Appraisals of Self in the Caregiver Role as Made by Married Custodial Grandparents

Abstract

Objective:

We examined how married custodial grandparents from the same household appraise the impact of caring for a grandchild on their sense of self, and how these appraisals are related to their psychological well-being

Background:

This study is important because there is scant information on how custodial grandmothers and grandfathers both experience their caregiver role, even though 70% of these grandparents are married.

Method:

Four competing measurement and structural models from the combined perspectives of the stress process model and the two-factor model of caregivers’ psychological well-being were tested with 193 married grandmothers and grandfathers.

Results:

The best fitting model was one in which two positive (personal gain and satisfaction) and two negative (loss of self and role captivity) appraisals emerged as distinct first-order constructs. Both negative appraisals correlated highest with negative affect, and both positive appraisals correlated highest with positive affect. The measurement and structural components of this model were largely invariant by grandparent gender. Mean comparisons showed that grandmothers reported significantly higher negative appraisals than grandfathers, with the latter reporting significantly greater perceived gain in the caregiving role. Bivariate correlations between grandmothers’ and grandfathers’ appraisals were nonsignificant for perceived gain and loss of self and of only moderate magnitude for caregiver satisfaction and role captivity.

Conclusions:

Given that appraisals are a core component of coping with stressors, our findings have important practice implications and point to meaningful directions for future research regarding custodial grandparent families.

Interest in custodial grandparent families, defined as families where grandparents provide full-time care without significant involvement by grandchildren’s biological parents, has soared over the past quarter century (for review, see Hayslip, Fruhauf, & Dolbin-MacNab, 2017). The most common reasons for these nonnormative care arrangements involve adverse events that lead to parental absence or inability to parent (e.g., substance abuse, child neglect, incarceration, physical and mental illness). Studies using large national samples have provided substantial evidence of risk for psychological difficulties among custodial grandmothers (CGM; Minkler, Fuller-Thomson, Miller, & Driver, 1997; Strawbridge, Wallhagen, Shema, & Kaplan, 1997; Szinovacz, DeViney, & Atkinson, 1999) who provide the bulk of care within these families. Although more than 70% of custodial grandparents (CGPs) are married (Kreider, & Ellis, 2011), scant research has focused the caregiving experiences of custodial grandfathers (CGFs), and there has been even less emphasis on how CGPs married to each other experience caring for a grandchild. To address these gaps in the literature, the present study examines how married CGMs and CGFs from the same household appraise the role of caring for their custodial grandchild and how these appraisals are related to their overall psychological well-being.

Given that CGPs find caregiving to be both stressful and rewarding (Giarrusso, Silverstein, & Feng, 1999; Hayslip & Kaminski, 2005), our first aim was to test and compare four models (see Figure 1) that are based conceptually on both the two-factor model of caregiving appraisal and psychological well-being (Lawton, Moss, Kleban, Glicksman, & Rovine, 1991) and the stress process model of family caregiving (Pearlin, Mullan, Semple, & Skaff, 1990). The second aim was to test whether the measurement and structural aspects of the best fitting model arising from these comparisons are equivalent by CGP gender. The final aim was to examine the means and zero-order correlations observed for CGMs and CGFs regarding their specific caregiving appraisals.

FIGURE 1.

Competing Models of Negative and Positive Appraisals of Self as a Caregiving Grandparent.

CGFs in the Caregiver Role

The few studies examining CGFs, to date, suggest that grandfathers are indeed involved with and impacted by providing care to a custodial grandchild (Patrick & Tomczewski, 2008). Bullock (2005) conducted interviews with 26 CGFs (aged 65 years and older) from rural North Carolina and reported that the grandfathers as feeling powerless, as exemplified by giving up leisure time and personal activities to maintain a schedule that was consistent with their parenting responsibilities. Kolomer and McCallion (2005) conducted focus groups with 33 CGFs in the New York City area and found these grandfathers to express such themes as missing freedom, experiencing childrearing differently from how they experienced it with their own children, and fearing how adverse health changes might affect the custodial grandchild’s future care. When these CGFs were later compared with 33 unrelated CGMs recruited from another study (matched on age, race, and marital status), CGMs reported significantly more depressive symptoms. There were no gender differences, however, in self-reported caregiving mastery.

Unlike the preceding studies, which relied on small, nonrandom samples, Park (2009) examined common and differential factors related to the psychological health of 1,781 CGMs and 209 CGFs present in three waves of the National Survey of America’s Families (NSAF; 1996–2001). Although CGFs had better overall mental health than CGMs, caring for custodial grandchildren with severe behavioral problems was nevertheless significantly related to lower psychological well-being regardless of CGP gender. Additionally, CGFs with spouses reported better mental health than their unmarried counterparts. Yet the generalizability of Park’s (2009) findings are limited given that the NSAF just included adults age 64 or younger and oversampled for low-income families.

In the only study with married CGFs and CGMs from the same family, Smith and Hancock (2010) found CGMs to report significantly greater anxiety, marital distress, use of harsh discipline, anger at the custodial grandchildren’s biological parent, and boundary ambiguity than did CGFs. In addition, self-rated health and attachment with custodial grandchildren were significantly lower for CGMs. Despite the mean level gender differences in the foregoing variables, the linkages among marital distress, grandparent psychological distress, dysfunctional parenting, and CGCs psychological difficulties were remarkably similar for CGFs and CGMs. Through structural equation modeling techniques, Smith and Hancock (2010) also found the key measurement properties of constructs examined in their study to be similar regardless of CGP’s gender.

