Health Profiles of Spouse Caregivers: The Role of Active Coping and the Risk for Developing Prolonged Grief Symptoms

Introduction

Currently in the United States there are 34 million informal caregivers providing care for adults with acute, chronic, or terminal illnesses, often to their own detriment (National Alliance for Caregiving, 2016). Dubbed the ‘hidden patients’ (Hill, 2003; Kristjanson & Aoun, 2004; Roche, 2009), family caregivers face health hazards that include serious physiological (e.g. faster cellular aging, cardiovascular risk factors, poorer immune function) (Fonareva & Oken, 2014; O’Donovan et al., 2012; Pinquart & Sorensen, 2007; Ross et al., 2017), functional (e.g. poorer cognitive function and self-care) (Hoffman, Lee, & Mendez-Luck, 2012; Vitaliano, Ustundag, & Borson, 2017), and mental health consequences (e.g. higher rates of anxiety disorder, depression, and suicidal ideation) (Adelman, Tmanova, Delgado, Dion, & Lachs, 2014; M. H. Kim et al., 2016; Pinquart & Sorensen, 2003a; Stansfeld et al., 2014; Vitaliano, Strachan, Dansie, Goldberg, & Buchwald, 2014). Yet, some caregivers report positive gains from caregiving, and may even experience physical and mental health benefits (Fauth et al., 2012; Pristavec, 2018).

Family caregivers are a heterogenous group, but are often considered as a single population within the scientific literature. The aggregation of potentially distinct sub-groups of caregivers may help explain the surprisingly small effect sizes found by meta-analyses of caregivers’ health outcomes when analyzed as single population in comparison to non-caregivers (Roth et al., 2019; Vitaliano, Zhang, & Scanlan, 2003). Several papers have called for future studies to look beyond a simple measure of caregiver status as a way to understand why some, but not all, caregivers experience poorer health and well-being associated with caregiving (Boerner, Shultz, & Horowitz, 2004; Kaschowitz & Brandt, 2017; Vitaliano et al., 2014). A major question remains as to whether some caregivers are more naturally resilient than others to the stress imparted by the caregiving role.

A number of studies have linked circumstances of the patient, such as changes in symptoms and declines in functional impairment, or circumstances of the caregiving role, such as duration, co-residential status, and daily demands of caregiving, or the relationship between caregiver and patient, as predictors of negative caregiver outcomes (Adelman et al., 2014; Gitlin & Schulz, 2012; Kaschowitz & Brandt, 2017; Rigby, Gubitz, & Phillips, 2009; Torti, Gwyther, Reed, Friedman, & Schulman, 2004). Other studies, though fewer in number, have emphasized characteristics of the caregiver, such as gender and age, genetic factors, personality, and what types of support and general coping strategies they use as possible predictors of caregiver outcomes (Goetter et al., 2019; H. Kim, Chang, Rose, & Kim, 2012; Uccheddu, Gauthier, Steverink, & Emery, 2019; Vitaliano et al., 2014). Those caregivers who adopt a more active, as opposed to avoidant, coping strategy may have better quality of life and less adverse outcomes associated with caregiving (Kershaw, Northouse, Kritpracha, Schafenacker, & Mood, 2004). Whether it is related to innate personality differences, effective coping behaviors learned earlier in the life course, or characteristics and resources of the caregiver, some persons appear to be more resilient, thrifty, or hardy as they negotiate the stressors associated with the caregiving role (Carver & Scheier, 2017; Connor-Smith & Flachsbart, 2007; McCrae & Costa, 1986). The concept of caregiver risk phenotypes, or a distinct subgroup of caregivers who are more vulnerable to negative stress-related outcomes, has emerged as one promising hypothesis (Vitaliano et al., 2014).

Our focus is on spousal caregiving for patients with terminal cancer, which is a specific but not uncommon context, given that cancer is among the top three leading causes of death in adults 65 years and older (Heron, 2019). Spousal caregiving represents the most intensive form of caregiving, due to its coresidential and often extended duration; spouse caregivers report higher rates of caregiver-related distress compared to adult child caregivers (Pinquart & Sorensen, 2011). End-of-life caregiving is particularly intense due to its close association with the care-recipient’s impending death (Penrod, Hupcey, Shipley, Loeb, & Baney, 2012). While patient death is the natural end of the caregiving role, it is also the start of a bereavement phase for the surviving spouse.

