The Implications of Being “In it Together”: Relationship Satisfaction and Joint Health Behaviors Predict Better Health and Stronger Concordance Between Partners

Abstract

Background

Extensive evidence shows that satisfying marriages boost physical health and longevity. A separate literature reveals strong concordance in couples’ health, but the relationship processes that contribute to health concordance remain poorly understood.

Purpose

The current study examined whether relationship satisfaction and joint health behaviors—the extent to which couples eat, sleep, and exercise together—are associated simultaneously with better health and greater health similarity between partners.

Methods

Heterogeneous variance multilevel models were applied to data from 234 married couples (M_age = 46, Range = 20–84) reporting on their relationship satisfaction, joint health behaviors, and four health indicators—health satisfaction, depressive symptoms, comorbidities, and medication use.

Results

More satisfied couples engaged in more joint health behaviors than less satisfied counterparts. When joint health behaviors and relationship satisfaction were examined as separate fixed effects, both predicted greater health satisfaction and fewer depressive symptoms. More joint health behaviors were also associated with less medication use. When both were modeled together, only relationship satisfaction predicted depressive symptoms. By contrast, in random effects, joint health behaviors predicted greater similarity in health satisfaction, depressive symptoms, and comorbidities. Relationship satisfaction only predicted more similar depressive symptoms.

Conclusions

Although more satisfied couples engaged in more joint health behaviors. relationship satisfaction and joint health behaviors uniquely predicted couples’ health quality and concordance, suggesting that distinct mechanisms may drive better health and stronger health resemblance.

Couples who slept, ate, and exercised together more frequently had better health and stronger health concordance than couples who engaged in fewer joint health routines.

Introduction

Marital quality has long been linked to couples’ physical and emotional health. According to a 126-study meta-analysis [1], satisfying marriages reliably predict better health with magnitudes similar to the benefits of daily exercise and healthy diet. In contrast, marital discord raises the odds of depression 25-fold [2] and elevates mortality risk by 20% [3]. Among longitudinal studies, marital strain foreshadowed declines in cardiovascular fitness [4] and increased depression and functional limitations 2–10 years later [5, 6].

A separate line of research has uncovered extensive evidence for spousal concordance—the degree to which partners’ health profiles resemble each other. For instance, a person’s Type 2 diabetes, heart disease, or depression doubles their partner’s risks for the same condition (e.g., [7]). However, the factors that drive health concordance remain understudied and poorly understood. One explanation for couples’ shared health fates is assortative mating, the tendency for people to choose a partner with similar characteristics, including the same health conditions, risk factors, and lifestyles [8]. Nevertheless, health concordance exists even after accounting for individuals’ own health risk factors, strengthens over time in some health dimensions, and varies in magnitude across couples [7]. Partners also share resources: they often cohabitate, pool finances, and overlap in their social networks. To varying degrees, spouses’ daily routines, including their physical activity and sleep [9, 10], are intertwined, and each partner’s stressors and moods can affect both spouses [11]. Indeed, aspects of couples’ shared lives may contribute to health similarity beyond couples’ initial congruence.

Relationship Processes and Couples’ Health Concordance

Features of the couple’s relationship that risk or promote health may also, in parallel, shape spouses’ behavioral and emotional interdependence and, thus, their health concordance [11]. In one prior study, couples who rated their relationship as closer and those who exhibited less hostility during a marital problem discussion also had more similar cardiometabolic health compared to their less close, more hostile counterparts [12]. In that study, the authors also examined health behavior concordance—the overlap in partners’ sleep, diet quality, and physical activity—as a putative mechanism, with the expectation that closer, more satisfied couples would have more congruent health behaviors [11]. However, health behavior concordance was unrelated to closeness and marital behavior, and in fact, controlling for it strengthened the effects of greater closeness and lower hostility on health concordance [12].

Associations between relationship quality and health behavior concordance have been mixed in other work as well. For example, concordance in daily physical activity was stronger among longer-married couples and those who spent more time together, but not among more satisfied couples [9]. In a study that tracked couples’ sleep minute-by-minute, marital satisfaction and sleep concordance were positively correlated, but adjusted models showed a more complex interaction between wives’ marital satisfaction and husbands’ attachment anxiety [10]. Taken together, even in ecological momentary studies that capture health behaviors in a rich, dynamic way, health behavior concordance has not neatly tracked with relationship quality.

This inconsistent link between relationship quality and health behavior concordance reflects the fact that a couple can sleep the same amount, eat the same foods, and maintain similar activity levels without engaging in those behaviors together or having a satisfying relationship. Likewise, joint health routines do not imply perfect health behavior concordance: couples who share a bed may sleep different amounts, and those who eat the same meals may consume different portions, both of which would decrease health behavior concordance.

