Spousal Associations between Frailty Status and Depressive Symptoms: Longitudinal Findings from the Cardiovascular Health Study

Abstract

Objectives

To determine whether older adult spouses’ frailty states and depressive symptoms are interrelated over time.

Design

Longitudinal, dyadic path analysis with the Actor Partner Interdependence Model.

Setting

Data were from baseline (1989–1990), wave 3 (1992–1993), and wave 7 (1996–1997), all waves in which frailty and depressive symptoms were measured, of the Cardiovascular Health Study (CHS), a multi-site, longitudinal, observational study of risk factors for cardiovascular disease in adults 65 years or older.

Participants

Each spouse within 1,260 community-dwelling, married couples.

Measurements

Frailty was measured using the CHS criteria, categorized as nonfrail, prefrail, or frail. Depressive symptoms were measured using the 10-item Center for Epidemiologic Studies Depression scale.

Results

Within individuals (actor effects), higher frailty status predicted greater subsequent depressive symptoms, and greater depressive symptoms predicted higher subsequent frailty status. Between spouses (partner effects), an individual’s greater frailty status predicted the spouse’s increased frailty status, and an individual’s greater depressive symptoms predicted the spouse’s increased depressive symptoms.

Conclusion

Frailty and depressive symptoms are interrelated among older adult spouses. For older couples, interventions for preventing or treating frailty and depression that focus on couples may be more effective than those that focus on individuals.

INTRODUCTION

Frailty is a syndrome that predicts vulnerability to disability, and it affects one in ten older adults.^1,2 According to Fried et al (2001), a person is considered frail if he or she has three or more of the following criteria: unintentional weight loss (10 lbs in past year), self-reported exhaustion, weakness (grip strength), slow walking speed, and low physical activity. One or two of these characteristics indicates that a person is prefrail. Frailty has been associated with an increased risk of disability, falls, hospitalization, and a shorter lifespan.^2,3 Research has also shown that frailty and depression are highly related.^2,4 However, when examining frailty and depression, information about social environmental influences is missing. For older married adults, spouses tend to have similar characteristics, engage in shared activities, and live in the same environment, making their health interdependent.⁵ There is evidence of effective interventions to prevent and treat frailty⁶ and depression⁷ separately in individuals; however, if spouses’ frailty and depressive symptoms are intertwined, there may be synergistic effects on health when interventions address both frailty and depressive symptoms and are designed for couples.

Although no research to our knowledge has examined spousal associations in frailty and depressive symptoms, studies have documented spousal associations in instrumental activities of daily living⁸, physical activity⁹, and depressive symptoms. It is proposed that one spouse’s frailty influences the other spouse’s depressive symptoms through multiple mechanisms. First, spouses of frail individuals frequently act as caregivers, and caregiving can lead to depression.¹⁰ Second, frail spouses are often depressed which can be contagious.¹¹ Depressive symptoms may lead to increased frailty through different mechanisms. Depression can reduce the couple’s social activity.^12,13 Also, depressed spouses may lack the energy to provide support to their partners exacerbating the partner’s health problems.¹⁴ Multiple relationship theories also suggest that one person’s frailty can impact their partner’s depressive symptoms directly. For example, Inclusion of the Other in the Self theory states that individuals incorporate the characteristics of their partners (e.g., frailty) into their own sense of well-being.¹⁵

In this study we first hypothesized that within each individual greater frailty status would predict increases in subsequent frailty status, and greater depressive symptoms would predict increases in subsequent depressive symptoms (H1a and b; actor effects). Second, we hypothesized that an individual’s greater frailty status would predict increases in the spouse’s frailty status; we expected the same for depressive symptoms (H2a and b; partner effects). Third, we hypothesized that within individuals there would be a bidirectional association between frailty and depressive symptoms (H3; actor effects). Fourth, we hypothesized a bidirectional association between an individual’s frailty status and the spouse’s depressive symptoms (H4; partner effects). We test our hypotheses using the Actor Partner Interdependence Model (APIM¹⁶) that takes into account how couple members’ data is related.