Spousal bereavement is considered among the most stressful of all life events (Holmes & Rahe, 1967). Overtime, spouse/partners gradually cope with the loss and most do not require intervention (Schut & Stroebe, 2010; M. S. Stroebe, Hansson, Schut, & Stroebe, 2008); however, there is growing recognition that a small minority of bereaved individuals experience an intractable and disabling form of grief (i.e. “prolonged grief” or “complicated grief”), where there is a failure to adjust to the loss (Maciejewski, Maercker, Boelen, & Prigerson, 2016; Shear et al., 2011). Factors associated with complicated grief include having avoidant or anxious attachment styles, greater preloss dependency on the deceased, difficulty in making sense of the loss or finding meaning, less wealth, low social support and higher social capital (Burke & Neimeyer, 2013; Kung, 2020). Individuals with compromised health status, particularly if they are older, also appear to be at greater risk for complicated grief due to coping reserves being drained (Lundorff, Holmgren, Zachariae, & Farver-Vestergaard, 2017). In addition, preloss characteristics and other experiences of the caregiver, particularly other losses, may impact and continue to reverberate into the bereavement period (M. S. Caserta, Utz, Lund, Supiano, & Donaldson, 2017; Lundorff et al., 2017; B. R. Williams, Sawyer, Roseman, & Allman, 2008).

This study adapts a theoretical framework based on the Stress Process Model (Pearlin, Mullan, Semple, & Skaff, 1990). The Stress Process Model is a well-established theoretical perspective that links life stressors and strains from caregiving to individual-level health and well-being outcomes, recognizing that each individual’s perception and reaction to stressors and strains will be associated with their personal context, coping styles, and appraisal of the stress. As shown in Figure 1, the proximal outcome in our model is caregiver mental and physical health, with three major caregiving-related stressors/strains, as well as background characteristics and active coping styles of the individual, as the predictors or antecedents of caregiver health. The model explicitly links “prolonged grief” as a bereavement-related outcome that is distal to the caregiver’s stressors/stains and health outcomes, yet is temporally and substantively relevant to the stressors/strains of caregiving. Our life course-enhanced Stress Process Model model explicitly recognizes that caregiving and bereavement are not isolated stressors, and thus provides a framework to explore whether there are subgroups of persons who are more or less resilient in the face of caregiving, which can then predict whether they have more or less favorable bereavement outcomes.

Fig. 1.

Adapted Stress Process Model.

The purpose of this study was to characterize distinct sub-groups of spouse/partner caregivers to persons with cancer that experience poorer mental and physical health, and then to examine whether belonging to one of these caregiver sub-groups was associated with grief symptoms, especially a risk for prolonged grief, after the spouse who was receiving care died. We sought to address three aims:

1. Characterize distinct profiles of caregivers’ physical and mental health during the end-of-life caregiving period (i.e., when the spouse was receiving hospice care for terminal cancer).

2. Identify the background and antecedent factors associated with the distinct profiles of caregivers.

3. Determine the relevance of caregiver profiles to later grief outcomes, including the risk for developing prolonged grief symptoms.

Method

Data

This study is a secondary analysis of data from a subsample of spouse/partner caregivers from the Cancer Caregiver Study (CCS). The CCS is a program-project study funded by the National Cancer Institute that consists of three interrelated projects focused on enhancing the well-being of cancer caregivers using hospice services (Mooney et al., 2013). One project tested the effectiveness of an automated telephone-delivered symptom management system on caregiver well-being; another project explored communication patterns between family caregivers and hospice staff; and a final project explored the possibility of individually tailored bereavement trajectories and interventions (Mooney et al., 2013). All three studies of the CCS used a single sample with individuals assigned to interventions using a randomized factorial design; each participant completed survey-based assessments upon enrollment (i.e., around time of patient’s hospice admission) and then at regularly scheduled intervals up to 15 months post-loss. The current longitudinal analyses focused on those randomized to the control group, who were not exposed to any of the CCS interventions. To capture the physical and mental health of caregivers and the characteristics of the caregiving role (Aims 1 and 2), we relied on data from the baseline survey of the CCS, which was collected pre-intervention and provided a present view of caregiving-related experiences and outcomes during the end-of-life period. For Aim 3, we used data from the follow-up surveys, which were collected from 6 to 15 months post-loss, providing a present view of how these same individuals were coping during bereavement. The prospective, longitudinal design of the CCS captures both time periods and is free from retrospective recall bias, which is inherent in other studies of caregiving and bereavement.

Sample

Potential participants were identified by hospices located in 4 metropolitan areas within different regions of the United States. All participants were identified as the primary caregiver to a family member with a late-stage cancer diagnosis who was being admitted to hospice care. Caregivers needed to speak English and be cognitively able to participate in all aspects of the original projects (surveys and interventions). Further details of the sampling procedures, including strategies used for recruitment and retention, can be found elsewhere (M. S. Caserta et al., 2017).The analytic sample was limited to spouse/partner caregivers only; other family caregivers such as siblings and adult-child caregivers were excluded. In total, 198 spouse/partner caregivers provided information on at least one of the key indicators of the baseline survey, comprising the sample used to characterize the health profiles of caregivers (Aim 1), as well as the background and antecedent factors associated with those profiles (Aim 2). And, 81 spouse/partners completed at least one follow-up survey between 6 and 15 months after the spouse/partner’s death and were in the “control” conditions of the CCS interventions. Participants randomized to a CCS intervention condition were excluded given that the interventions’ foci were to improve caregiver well-being and bereavement outcomes (Utz et al., 2017). This second subsample was used to capture one’s grief outcomes at or after 6-months post-loss (Aim 3). Together, these two analytic subsamples can identify naturally-occurring differences that occur among spouse/partners during the caregiving and bereavement period, free from the effect of any interventions.

