Maternal caregiving strain as a mediator in the relationship between child and mother mental health problems.
Key words: burden; caregiving; children; mental health; women
The strain on families dealing with an emotionally troubled child has received greater attention in recent years. In contrast to earlier psychiatric research that focused on the etiology of the child's problems, including parental pathology, more recent studies have examined the impact of child pathology on parental well-being. These studies delineated the array of difficulties experienced by parents of mentally ill children, especially by the mother (Marsh, 1996; Schene, Tessler, & Gamache, 1996). An overdue counterpoint to the more traditional psychiatric perspective, this revisionist view has emphasized the importance of understanding parenting stress related to children's problems and disabilities, has documented some of the difficulties experienced by parents, and has acknowledged the strong bidirectional influences in the parent-child dyad (Hammen, Burge, & Stansbury, 1990; McDonald, Poertner, & Pierpont, 1999).
Although this literature has increased in scope and sophistication, it has remained largely isolated from the broader body of research and theory relating to the burdens of caregivers (McDonald et al., 1999). The present study applies the conceptual and research lens of family caregiving to mothers caring for children with emotional problems. We examined a model of caregiving strain derived from Biegel, Sales, and Schulz (1991), which suggests that a mother's caregiving strain may mediate or serve as a pathway between a child's behavioral problems and a mother's mental health.
FAMILY CAREGIVING BURDEN
The study of caregiving burdens for family members has focused almost exclusively on adult's physical and mental illness, especially on people with Alzheimer's disease (George & Gwyther, 1986; Noelker & Townsend, 1987; Poulshock & Deimling, 1984; Zarit, Reever, & Bach-Peterson, 1980) and adults with mental illness, mainly schizophrenia (Cook & Pickett, 1987/1988; Lefley, 1996; Montgomery, Gonyea, & Hooyman, 1985; Schene et al., 1996; Thompson & Doll, 1982). A strong and consistent relationship between illness severity and caregiving strains or burden has been shown across these caregiving areas (Baronet, 1999; Biegel et al., 1991). In recent years this literature has been reframed within a general life stress conceptual model, which views the stresses of caring for an ill family member as affected by cognitive appraisals of the difficulties of the demands, taking into account the resources available (Lazarus & Folkman, 1984; Monat & Lazarus, 1991).
Family caregiving burden or strain has been viewed typically as encompassing the array of challenges experienced by families as a consequence of someone's illness. Researchers more or less agree that two central dimensions, objective and subjective burdens, define the family burden concept (Biegel et al., 1991; Lefley, 1996; Maurin & Boyd, 1990; Montgomery et al., 1985; Schene et al., 1996). Objective burden stems from the heightened caregiving tasks because of the illness. Subjective burden is the emotional cost or distress experienced by the caregiver in caring for an ill family member.
CAREGIVING STRAINS FOR PARENTS OF CHILDREN WITH EMOTIONAL PROBLEMS
Although the impact of caregiving for parents of adult children with serious mental illness, especially schizophrenia, has been studied extensively, there is limited literature on the caregiving strains of parents of minor-age children with emotional problems (Goldberg-Arnold, Fristad, & Gavazzi, 1999; Marsh, 1996). Perhaps because previous eras of research had focused on parental precipitants of children's psychiatric problems (Marsh; Schene et al., 1996), an extensive body of research documents the strong relationship between mothers' and childrens' mental health problems, with mothers' psychiatric problems often viewed as a risk factor for the child (Hammen, Gordon, & Burge, 1987; Hammen et al., 1990; Weissman, Leckman, Merikangas, Gammon, & Prusoff, 1984). However, it required a paradigm shift to recognize that parents, especially mothers, often experienced severe stresses as they attempt to cope with their psychiatrically troubled children (Floyd & Gallagher, 1997; Turnbull, Brotherson, & Summers, 1986). Yet, McDonald and colleagues' (1996) study of 259 family caregivers, predominantly mothers, of children with severe emotional or behavioral disabilities revealed that 62% viewed their child's behavior as the most stressful event they had experienced in the past year.
A number of researchers now have found associations between the child's behavioral problems, caregiving strains, and maternal distress and depression (Brannan & Heflinger, 2001; Early & Poertner, 1993; Goldberg-Arnold et al., 1999; McDonald et al., 1996; Suarez & Baker, 1997). Early and Poertner's (1993) review of 44 studies of families with children identified as having serious mental health problems concluded that the child's illness affected a variety of family relationships and interactions, parental distress, as well as the family's linkages to the larger ecological environment. Angold and colleagues (1998) found that both parents' and children's reports of symptoms, which closely corresponded to each other, correlated with parents' experiences of burden. More of this stress appears to be subjective rather than objective (Brannan, Heflinger, & Bickman, 1997), and parents with higher burden were more likely to experience psychological symptoms (Brannan et al.; Brannan & Heflinger, 2001). McDonald and colleagues' (1999) findings showed that although internalizing and externalizing child behaviors predicted significantly to life stress, internalizing child behaviors were a much stronger predictor once other factors were controlled.
Despite much progress in recent years, many studies of caregiving burden are still limited by problems of restricted sample sizes, special child populations (most commonly developmentally delayed children), and perhaps most important, the absence of a clear theoretical model. The recent work of McDonald and colleagues (1999) and Brannan and Heflinger (2001) has acknowledged and responded to the dearth of theoretically derived research in this area.