METHODS

Data and Sample Characteristics

Data was from the Cardiovascular Health Study (CHS), a population- based longitudinal study designed to determine risk factors for cardiovascular disease in adults 65 years or older. The CHS recruited participants from four U.S. communities: Forsyth County, NC; Sacramento County, CA; Washington County, MD; and Pittsburgh, PA. Participants underwent annual clinical examinations and structured interviews. The present study includes data from three waves: baseline (1989/1990), wave 3 (1992/1993), and wave 7 (1996/1997). The CHS sample included 5,201 individuals enrolled in 1989/1990, with an additional cohort of 687 African Americans enrolled in 1992/1993. See Fried et al. (1991) for more information about the design and sample.¹⁷

Measures

Frailty

Frailty was defined using CHS criteria, which include deficits in five areas: low weight, physical inactivity, exhaustion, weakness, and slowness.² Low weight was defined as a self-reported or calculated loss of 10% or more in body mass index (BMI) since the previous wave or as a current BMI of <18.5 kg/m2. Physical activity was calculated as the average frequency of three activity intensities weighted according to average metabolic equivalency of task (MET) scores (mild, 1–3 MET; moderate, 3–6 MET;and vigorous, 6–10 MET). Physical inactivity was defined as being in the lowest 20% on the physical activity score stratified according to sex. Exhaustion was defined by endorsement of one of two items from the Center for Epidemiologic Studies—Depression (CES-D) scale.¹⁸ Weakness was defined as having dominant hand grip strength below sex- and BMI-specific cut-points as established in the CHS. Slowness was defined as a speed < 0.762 m/s for women taller than 159 cm or men taller than 173 cm and as < 0.653 m/s for women 159 cm tall or less or men 173 cm tall or less, measured on a 2.5-m course.^2,19 Individuals were characterized as nonfrail, prefrail, or nonfrail at each wave.

Depressive symptoms

A modified version of the CES-D assessed self-reported depressive symptoms experienced during the preceding 7 days of the clinic visit¹⁸. The scale consists of 10 symptoms, each scored 0–3, for a maximum of 30 points. Higher scores indicate greater frequency of symptoms.

Covariates

Age, sex, race, and education were examined as potential covariates. Perceived social support was also examined as a covariate because of its theoretical relevance to frailty, depression, and disability.²⁰ The Interpersonal Support Evaluation List (ISEL²¹) was used to assess the perceived availability of belonging (emotional), appraisal (informational), and tangible (instrumental) support. Example items assessing belonging (2 items) and appraisal support (2 items) respectively are: “When I feel lonely, there are several people I can talk to,” and “When I need suggestions on how to deal with a personal problem, I know someone I can turn to.” Tangible support was measured with: “If I were sick I could easily find someone to help me with my daily chores” (only one item used for tangible support because of low reliability with the two items). All items were rated on a 4 point scale from 1 (definitely false) to 4 (definitely true). Scores were averaged.

Analysis

First, we examined correlations between baseline depressive symptoms and potential covariates and differences in covariates by baseline frailty status using ANOVAs. Covariates were included in the main models at a p< .10 threshold. Next, we described how each spouse’s frailty and depressive symptoms changed and the extent to which frailty statuses matched at each wave.

To test the main hypotheses, path analyses were conducted with Mplus 7.3, which employs full information maximum likelihood (FIML) estimation to use all available data.²² The analysis of dyadic data is often problematic because of non-independence among responses (e.g., spouses’ depressive symptoms are often highly related). Analyzing members of the dyad separately is one way to avoid statistical problems due to non-independence. However, this approach fails to incorporate the interdependence of dyad members. The Actor–Partner Interdependence Model (APIM) incorporates this interdependence by combining data from both partners into one analysis and providing information about whether wives influence husbands and husbands influence wives.^9,16,23

Because past research suggests that the association between frailty and depressive symptoms is bidirectional⁴, we ran two models. In Model 1, frailty states of both spouses were set to predict changes in both partner’s depressive symptoms at wave 3 and wave 7 (controlling for covariates). Our focus was on the influence of one's frailty state on changes in his or her subsequent depression. We did not hypothesize that these relationships would be different at each wave. Thus, all like effects (baseline to wave 3; wave 3 to wave 7) were set to be equal. See Figure 1 for the proposed model with actor and partner effects for frailty predicting depressive symptoms. Model 2 examined actor and partner effects for depressive symptoms predicting frailty states. For more information about including ordinal, or ordered categorical, outcomes (e.g., frailty) in path analysis refer to Holgado-Tello et al. (2010)²⁴ and Millsap & Yun-Tein (2004).²⁵

Figure 1.