Measures

Key variables used in these analyses were conceptualized in Figure 1. Outcome measures include a latent measure of caregiver health, which comprises both physical and mental health, as well as a distal measure of grief symptoms, including a risk for prolonged grief. Independent variables capture the stressors and strain associated with caregiving, as well as the caregiver’s general coping style. Background variables including sociodemographic characteristics are included as control variables.

Caregiver Health – Mental Health.

We used the anxiety subscale of the Hospital Anxiety and Depression Scale (HADS, (Zigmond & Snaith, 1983). This subscale consists of 7 Likert-type items each ranging from 0 to 3 that are summed for a total score with a possible range of 0 to 21, where a higher value is indicative of a greater level of anxiety. Multiple studies have demonstrated evidence of reliability (mean α = 0.83) (Bjelland, Dahl, Huag, & Neckelmann, 2002). Depressive symptoms were assessed with the short form of the Geriatric Depression Scale (GDS-SF) (Parmelee, Lawton, & Katz, 1989; Sheikh & Yesavage, 1986). The GDS-SF consists of 15 yes/no items that were summed for a total score ranging from 0 to 15. A higher score indicates a greater level of depressive symptomatology. This scale has been found to be a reliable measure of depression among a community-based sample of bereaved spouses/partners, including those who experienced losses due to cancer (Cronbach’s α = 0.84) (M. Caserta, Utz, & Lund, 2013).

Caregiver Health – Physical Health.

The physical health of caregivers was assessed using three measures. First, overall health was measured with a single self-report item, rated as excellent, very good, good, fair, or poor health. Second, the health subscale of Caregiver Reaction Assessment (Given et al., 1992) includes 4 items, with higher scores indicating greater impact of caregiving on physical health (Cronbach’s α = 0.72 in this sample). Third, we used the Meeting Physical Demands subscale of the Perceived Self-Care and Daily Living Competencies Scale (M. Caserta, Lund, & Obray, 2004; Utz, Lund, Caserta, & de Vries, 2011) with two items each ranging from 1–3 that are summed for a total of 2–6, where higher score indicates a greater level of one’s perceived ability to meet physical demands. For this measure, Utz et al. (2011) documented high reliability (Cronbach’s α = 0.85).

Grief.

The Prolonged Grief instrument (PG-13), (Prigerson et al., 2009) consists of 11 self-report items describing the presence and frequency of common grief symptoms. The 11 items, each measured on a 5-point Likert scale, were summed for a total score ranging from 11 to 55, with higher scores corresponding to a greater level of grief symptoms (Thomas, Hudson, Trauer, Remedios, & Clarke, 2014; Tomarken et al., 2012). Prigerson et al. (2009) reported a Cronbach’s α = 0.82 for these items. Each participant completed the PG-13 instrument up to 5 times after the loss: approximately 6 weeks, 2 months, 6 months, 9 months, and 15 months. Given our focus on identifying caregivers who are most at risk for adverse bereavement outcomes, such as complicated grief (defined as the persistence of intense grief feelings six or more months after the loss), we used the highest or max PG-13 score for each participant recorded at or after six months’ post-loss.

Stressors and Strains of Caregiving.

Perceived strains and stressors associated with caregiving were captured with three separate scales. Social impact from caregiving was measured with a subscale of the Caregiver Reaction Assessment (5 items: higher is a larger deficit of social support, Cronbach’s α = 0.80 in this sample) (Grov, Fossa, Tonnessen, & Dahl, 2006). A single item from the Social Relationship Index measured spousal relationship strain, and asked on a scale of 1 to 6: “Ignoring any helpful or positive aspects of the relationship, how upsetting is your spouse/partner to you?” (Campo et al., 2009). The Preparedness for Caregiving scale (Archbold, Stewart, Greenlick, & Harvath, 1990) consists of 8 items ranging from 0 – 4 and averaged for the summary score, in which higher scores indicate more preparedness for the caregiving role (standardized Cronbach’s α = 0.89 in this sample).

Coping Styles.

A subscale of the Perceived Self-Care and Daily Living Competencies Scale (M. Caserta et al., 2004; Utz et al., 2011) measured active coping, which refers to ability to address the challenges of daily living such as adapting to change, organizing time, planning ahead, and identifying and utilizing sources of help to meet one’s needs. This subscale consists of 7 items ranging from 7–21 where a higher score is associated with a greater level of active coping. Utz et al. (2011) documented high reliability (α = 0.85).

Background Variables.

Sociodemographic characteristics assessed in this study included age, gender, ethnicity/race, income, education, and employment.

