error term must be estimated to simultaneously examine individual and couple-level processes using this analytic framework. The key couple-level effects were similar in both their magnitude and significance when functional limitations were modeled as a count variable. Husbands and wives’ covariances and paths from individual parameters were tested for gender invariance using the log likelihood test.
To complete Aim 3, covariates were added sequentially in blocks to test explanations for couples’ individual and shared processes. All exogenous observed variables (all covariates) were correlated with one another. We concluded analysis by testing the effect of individuals’ and couples’ beliefs about aging on residual change in functional limitations. In these models, both beliefs about aging in 2008 and functional limitations in 2014 were adjusted for functional limitations in 2008. Residual change was examined first as unadjusted and then sequentially adjusted by the partner selection, shared health experiences, and age similarity/differences blocks. Data were analyzed using MPLUS.
Results
Couples’ Beliefs About Aging and Functional Limitations
Descriptive statistics of study covariates are presented in Table 1. Chi-squared tests and analyses of variance showed that husbands tended to be older and have higher educational attainment, greater disease burden, greater physical activity, negative beliefs about aging, and fewer functional limitations than wives. To accomplish our first aim, we examined the extent to which beliefs about aging were similar within couples and predicted couples’ future functional limitations. In total, 35% of the variation in beliefs about aging, 21% of variation in functional limitations, 17% of variation in physical activity, and 19% of variation in disease burden was attributed to couple membership (see Supplementary Table 2). Couples’ beliefs about aging predicted couples’ functional limitations at follow-up, (β = −.60, SE = 0.03, p < .001). Couples with more positive beliefs tended to have fewer functional limitations at follow-up. The likelihood test (Gonzalez & Griffin, 2001) found this path to be significantly different from zero (χ2(1) = 26.78, p < .001). Individual beliefs about aging were also correlated with individual functional limitations at follow-up; (rhis(beliefs, limitations) = −.29, SE = 0.03, p < .001; rher(beliefs, limitations) = −.35, SE = 0.04, p < .001). Although allowing variation in individual effects across gender did not significantly improve model fit (χ2(1) = 1.40, ns), the following models remained unconstrained across gender due to our interest in within-couple differences.
Descriptive Characteristics of Study Covariates Across Gender (N = 1,231 Couples)
| Men | Women | ||||||
|---|---|---|---|---|---|---|---|
| M | % | SD | M | % | SD | p | |
| Age | 67.94 | 8.07 | 65.00 | 7.88 | <.001 | ||
| 51–64 | 34% | 48% | <.001 | ||||
| 65–80 | 58% | 49% | |||||
| 80+ | 8% | 4% | |||||
| Race | |||||||
| White | 81% | 80% | |||||
| African American | 8% | 8% | |||||
| Hispanic/other | 12% | 12% | |||||
| Some college | 35% | 30% | =.003 | ||||
| Disease burden | 1.39 | 1.13 | 1.25 | 1.04 | =.002 | ||
| Diabetes | 23% | 14% | <.001 | ||||
| Cancer | 13% | 9% | =.002 | ||||
| Lung disease | 8% | 8% | |||||
| Heart disease | 28% | 16% | <.001 | ||||
| Stroke | 5% | 3% | =.01 | ||||
| Arthritis | 55% | 63% | <.001 | ||||