The Role of Joint Health Behaviors in Couples’ Health and Health Concordance

Instead of health behavior concordance, we posit that joint health behaviors—e.g., sleeping in the same bed, eating the same meals together, and exercising together—are key characteristics of more satisfying relationships. Joint health behaviors may also predict both greater health concordance (i.e., what we refer to as the convergence hypothesis) and better health (the enhancement hypothesis). According to the convergence hypothesis, sharing meals, exercising together, and co-sleeping can facilitate mutual support and encouragement, as well as overt pressure and persuasion, to maintain healthy routines. For example, a person’s new exercise regimen increases the likelihood that the partner will adopt the same healthy plan [13]. Likewise, partners can undermine one another’s attempts to maintain healthy behaviors [14] and reinforce unhealthy habits. Beyond direct influences on health behaviors, joint health routines allow dedicated time for discussion of daily hassles and uplifts—creating opportunities for collaborative problem-solving [15], emotional spillover [16], and physiological synchrony [17]. Whether partners encourage healthy or risky behavior, and transmit positive or negative emotions, joint engagement in health behaviors likely contributes to convergence in health profiles through behavioral, emotional, and physiological interdependence.

The enhancement hypothesis suggests that, on average, joint health behaviors may also benefit health in a variety of ways. According to the Core and Balance Model of Family Leisure Functioning [18], wherein leisure refers to time outside of work, “core” couple activities such as shared exercise and meals (among other joint activities such as shared recreation and household chores, [19]) serve to maintain the relationship by fostering closeness, communication, support, understanding, and shared rejuvenation. Indeed, among community couples, daily activities such as eating, relaxing, and sleeping are rated as more enjoyable with the partner rather than alone, and couples are happier in moments spent with the spouse than without [20, 21]. In another study, couples rated their marriage as more satisfying on days when they exercised together [22]. Parallel theories on couples’ sleep view sharing a bed as a key attachment behavior: couples who co-sleep report better subjective sleep quality compared to those who sleep separately [23]. To our knowledge, no prior work has integrated these perspectives to examine joint health behaviors as a class of shared health routines. Taken together, these “core” joint health routines may benefit health by bolstering relationship satisfaction [1], buffering stress [24], and boosting positive mood [25], beyond the possible advantages of reinforcing healthy behaviors like exercise and enhancing subjective sleep quality.

Current Study

In a cross-sectional, dyadic sample of 234 heterosexual couples, we examined relationship satisfaction and couples’ joint health behaviors as predictors of better health and stronger health concordance. Because we expected all three joint health behaviors—exercising together, co-sleeping, and eating the same meals together—to be associated in the same direction, we aggregated them into a composite index to minimize the number of statistical tests. To reflect the breadth of health outcomes represented in the marriage-health and spousal concordance literatures, we broadly construed health with four indicators—health satisfaction, depressive symptoms, comorbidities, and medication usage (e.g., Refs. [1, 5, 7, 26, 27]).

Heterogeneous-variance models treated relationship satisfaction and joint health behaviors as predictors of fixed effects to evaluate associations with health and used them as predictors of random effects to test links to health concordance. The current study represents one of the first efforts to capture relationship-relevant processes associated with better health and greater health concordance at once. This integrated approach aligns with the conceptualization that relationship satisfaction and joint health behaviors shape health and health concordance simultaneously. Indeed, couples’ health represents the net effect of factors that risk, promote, and intertwine. From a statistical perspective, models that examine the fixed effect of relationship satisfaction on health but ignore systematic differences in health concordance may violate the assumption of variance homogeneity and overestimate effects’ magnitude, increasing Type I error. Further, fixed effects can augment the amount of random variability to be explained, and thus can impact health concordance results. All in all, it is both conceptually valuable and statistically sound to consider health and health concordance simultaneously when investigating relationship processes that may affect both.

As depicted in Fig. 1, we first hypothesized that higher relationship satisfaction would be associated with more joint health behaviors (H1). Given the cross-sectional nature of our data and the likely bidirectional association between relationship satisfaction and joint health behaviors [28, 29], we evaluated this link with a correlational test. Next, associations with relationship satisfaction and joint health behaviors were tested in a sequential fashion. We predicted that higher relationship satisfaction would be associated with better health outcomes (H2a) (i.e., greater health satisfaction, fewer depressive symptoms, fewer comorbidities, and less medication usage) as well as more similar health between partners on the four indicators (H2b). We also conjectured that more joint health behaviors would predict better health (H3a, i.e., the enhancement hypothesis) and more similar health (H3b, i.e., the convergence hypothesis). Finally, we explored whether joint health behaviors and relationship satisfaction each were associated with better and more similar health when examined in the same model (R4a and R4b, respectively), to tease apart the relative contribution of the relationship’s affective quality from the role of coordinated health routines.

Fig. 1.

Conceptual model linking relationship satisfaction and joint health behaviors to better health outcomes and greater health concordance.