Analytic Approach

Missing Data.

Data missing at the item level were handled with mean item substitution if less than 30% of items were missing. If greater than 30% of items were missing, the scale score was assigned as missing. Data missing at the scale level were handled as described below per study aim.

Aim 1.

We sought to characterize caregiver health by identifying distinct profiles using latent class mixture modeling with the software program Mplus v 8.2 (Los Angeles, CA, USA). One advantage of performing this type of analysis is the ability to examine heterogeneous data for potential underlying subgroups that fit into distinct profiles. If more than one pattern is observed, then subgroups who belong to those profiles can be described and interventions tailored to their unique characteristics and needs. Our selection of the most appropriate model was guided by a model-building approach beginning with a one-class solution, comparing model fit through the 5-class solution, and by the following indicators of model performance and fit: model convergence (entropy nearest 1.0), the size of the observed patterns (pattern sizes over 10% of sample), average posterior probabilities for most likely pattern near 1.0, a significant (p < .05) Vuong-Lo-Mendell-Rubin adjusted likelihood ratio test, the Parametric Bootstrap Likelihood Ratio Test (p-value NS), and by comparing the information criteria between models (see Table 1). Rho parameters were considered when deciding whether or not to retain each indicator in the latent profile. Missing data among health indicator variables (1% to 2% missing data) were handled using the MLR estimator in MPlus (essentially full information maximum likelihood estimation with missing at random assumptions).

Table 1.

Latent profile analysis model selection fit statistics (n = 198).

p for VLMR Adjusted Likelihood Ratio Test	# Classes	AIC	BIC	aBIC	Entropy	Proportion of all Classes > 10% of sample?
-	1	4972.63	5012.09	4974.07	-	-
<0.001	2	4798.06	4860.53	4800.34	0.866	Yes
0.029	3	4735.75	4821.25	4738.88	0.839	Yes
0.038	4	4692.03	4800.54	4696.00	0.823	Yes
0.366	5	4680.38	4811.91	4685.19	0.818	No

VLMR= Vuong-Lo-Mendell-Rubin; AIC= Akaike Information Criteria; BIC= Bayesian Information Criteria; aBIC= Sample-Size Adjusted BIC

Aim 2.

After determining the most appropriate number of observed profiles, we distinguished between caregivers that were more likely to belong to each profile using t-tests and χ², and calculated effect sizes using Cohen’s d for comparing the relative separation of means across latent profiles by indicator variable. A multivariable (adjusted) logistic regression analysis was then conducted in Mplus using a 3-step approach to identify factors associated with the poorer health profile (Asparouhov & Muthén, 2014). Covariates for the final multivariable logistic regression model were selected based upon the Stress Process Model and variables with statistical significance (p < 0.05) in unadjusted models.

Aim 3.

We used multiple regression to explore whether caregiver profiles are predictive of at-risk grief outcomes in a sub-sample of caregivers who remained in the study for up to 15 months after the loss of their spouses (and who did not receive a bereavement intervention). Modeling missing data was not attempted for this aim due to the smaller proportion of caregivers (42% of the sample from Aim 1) who fit these criteria for Aim 3. A mediation analysis was performed with procedures for a dichotomous mediator using Stata, version 15 (Statacorp, 2017), and using the SPost13 command to calculate a fully standardized logit coefficient (Long & Freese, 2014). To determine the mediating effect of caregiver health on the relationship between active coping and prolonged grief symptoms, the fully standardized coefficient of Path a (logit model) was multiplied by the standard deviation of active coping, and then divided by the standard deviation of the caregiver health profile (see Figure 2).

Fig. 2.

Mediation analysis: direct and indirect effects of active coping on prolonged grief symptoms.

*Fully standardized coefficient calculated for model with dichotomous mediator.

Results

Spouse/partner caregivers were, on average, 65 years of age, of predominantly non-Hispanic white ethnicity/race (94%), a majority were female (61%), and, by study design, all were caring for a spouse with a terminal cancer diagnosis. The mean time that elapsed from the patient entering hospice until death was 68 days, with a median of 37 days. Additional descriptive data can be found in Table 3.

Table 3.

Differences in spouse caregivers’ characteristics (unadjusted) between health profiles (t-tests/χ²).

Variable	Profile 1: Better Health (n =149 pre-death)	Profile 2: Poorer Health (n = 49 pre-death)	p
Age in years (mean ± SD; sample range 36–90)	66.29 ± 11.44	62.51 ± 9.75	0.039
Gender (% female)	94 (64%)	28 (58%)	0.300
Race/Ethnicity (% non-Hispanic white)	5 (3%)	2 (4%)	0.813
Education level (mean ± SD; sample range 2–7)	4.73 ± 1.42	4.39 ± 1.48	0.144
Income (mean ± SD; sample range 1–7)	4.84 ± 1.57	4.00 ± 1.73	0.002
Preparedness for caregiving (sample range .33 – 4)	2.79 ± 0.66	2.33 ± 0.61	<0.001
Spousal relationship strain (sample range 1 – 6)	2.19 ± 1.30	2.92 ± 1.51	0.002
Constriction of social life (sample range 1 – 5)	1.72 ± 0.69	2.25 ± 0.84	<0.001
Active coping (sample range 9 – 21)	18.51 ± 2.31	14.69 ± 2.83	<0.001
PG-13 post-death max score (sample range 12–52)*	25.02 ± 8.78	34.00 ± 8.02	<0.001

^*sample size for the post-death PG-13 prolonged grief symptoms variable was n = 81.