Actor Partner Interdependence Model with Frailty Predicting Depressive Symptoms (Actor and Partner Effects)

Notes. W1= baseline, W3= wave 3, W7= wave 7. This figure represents Model 1 with frailty predicting depressive symptoms. Bold lines indicate significant paths. Model 2 differs from Model 1 in that there are arrows from W1 depressive symptoms to W3 frailty and from W3 depressive symptoms to W7 frailty instead of arrows from W1 frailty to W3 depressive symptoms and from W3 frailty to W7 depressive symptoms.

R-square values were calculated to determine the amount of variance that the predictors accounted for in the outcomes. We report three model fit indices: the confirmatory fit index (CFI), the Tucker Lewis Index (TLI), and the root mean squared error of approximation (RMSEA).²⁶ For the CFI values of >.95, for the TLI values of >.90²⁷, and for the RMSEA values of <.08²⁷ reflect good fit of a specified model to the data.

RESULTS

Description of the sample

Of the original CHS sample, 2,524 reported being married, and data was available for 1,260 couples. One hundred and sixteen husbands and 39 wives were deceased at wave 3, and 351 husbands and 132 wives were deceased at wave 7. Two husbands and two wives indicated they were caregivers. Couples were evenly distributed among the CHS sites (Forsyth County, NC: n=306; Sacramento County, CA: n=334; Washington County, MD: n=348; Pittsburgh, PA: n=272). Five hundred ninety nine (49.6%) couples reported an annual income of less than $25,000, 397 (33.2%) reported between $25,000 and $49,999, and 206 (17.2%) reported greater than $50,000.

The mean age of husbands and wives were 73.6 (SD=5.3) and 71.2 (SD=4.7), respectively. For husbands, 1125 (97%) were White, 30 (2.4%) were Black, 3 (0.2%) were American Indian or Alaska Native, and 2 (0.2%) were Other. Thirteen (1%) were Hispanic. For wives, 1229 (97.5%) were White, 27 (2.1%) were Black, 2 (0.2%) were Asian or Pacific Islander, 1 was American Indian/Alaska native, and 1 (0.1%) was Other. Nine (0.7%) were Hispanic. Husbands had a mean of 14.22 (SD=5.0) years of education; wives had a mean of 13.9 (SD=4.2).

Description of missing data

ANOVAs examining differences in baseline characteristics between participants who had data from all three, two, or one wave showed that participants who had more missing data were older (husbands: F=54.77, p=.000; wives: F=46.08, p=.000), less educated (husbands: F=22.57, p=.000; wives: F=17.42, p=.000), and more depressed (husbands: F=10.87, p=.000; wives: F= 5.79, p=.003). More missing data was associated with more support for husbands (F=3.17, p=.042). The more missing data, the more husbands were likely to be frail (no missing: not frail n=334, prefrail n=243, frail n=15; one wave missing: not frail n=155, prefrail n=195; frail n=18; two waves missing: not frail n=75, prefrail n=156, frail n=37; χ²=91.83, p<.0001). For wives, the more missing data, the more likely they were to be prefrail (no missing: not frail n=295, prefrail n=268; frail n=30; one wave missing: not frail n=156, prefrail n=174, frail n=37; two waves missing: not frail n=86, prefrail n=154; frail n=frail; χ²=30.75, p<.0001).

Covariates

Older husbands were more depressed (r=.07, p<.05) and frail (F=25.84, p<.001) than younger husbands. Older wives were not more depressed (r=.02, p=.45) but more frail (F=39.84, p<.001) than younger wives. Non-frail husbands and wives had the highest education (F=11.27, p<.001; F=4.60, p<.01, respectively). Support was associated with greater depressive symptom in both husbands (r=.29, p<.001) and wives (r=.32, p<.001). Wives who were more frail reported greater support (F=4.18, p<.05); there was a similar trend for husbands (F=2.72, p=.07). In our final models we included both spouses’ age, education, and support as covariates.