This conceptual model details the theorized associations of relationship satisfaction and joint health behaviors with more favorable health outcomes and stronger health concordance between partners. According to prior work, relationship satisfaction and joint health behaviors are likely to bolster each other: more satisfied couples are more likely to choose to engage in shared routines [28], which then can feedback to reinforce their bond [29]. Given the cross-sectional nature of our data, we tested this association with a correlation. In turn, relationship satisfaction and joint health behaviors were hypothesized to predict better health outcomes (Hypothesis 1a and 2a, respectively) and stronger health concordance (H1b and H2b, respectively).

Supplemental models explored the associations of health behavior concordance with joint health behaviors and relationship satisfaction, as well as its links to health and health concordance. Consistent with prior work, we expected these effects to be weak or nonsignificant. We also explored whether results were robust to changes in covariates.

Methods

Procedure

Qualtrics Panels were used to recruit participants across the United States (Qualtrics, Provo, UT). An email invitation was sent via Qualtrics Panels containing a description of the study, informed consent, and details of the study length and nature of the questions. Participants were able to choose from many incentives, for example, sweepstakes entrance or vouchers, airline miles, cash, gift cards, or redeemable points; participants’ payment choices were not disclosed to the researchers. The payment of participants varied depending on the acquisition difficulty, panelist profile, and length of survey, according to Qualtrics Panels’ regulations.

The email link was sent to one partner of a married couple with instructions for partners to take the survey back-to-back and to give one another privacy during each partner’s survey completion. Each partner separately signed the informed consent. Out of 1,003 participants who began the study, only 234 complete couples met criteria and took the survey (a 23% participation rate). All others were filtered out because they failed embedded attention checks (standard protocol for internet research to help ensure quality data), were not able to complete the survey in English, were not in a committed relationship of at least 3 years (not casually dating) or not living with their partner, or their partner did not complete his/her portion of the survey. The average time of survey completion was 28.54 min per couple. Among the final sample of 234 heterosexual couples, relationship length averaged 21 years (SD = 15), and their ages ranged 20–84 years (M = 46, SD = 15). Couples had 1–2 children living in the home (M = 1.69, SD = 1.32). Most participants were White, held a bachelor’s degree, and had health insurance (see Table 1).

Table 1.

Demographic Information for All Couples

	Men (N = 234)			Women (N = 234)
Variables	N or M (SD)			N or M (SD)
Age	47.63 (15.37)			44.96 (15.00)
Race
African/Black	7			4
Asian American/Pacific Islander	8			12
Caucasian	205			192
American Indian/Alaskan Native	2			4
Latino(a)	8			11
Mixed/Biracial	2			3
Other	2			5
Education
Less than high school	5			6
High school or GED	49			47
Some college, not graduated	44			40
Associate’s degree	28			37
Bachelor’s degree	66			61
Graduate or professional degree	42			40
Health insurance
Yes	217			215
No	16			15
Tobacco usage
No	174			183
Yes	60			48
Alcohol problems
No alcohol problem	163			167
Meets clinical cutoff	71			67
	Both (couple variable)
Income
Under $20,000	7
$20K–$39,999	43
$40K–$59,999	46
$60K–$79,999	51
$80K–$99,999	35
$100K+	52
Children	0	1	2	3	4	5 or more
Children <2 years	198	33	3	0	0	0
Children 3 to 5 years	196	33	5	0	0	0
Children 6 to 13 years	180	34	18	1	1	0
Children >14 years	121	38	47	13	10	5
Total sample (%)	20.9	24.8	32.5	12.4	5.6	3.9

Measures

For brevity, we describe the scoring and interpretation of well-established scales below and report more detailed information (e.g., full response scales, example items, psychometric information, and use in prior work) in Supplemental Material (S2–S4).

Predictors

Relationship satisfaction.

Relationship satisfaction was measured with the 4-item Couple Satisfaction Index (CSI) [30]. Each person’s responses were summed then averaged between partners (r = .79) to create a couple-level score for links to health and health concordance.

Joint health behaviors.

The frequency that participants engaged in health behaviors with their partner was assessed with three items for sleep, diet, and exercise. For joint sleep, participants reported how frequently they slept in the same bed with their partner (1 never to 5 always). For joint diet, participants answered “How often do you and your partner eat the same meal when you are together?” (1 never to 5 always). For joint exercise, participants reported how many of the last 7 days they exercised with their partner.

To examine associations with partners’ health and health concordance, joint health behaviors were aggregated at the couple level. First, each joint health behavior was averaged between partners. Within-couple correlations were high for all three: exercising together (r = .86), co-sleeping (r = .94), and eating the same meal together (r = .69). Then, they were z-scored and averaged to form a single couple-level composite score.

Outcomes

Health satisfaction.

One item assessed how satisfied individuals were with their health. Responses ranged from 1 extremely dissatisfied to 7 extremely satisfied so that higher scores indicated greater health satisfaction.

Depressive symptoms.