GOALS OF PRESENT STUDY
We examined a mediational model presented in Biegel and colleagues (1991), who synthesized findings from empirical research on the caregiving strains of families dealing with a variety of chronic illnesses of adult family members, and embedded these findings in a general theoretical model that delineates the network of predictors and consequences of family caregiving strains in chronic illness. This model was derived by Schulz and colleagues (1988) from family stress models (Hill, 1958; McCubbin & Patterson, 1983) and Lazarus and Folkman's (1984) general stress model. Schene and colleagues (1996) viewed Biegel and colleagues' model as one of the few incorporating both the predictors and longer range health outcomes of family caregiving burden. In this model caregiving strain is viewed as a mediating variable affected by the severity of the person's illness demands and as affecting the mental and physical health of the caregiver (Biegel et al.; McDonald et al., 1999; Song, Biegel, & Milligan, 1997). Thus, the strain experienced by the caregiver is viewed as an intervening variable between the patient's level of impairment and the caregiver's longer-term psychosocial outcomes.
As far as we are aware, this model had not been imported into the child and family arena. Although the studies reviewed earlier examined components relevant to this model as they may apply to parents of emotionally troubled children, very few viewed caregiver strain as a possible pathway between child and mother symptoms (Heller, Baker, Henker, & Hinshaw, 1996; Suarez & Baker, 1997), and these have not linked their work to the broader caregiving literature. Most, including recent model-testing studies of McDonald and colleagues (1999) and Brannan and Heflinger (2001), viewed caregiver burden as an outcome variable rather than as a mediator affecting more enduring mental health consequences such as depression (Biegel et al., 1991; Schulz & Williamson, 1991; Song et al., 1997).
Brannan and Heflinger's (2001) findings are, however, germane to our study in that they found distinct variables predicting to caregiver strain and psychological distress, although confirming the interrelationship between the two variables. In fact, they suggest that a mediational model viewing caregiving strain as a pathway between child symptomatology and caregiver mental health symptomatology may be the best way to interpret the data patterns revealed in their analyses.
We examined just such a mediational model, derived from Biegel and colleagues' (1991) general stressor-strain-outcome model, in which the relationship between the child's problem severity and maternal mental health outcomes is mediated by the amount of caregiving strain the mother experiences. We expected that mothers whose children manifest more behavioral problems would experience greater strain in their caregiving role, which would contribute to their having more mental health symptoms. Because McDonald and colleagues (1999) found consequential differences in the correlates of internalizing and externalizing child behavior problems, we also examined the differential effect of internalizing and externalizing child behavioral symptoms on maternal mental health.
Participants and Procedures
All mothers who brought their children to a rural mental health base service unit for psychiatric assessment between June 1998 and December 2000 were screened by clinic staff for study eligibility. Children had to be between ages six and 17, nonpsychotic, not mentally retarded, not viewed as being at risk of child abuse, and not needing hospitalization. Mothers were included if they were the primary custodial parent, lived with the child, and were not viewed as needing immediate psychiatric intervention. The 475 mothers who met eligibility criteria were informed of the research by their clinician and asked to give permission for subsequent contact. A project coordinator attempted to arrange a time for interviews for the 431 mothers who agreed to further contact (91% of those eligible). Despite repeated efforts, no data were obtained from 154 (36%) of these women, mostly because they could not be contacted (n = 112), with a few (n = 37) refusing to participate further when contacted. Of the 277 scheduled interviews, 222 (80%) were completed. The remaining 57 mothers either did not show up (n = 37) or refused when the interviewer came to their homes (n = 20).
On initial contact with the mothers, usually in their homes, a trained research clinician explained the project and obtained written consent from the mothers and children. A battery of clinical interviews and questionnaires, taking up to three hours, were administered to the mothers, who received $50 checks for their participation in baseline data collection. Clinical data from the children were obtained simultaneously by another trained research clinician. Children were given $20 gift vouchers from a local mall.
Mothers were, on average, 36.2 years old (SD = 6.2), mainly white (82%, n = 182), with 12.1 years of education (SD = 2.1). Among the 62% (n = 137) who were not married, 37% (n = 51) lived with a partner. Most had either two (36%, n = 80) or three (31%, n = 69) children, for an average of 2.7 (SD = 1.22) children. The fifty-seven percent who were employed averaged 33.2 (SD = 10.46) hours of work per week. Half the women had an annual family income of $16,000 or less, with 31% (n = 69) receiving between $16,000 and $30,000, and the remaining 19% (n = 42) having incomes over $30,000.
The 130 boys and 92 girls of these mothers had a mean age of 11.8 (SD = 3.4) years, with boys (M age = 11.3, SD = 3.25)being significantly younger than girls (M age = 12.6, SD = 3.39)(t = -2.81, df = 220, p = .006).
The limited data we were able to obtain on 140 nonparticipating, though eligible, children indicated that there were no significant differences by gender or race between participants and nonparticipants. However, nonparticipating children were significantly older (M age = 12.81, SD = 3.50) than participants (M age = 11.80, SD = 3.36) (t = -2.79, df = 376, p = .006).
Caregiver Strain Questionnaire (CGSQ). This 21-item scale developed by Brannan and colleagues (1997) for the Fort Bragg evaluation of children's mental health services (Bickman et al., 1995) measured objective and subjective parental burdens experienced in caring for their emotionally distressed children over the preceding six months. Originally called the Burden of Care Questionnaire, it consists of 11 items measuring objective strain and 10 items measuring subjective strain, each answered on a five-point Likert scale, with higher scores indicating more strain. Reliability is high ([alpha]=.93) (Brannan et al.). In our study the overall alpha was .92, with .91 for objective strain and .83 for subjective strain.
Mothers in our study had a significantly higher level of caregiving strain (M = 2.64, SD = .80) than those in the Fort Bragg study (M = 2.48, SD = .82) (t = 3.02, df = 218, p = .003). These differences seem primarily attributable to the significantly higher objective strain scores for women in our sample (M = 2.31, SD = .94) compared with Fort Bragg mothers (M = 2.02, SD = 2.97) (t = 4.60, df = 218, p = .000). In our group, as in Fort Bragg's, subjective strains were higher (M = 3.01, SD = .81) than objective strains.
