| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Victoria B. Mitrani, PhD, Center for Family Studies, Department of Psychiatry and Behavioral Sciences, University of Miami School of Medicine, 1425 NW 10th Avenue, Miami, FL 33136. E-mail: vmitrani{at}med.miami.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: REACH • Assessment • Multidimensional • Ethnicity
The family is the most proximal and influential interpersonal context throughout the life span and clearly plays a role in caregiver stress processes. Moreover, evidence is mounting that family factors such as marital cohesion (Rankin, Haut, & Keefover, 2001), communication patterns (Speice, Shields, & Blieszner, 1998), conflict (Semple, 1992), boundary ambiguity (Boss, Caron, Horbal, & Mortimer, 1990), and adaptability (Deimling, Smerglia, & Schaefer, 2001; Majerovitz, 1995) are related to the emotional functioning of caregivers. Despite such cumulative support for the relevance of the family, no theoretical model elucidates the mechanisms by which the family affects the stress process of caregiving.
One theoretical framework of family functioning that can be applied to caregivers is structural family theory (Minuchin, 1974), which views the family as an organism regulated by "structures" (i.e., repetitive patterns of interaction). Structures influence the manner in which family members interact across multiple dimensions such as managing disagreements, negotiating distance or closeness, and distributing roles. According to structural theory, problematic structures are the root cause of symptoms in individual family members. Whereas some structures are inherently maladaptive (e.g., the routing of conflict through a weak or subordinate third party), others become problematic because they are obsolete, representing the family's failure to adjust to changing circumstances. A family faced with a disruptive stressor such as dementia can either adapt in response to the changing needs of family members, thus protecting family members from experiencing negative sequelae, or fail to adapt and expose vulnerable family members to risk. This formulation maps on to the stress-process model in that the family's structural functioning is an intervening factor that mediates the relationship between the stressors experienced by the caregiver and his or her emotional functioning.
At the Miami site of the REACH (Resources for Enhancing Alzheimer's Caregiver Health; Eisdorfer et al., 2003) study, our team conducted a clinical trial examining the efficacy of Structural Ecosystems Therapy (SET; Mitrani, Szapocznik, & Robinson, 2000), a structural family intervention, in reducing distress among family caregivers of dementia patients. The aim of structural family therapy is to identify and correct problematic structures by observing and reshaping them as they occur in family therapy sessions. The focus is on process (i.e., how family members interact) rather than content (i.e., the issues they are dealing with). Structural family interventions have been found to be efficacious with other clinical populations (Chamberlain & Rosicky, 1995; Szapocznik et al., 2004) but had not been previously tested with dementia caregivers.
In the REACH trial, our Miami clinical team aimed to restructure family interaction patterns that were believed to contribute to caregiver distress such as emotional detachment or overinvolvement, negativity, and ineffective conflict resolution (Mitrani & Czaja, 2000). Our second aim was to strengthen protective interaction patterns such as family support, validation of the caregiver's leadership, and collaborative decision making. We randomly assigned 225 White non-Hispanic American and Cuban American caregivers to SET, SET augmented by a computer technology intervention that facilitated communication between caregivers and support persons, or a minimal support control condition. Caregivers in the combined family therapy and technology intervention experienced a significant reduction in depressive symptoms at the end of treatment, whereas those in the SET-only condition did not (Eisdorfer et al., 2003).
The Miami REACH trial demonstrated that a technology-enhanced structural family intervention can alleviate caregiver distress, but that the structural family therapy model should be improved. One avenue for treatment development is the identification of specific family patterns that are related to caregiver outcomes, so that family therapy can be aimed precisely at reinforcing those patterns known to be protective and restructuring those patterns known to be problematic. This was the impetus for undertaking a dementia caregiver adaptation of the Structural Family Systems Ratings (SFSR; Mitrani, Feaster, McCabe, Czaja, & Szapocznik, 2005), the measure of family functioning we used in the current study. Another approach for refining structural family interventions is a more complete understanding of the mechanisms by which family structure affects the stress process.
Our aim in the current study is to begin to test structural family theory in dementia caregiver families and expand our understanding of the role of family functioning in the caregiver stress process. We test a simple model that includes family functioning as an intervening factor in the relationship between objective burden reported by a caregiver (stressor) and the caregiver's self-report of distress (outcome). See Figure 1 for a graphical presentation of the hypothesized stress-process model. We hypothesize that family functioning will partially mediate the relationship between objective burden and distress. We also hypothesize that family functioning will provide substantial incremental prediction of distress over and above well-established predictors and correlates of distress such as objective burden and demographic factors. Post hoc analyses of moderated mediation (Frazier, Tix, & Barron, 2004) test the fit of the model for different caregiver subgroups and examine whether demographic variables affect the relationships uncovered in the overall model.
|
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
We used the 20-item Center for Epidemiologic Studies–Depression scale (CES-D; Radloff, 1977) to measure the presence of depressive symptoms in the caregiver. The CES-D has been widely used in the mental health literature, including intervention studies with family caregivers. A score of 16 or above has been identified as discriminating between groups with clinically relevant and nonrelevant depressive symptoms (Radloff & Teri, 1986). Cronbach's alpha for the CES-D in this sample was 
= 0.88.