The 2-item Patient Health Questionnaire-4 assessed depressive symptoms over the past 2 weeks [31]. Responses ranged from 0 not at all to 3 nearly every day, and scores were summed so that higher scores indicated greater depressive symptoms.

Comorbidities.

The Charlson Comorbidity Index [32] assessed a wide range of comorbid health conditions, including cardiovascular conditions (heart condition, circulation problems, high blood pressure), type 2 diabetes, stomach ulcers, joint or mobility problems (osteoporosis, bone issues), kidney or urinary problems, digestive issues, and cancer. These conditions were of interest because all share inflammation as a hallmark and, thus, are susceptible to psychological stress. The total score reflected a summed count of comorbid health problems.

Medication use.

Participants reported how often they take prescription medication with one item [27]. Responses ranged never to four or more times a day.

Covariates

Individual health behaviors.

The International Physical Activity Scale, short form [33] measured the average duration of sedentary behavior in the past week with one item. Participants rated their overall sleep quality with one item from the Pittsburgh Sleep Quality Index [34], 0 very good to 3 very bad. Poor eating habits were measured using eight items from the Starting the Conversation survey [35]. Participants reported how often they ate various foods (snacks, fruits/vegetables, desserts/sweets, sodas or sweet teas, etc.) over the past few months. Responses ranged from less than 1 time to 5 or more times, and higher scores indicated poorer diet.

Substance use.

The 3-item Audit-C assessed frequency of alcohol use. Responses ranged from 0 to 12, with scores of four or more for men and three or more for women indicating problematic use [36]. Smoking and tobacco use frequency were also measured, ranging from 1 never to 5 daily or almost daily, and was dichotomized given the infrequent use in the sample.

Analytic Plan

Model parameters and sequence

We fit heterogeneous-variance multilevel models (MLMs) [37] in SAS PROC MIXED with restricted maximum likelihood (REML) to evaluate our research questions. REML uses all available observations and assumes the data are missing at random. An extension of traditional MLMs, these models allowed us to test predictors of both average levels (i.e., fixed effects) and the random variance—in this case, the variance within couples. Thus, we were able to simultaneously evaluate the roles of joint health behaviors and relationship satisfaction in both the average levels and within-couple variance (i.e., partners’ similarity or concordance) of the four health-related outcomes—health satisfaction, depressive symptoms, comorbidities, and medication use. See Supplemental Material for an example equation, syntax, and detailed explanation of the model. Comorbidities and medication use were square-root-transformed to correct residuals. There was no evidence of multicollinearity (variance inflation factor = 1.25–1.29, Tolerance = 0.77–0.80).

A correlation evaluated the association between couple-level relationship satisfaction and joint health behaviors (H1). For subsequent hypotheses, in the fixed part of each model, we controlled for gender, age, education, and body mass index (BMI) given their known associations with health satisfaction, depressive symptoms, health conditions, and medication use [26, 38–40]. We also controlled for poor diet quality, sleep problems, and sedentary behavior because the joint health behavior items were not fully parallel with each other as written. On the one hand, co-sleeping and co-eating items did not inquire about the quality of sleep or diet, and thus could not be construed as health-compromising or health-promoting. On the other hand, exercising together more frequently inherently includes more exercise, a health-promoting activity. Although more sedentary behavior is not the opposite of more joint exercise, accounting for the quality of individual health behaviors conceptually distinguished sedentary behavior from a lack of exercising together. By teasing apart the joint engagement in behaviors from the quality of individual behaviors, we attempted to make the interpretation more comparable for the three behaviors. Moreover, theorized mechanisms linking joint health behaviors to better health outcomes extend far beyond the benefits of healthy behaviors. To assess Hypothesis 2a, couple relationship satisfaction was tested as a fixed effect on the four health outcomes. To predict the variance within couples in the random part of the model, relationship satisfaction was entered into the LOCAL = EXP() command of the REPEATED statement (H2b). Log-likelihood ratio tests evaluated whether its addition as a random effect significantly improved fit. To evaluate H3a and H3b, we repeated these steps with the joint health behavior index as the focal predictor. Finally, we added the joint health behavior index as a fixed and random effect in the models with relationship satisfaction to assess whether each explained unique variance in health (R4a) and health concordance (R4b). Predicted values of health and health concordance at representative values of relationship satisfaction and joint health behaviors are presented for significant effects in Supplemental Material.

In follow-up sensitivity tests, we examined whether findings were robust to the exclusion of health behavior covariates, and to the inclusion of alcohol problems, tobacco use, and having young children. We also explored the role of health behavior concordance in associations with relationship satisfaction, joint health behaviors, as well as health and health concordance. A health behavior concordance composite variable was created by taking the absolute difference between partners for sedentary behavior, poor diet, and sleep problems, then averaging the z-scored variables. Higher scores reflected larger differences between partners.