Child Behavior Checklist (CBCL). This 113-item checklist version of Achenbach and Edelbrock's (1983) measure assessing emotional and behavioral problems of children ages four through 16 asks parents to rate the extent of each problem in the past six months on a scale from 0 to 2, with higher scores indicating greater problem severity. In addition to computing an overall score for total items, we examined scores for the 31 items composing internalizing behaviors (for example, depressed, anxious) and the 33 items tapping externalizing behaviors (for example, aggressive, delinquent). Reliabilities are generally over .90 (Achenbach, 1991), with evidence of convergent validity with other scales and an ability to discriminate between clinical and nonclinical groups. Alpha in this study was .94 for the total scale, .84 for the 31 items on the internalizing subscale, and .91 for the 33 items on the externalizing subscale.
The children in this study had a mean total CBCL score of 61.73 (SD= 27.0) (M = 65.46 for boys, M for girls = 56.31), which is significantly higher than Achenbach's (1991) normative (24.3 for boys, 23.1 for girls) and clinic-referred samples (54.5 for boys, 52.1 for girls). The externalizing CBCL score mean in this study was 22.96 (SD = 11.7), and the internalizing CBCL mean was 16.8 (SD = 10.25). Boys were significantly higher than girls on total CBCL scores (t= 2.41, df = 214, p = .017) and on externalizing scores (boys = 25.21, girls = 19.74) (t = 3.45, df = 214, p = .001).
Beck Depression Inventory (BDI). This 21 -item scale asks respondents to rate the intensity of each depressive symptom experienced in recent weeks on a 0 to 3 Likert response scale, with higher scores indicating more symptoms (Beck, Ward, Mendelson, Mock, & Erbaugh, 1961). It has been shown to have good reliability and convergence with other measures of depression (Beck & Steer, 1993). In this study coefficient alpha was .92 and the score mean was 15.08 (SD = 10.87).
Beck Anxiety Inventory (BAI). This scale (Beck, Epstein, Brown, & Steer, 1988), measuring anxiety symptoms, also has 21 Likert-format items, and had a coefficient alpha of .93 and a score mean of 13.18 (SD = 11.62) in this study.
SF-36. This 36-item structured self-report scale measures various dimensions of physical and mental health (MH) functioning experienced in the preceding four weeks, with higher scores indicating better health (Ware & Sherbourne, 1992). Reliability ranges between .81 and .88 (McHorney, Ware, & Raczek, 1993). In our study only the five-item MH transformed score, measured on six-point Likert scale, was examined (M = 55.2, SD = 22.7).
Establishing the Correlational Preconditions for Mediation
Our study included three maternal mental health measures, the BDI, BAI, and SF-36 Mental Health scales, which measure various facets of psychological distress. As we would expect, these three indicators of a mother's mental health status all intercorrelated highly (see Table 1), with BDI and SF-36MH scores sharing the greatest variance (r = -.78, p < .001).
The child's CBCL scores were associated significantly with all maternal health indicators. Mothers of more emotionally troubled children experienced more caregiving strain (r = .47, p < .001) (Table 1). When the internalizing and externalizing dimensions of CBCL were isolated, externalizing CBCL scores were much more closely associated with caregiving strain (r = .52, p < .001) than were internalizing scores (r = .28, p < .001). Finally, there were significant relationships between caregiving strain and each maternal health indicator. This relationship was strongest for the BDI (r = .49, p < .001).
Inasmuch as differences in the correlates of objective, subjective, and total caregiving strain were minimal in these analyses, and to simplify the subsequent presentation, only the measure of total caregiving strain is reported in the remainder of this article.
Test of the Mediational Model
To test the mediational model, we used a path analytic strategy of regressing the mediator-caregiving strain--on the independent variable of child behavioral problems, and regressing each of the three dependent variables (BDI, BAI, SF-36MH) on child behavioral problems (independent variable) and caregiving strain (mediator). We conducted Sobel (1982) tests to formally examine the significance of the indirect effect of caregiving strain on the relationship between child behavioral problems and maternal mental health (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002). In each analysis, seven demographic variables--child gender, child age, mother's age, mother's years of education, race, number of hours employed, and income--were entered as controls. Inasmuch as scores on the BDI and BAI were very skewed, a square root transformation, which successfully normalized the scores for these two variables, was used in all multivariate analyses.
Table 2 presents the results of the regression analysis examining child behavioral problems predicting to the mediating variable of caregiving strain, controlling for seven demographic variables. This regression explained 28% of the variance in caregiving strain. By far the strongest predictor of caregiving strain was the perceived severity of the child's problems ([beta] = .45, p < .001).
In each analysis the mediator, caregiving strain, was by far the most significant predictor to the mental health measure. CBCL scores contributed little to the prediction, in contrast to the significant bivariate correlations between CBCL and each maternal mental health outcome variable (see Table 3). Caregiving strain contributed somewhat more to SF36 MH ([beta] = -.52, p < .001) and BDI ([beta] = .50, p < .001), than to BAI ([beta] = .36, p < .001). Sobel (1982) tests of whether caregiver strain mediated between child behavioral problems and each maternal health symptom variable allowed us to compute a test statistic of the indirect effect of the independent variable on the dependent variable by way of the mediator. Sobel test statistics revealed significant p values for the indirect effect of caregiving strain as a me diator between CBCL scores and each mental health indicator (BDI: z = 4.84, p < .001; BAI: z = 3.60, p < .001; SF-36MH: z = -4.68, p < .001). Figure 1 summarizes the direct and mediating patterns found for the central study variables, with demographic controls. These data support our mediational model, with maternal caregiving strain appearing to mediate the relationships of total CBCL scores with each maternal mental health indicator.