We used the 10-item State–Trait Anxiety Inventory (STAI; Spielberger, 1979), taken directly from the State–Trait Personality Inventory (Spielberger), to measure caregiver anxiety. Respondents are asked how much they agree with statements such as "I felt calm" and "I was worried" by indicating on a 4-point Likert scale the intensity of the item. Responses range from 1 (not at all) to 4 (very much). Cronbach's alpha for this scale in the current sample was = 0.88.
The Perceived Physical Health (PPH; Archbold, Stewart, Harvath, & Lucas, 1986) measure consisted of four items to assess self-perceived level of physical health in the caregiver. These items measure the respondent's overall assessment of his or her health, tendency to get sick, comparison with others' health, and the expectation of future health decline. Response options for the first item range from 1 (poor) to 5 (excellent) on a 5-point scale. The other items are statements on a 5-point scale with response options ranging from 1 (definitely true) to 5 (definitely false). The PPH measure had a Cronbach's alpha in this sample of = 0.75.
The Screen for Caregiver Burden (SCB; Vitaliano, Russo, Young, Becker, & Maiuro, 1991) is a 25-item inventory that includes statements such as "My spouse doesn't recognize me all the time" and "It is exhausting having to groom and dress my spouse every day." The total number of items endorsed is used to measure objective burden. The SCB objective burden scale had a Cronbach's alpha in this sample of = 0.90.
The Structural Family Systems Ratings–Dementia Caregiver (SFSR-DC; Mitrani, Feaster, McCabe, Czaja, & Szapocznik, 2005) is an adaptation of the SFSR (Robbins, Hervis, Mitrani, & Szapocznik, 2001; Szapocznik et al., 1991). The SFSR-DC is a multidimensional, process-oriented, and observational instrument that is based on structural family theory. The SFSR-DC uses a standardized procedure, the Wiltwick Family Tasks (Minuchin, Rosman, & Baker, 1978), for eliciting family interaction on three family exercises (planning a menu, discussing likes and dislikes about each other, and discussing a recent family argument).
Whereas the SFSR was intended to assess interactions in families with behavior-problem adolescents and has never been used with caregiver populations, the SFSR-DC was designed to measure family structures that are germane to the psychological functioning of dementia caregivers. These family interaction patterns are represented by six subscales: enmeshment–cohesion, care-recipient disengagement, identified patienthood, conflict resolution, expressed positive affect, and expressed anger. The current study is an extension of the study that developed the SFSR-DC, and it uses a subsample from that study (Mitrani et al., 2005).
To code family functioning, a trained rater observes videotapes of the family as they perform the three family tasks, and the rater notes the presence of behaviors that reflect the corresponding interaction pattern. The rater sums the behaviors that are noted within each subscale to yield the subscale score. A total of 47 indicators on the SFSR-DC represent 23 different behaviors, 3 of which are coded on only one family task, 16 of which are coded on two tasks, and 4 of which are coded on all three tasks. The subscale indicators are presented in Table 2. Table 3 shows each subscale's composite reliability and interrater reliability, using intraclass correlation (n = 37). Composite reliability ranged from.691 to.974 and interrater reliability (kappa) ranged from.617 to.937 (Mitrani et al., 2005).
|
|
Procedure
SFSR-DC Rating
A trained rater and V. Mitrani coded the family task videotapes. In the rater training, the two raters reviewed the coding manual, rated five tapes together, and rated five tapes independently; they then had meetings to reconcile discrepancies. Calibration between raters continued throughout the study. The 181 families in the current study include 147 families whose first SFSR-DC was conducted at baseline and 34 whose first SFSR-DC was at a later time point. These 34 families were enrolled in the study since baseline but had missing baseline SFSR-DC's because of unavailability of the care recipient, poor videotape quality, or breaches in administration of the family tasks.
The average number of participants in the family tasks was 3.04 (SD = 0.727). Although eligibility for the parent study required that a third family member (in addition to the caregiver and care recipient) be willing to participate, we administered family tasks even if no family other than the caregiver and care recipient were present. There were 36 caregiver–care recipient families in this sample. Other participants included caregivers' daughters (n = 47), sons (n = 33), husbands (n = 22), sisters (n = 15), daughters-in-law (n = 11), brothers (n = 10), sisters-in-law (n = 8), other extended family (n = 28), friends or neighbors (n = 11), and hired aides (n = 3).