Aim 1.

From the overall sample of 198 spouse/partner caregivers, two distinct health profiles were identified (Vuong-Lo-Mendell-Rubin adjusted likelihood ratio = p < .001; entropy = 0.87; posterior probabilities > 0.90), one of which was comprised of a minority of caregivers (n= 49; 25%) who exhibited significantly higher anxiety and depressive symptoms, greater health impact from caregiving, more self-reported health problems, and greater difficulty meeting physical demands of daily life. See Table 2 for differentiation of caregiver health profiles by indicator variables and effect sizes. The two health profiles were highly divergent, with a wide separation between means observed, especially for depressive symptoms (2.56 ± 0.18 in the “low risk” profile vs. 8.77 ± 0.60 in the “high risk” profile; Cohen’s d = 3.58), anxiety (6.73 ± 0.39 in the “low risk” profile vs. 12.43 ± 0.63 in the “high risk” profile; Cohen’s d = 1.49), and meeting physical demands of caregiving (5.06 ± 0.09 in the “low risk” profile vs. 4.02 ± 0.13 in the “high risk” profile; Cohen’s d = 1.16).

Table 2.

Caregiver health profiles: Differentiation by indicator variable (Latent Profile Analysis).

Mental & Physical Health Indicators	Profile 1: Low Risk (n = 149) ± SE	Profile 2: Higher-Risk (n = 49) ± SE	p	Effect Size (Cohen’s d)
Anxiety (HADs, range 0 – 21)	6.73 ± 0.39	12.43 ± 0.63	< .0001	1.49
Depressive symptoms (GDS, range 0–14)	2.56 ± 0.18	8.77 ± 0.60	< .0001	3.58
Health Impact of Caregiving (range 2–4.75)	3.28 ± 0.04	3.65 ± 0.06	< .0001	0.85
Meeting Physical Demands (range 2–6)	5.06 ± 0.09	4.02 ± 0.13	< .0001	1.16
Self-Reported Health Problems (range 1–5)	2.15 ± 0.07	2.57 ± 0.12	= .002	0.53

Aim 2.

Unadjusted differences in characteristics of caregivers observed in the two health profiles are presented in Table 3. Notably, spouse/partner caregivers observed in the poorer health profile were significantly (at the p < 0.05 level) younger, had lower incomes, had higher preloss grief symptoms, were less prepared for the caregiving role, had higher amounts of relationship strain with their spouse, had greater constriction of their social life due to caregiving, and had lower active coping. However, in the multivariable logistic regression model (see Table 4), only lower active coping remained significant (OR 0.59; 95% CI 0.48:0.72) as a risk factor for belonging to the poorer health profile.

Table 4.

Multivariable logistic regression for risk of belonging to the poorer health profile

Predictor	Odds Ratio	SE	z	P	95% Confidence Interval
Age	0.98	0.02	−0.73	0.46	[0.95; 1.03]
Income	0.80	0.10	−1.72	0.09	[0.62; 1.03]
Constriction of social life	1.35	0.40	1.03	0.30	[0.76; 2.40]
Spousal relationship strain	1.20	0.19	1.16	0.25	[0.88; 1.63]
Preparedness for caregiving	1.27	0.52	0.59	0.56	[0.57; 2.84]
Active coping	0.59	0.06	−5.10	<0.01	[0.48; 0.72]

Pseudo R² = 0.34.

Aim 3.

Using the subsample of participants who were eligible for the bereavement analyses (n=81), Tables 5 and 6 show the factors that were associated with bereavement outcomes. Younger age, lower levels of active coping, and belonging to the poorer caregiver health profile were significant predictors of higher levels of grief symptoms in the unadjusted regression models (at the p < 0.05 level; See Table 4). Table 6 presents the multivariable regression analyses with all three predictors, illustrating that caregivers’ pre-loss health profile remained a significant predictor of developing prolonged grief symptoms (p = 0.018), controlling for caregivers’ age (p = 0.040) and amount of active coping (p = 0.049). Figure 2 illustrates there was an indirect effect of caregiver health, which mediated the relationship between active coping and prolonged grief symptoms.

Table 5.

Unadjusted regression models predicting maximum PG-13 grief score captured at 6–15 months post-death (n=81).