[FIGURE 1 OMITTED]
In addition, the regression analysis of maternal caregiving strain (Table 2) revealed only one demographic variable to contribute significantly. The number of hours that women were employed was negatively related to their caregiving strain ([beta] = -.23, p < .01), with women employed for more hours having significantly less caregiving strain than those employed fewer hours. When examining the demographic variables predicting to maternal mental health symptoms (Table 3), older mothers had worse mental health scores across all three dependent variables, and mothers of younger children had higher BDI scores.
Subanalyses of Internalizing and Externalizing CBCL Scores
The analyses thus far strongly suggest that caregiving strain serves as a pathway between the child's overall symptom severity and maternal mental health problems. Next, we explore the contributions of the 31 item internalizing behavior and the 33-item externalizing behavior subscales, because some studies that examined just these subscales have suggested that these problem clusters may have differential effects for caregivers (McDonald et al., 1999; Suarez & Baker, 1997). We followed the analytic strategy of McDonald and colleagues by entering both internalizing and externalizing CBCL scores, rather than the total 113-item CBCL scale, as independent variables in each regression. These two CBCL subscales shared only a modest amount of variance in our data (r= .405, p < .001).
These analyses revealed more complex patterns of contrasting contributions for internalizing and externalizing child problem dimensions. Externalizing CBCL scores, but not internalizing CBCL scores, was a significant contributor to caregiving strain (Table 4). In contrast to total CBCL scores, which did not contribute to maternal mental health indicators once caregiving strain entered, some relationships between internalizing and externalizing subscales and the maternal mental health &pendent variables remained significant, although in somewhat complex patterns. Higher child internalizing problems predicted significantly to higher BDI and BAI scores, but fell short of significance in its contribution to SF36-MH scores (Table 5). Sobel tests revealed no evidence of maternal caregiving strain mediating between internalizing CBCL and any maternal mental health outcome indicator.
Severity of a child's externalizing problems predicted significantly to both BDI and SF-36MH, but not to BAI scores, with caregiving strain entered; however, these relationships were the reverse of what was anticipated, with externalizing CBCL scores relating significantly to lower BDI scores and higher SF-36MH scores. Thus, in contrast to internalizing child behavior, externalizing behavior predicted more positive maternal health once caregiving strain was controlled. The Sobel tests for caregiving strain as a mediator between externalizing CBCL and both BDI and SF36-MH fell just short of significance (BDI: z = -1.94, p = .053; SF-36 MH: z = 1.87, p = .06), whereas caregiving strain did not appear to mediate between externalizing CBCL scores and BAI.
Once again, the demographic variables contributed little to either maternal caregiving strain or mental health symptoms. As before, the only demographic variable predicting to caregiving strain was employment, with mothers who worked more hours having lower strain. Employment also contributed significantly to a mother's BAI scores (but not BDI or SF-36 MH scores), with those working more hours having lower anxiety. Mothers of younger children once again had significantly higher scores on the BDI, but neither of the other two mental health indicators. In addition, lower income mothers had higher BDI scores.
The data analyses presented here relating to overall CBCL scores support the illness stressor caregiver strain-health outcome mediational model being tested. Overall, these analyses confirm the prediction that maternal caregiving strain may serve as a pathway between the child's total emotional problems and a mother's mental health. Although CBCL score related directly to each maternal mental health indicator before caregiving strain was considered, maternal caregiving strain became the dominant predictor to a mother's mental health status in our regression analyses whereas total CBCL scores no longer contributed significantly. Furthermore, Sobel tests for mediation supported the intervening variable status of caregiving strain. Thus, our data suggest a possible intervening mechanism for the findings of studies indicating a main effect of severity of her child's behavioral problems on a mother's mental health (Angold et al., 1998; Greenberg, Seltzer, Krauss, & Kim, 1997; McDonald ct al., 1999).
These analyses support the more general family caregiving model of chronic illness presented by Biegel and colleagues (1991) and suggest that this model may be applicable as well to families of minor-age children. Our analyses also provide empirical confirmation for Brannan and Heflinger's (2001) speculation that caregiving strain may mediate the relationship between child and maternal distress. This mediation role suggests that the severity of the child's behavioral problems affects the mother's mental health through the caregiving demands that she experiences. Without these strains of caregiving eroding the mother's health, the impact of a child's problem may not be as consequential.
This consistency with the more general family caregiving literature suggests that at least some patterns found in the adult caregiving literature also may apply to minor-age children and their families. This extension cannot be taken for granted, because it has long been assumed that the issues for parental caregivers of young children are different because caregiving is a normal role expectation for them and because it has been difficult to disentangle normal caregiving strains from those engendered by a child's emotional difficulties. In fact, many researchers studying family caregiving, including Biegel and colleagues (1991), explicitly excluded studies of parental caregiving strains associated with children's illnesses because they believed that different processes might be involved.
The data also suggest that although overall severity of a child's symptoms may be mediated by caregiver strain, the CBCL subscale behavioral clusters of internalizing and externalizing child behavior problems may have divergent patterns relating to both caregiving strain and maternal mental health. We found that a child's internalizing behavioral problems were more directly and more closely related to maternal mental health problems than were externalizing behaviors. Our findings may parallel those of McDonald and colleagues (1999), whose measure of caregiver stress was more akin to general psychological distress than to the specific stress of caregiving. This shared variance between mothers' mental health symptoms and children's internalizing disorders suggests that these psychiatric problems may be directly linked, either genetically or through child-mother interaction patterns.