Although the preceding studies offer preliminary evidence of how CGFs perceive and are meaningfully affected by providing full-time care to a custodial grandchild, this critical area of research is in its infancy. Importantly, whether the influence of this enduring caregiving role on one’s sense of self is the same or different for CGMs and CGFs has yet to be examined.

Conceptual Framework

Under the umbrella of family systems theory (Cox & Paley, 1977), the present study is conceptually derived from a combination of the stress process model of family caregiving (Pearlin et al., 1990) and the two-factor model of caregiver psychological well-being (Lawton et al., 1991). Both approaches are rooted in general models of stress and coping and postulate that subjective appraisals of being in the caregiver role mediate the relationship between the objective demands of caregiving and the caregiver’s psychological well-being. Although both of these models were developed with family caregivers of older adults in mind, they are broad in scope and dynamic in nature, making them easily adapted to persons across a variety of caregiving situations (Whitlatch, Schur, Noelker, Ejaz, & Looman, 2001). In this regard, Pearlin et al. (1990) defined informal caregiving generally as “activities and experiences involved in providing help and assistance to relatives or friends who are unable to provide for themselves” (p. 583).

Stress Process Model

Pearlin et al. (1990) described several factors in the stress process model that contribute to outcomes for family caregivers of older adults that may apply to any caregiver who provides full-time unanticipated and nonnormative care to a family member. In particular, Skaff and Pearlin (1992) stated that “the profound changes in the organization of people’s lives that are often demanded by long-term caregiving can come to reshape self-concepts” (p. 656). They further held that changes in self-concept due to self-evaluations anchored in the caregiver role are likely to covary with psychological well-being and that the perception of caregiving as being relevant to one’s self-concept is more important than the “objective characteristics of the situation.” However, the extent to which these processes are true of CGFs and CGMs is currently unknown.

One key change in self-concept experienced by family caregivers noted by Skaff and Pearlin (1992) is a potential “loss of self,” which occurs when the constriction of normal activities resulting from caregiving lead to a shrinkage of self. This possibility is well exemplified by the themes of powerlessness and loss of freedom that were found in the previously mentioned qualitative studies of CGFs (Bullock, 2005; Kolomer & McCallion, 2005). Similarly, Landry-Meyer and Newman (2004) conducted in-depth interviews with 25 custodial CGMs and found that most saw their enactment of the caregiving role as being “off-time” with respect to their life-cycle timing and accompanied by loss of the traditional grandparent role that they had envisioned. Like Skaff and Pearlin (1992), they concluded that such feelings of loss and significant lifecycle disruption accompanying the caregiver role could lead to change in one’s sense of self.

Family caregiving can lead also to feelings of self-gain as exemplified by a sense of personal enrichment or character building that comes from having coped with a difficult situation (Lawton et al., 1991; Skaff & Pearlin, 1992). It is common for caregivers to report positive feelings, including the satisfaction of knowing that their loved one is getting excellent care, sensing personal growth and increased purpose in one’s life, as well as gratification from passing on a tradition of care and modeling caregiving to their custodial grandchildren. Similarly, love for and commitment to a custodial grandchild can lend value and satisfaction to the caregiving role, as does CGPs’ knowledge that they are contributing to their grandchild’s well-being (Hayslip & Kaminski, 2005). Self-gain is like self-loss given that it requires one to evaluate the personal meaning of the caregiving experience to one’s sense of self (Skaff & Pearlin, 1992).

Two-Factor Model

Like Skaff and Pearlin’s (1992) assertion that changes in how the self is appraised in the caregiver role covary with psychological well-being, the two-factor model maintains that one’s appraisal of the quality of his or her caregiving is the key mediator between the objective stressful demands of caregiving and the outcome of psychological well-being (Lawton et al., 1991). The two-factor model expands the stress process model, however, by advancing that two parallel processes occur whereby two kinds of appraisals differentially affect two types of well-being in ways that are congruent with their emotional valence and partially independent of each other: Positive appraisals yield positive affect, whereas negative appraisals yield negative affect (Lawton et al., 1991). Importantly, however, positive emotions can co-occur with negative emotions with relatively high frequency, even in the presence of personally significant stress (Ong, Bergeman, & Bisconti, 2006).

The distinction between caregivers’ positive and negative emotionality advanced by the two-factor model is imperative in view of evidence from the general stress and coping literature that positive emotions have important functions in the stress process and encompass coping processes that are separate from those that regulate distress. For example, Folkman (2008) maintains that positive emotions and appraisals are normative aspects of the stress process that facilitate coping by activating effective coping strategies, giving meaning and purpose to stressful situations, and fostering positive biological responses to stress. Regarding family caregivers per se, Ong, Bergeman, and Boker (2009) noted that both positive and negative appraisals are normative reactions to the demands of caregiving with positive appraisals being associated with resilience.

Applying the two-factor model to CGPs is important because the experience of raising a custodial grandchild is not entirely negative or positive to one’s selfhood, as evidenced by Giarrusso et al. (1999) who observed that 54% of 162 caregiving grandparents from the larger Study of Intergenerational Linkages II found caregiving to be both stressful and rewarding. Yet to date, research on CGPs has focused predominantly on the negative aspects of caregiving. An exception, however, was Pruchno and McKenney’s (2002) study of 867 Black and White CGMs, which showed that the principles guiding a two-factor model of psychological well-being operate at the level role appraisals among CGMs of both races. In a study of 733 Black and White CGMs, Smith and Dolbin-MacNab (2013) similarly found that positive and negative appraisals were significantly related to both psychological distress and positive well-being.