| Depression | 10% | 14% | =.002 | ||||
| Physical activity | 2.43 | 1.97 | 2.47 | 1.91 | |||
| Self-perceptions of aging | 4.20 | 1.06 | 4.30 | 1.03 | =.05 | ||
| N functional limitations 2008 | 1.58 | 1.98 | 2.04 | 2.32 | <.001 | ||
| N functional limitations 2014 | 2.25 | 2.60 | 2.51 | 2.70 | =.02 | ||
| Walking several blocks | 28% | 30% | |||||
| Jogging 1 mile | 66% | 66% | |||||
| Walking 1 block | 14% | 13% | |||||
| Sitting for 2 hr | 14% | 17% | |||||
| Getting up from chair | 37% | 37% | |||||
| Climbing flights of stairs | 41% | 47% | =.002 | ||||
| Climbing 1 flight of stairs | 14% | 18% | =.005 | ||||
| Stooping | 45% | 45% | |||||
| Reaching arms above shoulders | 16% | 15% | |||||
| Pulling/pushing large objects | 21% | 27% | <.001 | ||||
| Lifting heavy objections | 16% | 25% | <.001 | ||||
| Picking up a dime | 9% | 8% | |||||
| Men | Women | ||||||
|---|---|---|---|---|---|---|---|
| M | % | SD | M | % | SD | p | |
| Age | 67.94 | 8.07 | 65.00 | 7.88 | <.001 | ||
| 51–64 | 34% | 48% | <.001 | ||||
| 65–80 | 58% | 49% | |||||
| 80+ | 8% | 4% | |||||
| Race | |||||||
| White | 81% | 80% | |||||
| African American | 8% | 8% | |||||
| Hispanic/other | 12% | 12% | |||||
| Some college | 35% | 30% | =.003 | ||||
| Disease burden | 1.39 | 1.13 | 1.25 | 1.04 | =.002 | ||
| Diabetes | 23% | 14% | <.001 | ||||
| Cancer | 13% | 9% | =.002 | ||||
| Lung disease | 8% | 8% | |||||
| Heart disease | 28% | 16% | <.001 | ||||
| Stroke | 5% | 3% | =.01 | ||||
| Arthritis | 55% | 63% | <.001 | ||||
| Depression | 10% | 14% | =.002 | ||||
| Physical activity | 2.43 | 1.97 | 2.47 | 1.91 | |||
| Self-perceptions of aging | 4.20 | 1.06 | 4.30 | 1.03 | =.05 | ||
| N functional limitations 2008 | 1.58 | 1.98 | 2.04 | 2.32 | <.001 | ||
| N functional limitations 2014 | 2.25 | 2.60 | 2.51 | 2.70 | =.02 | ||
| Walking several blocks | 28% | 30% | |||||
| Jogging 1 mile | 66% | 66% | |||||
| Walking 1 block | 14% | 13% | |||||
| Sitting for 2 hr | 14% | 17% | |||||
| Getting up from chair | 37% | 37% | |||||
| Climbing flights of stairs | 41% | 47% | =.002 | ||||
| Climbing 1 flight of stairs | 14% | 18% | =.005 | ||||
| Stooping | 45% | 45% | |||||
| Reaching arms above shoulders | 16% | 15% | |||||
| Pulling/pushing large objects | 21% | 27% | <.001 | ||||
| Lifting heavy objections | 16% | 25% | <.001 | ||||
| Picking up a dime | 9% | 8% | |||||
Note: Gender differences tested using chi-squared analysis for categorical variables and analysis variance for continuous variables. N = number.
Contributions of Individual Physical Activity and Disease Burden
To accomplish Aim 2, individual processes of physical activity (Table 2) and disease burden (Table 3) were added to the models. Spouses’ individual physical activity was significantly linked to more positive couples’ beliefs about aging and predicted lower couples’ functional limitations at follow-up. Individual disease burden was associated with more negative couples’ beliefs about aging and predicted greater couples’ functional limitations at follow-up. The likelihood test found neither individual physical activity nor disease burden to vary significantly by gender in their contributions to couples’ shared processes (see Supplementary Table 3). At the individual level, the covariance of chronic conditions and functional limitations was stronger for wives than husbands (χ2(1) = 8.58, p < .01).