Interpretation of the within-couple variance estimates

In these models, a positive variance coefficient indicates greater within-couple variability (i.e., less similarity), whereas a negative coefficient reflects smaller within-couple variance (i.e., greater similarity). Thus, relationship satisfaction and joint health behaviors were expected to negatively predict within-couple variability, such that greater satisfaction and more shared health behaviors would be associated with smaller variance, that is, greater similarity, in the health-related outcomes.

Results

Descriptive Statistics

Most couples ate the same meal most or all of the time (77.9%) and reported always sleeping in the same bed (65.3%). Roughly one-third of couples reported exercising together in the past week (32.6%). Relationship satisfaction was high on average but varied widely, with 23.5% of couples reporting clinically significant relationship distress (CSI < 13.5).

Table 2 displays means, standard deviations, and intercorrelations among primary study variables. The four health outcomes correlated in the expected direction, with small to moderate magnitudes. The three health behaviors, diet quality, sedentary behavior, and sleep quality, were surprisingly not interrelated. They also shared only small and inconsistent correlations with joint health behaviors.

Table 2.

Descriptive Information for Primary Study Variables

	1	2	3	4	5	6	7	8	9	10	11	12	M (SD)
1. Relationship satisfaction	–												15.9 (4.6)
2. Eat together	0.29*	–											3.9 (1.0)
3. Exercise together	0.20*	0.10	–										1.0 (1.7)
4. Co-sleep	0.36*	0.27*	0.14*	–									4.2 (1.3)
5. Joint index	0.42*	0.70*	0.63*	0.69*	–								0.0 (0.7)
6. Poor diet quality	−0.07	−0.11*	0.13*	−0.04	−0.003	–							13.9 (2.4)
7. Sedentary behavior	−0.004	0.10*	−0.002	0.005	0.04	0.0003	–						900.1 (1128.0)
8. Poor sleep quality	−0.11*	−0.10*	−0.22*	−0.07	−0.19*	0.004	0.04	–					2.1 (0.7)
9. Health behavior concordance	−0.11	0.02	−0.15*	0.02	−0.07	0.05	0.30*	0.14*	–				0.003 (0.606)
10. Health satisfaction	0.22*	0.11*	0.27*	0.15*	0.27*	0.01	−0.06	−0.42*	−0.11*	–			5.2 (1.6)
11. Depressive symptoms	−0.28*	−0.13*	−0.09	−0.13*	−0.16*	0.09*	−0.0003	0.33*	−0.22*	0.09	–		1.1 (1.6)
12. Comorbidities	−0.12*	−0.08	−0.14*	−0.09	−0.16*	−0.01	0.03	0.15*	−0.45*	0.01	0.05	–	0.5 (0.8)
13. Medication use	−0.10*	−0.10*	−0.09	−0.18*	−0.18*	−0.05	0.07	0.13*	−0.36*	0.02	0.10*	0.48*	2.5 (1.5)

Variables 1–5 are dyadic; variables 6–12 are reported at the individual level. Ns for dyadic variables = 219–234; Ns for individual variables = 438–468. With the exception of health satisfaction, health and health behavior variables were scaled so that higher scores indicated poorer quality. Health behavior concordance was scaled so that higher values indicate larger differences.

*p < .05.

Do More Satisfied Couples Engage in More Joint Health Behaviors (H1)?

Shown in Table 2, more satisfied couples engaged in a greater variety of health behaviors together compared to less satisfied couples. They also ate meals together more frequently, exercised together more often, and slept in the same bed more regularly. The joint behavior index shared moderate to strong correlations with each of the individual shared health behaviors. Correlations among each of the shared health behaviors were weaker.

Does Relationship Satisfaction Predict Health (H2a) and Health Concordance (H2b)?

Health

More satisfied couples reported greater health satisfaction (Estimate = 0.054, SE = 0.016, p = .001) and fewer depressive symptoms (Estimate = −0.100, SE = 0.019, p < .0001). The effect was not significant for medication use or comorbidities (ps > .070).

Health concordance

More satisfied couples had more similar depressive symptoms (Estimate = −0.115, SE = 0.019, p < .0001, Χ²(1) = 40.3, p < .0001). The associations were not significant for similarities in health satisfaction, comorbidities, or medications (ps > .053).

Do Joint Health Behaviors Predict Health (H3a) and Health Concordance (H3b)?

Health

Couples who shared more health behaviors also were more satisfied with their health (Estimate = 0.359, SE = 0.111, p = .001), less depressed (Estimate = −0.299, SE = 0.121, p = .014), and took fewer medications (Estimate = −0.074, SE = 0.035, p = .032) compared to couples who shared fewer health behaviors. There was no association with comorbidities (p = .162).

Health concordance

Couples who jointly engaged in a greater number of health behaviors also shared more similar levels of health satisfaction (Estimate = −0.451, SE = 0.122, p = .0002, Χ²(1) = 13.5, p = .0002), depressive symptoms (Estimate = −0.383, SE = 0.130, p = .003, Χ²(1) = 9.0, p = .003), and comorbidities (Estimate = −0.298, SE = 0.112, p = .008, Χ²(1) = 7.0, p = .008), but not medication use (p = .171).