Study Aims and Hypotheses

Examining how married CGPs from the same household appraise the role of caring for a custodial grandchild and how such appraisals are related to their psychological well-being is important in view of the family systems theory prediction that the emotions and behaviors of each subsystem are related to the emotions and behaviors of other systems (Cox & Paley, 1997). Thus, it would be expected that the emotions of one marital partner would influence the emotions of the other in their interactions (Cummings, Goeke-Morey, & Papp, 2002). Our explicit focus on CGMs and CGFs who are married to each other permits us to control for potential confounding factors that might otherwise be present if CGPs from both genders were recruited separately.

Given that CGPs find caregiving to be both stressful and rewarding (Giarrusso et al., 1999; Hayslip & Kaminski, 2005), our first aim was to test and compare the four models in Figure 1. These models were derived conceptually from both the stress process model (Pearlin et al., 1990) and two-factor model (Lawton, et al., 1991). All four models reflect the main tenet of the two-factor model that two parallel processes occur in which positive and negative appraisals of being a caregiver differentially affect psychological distress and positive well-being in ways that are congruent with their positive and negative valence (Lawton et al., 1991). All four models also involve appraisals of self as emphasized by the stress process model. The four models differ, however, regarding the extent to which negative and positive appraisals of self as a caregiver exist as first or second order constructs. This differentiation is important in light of the observation that “many indicators of positive and negative affect of Black and White grandmothers are similar to one another, including levels of caregiver satisfaction, life satisfaction, positive affect, and depression” (Pruchno & McKenney, 2002, p. 451). Thus, an important step is to determine whether these constructs are truly distinct from one another.

Model 1 depicts all four of the appraisals examined here as being separate and distinct first order constructs, with each having its own independent direct effects on CGP’s positive and negative affect. Model 2 depicts the two negative caregiving appraisals (role captivity and loss of self) as first-order constructs, whereas the two positive caregiving appraisals (gain and satisfaction) are regarded as indicators of a second-order positive appraisal factor. Model 3 alternatively posits the two positive appraisals as separate first order constructs, while the two negative appraisals are regarded as indicators of a second-order negative appraisal factor. Finally, Model 4 postulates the simultaneous presence of second-order positive and negative caregiving appraisal constructs, respectively. Across all models, we hypothesize that appraisals (modeled as either first- or second-order constructs) will be more strongly associated with the affect state of similar valence than with that of the opposite valence (Lawton et al., 1991).

Our second aim was to test whether the measurement and structural aspects of the best fitting model arising from the preceding comparisons are equivalent by CGP gender. This test was important because it cannot be inferred that the meaning of constructs in the model are similar for CGFs and CGMs in the absence of true measurement invariance (Adamsons & Buehler, 2007; Byrne, 2012). Even if the measurement model is equivalent by gender, it may be that invariance does not similarly exist by gender for the pathways, which connected the latent constructs.

Our final aim was to examine mean differences between CGMs and CGFs in their caregiving appraisal scores, as well the zero-order correlations between CGM and CGF appraisals. Given that CGMs are far more like to be primary caregivers and more invested than CGFs (Hayslip et al., 2017), we expect their caregiving appraisals mean scores to be higher than those of CGFs. In addition, because of their shared caregiving environment within the same family system (Cox & Paley,1997; Cummings et al., 2002), we expect moderate to high correlations between the caregiving appraisals reported by respectively by CGMs and CGFs.

Method

Participants

Participants were 386 married custodial grandparents (193 married custodial CGMs [M_age = 55.7 years, SD = 7.61] and 193 CGFs [M_age = 58.2 years, SD = 9.3]) residing in the same household, and providing full-time care to a custodial grandchild in absence of the child’s birth parents for at least 3 months, M = 6.0 years, SD = 3.4. These 193 married CGPs couples were from a larger sample of 733 CGMs who participated in an earlier study by Smith, Palmieri, Hancock, and Richardson (2008) and were recruited across the 48 contiguous states via a combination of convenience (social service agencies; Internet, radio, and newspaper ads) and population-based (recruitment letters sent to randomly generated lists of households with children under age 18 years) approaches. All 733 CGMs from the larger sample were asked at screening if they were living with and married to their custodial grandchild’s grandfather. If they answered yes, then the CGF was invited to participate. This approach yielded the 193 CGM–CGF couples examined in the present study. Grandparents were excluded if care provision was due to the death of their own offspring. If a dyad was caring for multiple custodial grandchildren, then a target custodial grandchild was objectively selected based on the most recent birthday. Target custodial grandchildren were 103 girls and 90 boys (M_age = 9.3 years, SD = 3.6, range = 4–16.4 years). Most CGPs (68.4%) gave care to a custodial grandchild born to a daughter or stepdaughter; were White (64 %); had incomes below $50,000 (69.5%); lived in large cities, small towns, or suburbs (55.9%); and reported some form of legal arrangements for the custodial grandchild (84.5%). Reasons for providing care to custodial grandchildren primarily concerned crisis or tragedy in the parent generation (e.g., 51.8% parental substance abuse; 47.2% parental incarceration).