Individual Physical Activity and Couples’ Shared and Individual Self-perceptions of Aging in 2008 Predicting Couples’ Shared and Individual Functional Limitations in 2014 (N = 1,231 Couples)
| Unadjusted | Sociodemographicsa | Healthb | Chronological agec | |||||
|---|---|---|---|---|---|---|---|---|
| β | SE | β | SE | β | SE | β | SE | |
| Couple level | ||||||||
| Couples’ SPA ← | ||||||||
| His physical activity | .17** | 0.06 | .08 | 0.06 | .04 | 0.06 | .04 | 0.06 |
| Her physical activity | .17** | 0.06 | .11† | 0.06 | .10 | 0.06 | .10 | 0.06 |
| Couples’ limitations ← | ||||||||
| Couples’ SPA | −.55*** | 0.09 | −.50*** | 0.12 | −.41** | 0.14 | −.46** | 0.17 |
| His physical activity | −.28** | 0.09 | −.26* | 0.11 | −.24* | 0.11 | −.25† | 0.14 |
| Her physical activity | −.26** | 0.08 | −.24* | 0.10 | −.21† | 0.11 | −.20 | 0.13 |
| Individual level (residual correlations) | ||||||||
| SPA ← | ||||||||
| His physical activity | .19*** | 0.04 | .20*** | 0.04 | .17*** | 0.04 | .16*** | 0.04 |
| Her physical activity | .11* | 0.04 | .09* | 0.05 | .04 | 0.05 | .05 | 0.05 |
| Limitations ← | ||||||||
| His SPA | −.27*** | 0.04 | −.28*** | 0.03 | −.22*** | 0.04 | −.21*** | 0.04 |
| Her SPA | −.31*** | 0.03 | −.30*** | 0.03 | −.21*** | 0.04 | −.22*** | 0.04 |
| His physical activity | −.21*** | 0.05 | −.20*** | 0.05 | −.17*** | 0.04 | −.16*** | 0.04 |
| Her physical activity | −.22*** | 0.04 | −.21*** | 0.04 | −.16*** | 0.04 | −.15*** | 0.04 |
| Model fit indices | ||||||||
| RMSEA | .04 | .03 | .02 | .01 | ||||
| CFI | .99 | .98 | .99 | .99 | ||||
| BIC | 25475.574 | 30556.105 | 49811.146 | 51982.25 | ||||
| χ2 (model vs. saturated) | χ2(4) = 9.95* | χ2(22) = 40.82* | χ2(52) = 69.73 | χ2(64) = 76.03 | ||||
| Unadjusted | Sociodemographicsa | Healthb | Chronological agec | |||||
|---|---|---|---|---|---|---|---|---|
| β | SE | β | SE | β | SE | β | SE | |
| Couple level | ||||||||
| Couples’ SPA ← | ||||||||
| His physical activity | .17** | 0.06 | .08 | 0.06 | .04 | 0.06 | .04 | 0.06 |
| Her physical activity | .17** | 0.06 | .11† | 0.06 | .10 | 0.06 | .10 | 0.06 |
| Couples’ limitations ← | ||||||||
| Couples’ SPA | −.55*** | 0.09 | −.50*** | 0.12 | −.41** | 0.14 | −.46** | 0.17 |
| His physical activity | −.28** | 0.09 | −.26* | 0.11 | −.24* | 0.11 | −.25† | 0.14 |
| Her physical activity | −.26** | 0.08 | −.24* | 0.10 | −.21† | 0.11 | −.20 | 0.13 |
| Individual level (residual correlations) | ||||||||
| SPA ← | ||||||||
| His physical activity | .19*** | 0.04 | .20*** | 0.04 | .17*** | 0.04 | .16*** | 0.04 |
| Her physical activity | .11* | 0.04 | .09* | 0.05 | .04 | 0.05 | .05 | 0.05 |
| Limitations ← | ||||||||
| His SPA | −.27*** | 0.04 | −.28*** | 0.03 | −.22*** | 0.04 | −.21*** | 0.04 |
| Her SPA | −.31*** | 0.03 | −.30*** | 0.03 | −.21*** | 0.04 | −.22*** | 0.04 |
| His physical activity | −.21*** | 0.05 | −.20*** | 0.05 | −.17*** | 0.04 | −.16*** | 0.04 |
| Her physical activity | −.22*** | 0.04 | −.21*** | 0.04 | −.16*** | 0.04 | −.15*** | 0.04 |
| Model fit indices | ||||||||
| RMSEA | .04 | .03 | .02 | .01 | ||||
| CFI | .99 | .98 | .99 | .99 | ||||
| BIC | 25475.574 | 30556.105 | 49811.146 | 51982.25 | ||||
| χ2 (model vs. saturated) | χ2(4) = 9.95* | χ2(22) = 40.82* | χ2(52) = 69.73 | χ2(64) = 76.03 | ||||
Note: Standardized β coefficients presented. BIC = Bayesian Information Criterion; CFI = Cumulative Fit Index; RMSEA = Root Mean Square Error of Approximation; SPA = self-perceptions of aging.