In contrast, the direct relationship between a child's externalizing behavior and a mother's mental health problems was more modest, yet, mothers experienced much greater difficulties in caring for children manifesting externalizing behavioral problems. Perhaps not surprising, mothers of children who manifested more aggressive, acting out behaviors found it harder to deal with their children's objective behavioral demands and expressed more negative emotional reactions. Such behaviors are more likely to get the attention of others and may result in earlier treatment referrals, as was found in this study. Also, although counter-intuitive, our data suggest that once a child's internalizing behavior and maternal caregiving strains are controlled, a child's externalizing behavior problem severity" may actually predict to fewer maternal mental health symptoms.
Consistent with Brannan and colleagues (1997), we found that the subjective strains of caring for an emotionally troubled child were higher than the objective strains. However, unlike some studies, we found few differences between either the predictors or maternal health consequences of objective and subjective strains (Brannan & Heflinger, 2001; Montgomery et al., 1985; Thompson & Doll, 1982).
STUDY LIMITATIONS AND STRENGTHS
The analyses and interpretations presented in this article may be seen as competing with the more traditional psychiatric model that views the child's emotional problems as the &pendent variable and the mother's psychiatric status as a key independent variable. We share the view of others (Hammen et al., 1990; McDonald et al., 1999) that these patterns are likely to be bidirectional. However, even with this more traditional model, caregiving strain may function as a pathway between the mother's distress and the child's behavioral problems. A mother with psychiatric problems may experience greater distress and heightened concern in her parenting role, and her emotional reactions may in turn exacerbate her child's symptomatology. Although this model appears less plausible to us, as well as less theory-driven, it is an alternative way of interpreting the findings of our study.
The maternal self-report measures used in our study may, in addition to the shared method variance, share variance as a result of conceptual or measurement overlaps in the instruments used. Schene and colleagues (1996) suggested that measures of caregiving strain max, be capturing some aspects of depression. Thus, depressed mothers may view the care of their children as more demanding or worrisome than other mothers or may exaggerate the severity or frequency of their child's symptoms. The use of multiple maternal mental health indicators in our study may provide somewhat greater assurance that measurement artifacts are not fully determining the relationships obtained.
One strength of our study lies in the broad sample-inclusion criteria used. The children in this study may represent a more typical cross-section of children seen in mental health settings, rather than more severe emotional problems, as in McDonald and colleagues' (1999) study, or the very different challenges to parents presented by retardation (Hadadian, 1994) or autism (Gill & Harris, 1991), which have been studied more commonly. Our large sample size also allowed more statistical power for hypothesis testing and a better ability to examine subgroup patterns. However, although our screening and recruitment procedures were rigorous, we were only able to locate and obtain the cooperation of slightly fewer than half of those eligible for inclusion. Furthermore, our limited data on nonparticipants, although showing gender and race comparability with our obtained sample, revealed that the children in our sample were significantly younger than those who, although eligible, did not participate. These factors suggest that generalizability of our findings is uncertain.
Perhaps the most important implication to be drawn from these analyses is that a mother's health status may be more closely linked to the magnitude of caregiving strain she experiences than to the severity of the child's emotional problems. Thus, the mental health of mothers who have children with psychiatric disorders may be improved substantially if their felt strain of caregiving is reduced. The mediating effect of caregiving strain may be especially important for mothers of children with externalizing disorders, inasmuch as these children are objectively more difficult to manage. These strains may be more pronounced for older mothers, especially in their child's early years.
How might we better serve families of children with emotional problems? At present, mental health services for children focus on diagnosing and treating the child's problems, involving parents primarily as they may facilitate their child's treatment. Such interventions may, over time, reduce the severity of the child's problem and ultimately alleviate parental strain. Yet, it is clear from this study that stress levels of many mothers are high and may have detrimental effects on their mental health. These mothers could benefit from services that address their caregiving concerns so that they may care for their disturbed children without experiencing as much emotional strain.
The traditional stress model suggests that cognitive restructuring or reframing of the mother's perceptions of her child's behavior, as provided in cognitive counseling approaches, may help reduce her subjective strain by normalizing her situation. Training mothers to better deal with their child's externalizing problem behaviors, which were the strongest source of caregiving strain, might also be effective. In addition, building on the finding from this study regarding greater work involvement predicting lower caregiving strain and better mental health, a worker might help a mother consider whether employment outside the home might provide a respite that could ameliorate caregiving strain. Other respite options, such as educational or training classes or child care provision also may be beneficial in allowing the mother time away from her caregiving role.
To further clarify the relationships between children's mental health problems and caregiving stress, McDonald and colleagues (1999) expressed a need for more studies based on large representative samples of children with emotional disorders. To our knowledge, this study represents one of the few investigations of the relationship between children's problem severity and maternal caregiving strain in a large nonspecialized child mental health outpatient population. The data patterns are compatible with the general stressor-caregiver strain-health outcome model, suggesting that this model may be appropriate for caregivers of minor-age children as well as for adult problem populations. Our findings support a mediational model that identifies a mother's caregiving strain as a pathway between her child's mental health problems and her own distress. It also directs attention of researchers and clinicians working with families with emotionally disturbed children to the caregiving strains of mothers as an important arena for intervention in the treatment process.