Measures

Interviewers at a major public university in Ohio conducted separate phone interviews with CGMs and CGFs. The measures described here were modeled as latent constructs (Figure 1). Parceling was used to obtain indicators of the negative and positive affect constructs to control for error in estimating relations among model constructs (MacCallum & Austin, 2000). All appraisal measures were preceded by this statement: “The following is a list of feelings that grandparents who are caring for their grandchildren may feel. How often have you experienced each of these feelings since you started caring for your grandchild?” All appraisal measures were derived from the stress process model except for caregiver satisfaction.

Affect.

Negative affect was measured using 16 items from the 20-item Center for Epidemiologic Studies Depression Scale (CES–D; Radloff, 1977) to derive three separate indicators of negative affect (i.e., depressive affect, somatic, and interpersonal) previously shown by Hertzog, van Alstine, Usala, Hultsch, and Dixon (1990) to represent distinct factors existing within the overall CES–D as purported by Radloff (1977). For each item, participants responded how often they had felt in the past week on a 4-point scale from 0 (rarely or none of the time) to 3 (most or all of the time). The items purported by Radloff (1977) for each CES–D negative affect factor were summed to create a total score for that factor. The following Cronbach’s alpha values were observed for total scores: CGM = .89, CGF = .83.

Positive affect was measured with 11 items including 10 from the Positive and Negative Affectivity Scale (Watson, Clark, & Tellegen, 1988) and a single item (“happy”). Respondents rated how often they experienced each emotion (e.g., interested, excited, enthusiastic), from 1 (not at all) to 6 (extremely). Potential scores ranged from 11 to 66, with higher scores reflecting greater positive affectivity. Cronbach’s alpha values were observed for total scores: CGM, α = .87; CGF, α = .85.

Positive appraisals.

Perceived gain was measured by summing four items (“Became more aware of your inner strengths,” “Became more self-confident,” “Felt as though you’ve grown as a person,” Learned to do things you didn’t do before”) developed by Pearlin et al. (1990) for use with family caregivers and rated from 1 (never) to 4 (very often). Potential scores range from 4 to 16. Higher scores indicated greater gain in the caregiver role Cronbach alpha values were .75 for CGM; .66 for CGF.

Caregiver satisfaction was assessed by the three-item Kansas Parental Satisfaction Scale (James et al., 1985; “I felt satisfied with my grandchild’s behavior,” “I felt satisfied with myself as a grandparent,” I felt satisfied with my grandchild”); items were rated on a 4-point scale from 1 (never) to 4 (very often). Items were tailored to indicate the target custodial grandchild as the reference child, and the rating scale was changed from level of agreement to frequency of occurrence. Potential scores range from 3 to 12, with higher scores indicating greater care-giver satisfaction. Cronbach’s alphas were .75 for CGM; .78 for CGF. This scale was selected because (a) it was validated for use with married parents, and (b) measures of caregiving satisfaction derived from the stress process model and two-factor model contain items that overlap with other positive appraisals of being in the caregiver role (Kramer, 1997).

Negative appraisals.

Role captivity was measured by three items (“Wished you could run away,” “Felt trapped by caring for your grandchild,” Wished you were free to lead a life of your own”) developed by Pearlin et al. (1990). Participants rated each item on a 4-point scale from 1 (never) to 4 (very often). Potential scores range from 3 to 12, with higher score reflecting greater perceived role captivity. Cronbach’s alphas were .73 for CGM and .68 for CGF.

Loss of self was measured by two items (“Felt that you’ve lost a sense of who you are” and “Felt that you lost an important part of yourself”) developed by Pearlin et al. (1990), which participants rated on a scale from 1 (never) to 4 (very often). Higher scores reflecting more perceived loss of self-attributed to caregiving Cronbach’s alphas were .75 for CGM and .82 for CGF.

Data Analysis Plan

Structural equation modeling with Mplus 6.11 (Muthén & Muthén, 1998–2010) was performed to test the four competing models. The best fitting model for both genders was first identified through a series of chi-square difference tests. The best fitting model was then tested for measurement and structural invariance by CGP gender in the logically ordered and increasingly restrictive sequence specified by Bryne (2012). Overall fit with the observed data was assessed using the comparative fit index (CFI < .90), the standardized root mean square residual (SRMR < .08), and the root mean square error of approximation (RMSEA < .06). The sample size of 193 CGMs and 193 CGFs (386 in total) was sufficient for these analyses in view of Nunally’s (1967) specification that 10 cases are needed for each observed variable.

Related samples Wilcoxon signed rank tests were conducted for the mean comparisons of caregiving appraisals by gender, and bivariate correlations were performed using Spearman’s rho. Both procedures were completed with SPSS version 24 software. Effect sizes (ES) for statistically significant mean comparisons were obtained by dividing the standardized test statistic by the square root of the total observations. Summed scores were used rather than latent factor scores because the former are more likely to be computed in future research and practice.

Results

To examine the first aim, each baseline model was tested for CGMs and CGFs to examine if it represented a well-fitting model for both groups. Table 1 summarizes, separately by CGP gender, the comparison of the four competing models where it can be seen that the respective fit indices (CFI = .937–.944, RMSEA = .048, SRMR = .056–.065) were within the acceptable ranges across genders. For both genders, the fit indices associated with Model 1 (all first order constructs) were superior to those for both Models 2 (with a second order positive appraisal constructs) and 4 (with both negative and positive second order appraisal constructs) as revealed by the statistically significant Δ between these respective models. Although the Δχ² between the fit indices for Models 1 and 3 (with a negative appraisal second order construct) was not statistically significant (Δχ² = 1.65, ns) the fit indices associated with Model 1 were slightly better than those for Model 3, across both genders. Thus, Model 1 was deemed the best fitting model. It is noteworthy that Model 1 was revised to include error covariance between one item on the role captivity scale (i.e., “wished you could run away”) and one loss of self item (i.e., “felt you lost an important part of yourself”) as suggested by the Mplus 6.11 modification indices.