Models are additive and adjusted for: aIndividual’s race and education, and sample quintiles of household wealth. bIndividual smoking status, use of alcohol (1 = drinks alcohol, 0 = does not drink alcohol), body mass index (<18.5–24.9 [ref.], 25–29.9, 30+), and number of chronic conditions. cIndividual age group (51–64, 65–79, 80+).
†p < .10. *p < .05. **p < .01. ***p < .001.
Individual Disease Burden and Couples’ Shared and Individual Self-perceptions of Aging in 2008 Predicting Couples’ Shared and Individual Functional Limitations in 2014 (N = 1,231 Couples)
| Unadjusted | Sociodemographicsa | Healthb | Chronological agec | |||||
|---|---|---|---|---|---|---|---|---|
| β | SE | β | SE | β | SE | β | SE | |
| Couple level | ||||||||
| Couples’ SPA ← | ||||||||
| His disease burden | −.21*** | 0.06 | −.13* | 0.06 | −.10† | 0.06 | −.09 | 0.06 |
| Her disease burden | −.32*** | 0.06 | −.26*** | 0.07 | −.23*** | 0.07 | −.21*** | 0.06 |
| Couples’ limitations ← | ||||||||
| Couples’ SPA | −.39** | 0.13 | −.33* | 0.15 | −.38* | 0.16 | .49** | 0.18 |
| His disease burden | .21*** | 0.11 | .25* | 0.12 | .22† | 0.13 | .02 | 0.13 |
| Her disease burden | .32*** | 0.11 | .24† | 0.13 | .16 | 0.13 | .02 | 0.14 |
| Individual level (residual correlations) | ||||||||
| SPA ← | ||||||||
| His disease burden | −.24*** | 0.04 | −.25*** | 0.04 | −.22*** | 0.04 | −.21*** | 0.04 |
| Her disease burden | −.26*** | 0.05 | −.25*** | 0.05 | −.24*** | 0.04 | −.24*** | 0.04 |
| Limitations ← | ||||||||
| His SPA | −.29*** | 0.03 | −.29*** | 0.03 | −.26*** | 0.03 | −.25*** | 0.03 |
| Her SPA | −.30*** | 0.04 | −.29*** | 0.04 | −.27*** | 0.04 | −.28*** | 0.04 |
| His disease burden | .29*** | 0.04 | .29*** | 0.04 | .27*** | 0.04 | .27*** | 0.04 |
| Her disease burden | .45*** | 0.03 | .44*** | 0.03 | .42*** | 0.03 | .41*** | 0.03 |
| Model fit indices | ||||||||
| RMSEA | .04 | .04 | .02 | .02 | ||||
| CFI | .99 | .97 | .98 | .98 | ||||
| BIC | 25149.31 | 30236.69 | 49828.50 | 52012.07 | ||||
| χ2 (model vs. saturated) | χ2(4) = 12.06* | χ2(22) = 64.21** | χ2(52) = 87.08** | χ2(64) = 105.85*** | ||||
| Unadjusted | Sociodemographicsa | Healthb | Chronological agec | |||||
|---|---|---|---|---|---|---|---|---|
| β | SE | β | SE | β | SE | β | SE | |
| Couple level | ||||||||
| Couples’ SPA ← | ||||||||
| His disease burden | −.21*** | 0.06 | −.13* | 0.06 | −.10† | 0.06 | −.09 | 0.06 |
| Her disease burden | −.32*** | 0.06 | −.26*** | 0.07 | −.23*** | 0.07 | −.21*** | 0.06 |
| Couples’ limitations ← | ||||||||
| Couples’ SPA | −.39** | 0.13 | −.33* | 0.15 | −.38* | 0.16 | .49** | 0.18 |
| His disease burden | .21*** | 0.11 | .25* | 0.12 | .22† | 0.13 | .02 | 0.13 |
| Her disease burden | .32*** | 0.11 | .24† | 0.13 | .16 | 0.13 | .02 | 0.14 |
| Individual level (residual correlations) | ||||||||
| SPA ← | ||||||||
| His disease burden | −.24*** | 0.04 | −.25*** | 0.04 | −.22*** | 0.04 | −.21*** | 0.04 |
| Her disease burden | −.26*** | 0.05 | −.25*** | 0.05 | −.24*** | 0.04 | −.24*** | 0.04 |
| Limitations ← | ||||||||