Statistical Analyses
We analyzed data by using SPSS 11.5 for Windows and AMOS 5.0 software. We imputed missing data by using SPSS missing value analysis. To examine preliminary relationships between family functioning, caregiver burden, and caregiver distress, we used Pearson product–moment correlations. We evaluated the stress-process model with a structural equation model (SEM; Kline, 1998) estimated by maximum likelihood in AMOS (Arbuckle, 2003). We assessed model fit by using the model chi square value (Raykov, Tomer, & Nesselroade, 1991), the Comparative Fit Index (CFI; Bentler, 1990), and the root mean square error of approximation (RMSEA; Joreskog & Sorbom, 1988), as suggested by Hu and Bentler (1999). We utilized a value of CFI 
0.95 and a value of RMSEA 
0.08 as criteria for acceptable fit according to Kline. The significance-level criterion for all statistical tests was = 0.05.
To assess whether family functioning mediated the relationship between objective burden and caregiver distress, we used the statistically powerful asymmetric distribution of products test (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002), which is based on the distribution of the product of the two model paths (i.e., objective burden to family functioning times family functioning to caregiver distress) that determines the mediating pathway. This procedure tests whether a significant effect of objective burden on caregiver distress works through family functioning. Finally, we used planned post hoc tests to examine overall model differences by ethnicity (Cuban American and White non-Hispanic American), relationship of caregiver to care recipient (spouse vs daughter and husband vs wife), and income (low and high) to determine if these variables were moderating the stress-process model. If the overall models were significantly different, individual paths were then tested. We provide standardized path coefficients for all subgroups as estimates of the effect sizes.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
2(4) = 8.6, p <.07; CFI = 0.98; RMSEA = 0.08. All of the path coefficients of the structural portion of the model were statistically significant. Table 5 displays the squared multiple correlation for caregiver distress, the fit indices, and the standardized path coefficients for this and all subsequent models. The asymmetric confidence interval test suggested that family functioning is a mediator of the effect of objective burden on caregiver distress (confidence interval or CI = 0.0223, 0.1362; p <.05).
|
Investigating the Moderating Effects of Demographic Variables
We examined the moderating effects of ethnicity, caregiver–care recipient relationship (spouses vs daughters and husbands vs wives), and income (< $30,000 vs $30,000), and we determined that these variables did not moderate the stress-process relationship in this sample. Fit indices were acceptable for all subgroups except wives. To test moderation, we first determined whether the model was significantly different across the subgroups. Then, in a similar fashion, we tested individual structural paths or variances suspected of being different on visual inspection (e.g., a path was significant in one subgroup model, but the same path was nonsignificant in the other subgroup).
For ethnicity, we determined observed and measurement variances and structural parameters between White non-Hispanic Americans and Cuban Americans to be nonsignificantly different. We conducted tests for differences in two paths, objective burden to family functioning and family functioning to caregiver distress, which showed no significant differences between the two ethnic groups.
For the comparison between spouses and daughters, the overall variance and structural parameters were similar. We conducted one additional analysis to determine if the individual path from family functioning to caregiver distress was different, but we found this path to be similar for both groups. Although the fit was acceptable for the husbands' model, it was not satisfactory for that of the wives. The measurement model and all pathways in the structural model were not statistically different between husbands and wives, but a test of the variance parameters revealed that wives were different from husbands for objective burden: 2(5) = 13.0, p =.02. The variance of objective burden for husbands was higher (19.37) than that for wives (16.45). The path from family functioning to caregiver distress for husbands (ß = –.32) was significantly different, 2(8) = 18.4, p =.02, from the path for the wives (ß = –.18).
For the low- versus high-income comparison, the variances between groups and the structural parameters were similar. We also tested the path from objective burden to caregiver distress and found it to be nonsignificantly different.
Assessing Potential Confounding Variables
Before conducting SEM analyses, we compared data from the baseline group (n = 147) with data from those whose first assessment was later in the study (n = 34). Only one of the five variables showed a significant difference. Family functioning was higher in the baseline group, M = 19.8, SD = 3.9, than it was in those families not assessed at baseline, M = 18.4, SD = 3.3, t(179) = 2.0, p =.05. To ensure that this mean difference would not confound subsequent SEM results, we compared the SEM models of the two groups. The structural paths were not different between groups. The variances, however, were different: 2(6) = 13.4, p =.04. Further examination showed that only the variance of objective burden for the baseline groups (variance = 19.97; SE = 2.34) was significantly higher than that for the nonbaseline group: 2(1) = 6.1, p =.01; variance = 9.68; SE = 2.36. Thus, no structural paths and only one variance term differed by time of assessment, indicating little confounding effect.