Do Joint Health Behaviors and Relationship Satisfaction Predict Health (R4a) and Health Concordance (R4b) Independently?

Health

Shown in Table 3, accounting for joint health behaviors, higher relationship satisfaction remained significantly associated with greater health satisfaction and lower depressive symptoms. Conversely, the link between joint health behaviors and depressive symptoms became nonsignificant with the inclusion of relationship satisfaction. However, the associations of more shared health behaviors with greater health satisfaction and lower medication use remained significant.

Table 3.

Joint Health Behavior Index and Relationship Satisfaction Predicting Health and Health Concordance Between Partners

	Health satisfaction			Depressive symptoms			Comorbidities			Medications
Fixed effects	Estimate	SE	p	Estimate	SE	p	Estimate	SE	p	Estimate	SE	p
Intercept	4.831	0.159	<.0001	1.086	0.158	<.0001	0.471	0.063	<.0001	1.540	0.051	<.0001
Female gender	−0.059	0.113	.599	0.280	0.104	.008	−0.059	0.047	.209	0.029	0.036	.423
Age	−0.009	0.005	.061	−0.019	0.005	<.0001	0.016	0.002	<.0001	0.012	0.002	<.0001
Education	0.126	0.044	.005	−0.053	0.043	.225	−0.009	0.017	.618	−0.011	0.014	.456
BMI	−0.050	0.009	<.0001	−0.020	0.009	.035	0.009	0.004	.018	0.003	0.003	.319
Poor diet quality	−0.007	0.027	.795	0.022	0.027	.413	0.012	0.011	.250	−0.0004	0.009	.965
Sedentary behavior	−0.002	0.006	.668	0.001	0.006	.889	0.001	0.002	.591	0.003	0.002	.079
Poor sleep quality	−0.747	0.090	<.0001	0.533	0.087	<.0001	0.158	0.036	<.0001	0.082	0.029	.005
Relationship satisfaction	0.046	0.018	.009	−0.093	0.020	<.0001	−0.010	0.007	0.112	0.001	0.005	.814
Joint behavior index	0.243	0.117	.039	−0.095	0.120	.430	−0.032	0.045	0.486	−0.079	0.038	.039
Random effects	Estimate	SE	p	Estimate	SE	p	Estimate	SE	p	Estimate	SE	p
Couple intercept σ 2	0.335	0.124	.003	0.518	0.120	<.0001	0.034	0.019	.036	0.042	0.013	.001
Relationship satisfaction σ 2	−0.021	0.019	.278	−0.118	0.022	<.0001	−0.007	0.019	.708	−0.004	0.020	.835
Joint behavior index σ 2	−0.432	0.126	.001	0.012	0.149	.938	−0.300	0.119	.011	−0.160	0.141	.256
Residual σ 2	1.414	0.138	<.0001	1.246	0.120	<.0001	0.239	0.023	<.0001	0.138	0.013	<.0001
	−2 LL	Χ 2	p	−2 LL	Χ 2	p	−2 LL	Χ 2	p	−2 LL	Χ 2	p
Model fit	1564.3	11.5	.001	1563.6	0.0	1.00	755.1	6.3	.012	564.1	1.3	.254

Comorbidities and medication use were square-root-transformed. With the exception of health satisfaction, all health and health behavior variables were scaled so that higher scores indicated poorer quality. Sedentary behavior estimates are presented in 100-unit increments to facilitate interpretation. Log likelihood ratio tests (LLRT, df = 1) compare the −2 log likelihood of the model with the random effect of joint health behaviors to the model without this term. This test examined whether the random effect of joint health behaviors improved the fit of the model above and beyond the fixed effects and the random effect of relationship satisfaction. Italicized effects were statistically significant in previous models but no longer significant in the fully adjusted model. See Supplemental Material for predicted health and health concordance values at representative values of relationship satisfaction and joint health behaviors.

BMI body mass index.

Health concordance

Depicted in Table 3, more joint health behaviors continued to significantly predict greater concordance in health satisfaction and comorbidities, and improved model fit. However, more joint health behaviors no longer predicted greater depressive concordance with relationship satisfaction in the model.

Supplemental Analyses: Ancillary Covariates and Health Behavior Concordance

Two sets of post-hoc sensitivity analyses were conducted, excluding individual health behaviors, then additionally controlling for the fixed effects of alcohol use problems, tobacco use, and having young children. All results remained the same. Supplemental models also explored whether health behavior concordance was associated with relationship satisfaction, joint health behaviors, health, or health concordance. As hypothesized based on our conceptual framework and prior studies, there was no correlation between health behavior concordance and relationship satisfaction or joint health behaviors (Table 2). Whereas health behavior concordance was not associated with better health on any dimension in MLMs (ps > .250), stronger health behavior concordance did significantly predict more concordant health satisfaction (Estimate = 0.321, SE = 0.129, p = .013) and more similar depressive symptoms between partners (Estimate = 0.569, SE = 0.140, p < .0001). Nevertheless, including health behavior concordance in the models with relationship satisfaction and joint health behaviors did not change any of the findings.