| His SPA | −.29*** | 0.03 | −.29*** | 0.03 | −.26*** | 0.03 | −.25*** | 0.03 |
| Her SPA | −.30*** | 0.04 | −.29*** | 0.04 | −.27*** | 0.04 | −.28*** | 0.04 |
| His disease burden | .29*** | 0.04 | .29*** | 0.04 | .27*** | 0.04 | .27*** | 0.04 |
| Her disease burden | .45*** | 0.03 | .44*** | 0.03 | .42*** | 0.03 | .41*** | 0.03 |
| Model fit indices | ||||||||
| RMSEA | .04 | .04 | .02 | .02 | ||||
| CFI | .99 | .97 | .98 | .98 | ||||
| BIC | 25149.31 | 30236.69 | 49828.50 | 52012.07 | ||||
| χ2 (model vs. saturated) | χ2(4) = 12.06* | χ2(22) = 64.21** | χ2(52) = 87.08** | χ2(64) = 105.85*** | ||||
Note: Standardized β coefficients presented. BIC = Bayesian Information Criterion; CFI = Cumulative Fit Index; RMSEA = Root Mean Square Error of Approximation; SPA = self-perceptions of aging.
Models are additive and adjusted for: aIndividual’s race and education, and sample quintiles of household wealth. bIndividual smoking status, use of alcohol (1 = drinks alcohol, 0 = does not drink alcohol), body mass index (<18.5–24.9 [ref.], 25–29.9, 30+), and frequency of moderate physical activity. cIndividual age group (51–64, 65–79, 80+).
†p < .10. *p < .05. **p < .01. ***p < .001.
Selection, Shared Experiences, and Age Similarities/Differences
To address Aim 3, we progressively adjusted for indicators of partner selection (sociodemographics), shared health experiences (health and health behaviors), and similarity and differences in age to explain how couples’ shared and individual processes were predictive of future functional limitations. We began with adjusting for sociodemographics of race, education, and wealth, which are known to be implicated in partner selection. As presented in Tables 2 and 3, the effect of couples’ beliefs about aging was dampened but continued to predict couples’ future functional limitations. Adjusting for partner selection rendered the effects of physical activity on couples’ beliefs about aging nonsignificant (Table 2, sociodemographics), but spouses’ physical activity continued to significantly predict couples’ future functional limitations. In contrast, the effects of spouses’ disease burden on couple processes remained robust, although more so for wives’ than husbands (see Table 3, sociodemographics).
In our second set of models, we adjusted for health and health behaviors—covariates that are representative of couples’ shared health experiences. The effect of couples’ beliefs about aging on future functional limitations was slightly reduced in both the physical activity (Table 2) and disease burden (Table 3) models. The remaining contribution of physical activity to couples’ functional limitations was rendered nonsignificant once shared health experiences, including disease burden, were also accounted for (Table 2, health). The contribution of husbands’ disease burden to couples’ beliefs about aging was explained by couples’ shared health experiences. The effect of spouses’ disease burden on couples’ functional limitations was rendered nonsignificant (Table 3, health).