Predictor	Coefficient	SE	t	P	95% Confidence Interval
Age	−0.27	0.09	−2.63	0.006	[−0.46; −0.08]
Income	−1.13	0.59	−1.93	0.057	[−2.30; 0.03]
Preparedness for caregiving	−2.63	1.41	−1.86	0.066	[−5.44; 0.18]
Spousal relationship strain	1.54	0.80	1.93	0.058	[−0.05; 3.13]
Constriction of social life	2.43	1.30	1.88	0.064	[−0.15; 5.01]
Active coping	−1.37	0.31	−4.42	<0.001	[−1.98; −0.75]
Health profile (class 1 vs. 2)	9.99	2.16	4.63	<0.001	[5.70; 14.29]

Table 6.

Adjusted regression predicting maximum PG-13 grief score captured 6–15 months post-death (n=81).

Predictor	Coefficient	SE	t	p-value	95% Confidence Interval
Age	−0.19	0.09	−2.10	0.040	[−0.36; −.01]
Active coping	−1.37	0.37	−2.00	0.049	[−1.47; −0.00]
Health profile (class 1 vs. 2)	6.40	2.64	2.42	0.018	[1.14; 11.67]

Discussion

The 34.2 million Americans who are unpaid family caregivers to aging adults are essential to the well-being of and care for disabled and chronically ill patients over the age of 50 (National Alliance for Caregiving, 2016), and especially for those patients facing end-of-life health transitions (Waldrop, Kramer, Skretny, Milch, & Finn, 2005). Caregivers, however, are not immune to encountering their own physical and mental health challenges, often as a result of ignoring their own health and wellness while tending to the many stressors that come with being the primary caregivers to an end-stage patient (Pinquart & Sorensen, 2003b). For this reason, family caregivers have been dubbed the “hidden patient” (Roche, 2009). Spouse/partner caregivers may be particularly at risk for caregiver-related health challenges, given the co-residential and extended nature of the caregiving tasks they provide to a spouse/partner, as well as the impact grappling with an impending loss of an intimate partner (Pinquart & Sorensen, 2011). The transition from caregiving to widowhood is challenging (W. Stroebe & Schut, 2001), and both roles have been associated with a number of physical and mental health consequences (Pinquart & Sorensen, 2007; Schulz & Sherwood, 2008; M. S. Stroebe, Schut, & Stroebe, 2007). Caregiving researchers have called for studies exploring the circumstances around, and the characteristics of, caregivers that are associated with more or less favorable health outcomes during this end-of-life period (Vitaliano et al., 2014; Vitaliano et al., 2003).

This study used a subsample of spouse/partner caregivers from the Cancer Caregiver Study (CCS), a National Cancer Institute (NCI) funded study that followed cancer caregivers from the time a patient entered hospice until 15 months post-loss (Mooney et al., 2013). These data allowed us to explore whether there is a particular profile of caregivers who are more or less resilient in the face of the stressors and strains of providing care to an end-stage spouse or partner, and whether caregiver health profiles extend to or perhaps mold the way that an individual will cope with the challenges of widowhood. This study is framed within Stress Process Model (Pearlin et al., 1990), and provides an extension of this theoretical framework by linking the factors associated with the physical and mental health outcomes of caregiving and bereavement into a single model.

Our analyses observed two profiles of caregivers – one that included a majority of caregivers (75%) who appeared to be weathering the stressors and strains of caregiving with relatively few mental and physical health consequences, and another group that exhibited higher levels of depression and anxiety, greater levels of health problems associated with caregiving, and reduced ability to meet the physical demands of daily life. This “higher-risk” profile group comprised about 25% of the sample, suggesting that 1 in 4 spouse/partner caregivers may be experiencing significant health problems during the time they are in an active caregiving role. Those in the higher-risk profile group also felt significantly less prepared to be a caregiver, had greater constriction of their social life as a result of caregiving, and lower levels of active coping, compared to those in the lower-risk profile group. These differences reveal that, in addition to worse physical and mental health outcomes, those in the higher-risk profile are more likely to hold negative appraisals of the caregiving role or of their ability to cope with or carry out the caregiving role.

It is unclear from these analyses whether the differences in health outcomes or the differences in their reported levels of caregiving-related stressors are caused by the caregiving role. It is possible that caregivers in poorer health may be less able to perform or function in the caregiving role and therefore have more negative appraisals of the caregiving-related stressors, rather than the stressors of caregiving causing poorer health outcomes. A future study that is able to capture both the caregivers’ health outcomes and their appraisals of the caregiving-related stressors over time, and throughout a more extended end-of-life period that captures the transition of the spouse from a healthy to non-healthy state, would provide more clarity on the causal time-ordering of these effects between caregiving and health. Regardless, our analyses observed two distinct profiles of caregivers, with one identifying 25% of the sample that experienced significantly greater health challenges, including both physical and mental health issues, and significantly greater caregiver-related stressors.