Discussion

In a community sample of married couples, those in more satisfying marriages shared more joint health behaviors than their less satisfied counterparts. In turn, when relationship satisfaction and joint health behaviors were examined as predictors of health and health concordance, relationship satisfaction predicted better health, particularly with outcomes involving participants’ affective appraisals. On the other hand, joint health behaviors were linked to greater health satisfaction, lower medication use, and stronger concordance in partners’ health across outcomes, providing evidence for both enhancement and convergence hypotheses. Taken together, results suggest that joint health routines may play a role in both better health and stronger health concordance—insights that can be leveraged to enhance dyadic health promotion and advance the theoretical integration of marriage-health and spousal concordance perspectives.

The Value of Examining Relationship Processes in Health and Health Concordance

The current study represents one of the first efforts to simultaneously examine health and health concordance. Heterogeneous variance MLMs allowed predictors of both average levels (i.e., fixed effects) and within-couple variance (i.e., random effects). This approach provides a way to synthesize our understanding of the processes that drive the health benefits of a satisfying marriage and shape couples’ health concordance, two rich areas of work that, to date, have progressed independently. Our findings identify joint health routines as a novel candidate to unite these lines of research. Models of marriage and health that ignore systematic differences in health concordance may violate variance assumptions and overestimate the magnitude of fixed effects. Likewise, models of health concordance should account for the fixed effects of relationship factors and other important covariates, as they augment the signal-to-noise ratio.

Relationship Satisfaction and Joint Health Behaviors (Hypothesis 1)

As predicted, more satisfied couples slept, ate, and exercised together more frequently than less satisfied counterparts, and they engaged in a wider array of health routines together. Correlations were small to moderate, evidence that the intertwinement of health routines is related to, but not synonymous with, the affective quality of the relationship. This finding dovetails with prior evidence on the moderate association between marital satisfaction and closeness [11]: indeed, satisfied couples are often close, but dissatisfied partners can also function interdependently. Likewise, relationship satisfaction and joint health behaviors shared some overlapping associations but also demonstrated unique effects on aspects of partners’ health levels and concordance, pointing to distinct pathways.

Associations With Couples’ Health (Hypotheses 2a and 3a; Research Question [RQ] 4a)

When examined separately, both relationship quality and joint health behaviors shared significant associations with better health. These results add to a large literature demonstrating positive associations between relationship function and health that are robust to traditional covariates (e.g., age, gender, BMI, education). The fact that associations held controlling for the quality of health behaviors suggests that relationship satisfaction and joint health behaviors do not operate strictly through their associations with healthy habits. Indeed, correlations of health behaviors with relationship quality and joint engagement were small and inconsistent.

In accordance with the enhancement hypothesis, couples who engaged in more joint health behaviors also had greater health satisfaction, lower depressive symptoms, and less medication usage. Associations with health satisfaction and medication use were robust to the inclusion of relationship satisfaction, suggesting that joint health behaviors may operate through other routes. For community couples, engaging in “core” daily health routines together may buffer against stress and boost positive mood [18], which in turn can bolster health and well-being. Sharing meals, coordinating bedtime routines, and exercising together foster closeness and provide opportunities for communication, mutual support, and shared rejuvenation. The effect on depressive symptoms was attenuated by relationship satisfaction, suggesting that joint health routines are linked to better mental health through greater satisfaction. Indeed, more satisfied couples may seek out opportunities to coordinate their routines, which in turn may reinforce their connection and maintain mental health. Some mechanisms may be unique to particular joint health routines, a fruitful direction for future work. For example, co-sleeping may facilitate sexual intimacy; eating meals together reserves time for communication; and exercising together may increase self-efficacy, boost affect and attraction through arousal misattribution [41], and fulfill relatedness, agency, and competency goals [42].

Associations With Couples’ Health Concordance (Hypotheses 2b and 3b; RQ 4b)

In concert with the convergence hypothesis, joint health behaviors were associated with stronger concordance in depressive symptoms, health satisfaction, and comorbidities. A prior study examined health behavior concordance in the same three dimensions—diet, activity, and sleep—and found that they improved model fit in three of the five cardiometabolic health outcomes [12]. However, health behavior concordance was unrelated to marital quality. Indeed, having the same diet and exercise routines may not track with relationship processes if behavioral concordance diverges from the coordination of joint routines. For instance, even if two partners sleep similar amounts, and thus have high sleep concordance, sleeping in separate beds reduces opportunities for emotional spillover during bedtime conversation and eliminates direct sleep disturbances from the partner’s restlessness or snoring. The present study replicated these patterns and extended the work by identifying joint engagement in health routines as a relationship-relevant process that predicted both health and health concordance. According to the convergence hypothesis, the effects of joint health behaviors on health concordance are likely multifaceted, operating through the spillover of stressors and uplifts, mood contagion, physiological synchrony, and dyadic coping efforts, in addition to direct influences on health behaviors. Indeed, the coordination of health routines may lead to convergence in health behaviors over time. However, this may be better captured longitudinally because we did not find a correlation between joint behaviors and behavioral concordance, and health behavior concordance did not explain the associations of joint health behaviors with health or health concordance.