The third model added adjustments for similarities and differences in spouses’ age. Couples’ beliefs about aging continued to predict couples’ future functional limitations in the fully adjusted models. Accounting for husbands’ and wives’ age rendered all remaining effects of physical activity on couples’ functional limitations nonsignificant (Table 2, chronological age). The effect of wives’ disease burden on couples’ beliefs about aging remained significant.
We concluded by testing the effects of couples’ beliefs about aging on couples’ shared processes of change in functional limitations (see Table 4). The contributions of couples’ beliefs about aging on couples’ shared processes of change in functional limitations was largely explained by indicators of shared health experiences.
Couples’ Shared and Individual Self-perceptions of Aging on Residual Change in Functional Limitations (N = 1,231 Couples)
| Unadjusteda | Sociodemographicsb | Healthc | Chronological aged | |||||
|---|---|---|---|---|---|---|---|---|
| β | SE | β | SE | β | SE | β | SE | |
| Couple level | ||||||||
| Δ Couples’ limitations ← | ||||||||
| Couples’ SPA | −.44** | 0.15 | −.38* | 0.15 | −.32† | 0.17 | −.41 | 0.28 |
| Individual level (residual correlations) | ||||||||
| Δ Limitations ← | ||||||||
| His SPA | −.14*** | 0.03 | −.15*** | 0.04 | −.13*** | 0.04 | −.13*** | 0.04 |
| Her SPA | −.14** | 0.05 | −.12** | 0.04 | −.09** | 0.04 | −.09* | 0.04 |
| Model fit indices | ||||||||
| RMSEA | .18 | .03 | .02 | .02 | ||||
| CFI | .85 | .99 | .99 | .99 | ||||
| BIC | 27942.66 | 32782.03 | 59314.97 | 61541.74 | ||||
| χ2 (model vs. saturated) | χ2(7) = 298.34*** | χ2(19) = 40.83** | χ2(43) = 67.83** | χ2(51) = 72.09* | ||||
| Unadjusteda | Sociodemographicsb | Healthc | Chronological aged | |||||
|---|---|---|---|---|---|---|---|---|
| β | SE | β | SE | β | SE | β | SE | |
| Couple level | ||||||||
| Δ Couples’ limitations ← | ||||||||
| Couples’ SPA | −.44** | 0.15 | −.38* | 0.15 | −.32† | 0.17 | −.41 | 0.28 |
| Individual level (residual correlations) | ||||||||
| Δ Limitations ← | ||||||||
| His SPA | −.14*** | 0.03 | −.15*** | 0.04 | −.13*** | 0.04 | −.13*** | 0.04 |
| Her SPA | −.14** | 0.05 | −.12** | 0.04 | −.09** | 0.04 | −.09* | 0.04 |
| Model fit indices | ||||||||
| RMSEA | .18 | .03 | .02 | .02 | ||||
| CFI | .85 | .99 | .99 | .99 | ||||
| BIC | 27942.66 | 32782.03 | 59314.97 | 61541.74 | ||||
| χ2 (model vs. saturated) | χ2(7) = 298.34*** | χ2(19) = 40.83** | χ2(43) = 67.83** | χ2(51) = 72.09* | ||||
Note: Standardized β coefficients presented. BIC = Bayesian Information Criterion; CFI = Cumulative Fit Index; RMSEA = Root Mean Square Error of Approximation; SPA = self-perceptions of aging.
Models are additive and adjusted for: aFunctional limitations in 2008. bIndividual’s race and education, and sample quintiles of household wealth. cIndividual smoking status, use of alcohol (1 = drinks alcohol, 0 = does not drink alcohol), body mass index (<18.5–24.9 [ref.], 25–29.9, 30+), and frequency of moderate physical activity. dIndividual age group (51–64, 65–79, 80+).
†p < .07. *p < .05. **p < .01. ***p < .001.