Description of husbands’ and wives’ frailty and depressive symptoms over time

Table 1 shows how frailty changed over time for husbands and wives and the extent to which frailty statuses matched. As previously reported, depressive symptoms increased for both husbands (F(1.008)=912.840, p<.001) and wives (F(1.008)=520.668, p<.001).⁹ The means for husbands’ depressive symptoms were 3.41 (n= 1260; SD=3.73) at baseline, 4.26 (n=1042; SD=4.11) at wave 3, and 5.09 (n=735; SD=4.66) at wave 7. The means for wives’ depressive symptoms were 4.60 (n=1259; SD=4.60) at baseline, 5.49 (n=1095; SD=4.94) at wave 3, and 6.09 (n=917; SD=4.95) at wave 7.

Table 1.

Husbands’ and wives’ frailty over time

		Wives
		Baseline						Wave 3						Wave 7
Husbands		Miss	Deceased	Non- frail	Pre- frail	Frail	Total	Miss	Deceased	Non -frail	Pre- frail	Frail	Total	Miss	Deceased	Non- frail	Pre- frail	Frail	Total
Baseline	Miss	1	0	1	1	1	4
	Deceased	0	0	0	0	0	0
	Nonfrail	0	0	296	256	25	577
	Prefrail	2	0	229	317	60	608
	Frail	0	0	25	36	10	71
	Total	3	0	551	610	96	1260
Wave 3	Miss							84	5	10	12	2	113
	Deceased							21	8	27	51	9	116
	Nonfrail							14	5	202	207	26	454
	Prefrail							21	17	180	230	43	491
	Frail							6	4	21	43	12	86
	Total							146	39	440	543	92	1260
Wave 7	Miss													121	18	16	24	4	183
	Deceased													66	55	78	121	31	351
	Nonfrail													12	18	111	123	16	280
	Prefrail													24	33	122	167	33	379
	Frail													5	8	25	25	15	67
	Total													228	132	341	460	99	1260

Note. Raw numbers of participants are reported.

Main hypothesis testing

Model estimates are presented in Tables 2 for each hypothesis. Supporting hypothesis 1, Models 1 and 2 indicated that for both husbands and wives, an individual’s greater frailty predicted increases in his or her own subsequent frailty (actor effects). The same was true for depressive symptoms. Supporting hypothesis 2, Models 1 and 2 indicated that greater frailty status in one spouse predicted a greater frailty status in the other spouse (partner effects). The same was true for depressive symptoms with one exception. In Model 1, wives’ depressive symptoms did no significantly predict husbands’ depressive symptoms. In line with hypothesis 3, in both models each individual’s greater frailty was associated with increases in their own depressive symptoms, and greater depressive symptoms was associated with increases in their own frailty (actor effects). Finally, hypothesis 4 was not supported. The direct associations between one partner’s frailty and the other partner’s depressive symptoms did not reach statistical significance in either model.

Table 2.

Estimates for overall APIM models: Hypothesized actor and partner effects and secondary associations

Model 1: Frailty predicting depressive symptoms				Model 2: Depressive symptoms predicting frailty
	Estimates	S.E.	Est./S.E.		Estimates	S.E.	Est./S.E.
Longitudinal associations between frailty
Actor effects (Hypothesis 1a)
Wives frailty→ wives frailty	0.809	0.045	17.983***	Wives frailty → wives frailty	0.702	0.045	15.631***
Husbands frailty→ husbands frailty	0.779	0.046	16.998***	Husbands frailty → husbands frailty	0.676	0.046	14.660***
Partner effects (Hypothesis 2a)
Wives frailty →husbands frailty	0.179	0.042	4.258**	Wives frailty →husbands frailty	0.131	0.044	2.961**
Husbands frailty→ wives frailty	0.228	0.042	5.408***	Husbands frailty→ wives frailty	0.140	0.042	3.366**
Longitudinal associations between depressive symptoms
Actor effects (Hypothesis 1b)
Wives depress→ wives depress	0.623	0.021	29.880***	Wives depress→ wives depress	0.661	0.019	34.425***
Husbands depress→ husbands depress	0.643	0.022	29.132***	Husbands depress→ husbands depress	0.689	0.021	32.862***
Partner effects (Hypothesis 2b)
Wives depress → husbands depress	0.044	0.025	1.801††	Wives depress → husbands depress	0.084	0.022	3.816***
Husbands depress → wives depress	0.115	0.025	4.624***	Husbands depress → wives depress	0.159	0.025	6.422**
Longitudinal associations between frailty and depressive symptoms
Actor effects (Hypothesis 3)
Wives frailty→ depress	0.267	0.126	2.119**	Wives depress → frailty	0.022	0.007	3.058**
Husbands’ frailty → depress	0.338	0.128	2.633***	Husbands’ depress → frailty	0.027	0.009	3.009**
Partner effects (Hypothesis 4)
Wives frailty → husbands depress	0.270	0.144	1.882†	Wives depress → husbands frailty	0.003	0.007	0.357
Husbands frailty → wives depress	0.254	0.135	1.886†	Husbands depress → wives frailty	0.013	0.009	1.419