More joint health behaviors predicted more similar rates of comorbidities and health satisfaction, whereas relationship satisfaction did not. This early evidence could suggest that behavioral coordination and its consequences are more important for partners’ health concordance than the affective quality of the relationship. Alternatively, the demand characteristics of self-reported relationship satisfaction may have attenuated its links to health concordance: one prior study found observer-rated hostility to be a more robust predictor of partners’ cardiometabolic similarity than self-reported marital satisfaction [12].

In one exception, more satisfied couples did share greater depressive concordance. This parallels the results of one study wherein more satisfied wives’ emotional distress mirrored their partner’s distress more closely than did less satisfied counterparts [43]. However, contrasting evidence showed that negative emotions were more contagious among unhappily married couples [44]. These mixed results raise interesting questions about the unique roles of empathy and emotion regulation in a cascade linking marital function to mood congruence and spillover.

Strengths, Limitations, and Directions for Future Research

Among its strengths, this study is one of the first to simultaneously capture processes relevant to both the marriage-health link and couples’ health concordance with data from more than 200 couples. Integrating these perspectives represents a valuable step to facilitate theoretical advancement. Moreover, understanding the multidimensional effects of joint health routines on health and health concordance may better equip primary care and mental health practitioners to effectively treat people in committed partnerships. Engaging in coordinated health behaviors could have intrinsic health benefits, at least in community couples, but these joint routines may present potential risks to both partners when a person struggles with chronic stress, health issues, or unhealthy habits. Our study used a cross-sectional design and, thus, could not distinguish whether health concordance resulted from long-established shared health routines or led to convergence in routines. Future intervention and experimental work in both community couples and distressed populations must tease apart the causal associations.

Standardized effect sizes could not be calculated because these methods have not been developed for heterogeneous variance MLMs [45]. Nevertheless, even effects of small magnitude (e.g., those of a happy marriage, regular exercise, and diet on health [1]) can accumulate over longer periods to have meaningful impacts on health and well-being. Additionally, future studies should replicate the findings with different sampling procedures. We used Qualtrics panels, which may be subject to selection bias and requires partners to take turns with survey completion. All measures were self-reported and brief, due to time restrictions; some were single-item surveys, which have psychometric limitations. Future work should incorporate direct measures of physiology, observation of couples’ functioning, and clinical interviews. Gold-standard assessments of health behaviors should be used to definitively assess the roles of health behavior quality and concordance. Follow-up studies must also replicate and extend this work with racially diverse samples that include same-sex and unmarried, cohabiting couples, as well as those without access to health insurance. The current study consisted of heterosexual, mostly White married couples who were insured.

Future work should investigate whether links between joint health behaviors and health outcomes vary by context. Indeed, some theories of interpersonal dynamics such as social allostasis and interdependence [11, 46] suggest that closeness and contagion can promote relationship quality and health in low-stress contexts but may be associated with poor outcomes in adverse contexts. In this view, joint health behaviors may strengthen the health-risky effects of poor health behaviors by reinforcing unhealthy habits, and exacerbate stress by giving distressed couples more opportunities to bicker. The ability to detect this amplification effect depends on the extremity and prevalence of stress and unhealthy behavior in the sample. Future research should target subsamples at the extremes to properly test the association between joint health behaviors and health across disparate contexts (e.g., sedentary vs. active, sleep-disordered vs. not, financially strained vs. not, maritally distressed vs. satisfied).

Conclusions

In conclusion, these novel findings reveal that relationship-relevant processes share important associations with both better health and stronger health concordance across a range of outcomes. This early evidence suggests that joint health behaviors may play a key role in a dynamic process linked to greater relationship satisfaction and better health, and may be particularly important in routes to stronger health concordance. Conversely, relationship satisfaction may be more central to better health, and may influence health concordance through the coordination of health routines. The current approach provides a means to integrate models of marriage and health with those of couples’ health concordance. We hope it will encourage researchers to consider whether the relationship processes that risk couples’ health also bring partners into closer proximity or drive them apart. These insights may also be leveraged to enhance the benefits of dyadic intervention, providing fertile ground for the promotion of relationship satisfaction and both partners’ health.

Compliance With Ethical Standards

Conflict of Interest The authors declare that they have no conflict of interest.

Informed consent Informed consent was obtained from all individual participants included in the study. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.