Those caregivers in the higher-risk profile group did not differ from those in the lower-risk profile group by gender, race, education, or employment status. They did, however, exhibit lower incomes and were younger than those in the lower-risk profile group. The reported income differences are consistent with the large bodies of literature that find a stronger stress-illness relationship among those with lower income (D. R. Williams, 1990), and also the association between lack of financial resources and lower levels of overall well-being among caregivers (Chen, Fan, & Chu, 2019). Age differences are reflective of developmental lifespan theories (Neugarten, Moore, & Lowe, 1965) that suggest it is more “off-time” to provide end-stage caregiving to a spouse at younger ages (Van Vilet, de Vugt, Bakker, Koopmans, & Verhey, 2010). Poorer health outcomes among the younger caregivers could reveal failure to adjust to the role and all the stressors and strains that are a part of being a younger-than-average spousal caregiver, especially since employment did not differ across the two profiles. The CCS data are limited to spouse/partners who were over the age of 50 (with an average age of about 65 in the sample), thus representing a mostly retirement-age and older population, where chronic conditions and disability are not uncommon and where young children in the household are relatively rare. Repeating this research with a sample of spouses/partners who are facing end-stage caregiving at an “off-time” age might reveal even greater health consequences among young-age caregiver spouses/partners, as they are likely juggling the challenges of end-stage caregiving with other life stage demands related to employment and child care.

In the adjusted analysis predicting membership in one of the two caregiver health profiles, active coping washed out the effects of other variables that were associated with caregiver health profiles in the unadjusted or bivariate associations (e.g., age, income, anticipatory grief, preparedness for caregiving, relationship strain, and constriction of social life). According to the Stress Process Model theoretical framework that guided these analyses (Figure 1), active coping was conceptualized as a potential mediator between the stressors/strains of caregiving and caregiver health. Thus, both theoretically and empirically, the measure of “active coping” appears to capture an individual’s overall resilience or ability to cope with the stressors and strains of providing end-stage caregiving to a spouse/partner. Caregivers who reported high levels of active coping did not eliminate caregiving-related stressors in their lives, as evidenced by the strong and consistent bivariate associations between caregiver health and caregiver-related stressors and strains; however, it is possible that those caregivers with higher levels of active coping may have developed the skills and strategies to most effectively adjust to living with caregiving stressors (Coon, 2012). This study therefore supports previous studies indicating that active coping is a protective risk factor against the adverse health outcomes associated with caregiving (Watson, Tatangelo, & McCabe, 2018).

Furthermore, active coping emerged as a prominent predictor of bereavement outcomes in both the bivariate and multivariate analyses, as did caregiver age and caregiver health. Caregivers who were younger, had lower levels of active coping, and belonged to the higher-risk caregiver health profile had significantly higher levels of grief symptoms, as measured by a self-report grief instrument collected at 6–15 months post-loss. Grief is a natural response to loss and rarely requires clinical intervention; however, individuals who express high levels of grief symptoms that persist beyond 6 months post-death are at risk for prolonged grief disorder, which is a clinically-relevant condition that impacts functioning and can have adverse long-term effects on one’s health (Burke & Neimeyer, 2013). Active coping had both a direct effect on prolonged grief symptoms measured at or after 6 months, as well as an indirect effect through its association with caregiver health profiles.

Strengths and Future Directions.

By linking caregiver and bereavement-related health outcomes in a single analysis, we reinforce the importance of not viewing these as isolated life events. The stressors and strains of caregiving, how well one copes with those, and also how poor one’s health is at the time of caregiving sets the stage for how well one is able to cope with the stressors associated with bereavement. These analyses, especially the identification of a subset of caregivers who comprise a higher-risk health profile (25%) associated with both lower levels of active coping and higher levels of prolonged grief symptoms, provide an empirical foundation for upstreaming the identification of those caregivers who are most at-risk for adverse grief reactions that may require clinical interventions. Identifying and then intervening with caregivers who fall in the higher-risk profile may be a promising avenue for bereavement intervention. For example, intervention could be aimed at increasing the active coping skills of this group during the caregiving period, rather than waiting until 6+ months post-death to identify those who have the most intense and persistent grief responses that require clinical intervention. This type of intervention may ameliorate the negative health consequences often associated with caregiving, and also set the path toward a more positive bereavement adjustment as well.

This analysis extends the Stress Process Model by linking two related but distinct health outcomes – during the caregiving period and during the bereavement period – and by showing how well one copes with the common caregiving stressors is associated with how well one will cope with the natural stressors of bereavement. Coping behaviors are learned earlier in life, often during times of stress. In the transition from caregiving to bereavement, spouse/partners who had developed a more active coping strategy during caregiving were more likely to be healthier. The caregiver’s health profile was also significant as the mechanism through which active coping influenced symptoms of prolonged grief after losing a spouse/partner. These findings suggest that the positive coping skills and resources that are gained and practiced throughout the life course have a significant impact on health and well-being during the caregiving and bereavement periods (Kung, 2020). Then again, some persons may simply be better at coping with stressful situations, perhaps as an innate or personality-linked trait. Both theoretical explanations are plausible given the analyses at hand. Future research may continue to explore the development of coping skills throughout the life course, whether they are associated with innate personality characteristics, and how they are modified during the experience of repeated stressors across the life course.