Table 1.

Summary of Model Comparisons by Custodial Grandparent Gender

Model	χ2	df	CFI	RMSEA	SRMR	Model comparison	Δχ2	Δdf	p
Grandmothers
1. With first-order factors onlya	196.03	136	.944	.048	.056
2. With positive appraisal second order factor	204.53	139	.939	.040	.061	1 vs. 2	8.36	3	<.05
3. With negative appraisal second order factor	198.28	139	.945	.057	.057	1 vs. 3	1.65	3	ns
4. With both negative and positive appraisal second-order factors	206.03	141	.940	.049	.062	1 vs. 4	9.95	5	<.05
Grandfathers
1. With first-order factors onlya	196.87	136	.937	.048	.065
2. With positive appraisal second-order factor	205.21	139	.932	.050	.071	1 vs. 2	13.28	3	<.01
3. With negative appraisal second-order factor	199.82	139	.937	.048	.066	1 vs. 3	2.84	3	ns
4. With both negative and positive appraisal second-order factors	207.89	141	.931	.050	.073	1 vs. 4	13.03	5	<.025

Note. CFI = comparative fit index; ns = not significant; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual.

^aModel 1 was revised to include error covariance between RC2 and LS2 as suggested by modification indices for both genders.

To examine our second aim, we conducted a series of sequential steps to investigate CGP gender invariance regarding the measurement and structural components of Model 1 (see Table 2). A hierarchical series of nested models were applied to Model 1, with each step building on and keeping constraints from the prior model. As a first step, the same parameters estimated in Model 1 separately by gender were estimated again within a multigroup representation of the baseline model with data from both genders. The fit of the resulting configural model (labeled as Invariance Model A) was good (CFI = .941, RMSEA = .048, SRMR = .061), thereby providing baseline values against which subsequent invariance models were compared.

Table 2.

Tests for Invariance of Model 1 by Custodial Grandparent Gender: Summary of Model Fit and χ² Difference Test Statistics

Model	χ2	df	CFI	RMSEA	SRMR	Model comparison	Δχ2	Δdf	p
Configural model
(Model A) No constraints	392.88	272	.941	.048	.061
Measurement parameters
(Model B) All factor loadings invariant	407.29	285	.940	.047	.068	A vs. B	15.75	13	ns
(Model C) All factor loadings invariant + error covariance between RC2 and LS2 invariant	405.80	286	.942	.047	.067	B vs. C	0.01	1	ns
Structural parameters
(Model D) All factor loadings invariant + error covariance between RC2 and LS2 invariant + all factor variances invariant	424.77	292	.935	.049	.087	C vs. D	15.75	6	<.025
(Model E) Model D with released constraint of factor variance for Loss of Self	413.30	291	.940	.047	.079	D vs. E	7.40	5	ns
(Model F) Model E + all factor covariances invariant	432.01	306	.938	.046	.092	E vs. F	19.11	15	ns

Note. CFI = comparative fit index; ns = not significant; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual.

To test for invariance by gender across all factor loadings, Invariance Model A was next compared with Invariance Model B in which all factor loadings were constrained to be equal for CGM and CGF. The corresponding chi-square difference test was not significant (Δχ² = 15.75, df = 13), suggesting that all factor loadings were similar by gender across all model constructs.

All standardized and unstandardized estimates for these factor loadings (available upon request) were statistically significant and of high magnitude. Measurement invariance was further tested by comparing Invariance Model B with Invariance Model C, with the latter constraining the error covariance between items RC2 and LC2 as equal by gender. The resulting chi-square difference test was not statistically different (Δχ² = 0.01, df = 1, ns), suggesting that error covariance between the two items was invariant by CGP gender.

Three steps were taken to test for structural invariance of Model 1 by gender. First, Invariance Model D, which constrained all factor invariances to be equal by gender, was compared with Invariance Model C. The resulting chi-square difference test was significant (Δχ² = 15.75, df = 6, p < .025), with modification indices suggesting the release of the constraint involving the factor variance for the loss of self. In turn, this constraint was released in Invariance Model E which did not differ significantly from Invariance Model D (Δχ² = 7.40, df = 5, ns). Although these results suggest that this particular error covariance was not equivalent for CGMs (β = .36) and CGFs (β = .15), the estimated values were statistically significant for both genders. To test for invariance of factor covariance, Invariance Model D was compared with Invariance Model E in which all factor covariances were constrained equal. The resulting chi-square difference test was nonsignificant (Δχ² = 19.11, df = 15), suggesting that all factor covariances were invariant by CGP gender. As summarized in Table 2, the preceding results show both the measurement and structural features of Model 1 to be highly similar for CGMs and CGFs, except for minor gender differences regarding the error covariance between one role captivity and one loss of self item.