TABLE 1--Correlations among Variables in Study of Mental Caregiving Strain of Mothers of Children with Mental Health Problems (N = 204) CBCL CGSQ Variable External Internal Total Obj. Subj. Total CBCL .81 *** .80 *** .47 *** .47 *** .44 *** Externalizing .40 *** .52 *** .50 *** .47 *** Internalizing .28 *** .26 *** .27 *** Total CGSQ .93 *** .89 *** Objective .67 *** Subjective BDI BAI Variable BDI BAI SF36-MH Total CBCL .28 *** .29 *** -.22 *** Externalizing .19 ** .19 ** -.18 ** Internalizing .28 *** .29 *** -.22 *** Total CGSQ .49 *** .41 *** -.45 *** Objective .42 *** .38 *** -.38 *** Subjective .48 *** .37 *** -.455 *** BDI .70 *** -.78 *** BAI -.71 *** NOTES: CBCL = Child Behavior Checklist (Achenbach & Edelbrock, 1983). CGSQ = Caregiver Strain Questionnaire (Brannan, Heflinger, & Bickman, 1997). BDI = Beck Depression Inventory (Beck, Ward, Mendelson, Mock, & Erbaugh, 1961). BAI = Beck Anxiety Inventory (Beck, Epstein, Brown, & Steer, 1988)). SF-36-MH = SF-36 Mental Health subscale (Ware & Sherbourne, 1992). ** p < .01. *** p < .001. TABLE 2--Regression Analysis for Demographic Variables and CBCL Scores Predicting Caregiving Strain among Mothers of Children with Mental Health Problems (N = 167) Variable [beta] SE [beta] t Demographic variable Child's gender -.17 .12 -.10 -1.40 Child's age .02 .02 .07 .87 Mother's age -.01 .01 -.63 -.75 Race .10 .17 .04 .56 Years of school .00 .03 .00 -.035 Hours employed -.01 .00 -.23 -3.17 ** Income .02 .125 .01 .14 Total CBCL .01 .00 .45 6.43 *** R = .53, [R.sup.2] = .28, df = 8/159, F = 7.87 *** NOTE: CBCL = Child Behavior Checklist (Achenbach & Edelbrock, 1983). ** p < .01. *** p < .001. TABLE 3--Regression Analyses of Demographic Variables, Total CBCL, and CGSQ Predicting to Three Maternal Mental Health Indicators (N = 167) BDI B SE [beta] t Demographic Variable Child's gender .10 .20 .04 .51 Child's age -.08 .03 -.20 -2.48 * Mother's age .04 .02 .18 2.13 * Race -.16 .29 -.04 -.57 Years of school -.03 .05 -.04 -.48 Hours employed .00 .01 .00 .07 Income -.37 .21 -.13 -1.78 (a) CBCL .00 .00 .01 .11 CGSQ .86 .13 .50 6.45 *** R .56 [R.sup.2] .32 f 8.19 p .000 BAI B SE [beta] t Demographic Variable Child's gender .24 .25 .07 .98 Child's age -.06 .04 -.13 -1.56 Mother's age .05 .02 .19 2.23 * Race -.13 .35 -.03 -.36 Years of school -.06 .07 -.07 -.91 Hours employed -.01 .01 -.15 -1.95 Income .25 .26 .08 .99 CBCL .01 .00 .12 1.48 CGSQ .71 .16 .36 4.31 *** R .50 [R.sup.2] .25 f 5.81 p .000 SF-36MH B SE [beta] t Demographic Variable Child's gender -1.06 3.1 -.02 -.34 Child's age .59 .51 .09 1.16 Mother's age -.60 .28 -.18 -2.15 * Race -2.34 4.45 -.04 -.53 Years of school .75 .84 .07 .90 Hours employed .09 .09 .07 1.02 Income 4.56 3.21 .11 1.42 CBCL .06 .06 .07 .97 CGSQ 13.53 2.04 -.52 -6.64 *** R .56 [R.sup.2] .32 f 8.08 p .000 NOTES: CBCL = Child Behavior Checklist (Achenbach & Edelbrock, 1983). CGSQ = Caregiver Strain Questionnaire (Brannan, Heflinger, & Bickman, 1997); BDI = Beck Depression Inventory (Beck, Ward, Mendeson, Mock, & Erbaugh, 1961); BAI = Beck A (Beck, Epstein, Brown, & Steer, 1988). (a) < .10. * p < .05. ** p < .01. *** p < .001. TABLE 4--Summary of Regression Analysis for Demographic Variables and Internalizing and Externalizing CBCL Predicting Caregiving Strain among Mothers of Children with Mental Health Problems (N = 167) Variable B SE [beta] t Child's gender -.12 .12 -.07 -1.00 Child's age .00 .02 .02 .22 Mother's age .00 .01 .00 .01 Race .05 .17 .02 .28 Years of school .00 .03 .01 .11 Hours employed -.01 .00 -.22 -3.06 ** Income .06 .12 .04 .53 Internalizing CBCL .01 .01 .11 1.34 Externalizing CBCL .03 .01 .44 5.23 *** R = .56, [R.sup.2] = .32, df = 8/158, F = 8.09 *** NOTE: CBCL = Child Behavior Checklist (Achenbach & Edelbrock,1983). ** p < .01. *** p <.001. TABLE 5--Regression Analyses of Demographic Variables, Internalizing CBCL, Externalizing CBCL and CGSQ Predicting to Three Maternal Mental Health Indicators (N = 167) BDI Variable B SE [beta] t Demographic variable Child's gender .00 .20 .00 -.01 Child's age -.08 .03 -.20 -2.49 * Mother's age .03 .02 .13 1.55 Race -.04 .29 -.01 -.15 Years of school -.04 .05 -.06 -.77 Hours employed .00 .01 .00 -.00 Income -.44 .21 -.16 -2.12 * Internalizing CBCL .03 .01 .19 2.38 * Externalizing CBCL -.02 .01 -.19 -2.13 * CGSQ .92 .13 .53 6.84 *** R .59 [R.sup.2] .35 F 8.35 p .000 BAI Variable [beta] SE [beta] t Demographic variable Child's gender .12 .24 .04 .49 Child's age -.06 .04 -.14 -1.62 Mother's age .04 .02 .14 1.63 Race .00 .35 .00 .01 Years of school -.07 .06 -.08 -1.10 Hours employed -.01 .01 -.16 -2.12 * Income .21 .25 .07 .85 Internalizing CBCL .04 .01 .25 3.00 ** Externalizing CBCL -.01 .01 -.10 -1.10 CGSQ .76 .16 .39 4.63 *** R .53 [R.sup.2] .28 F 6.11 p .000 SF-36MH Variable [beta] SE [beta] t Demographic variable Child's gender -.41 3.1 .01 .13 Child's age .49 .51 .08 .96 Mother's age -.45 .28 -.13 -1.60 Race -3.60 4.46 -.05 -.81 Years of school .95 .83 .08 1.15 Hours employed .09 .09 7.00 1.05 Income 5.28 3.20 .12 1.65 Internalizing CBCL -.26 .17 -.12 -1.54 Externalizing CBCL .33 .16 .18 2.01 * CGSQ -13.84 2.06 -.53 -6.73 *** R .58 [R.sup.2] .33 F 7.79 p .000 NOTES: CBCL = Child Behavior Check List (Achenbach & Edelbrock, 1983). CGSQ = Caregiver Strain Questionnaire (Brannan, Heflinger, & Bickman, 1997). BDI = Beck Depression Inventory (Beck, Ward, Mendelson, Mock, & Erbaugh, 1961). BAI = Beck Anxiety Inventory (Beck, Epstein, Brown, & Steer, 1988). SF-36-MH = SF-36 Mental Health subscale (Ware & Sherbourne, 1992). * p < .05. ** p < .01. *** p < .001.