Limitations.

There are some limitations to our analyses. The CCS sample is comprised only of cancer caregivers, who may differ qualitatively from caregivers of someone with dementia or other chronic condition in both caregiving and bereavement responses. Research, however, has found that bereavement from cancer death is equally as distressing as from many other causes of death (M. Caserta et al., 2013). Furthermore, the CCS sample was focused on the end-stage caregiving experience and restricted eligibility to those who were caring for a patient who was admitted to hospice. Hospice-users may differ from end-stage caregivers who do not use hospice, with the potential for superior adjustment to the loss with hospice support. It is also possible that caregivers who are just beginning caregiving have different perceived stressors and reactions to caregiving than those who are at a clear end-stage of caregiving. Yet, many families begin their experience with cancer uncertain if it will be terminal, and entrance to hospice is often the beginning of a new reality for both patient and caregivers.

The range of time from hospice entrance to death varies widely. Some families may stay in hospice for a lengthy decline or even transition out of hospice when death does not occur as anticipated (Russell et al., 2017), while other deaths occur within hours of entering hospice. The opportunity to gain the palliation and support from hospice would, in theory, vary accordingly. However, previous research has indicated no significant differences in caregiver outcomes associated with length of hospice care (Kris et al., 2006), nor differences in bereavement outcomes associated with whether the loss was anticipated (Carr, House, Wortman, Nesse, & Kessler, 2001). Although the cancer caregiver sample used here represents a distinct set of caregivers who have received the palliation and support from hospice, it is likely indicative of other caregivers’ experience. Future research should further explore the benefits of longer term hospice use for caregivers.

We were also not able to include all potentially salient characteristics of caregivers in this secondary analysis. For example, personality traits (particularly neuroticism) has been linked to health outcomes and to bereavement outcomes (Goetter et al., 2019). Including personality characteristics as moderators in future research would be important in furthering our understanding of the longitudinal caregiving and bereavement trajectories. Furthermore, replicating these findings in other caregiving relationships is also recommended. We limited our analyses to spouse/partner caregivers, since they often have the most intense caregiving experiences and bereavement responses, but other caregiver groups, such as adult-child caregivers, likely enact similar patterns of coping strategies and outcomes in response to the stressors of caregiving and bereavement. Additionally, future research should explore whether the health profile groups are perhaps predictive of other types of outcomes such as overall life satisfaction or personal growth. Those findings would continue to enhance our understanding of how common life stressors, including caregiving and bereavement, affect one’s health, and how one’s coping strategies may minimize the adverse health effects of those stressors.

Finally, there are important methodological limitations inherent in the dichotomy created by identifying a 2-class solution (essentially a “good health” versus “poor health” split), when it is plausible that in the larger population of caregivers there exists more of a gradient between good and poor health. The upside of the 2-class solution is that in health care, and practically speaking, clinicians are attempting to discover a similar dichotomy in a highly diverse patient population: those at high risk for poor outcomes and in need of an intervention versus those who will likely be okay with relatively little intervention. We realize that by attempting to identify, characterize, and explore the relevance of this dichotomy, we are potentially losing the nuance and complexity of a continuum. Future research should further explore these methodological issues in order to move the science forward.

Conclusions

Caregiving for a spouse at the end-of-life is a role that is naturally linked to the transition to widowhood, and yet the two periods are often considered in isolation in research. Although the distinct challenges and risk factors for poor outcomes should be considered for each time period separately, this study highlights the importance of considering the linkages across caregiving and bereavement. The majority of caregivers (75%) were observed as belonging to the healthier profile in this study. While this sub-group still exhibited some degree of mental and physical health problems, on average, it was consistent with the general population for their age group (Miller, 2012; Simon, 2015). The caregivers observed in the poorer health profile (25%) represent the target population for future interventions to improve health and well-being, especially as they confront the additional transition into the bereavement period. This has important implications for conserving scarce resources and creating interventions that are more effective in supporting the higher-risk caregivers. For practitioners who encounter families grappling with a terminal illness, this study suggests that a distinct subset of spouse/partner caregivers may already be experiencing poor health and will be at high risk of needing health care interventions themselves. Considering the importance of the family caregiver to the care of the patient, it may be most effective for practitioners to assess the family’s current or imminent health care needs to achieve quality end-of-life and bereavement care.

Highlights.

• Many spouses appeared to be weathering the stressors of end-of-life caregiving well.

• 1 in 4 spouse/partner caregivers exhibited significant health problems.

• Better active coping may help spouses adapt to caregiver role and preserve health.

• Coping style and health during the caregiving period may impact the grief process.

• Caregiving and bereavement should not be considered as isolated life phases.