Findings regarding the structural aspects of Model 1 are summarized in Table 3, which shows the estimated factor variances and factor covariances for all model latent constructs by CGP gender. Factor covariances are displayed according to positive and negative valance from the perspective of the two-factor model of care-giver well-being (Lawton et al., 1991). Inspection of the covariance estimates reveals that, for CGPs of both genders, all model constructs correlate positively with all constructs of the same valance while correlate negatively with all constructs of the opposite valence. As hypothesized, the magnitude of the covariance between same valance constructs is higher than that of the covariance between opposite valance constructs. Finally, only three of the 15 covariances between model constructs were not significantly significant. Each of these nonsignificant covariances involved the caregiving appraisal of gain with model constructs of the opposite valance (i.e., negative affect, role captivity, and loss of self).

Table 3.

Summary of Final Structural Results by Grandparent Gender

	Estimates for grandmothers			Estimates for grandfathers
	Standardized	Unstandardized	p	Standardized	Unstandardized	p
Factor variances
Positive affect		3.99	<.001		3.99	<.001
Negative affect		6.67	<.001		6.67	<.001
Gain		.37	<.001		.37	<.001
Satisfaction		.25	<.001		.25	<.001
Role captivity		.23	<.001		.23	<.001
Loss of selfa		.36	<.001		.22	<.001
Factor covariances
Opposite valances
Positive with negative affect	−.26	−1.33	.001	−.26	−1.13	.001
Negative affect with gain	−.06	−.10	ns	−.06	−.10	ns
Negative affect with satisfaction	−.23	−.30	.005	−.23	−.30	.005
Positive affect with role captivity	−.29	−.28	<.001	−.29	−.28	<.001
Positive affect with loss of self	−.23	−.28	<.001	−.30	−.28	.001
Gain with role captivity	−.10	−.03	ns	−.10	−.03	ns
Gain with loss of self	−.09	−.03	ns	−.12	−.03	ns
Satisfaction with role captivity	−.26	−.06	<.001	−.26	−.06	<.001
Satisfaction with loss of self	−.25	−.07	<.001	−.32	−.07	<.001
Same valances
Positive affect with gain	.48	.59	<.001	.48	.59	<.001
Positive affect with satisfaction	.33	.33	<.001	.33	.33	<.001
Negative affect with role captivity	.32	.40	<.001	.32	.40	<.001
Negative affect with loss of self	.37	.58	<.001	.48	.58	<.001
Gain with satisfaction	.31	.09	<.001	.31	.09	<.001
Role captivity with loss of self	.67	.20	<.001	.86	.20	<.001

Note. ns = not significant.

^aAll factor variances were invariant by gender except for loss of self.

Regarding Aim 3, the related-samples Wilcoxon signed rank test comparisons of each caregiving appraisal by gender are summarized in Table 4. There was no statistically significant difference by gender on perceived gain, although the observed mean was slightly higher for CGMs. Compared to CGFs, CGMs reported statistically significant higher perceived role captivity and loss of self (p < .001). CGFs reported higher caregiver satisfaction compared to CGMs (p < .05). The corresponding Effect sizes (ES) were in the small to moderate range.

Table 4.

Related-Samples Wilcoxon Signed Rank Test Mean Comparisons and Spearman’s Rho Bivariate Correlations for CGM and CGF Caregiving Appraisals

Appraisals	CGMM (SD)	CGFM (SD)	Test statistic	ES	Bivariate correlations	Types of CGM–CGF differences (n)
						+	−	Tie
Perceived gain	12.10 (2.94)	11.64 (2.81)	1.48	–	.06	93	72	28
Caregiver satisfaction	9.98 (1.93)	10.36 (1.81)	−2.47 *	−.13	31 **	53	83	57
Role captivity	5.27 (2.06)	4.15 (1.66)	6.39 ****	.33	.31 **	103	29	61
Loss of self	3.25 (1.48)	2.58 (1.07)	5.07 ****	.26	.08	91	35	67

Note. CGF = custodial grandfather; CGM = custodial grandmother; ES = effect size.

^*p < .05.

^**p < .01.

^***p < .005.

^****p < .001.

Also shown in Table 4 are the numbers of positive differences, negative differences, and ties on each caregiving appraisal when CGF scores were subtracted from CGM scores. In general, differences between each paired CGM-CGF dyad exceed similarities. In all cases, total differences between CGM and CGF scores far exceeded the number of tie scores. Except for caregiver satisfaction, CGMs had higher scores than did CGF on all other appraisals.

Bivariate correlations between CGM and CGF appraisals were nonsignificant for personal gain and loss of self. Correlations between CGM and CGF appraisals of caregiver satisfaction and role captivity were both statistically significant and moderate in magnitude (see Table 4).

Discussion

We examined how married CGPs from the same family appraise the impact of caring for a custodial grandchild on their sense of self as caregivers and how these appraisals were related to their psychological well-being from a conceptual framework combining elements of both the two-factor model (Lawton et al., 1991) and stress process model (Pearlin et al., 1990). In doing so, we applied measures of two positive (gain and satisfaction) and two negative (role captivity and loss of self) appraisals of caregiving that were originally developed for use with family caregivers of frail older adults (Lawton et al., 1991; Pearlin et al., 1990). Specifically, we compared four competing models via SEM with observed data from both CGMs and CGFs that varied regarding the extent to which their negative and positive caregiving appraisals were represented as either distinct first-order latent constructs or higher order constructs in which appraisals were fused.

The best fitting model for both CGM and CGF was one in which all four appraisals were represented as distinct first-order constructs, with no cross-loadings between the indicators for each type of appraisal apart from one statistically significant error covariance between one loss of self and one role captivity item. As predicted, the covariances between the four appraisal constructs were in the low to moderate range. A minor exception was the high covariance observed between the role captivity and loss of self-constructs.