This research was supported by grant no. R24 MH-56848 from the National Institute of Mental Health (M. K. Shear, principal investigator). The authors thank Dr. Gary Koeske, Jee Sook Lee, and the anonymous reviewers for their helpful comments on an earlier version of this article.
Achenbach, T. M. (1991). Manual for the Child Behavior Checklist/4-18 and 1991 Profile. Burlington: University of Vermont, Department of Psychiatry.
Achenbach, T. M., & Edelbrock, C. S. (1983). Manual for the Child Behavior Checklist and revised Child Behavior Profile. Burlington: University of Vermont, Department of Psychiatry.
Angold, A., Messer, S. C., Stangl, D., Farmer, E. M., Costello, E. J., & Burns, B. J. (1998). Perceived parental burden and service use for child and adolescent psychiatric disorders. American Journal of Public Health, 88, 75-80.
Baronet. A, (1999). Factors associated with caregiver burden in mental illness: A critical review of the research literature. Clinical Psychology Review. 19, 819-841.
Beck, A.T., Epstein, N., Brown, G., & Steer, R. A. (1988). An inventory for measuring clinical anxiety: Psychometric properties. Journal of Consulting and Clinical Psychology, 56. 893-897.
Beck, A. T., & Steer, R. A. (1993). Manual for the Beck Depression Inventory. San Antonio, TX: Psychological Corporation.
Beck. A. T., Ward. C. H., Mendelson, M., Mock. J., & Erbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatry. 12, 63-70.
Bickman, L., Guthrie, P.R., Foster, E, M., Lambert, E. W., Summerfelt, W. T., Breda, C. S., & Heflinger, C. A. (1995). Evaluating managed mental health services: The Fort Bragg experiment. New York: Plenum Press.
Biegel, D., Sales. E., & Schulz, R. (1991). Family caregiving in caregiver illness. Newbury Park, CA: Sage Publications.
Brannan, A. M., & Heflinger, C. A. (2001). Distinguishing caregiver strain from psychological distress: Modeling the relationships among child, family, and caregiver variables. Journal of Child and Family Studies, 10, 405-418.
Brannan, A., Heflinger, C., & Bickman, L. (1997). The caregiver or strain questionnaire: Measuring the impact on the family of living with a child with serious emotional disturbance. Journal of Emotional and Behavioral Disorders, 5, 212-222.
Cook, J. A., & Pickett, S. A. (1987/1988). Feelings of burden and criticalness among parents residing with chronically mentally ill offspring. Journal of Applied Social Science. 12(1), 79-107.
Early, T. J., & Poertner, J. (1993). Families with children with emotional disorders: A review of the literature. Social Work, 38, 743-751.
Floyd, F. J., & Gallagher. E. M. (19971. Parental stress, care demands, and use of support services for school-age children with disabilities and behavior problems. Family Relations, 46, 359-371.
George. L., & Gwyther, L. P. (1986). Caregiver well-being: A multidimensional examination of family caregivers of demented adults. Gerontologist, 26, 253-259.
Gill, M. J., & Harris, S. L. (1991). Hardiness and social support as predictors of psychological discomfort in mothers of children with autism. Journal of Autism and development Disorders, 21 407-416.
Goldberg-Arnold. J., Fristad, M. A., & Gavazzi, S. M. (1999). Family psychoeducation: Giving caregivers what they want and need. Family Relations. 48, 411-417.
Greenberg, J. S., Seltzer, M, M., Krauss, M. W., & Kim, H. (1997). The differential effects of social support on the psychological well-being of aging mothers of adults with mental illness or mental retardation. Family Relations, 46, 383-394.
Hadadian. A. (1994). Stress and social support in fathers and mothers of young children with and without disabilities. Early Education and Development, 5, 226-235.
Hammen, C., Burge, D., & Stansbury, K. (1990). Relationship of mother and child variables to child outcomes in a high-risk sample: A causal modeling analysis Development Psychology, 26(1), 24-30.
Hammen. C., Gordon, D., & Burge, D. (1987). Maternal affective disorders, illness, and stress-risk for children's psychopathology. American Journal of Psychiatry, 144, 736-741.