Note. All like effects (e.g. baseline to wave 3; wave 3 to wave 7) were set to be equal, giving us an overall test of whether previous scores influenced subsequent scores.

^***p<.001,

^**p<.01,

^*p<.05,

^†p=.06,

^††p=.07. The fit indexes for Model 1 were: CFI=.913; TLI=.859; RMSEA=.057. For Model 2, they were: CFI=.914; TLI=.860; RMSEA=.057. Model 1 accounted for 37% and 44% of the variance in wives’ wave 3 and wave 7 depressive symptoms, respectively; 22% and 49% for wives’ frailty; 38% and 38% for husbands’ depressive symptoms; and 20% and 45% for husbands’ frailty. Model 2 accounted for 40% and 45% of the variance in wives’ wave 3 and wave 7 depressive symptoms, respectively; 21% and 44% for wives’ frailty; 41% and 38% for husbands’ depressive symptoms; and 18% and 41% for husbands’ frailty. In both Models, frailty is entered as an ordinal variable with possible values of 0 (nonfrail), 1 (prefrail), and 2 (frail). Covariates in each model include both partners’ age, social support, and education.

DISCUSSION

This study examined associations between older spouses’ frailty and depressive symptoms over time. As hypothesized spouses’ frailty statuses were associated, such that greater frailty in one spouse was related to greater subsequent frailty in the other spouse. This supports past findings that older spouses’ physical health is interdependent^8,28 and extends this to frailty specifically. Because clinicians use frailty criteria to classify individuals as vulnerable to disability and as a basis for designing interventions, it is important to take into account that spouses’ frailty statuses are linked.

We also found that an individual’s own greater frailty status was associated with his or her own greater subsequent depressive symptoms. It was also the case that an individual’s own depressive symptoms was associated with his or her own greater subsequent frailty. These findings provide further evidence for bi-directional pathways and potentially the diagnostic overlap⁴, of frailty and depressive symptoms.⁴

Our results for the direct associations between an individual’s frailty and the spouse’s depressive symptoms did not reach statistical significance. This finding does not support relational identity theories¹⁵ and findings from previous research on disability and depressive symptoms⁸. However, in general we found that spouses’ depressive symptoms were linked, supporting affect contagion theory^11,29. Thus, it seems that frailty is associated with depressive symptoms within an individual and the resulting depressive symptoms have influences on spouses. We also found that effects did not differ for husbands and wives. This contrasts with our previous work on physical activity⁹ and the theory that women have more interdependent identities than men³⁰.

Strengths of this study include, first, that we used a large sample of older couples over a span of approximately eight years. Second, we used the APIM¹⁶, the state of the art approach for analyzing couples data. Third, the frailty measure included both subjective and objective health indicators. Limitations include that the CHS does not have variables related to relationship functioning and caregiving burden that would have shed light on mechanisms through which frailty related to depressive symptoms. Additional limitations include potential comorbidity, as we were not able to control for all possible medical conditions in the model. Future research may benefit from using other frailty measures and examining associations in more ethnically diverse populations.

CONCLUSION

Taken together, this study shows that spouses’ frailty and depressive symptoms are intertwined, highlighting the need for clinicians to take into account older adults’ close social environment when diagnosing patients and when designing interventions to treat and prevent frailty. Accounting for close partners’ shared activities and environment may have additive or synergistic effects on physical and psychological health. In addition, this work suggests that senior living communities may benefit from increasing incentives and opportunities for couples to engage in physical activity, socialization, and mutual support together.