Discussion
Signs of one’s own aging are ubiquitous in daily life. Coded signals from institutions, the media, and interactions with others make changes in physical ability and appearance, relationships, and roles salient in an age-structured society (Levy, 2009; Settersten et al., 2015). Aging is inescapable, but beliefs about aging are malleable (Stephan et al., 2013), can remain positive until the end of life (Kotter-Grühn, Kleinspehn-Ammerlahn, Gerstorf, & Smith, 2009), and are predictive of future health, well-being, and longevity (Westerhof et al., 2014). Our research extended the proposition that aging is not experienced alone, but in concert with close others (Antonucci et al., 2010; Settersten et al., 2015). We examined the extent and implications of older couples’ shared beliefs about their own aging. We found couples to be similar in beliefs about aging, their shared beliefs to predict future functional limitations, and their individual experiences of health and health behaviors to shape these processes.
Beliefs about aging develop over a lifetime of experiences with age stereotypes, age-graded social institutions, and interactions with others. Our research showed that a striking 35% of the variation in beliefs about aging is shared within couples and hold enduring implications for future health. The couple’s shared environment represents co-constructed couple norms, rituals, and tendencies that shape how spouses interact with and support one another (Rook, 2015; Tucker & Mueller, 2000). Our research identified a unique and separate pathway by which beliefs about aging are linked to future functional limitations—that the shared environment exerts an independent influence on future health. Beliefs about aging influence health through behavioral and cognitive pathways (Levy, 2009) and shared beliefs may shape how couples jointly appraise, interpret, and respond to experiences of aging (Rook, 2015; Tucker & Mueller, 2000). In designing policies and interventions to improve individuals’ beliefs about aging, our findings highlight the importance of recognizing individual beliefs as part of a larger whole. Policies and interventions are most often directed at the individual, and attention to the couple environment provides an opportunity to amplify a program’s intended effects (Martire et al., 2010). Our findings justify further research on couple-level effects. For example, cognitive framing interventions could include components where couples discuss aging beliefs together. Alternatively, an estimation of spouses’ shared beliefs could assist practitioners in placing assessment of an individual within the context of the couple’s norms.
Individuals’ beliefs, resources, and behaviors are known to be important for both partners’ future health (e.g., Drewelies et al., 2016; Franks et al., 2004; Kim et al., 2014; Westerhof et al., 2014). We expanded this body of research and illustrated how individual processes of physical activity and disease burden contribute to couples’ co-constructed norms of beliefs about aging and future functional health. Relationship processes are known to be essential to health behavior and disease management (Martire et al., 2010; Rook, 2015). In addition to shaping individual beliefs about aging and future health, we found husbands’ physical activity to predict couples’ future functional health, whereas wives’ disease burden shaped couples’ concurrent shared beliefs about aging. These findings suggest that first, individual processes are important—they shape one’s own beliefs, which then form the couples’ shared beliefs. Therefore, an intervention that “treats” an individual has the potential for broader impact through its contribution to shared norms. However, these shared norms can also dampen the effects of individual treatments if not accounted for (Martire et al., 2010). Second, although husbands and wives were found to be similar in how their own physical activity and disease burden contributed to their own beliefs about aging, our research suggests that contribution of individual experiences to couples’ norms is gendered (Calasanti, 2010; Kornadt et al., 2013). Our research suggests that interventions aimed at modifying married couples’ beliefs about aging should be sensitive to how husbands and wives perceive their disease burden. For example, previous research has shown how individuals interpret the experience of chronic conditions—as actionable or an unavoidable consequence of aging (Stewart et al., 2012). Based on the findings from this study, these beliefs are expected to be amplified through their contribution to the couples’ shared beliefs. The potential for couples’ norms to amplify individual beliefs and behaviors would be a fruitful topic for further research.