Heller. T. L., Baker, B. L., Henker, B., & Hinshaw, S. P. (1996). Externalizing behavior and cognitive functioning from preschool to first grade: Stability and predictors. Journal of Clinical Child Psychology, 25, 376-387.
Hill, R. (1958). Generic features of families under stress. Social Casework, 49, 139-150.
Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal, and coping. New York: Springer.
Lefley, H. P. (1996). Family caregiving in mental illness. Thousand Oaks, CA: Sage Publications.
MacKinnon, D. P., Lockwood, C. M., Hoffman. J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7(1), 83-104.
Marsh, D. T. (1996). Families of children and adolescents with serious emotional disturbance. In C. A. Heflinger & C. Nixon (Eds.), Families and the mental health system for children and adolescents (pp. 75-95). Thousand Oaks, CA: Sage Publications.
Maurin, J., & Boyd, B. (1990). Burden of mental illness on the family: A critical review. Archives of Psychiatric Nursing, 4, 99-107.
McCubbin. H. I., & Patterson, J. M. (1983). The family stress process: The double ABCX model of adjustment and adaptation. In H. I. McCubbin, M. Sussman, & J. M. Patterson (Eds.), Social stress and the family: Advances and developments in family stress theory and research (pp. 7-37), New York: Haworth Press.
McDonald. T. P., Couchonnal, G., & Early, T. (1996). The impact of major events on the lives of family caregivers of children with disabilities. Families in Society, 77, 502-514.
McDonald, T. P., Poertner, J., & Pierpont, J. (1999). Predicting caregiver stress: An ecological perspective. American Journal of Orthopsychiatry, 69, 100-109.
McHorney, C. A., Ware, J. E., & Raczek, A. E. (1993). The MOS 36-Item Short Form Health Survey (SF-36) II: Psychometric and clinical tests of validity in measuring physical and mental health constructs. Medical Care, 3, 247-263.
Monat, A., & Lazarus. R. S. (Eds.). (1991). Stress and coping, New York: Columbia University Press.
Montgomery, R., Gonyea, J., & Hooyman, N. (1985). Caregiving and the experience of subjective and objective burden. Family Relations, 34, 19-26.
Noelker, L., & Townsend, A. (1987). Perceived caregiving effectiveness: The impact of parental impairment, community resources and caregiver characteristics. In T. Brubaker (Ed.), Aging, health, and family (pp 58-79). Newbury Park, CA: Sage Publications.
Poulshock, S. W., & Deimling, G. T. (1984). Families caring for elders in residence: Issues in the measurement of burden. Journal of Gerontology, 39, 230-239.
Schene, A. H., Tessler, R. C., & Gamache, G. M. (1996). Caregiving in severe mental illness: Conceptualization and measurement. In H. C. Knudsen & G. Thornicroft (Eds.), Mental health service evaluation (pp. 296-316). New York: Cambridge University Press.
Schulz, R., Tompkins, C. A., & Rau, M. T. (1988). A longitudinal study of the psychosomatic impact of stroke on primary support persons. Psychology and Aging, 3, 131-141.
Schulz, R., & Williamson, G. M. (1991). A 2-year longitudinal study of depression among Alzheimers' caregivers. Psychology and Aging, 6, 569-578.
Sobel, M. E. (1982). Asymptotic intervals for indirect effects in structural equation models. In S. Leinhart (Ed.), Sociological methodology 1982 (pp. 290-312). San Francisco: Jossey-Bass.
Song, L., Biegel, D., & Milligan, S. (1997). Predictors and depressive symptomatology among lower social class caregivers of persons with chronic mental illness. Community Mental Health Journal, 33, 269-285.
Suarez, L. M., & Baker, B. L. (1997). Child externalizing behavior and parents' stress: The role of social support. Family Relations, 46, 373-381.
Thompson, E. H., & Doll, W. (1982). The burden of families coping with the mentally ill: An invisible crisis. Family Relations, 31, 379-388.
Turnbull, A. R, Brotherson, M. J., & Summers, J. A. (1986). Family life cycle: Theoretical and empirical implications and future directions for families with mentally retarded members. In J. J. Gallagher & P. M. Vietze (Eds.), Families of handicapped persons: Research, programs, and policy issues (pp. 45-66). Baltimore: Paul H. Brookes.
Ware, C. A., & Sherbourne, C. D. (1992). The MOS 36-Item Short Form Health Survey (SF-36). Medical Care, 30, 473-481.
Weissman, M. M., Leckman, J. F., Merikangas, K. R., Gammon, G. D., & Prusoff, B. A. (1984). Depression and anxiety disorders in parents and children: Results from the Yale Family Study. Archives of General Psychiatry, 41, 845-852.
Zarit, S. H., Reever, K. E., & Bach-Peterson, J. (1980). Relatives of the impaired elderly: Correlates of feelings of burden. Gerontologist, 20, 649-655.
Original manuscript received June 9, 2003
Final revision received February 23, 2004
Accepted June 29, 2004
Esther Sales, PhD, is professor, School of Social Work, University of Pittsburgh, 2217F Cathedral of Learning, Pittsburgh, PA 15260; e-mail: email@example.com. Catherine Greeno, PhD, is assistant professor, School of Social Work, University of Pittsburgh. M. Katherine Shear, MD, is professor, Department of Psychiatry, and Carol Anderson, PhD, is professor, Department of Psychiatry and Western Psychiatric Institute and Clinic, University of Pittsburgh.
|Printer friendly Cite/link Email Feedback|
|Author:||Sales, Esther; Greeno, Catherine; Shear, M. Katherine; Anderson, Carol|
|Publication:||Social Work Research|
|Date:||Dec 1, 2004|
|Previous Article:||Child custody loss among women with persistent severe mental illness.|
|Next Article:||Factors influencing self-esteem among individuals with severe mental illness: implications for social work.|