Consistent with the two-factor model (Lawton et al., 1991), all four appraisals correlated more strongly with constructs of the same emotional valance than with those of the opposite valence. Most important, both negative appraisals covaried significantly and most strongly with negative affect while both positive appraisals correlated significantly and most strongly with positive affect. Thus, both the positive and negative dimensions of caregiving appear to contribute uniquely to CGM and CGF well-being as predicted by the two-factor model (Lawton et al., 1991).

This pattern of findings leads us to conclude that negative and positive appraisals of caregiving made by CGPs of both genders are more than just opposite ends of the same continuum and can co-occur with relatively high frequency (Gable & Reiss, 2010; Ong et al., 2006; Quinn et al., 2019). This distinction is critical given evidence from the stress and coping literature that positive emotions have important functions in the stress process and encompass coping processes that are unique from those that regulate distress. As noted previously, Folkman (2008) asserted that positive emotions and appraisals are a normative aspect of the stress process that facilitate coping by activating effective coping strategies, giving meaning and purpose to stressful situations, and fostering positive biological responses to stress.

In terms of significance for future research, our results suggest that the four appraisals of self in the caregiver role examined here can be used confidently in future studies on how CGM and CGF experience the role of providing full time care to a custodial grandchild. Our finding that the measurement aspects of these appraisals, which were originally developed with caregivers of older adults with dementia in mind (Lawton et al., 1991; Pearlin et al., 1990), held up with CGPs of both genders. This finding suggests the possibility that these four appraisals have universal meaning across diverse types of nonnormative caregiver populations (e.g., dementia, cancer, stroke, severe mental illness, or mental deficiencies). This underscores the assertion made within the American Psychological Association’s Caregiver Briefcase that common problems exist across all types of caregivers.

As pointed out by the researchers with the Rosalynn Carter Institute, “what has been overlooked in the past are caregiver similarities, universal psychosocial difficulties, and shared rewards associated with all types of caregiving” (Haigler, Bauer, & Travis, 2004, p. 96). A key goal of future research, then, is to identify which objective demands faced by different types of family caregivers are most influential in shaping their appraisal of self as a caregiver. For example, whereas feelings of role captivity among CGFs may be associated with having to give up valued leisure pursuits as suggested by Bullock (2005), CGMs may feel captive due to their direct care duties.

The need to recognize individual differences among family caregivers is further indicated by our findings regarding mean differences and bivariate correlations between CGMs and CGFs on all four appraisals. For example, the bivariate correlations between CGM and CGF scores were extremely low for perceived gain (r =.06) and loss of self (r = .08), while falling only within the moderate range for caregiver satisfaction (r = .31) and role captivity r = −.31). Thus, at best, less than 10% of the variance in any of these appraisals was attributable to the scores of the other member of the CGM-CGF couple, even though both members experienced the caregiver role within a shared family environment. This lack of CGM–CGF congruence was likewise revealed by the Wilcoxon matched-pair mean comparisons, which showed that the number of CGM–CGF positive and negative differences far exceeded the number of tie scores across all four appraisals.

Our hypothesis that CGMs would report higher mean scores on all four appraisals due to their likely greater involvement and investment in the caregiver role was only partially supported. Instead, an interesting pattern emerged whereby mean differences in positive appraisals were either not significant (perceived gain) or CGFs had higher scores (caregiver satisfaction) with a corresponding small ES. In contrast, CGMs showed significantly higher mean scores than CGFs with corresponding moderate ES on both negative appraisals. It may be that CGMs are more inclined than CGFs to make negative appraisals while both genders are equally prone to make positive appraisals. These findings clearly illustrate the need to differentiate between negative and positive appraisals and how this may vary by caregiver gender.

Implications for Practice

Our findings that measurable positive and negative appraisals of caregiving contribute to the well-being of CGMs and CGFs alike, implies that interventions aimed at improving their emotional well-being need to consider the simultaneous influence of both positive and negative dimensions of caregiving. As asserted in the coping literature, attending to appraisals is critical because they often precede emotional response and ultimately influence the particular coping strategies used in response to stress at both the individual (Folkman, 2008) and dyadic level (Falconier & Kuhn, 2019). Yet we are aware of no interventions for CGPs in which either positive or negative appraisals of self in the caregiver role have been the specific therapeutic targets (for review, see McLaughlin, Ryder, & Taylor, 2017). Indeed, the family caregiving intervention literature as a whole has primarily focused on stress reduction and far less so on strategies for enhancing a positive appraisal of the caregiving experience (Schulz, 2019), although both are required to optimize the psychological well-being of family caregivers.

It is important for helping professionals to validate both the positive and negative emotions expressed by CGMs and CGFs, explore the sources of these emotions, and help grandparents to generate and sustain future positive emotions (Folkman, 2008). In this regard, not only is it necessary to help them meet the objective demands of their role, but it is also essential to help them modify their interpretations of these caregiving demands and of their abilities to respond to them (Rapp & Chao, 2000). This approach is especially critical when the objective demands of caregiving are not modifiable (Quinn et al., 2019). For example, even if CGPs are caring for a grandchild with a severe developmental disability, they could be encouraged to reflect on how they are contributing to that child’s well-being.