The source of couples’ shared processes is important because it assists researchers, practitioners, and policy makers in positively affecting older adults’ health and well-being. Couple processes are grounded in the distal history of how couples selected one another, their shared historical and current experiences and behaviors, and also by their similarities and differences in chronological age (Choi & Vasunilashorn, 2014; Meyler et al., 2007). In this study, the contribution of spouses’ physical activity to couples’ shared beliefs was entirely explained by indicators of partner selection. Distal statuses such as race and education are known to open and constrain access to health-promoting environments (Calasanti, 2010; Settersten et al., 2015). Building on this earlier work, our findings suggest that interventions aimed at improving couples’ shared beliefs through increased engagement in physical activity must be sensitive to the constraints couples face. Additionally, we found couples’ shared health experiences to explain the contributions of not only husbands’ disease burden to functional limitations, but also the contribution of couples’ beliefs about aging to couples’ change in functional limitations. This is important because health and health behaviors are accessible to intervention and are also known to be deeply influenced by couples’ relationships processes (Martire et al., 2006; Rook, 2015; Tucker & Mueller, 2000). Finally, adjusting for age did not further explain how couples’ physical activity and disease burden contribute to couples’ beliefs about aging. This suggests that the relevance of similarities and differences in age for couples’ beliefs about aging is encompassed by age-related variation in health and health behaviors, rather than characteristics of age such as age-graded roles, responsibilities, and experiences.
The results from this study must be interpreted within the context of its limitations. First, beliefs about aging and functional limitations maintain a reciprocal relationship. Although this study was longitudinal, its design prohibits conclusions about the direction of causality. We acknowledge that physical activity and functional limitations are reciprocally related. Within the context of this study, husbands’ and wives’ frequency of moderate activity—actual activity—was found to be predictive of future functional limitations—the ability to do activities—at both the individual and couple level. Additionally, couples’ health and beliefs about aging contributed to selective attrition. It is therefore important for future research to examine the implications of couple-level constructs on survival. We also acknowledge that with only two points of measurement, it was not possible to examine proximal within-couple relationship processes. This is important because aging is a gendered process, where husbands and wives have distinct experiences as care givers and care receivers (e.g., Calasanti, 2010). Additional time points would allow examination of these processes within the context of couples’ shared beliefs. Additionally, modeling functional limitations as a continuous variable estimates a linear function on the raw scale, rather than a linear function on the log scale. This limits the ability to think of our independent variables as risk multipliers. However, this limitation is most relevant for the individual effects, as the couple-level effects were comparable in the count model (see Supplementary Table 4). This research does illustrate how the couple environment encompasses previous unexplored variance in the relationship between beliefs about aging and future functional limitations. The findings from this study lay the groundwork to further examine individual beliefs, health, and behaviors as parts of a larger whole. Additionally, due to the relatively small sample size, our models assumed invariance across couples’ age difference and marriage tenure. The heterogeneity of effects across subgroups of within-couple age differences, marriage tenure, and relationship intensity is an important avenue for future research. Finally, due to constraints in the available data, this study focused on married heterosexual couples in older adulthood. Future research should examine similarity and differences in shared beliefs about aging in other couple formations as well as consider the lingering effect of previous relationships and critical periods of shared time together.
In conclusion, couples are not only similar in their health and behavior, but also in their beliefs about aging. The quality of couples’ shared beliefs predicts both individual and couple-level differences in functional limitations 6 years in the future. Knowledge of the couples, including their norms, rituals, beliefs, and general health, will assist gerontologists and policy makers in supporting couples in their individual and collective efforts to age successfully together.
Supplementary Material
Supplementary data are available at The Gerontologist online.
Conflicts of Interest
The authors S. T. Mejía and R. Gonzalez certify that they have no affiliations with or involvement in any organization or entity with any financial or nonfinancial interest in the subject matter or materials discussed in this manuscript.
Funding
The authors’ contributions were partially funded by the National Institute on Aging (R01AG040635). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funders. This paper was published as part of a supplement sponsored and funded by AARP. The statements and opinions expressed herein by the authors are for information, debate, and discussion, and do not necessarily represent official policies of AARP.