Although reducing negative appraisals is typically addressed by means of cognitive behavioral therapy approaches, which target the alleviation of unhelpful thoughts (Gallagher-Thompson et al., 2002), strategies for enhancing positive appraisals associated with caregiving are less evident. Benefit-finding is one approach mentioned in the general stress and coping literature (Folkman, 2008) that might be used to help facilitate positive appraisals among CGPs. This could be done by encouraging them to explore possible benefits within the caregiver role (e.g., serving as a role model for their grandchild) that they have overlooked (Folkman, 2008). A similar approach is benefit-reminding in which, for example, CGPs might be asked to keep a log of each time they have recognized a benefit associated with caring for their grandchild (Folkman, 2008). Both benefit-finding and benefit-reminding are directed toward helping caregivers derive positive meaning from their overall caregiving experience.

Fostering positive appraisals like gain and satisfaction with CGM–CGF couples may be particularly important in light of social psychological findings that (a) the frequency of positive events typically outweighs that of negative events; (b) positive events are linked to positive emotions; (c) positive emotions must outnumber negative emotions for optimal well-being to occur; and (d) one important ways that people react to positive events is to share them with close others, which enhances their overall impact on well-being (Gable & Reiss, 2010). The activity of sharing positive events with a close other is known as capitalization, and for this to occur successfully, the partner’s reaction must be perceived as responsive. A partner’s enthusiastic response not only fosters more interactions where positive emotions are shared but also signals the partner’s understanding and appreciation of the event’s good news for the discloser (Gable & Reiss, 2010).

The literature on dyadic coping (Falconier & Kuhn, 2019) suggests that practitioners can foster enthusiastic partner responses to positive events by encouraging partners to ask questions about the event, seek additional details about the event, elaborate on the possible benefits of the event for the discloser, and suggest why the event is meaningful to the discloser. For example, a CGF who is told by his wife that she was praised by their grandchild’s teacher might say the following to her: “You always do such a great job helping Susan with her homework, and I know how that makes you feel closer to her.” According to Gable and Reiss (2010), active and constructive responses of this nature are likely to result in increased memory for the positive event, greater importance attributed to the event, and perceptions that the partner understands and cares for the discloser; they may also ultimately lead to an upward spiral of positive emotions.

The present finding of measurement equivalence between CGM and CGF on the four appraisals we examined also has applied significance within the context of dyadic coping, where communication of the experience of stress between partners and positive empathic responding to each other’s perceived stress are critical components (Falconier & Kuhn, 2019). Specifically, by knowing that appraisals such as personal gain, satisfaction, loss of self, and role captivity have similar meanings for CGM and CGF, helping professionals can use them as starting points for helping married CGPs to begin communicating and responding empathically to each other’s appraisals. Over time, practitioners may then move on to encouraging disclosures that may be uniquely meaningful to only one member of the CGM-CGF couple.

Our findings regarding mean differences between married CGM and CGFs and the low magnitude of bivariate correlations across their respective caregiving appraisals also have practice implications. In particular, when married grandparents have differing appraisals, they might be encouraged by practitioners to first validate each other’s feelings and then work as a team to help each other reduce their negative appraisals while increasing positive appraisals. Clinicians should also determine whether there are different typologies among CGMs-CGFs in terms of how both members appraise the care-giver role. Families where both the CGM and CGF have high negative appraisals may need different intervention approaches than those in which only one member has highly negative appraisals (Robertson, Zarit, Duncan, Rovine, & Femia, 2007).

Limitations and Future Directions

One major shortcoming of the present study was that we are unable make firm conclusions regarding the causal direction between the appraisal of caregiving made by CGM and CGF and their psychological well-being given our cross-sectional data. Nevertheless, the covariance models that we tested were well supported in the coping literature, where appraisals are universally regarded as antecedents of emotional well-being (Folkman, 2008; Lawton et al., 1991). However, we were unable to investigate the further contention in the coping literature that the impact of appraisals on emotional well-being is mediated by the types of coping behaviors precipitated by appraisals. Studying the relationship between custodial grandparents’ appraisals and their coping behaviors in the caregiver role is an important direction for future research.

Another limitation was that we examined only a small number of appraisals. Indeed, there are other known appraisals of self in the caregiver role that can be either positive (e.g., caregiving mastery, sense of commitment, improved relationship with the care recipient) or negative (social isolation, adverse change in health, financial strain) in valence that remain to be rigorously examined with custodial grandparents. In addition, there were many measures of positive and negative affect beyond those used in the present study, inclusion of which might have yielded different findings (Yamasaki & Uchida, 2016). The generalizability of our findings may be restricted by some characteristics of the sample. One concern was that initial recruitment was of CGM; only those CGF who agreed to participate when asked by CGM were in the study. This may explain why scores for both negative and positive appraisals were skewed in positive directions. In addition, our sample of only 193 CGP couples was not large enough to look at potential differences associated with race or ethnicity.

Conclusion

Despite its limitations, this study is the first to examine how married CGPs appraise their mutual involvement within the caregiver role. Distinct positive and negative appraisals were found to be equivalent by gender in terms of both their measurement properties and how they covaried with general indices of positive and negative emotional well-being. Because appraisals are a core component of coping with stressors at both the individual and couple level, our findings have important practice implications and point to meaningful directions for future research. From a policy perspective, the present findings suggest that efforts such as the Advisory Council enacted recently by Congress under the Supporting Grandparents Raising Grandchildren Act (Govtrac.US, 2019) should address the concerns of CGFs and CGMs alike.