In addition to assessing time point, we also evaluated the influence of intervention group assignment on the measurement of the data. According to chi-square analysis, time point by intervention assignment was nonsignificant: 2 =.12(2), p =.94. Next, we conducted an analysis of variance with SFSR-DC score as the dependent variable and time point and intervention assignment as the independent variables. The interaction was nonsignificant, F =.59(2, 170), p =.55, thus demonstrating that intervention assignment was not likely to confound subsequent results.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This study suggests that interventions aimed at transforming problematic family interaction patterns and supporting protective family patterns can help to prevent or alleviate the distress that caregivers experience in response to caregiving stress. Specifically, these results support the adoption of family interventions that promote family cohesion, involvement of the care recipient in family activities, resolution of disagreements, and expressions of affection and levity; and that reduce expressed anger and negativity about the care recipient. Within a structural approach, such an intervention would focus on changing family processes regardless of content (i.e., sessions would not always focus on caregiving issues). For example, a session could revolve around having family members plan an outing together. Having family members agree on the details of the outing could be an avenue to work on conflict resolution; gathering the family for the outing would promote cohesiveness; and planning for an outing that actively includes the care recipient would reduce care-recipient disengagement.
The finding that the stress-process model had adequate fit across nearly all of the demographic subgroups suggests that family functioning is a robust factor among a wide range of caregiver characteristics. Nonetheless, the tests of invariance within the multiple-method framework of SEM had small sample sizes, which may have resulted in insufficient power to detect moderation effects. Examining the absolute differences in standardized pathways by each comparison group (e.g., Cuban-Americans vs White non-Hispanic Americans; see Table 4) shows that a majority of the values fell between what would be considered small to medium effect sizes (Cohen, 1988), suggesting that demographic factors may be moderating the stress process involving family functioning. Moreover, we found notable differences in the amount of variance that caregiver distress accounted for by the model. The highest squared multiple correlation coefficients were for husbands (.33), daughters (.25), White non-Hispanic Americans (.25), and low-income families (.22); the lowest were for wives (.11), Cuban Americans (.12), and relatively high-income families (.15). This would indicate that additional variables may be important in determining distress for wives, Cuban Americans, and high-income caregivers that are not as relevant for the other groups.
The intersection of wives and Cuban Americans is particularly interesting given the central role that women play in Hispanic families. It is possible that this central role causes the Cuban wives to have a strong influence on their families, but that this effect is not entirely reciprocal; that is, the family is affected by the woman's distress, but they, in turn, do not have as much of an impact on her. This supposition is consistent with the comparisons of structural path coefficients in Cuban Americans versus White non-Hispanic Americans and husbands versus wives. In the Cuban and wife caregiver families, a significant coefficient is found in the pathway between objective burden and family functioning. This pathway is not significant for the White non-Hispanic American and husband caregiver families. These findings suggest that the caregiving stress experienced by the caregiver had an impact on family functioning in Cuban and wife caregiver families but not in White non-Hispanic and husband caregiver families. Conversely, the coefficients for the family functioning to caregiver distress paths are not significant for the Cuban and wife caregiver groups, but are significant for the White non-Hispanic and husband caregiver groups, suggesting that family functioning had a significant influence on caregiver distress among White non-Hispanics and husbands, but not among the Cuban and wife caregiver groups. Further research is needed to understand how family functioning interacts with demographic characteristics, why the model is more predictive for some groups than others, and to identify the other factors that can fill in the stress-process picture for the various subgroups.
Limitations
Limitations in this study warrant caution when one is interpreting results. First, the exclusion of non-Cuban Hispanics and African Americans limits the generalizability of the findings. Another limitation related to sampling is that the definition of "family" in the parent study was deliberately left to the discretion of the caregiver, and the sample includes only those who chose to attend after being invited by the caregiver. Therefore, the family constellations are not standardized and the sample may be overrepresented by family members who have positive relationships with the caregiver. The complexity of defining who constitutes a family unit and the difficulty of recruiting all members of a family represent inherent difficulties in conducting family research. These issues are particularly complex in late-life families because of developmental transitions associated with age, such as relocations that make it difficult to include geographically distant but nevertheless influential family members, the recent losses of significant family members to death or disability, and the importance of recently formed informal kinship networks that may include neighbors and hired aides.
A further limitation associated with family definition and recruitment is that 36 families consisted of only the caregiver and care recipient. Although mediation analyses could not be conducted with such a small subsample, we did examine the psychometric properties of the SFSR-DC subscales in these two-person families. Alpha reliabilities were satisfactory for all subscales tested (conflict resolution is not amenable to this analysis because items represent mutually exclusive categories), ranging from = 0.617 (expressed anger) to = 0.880 (care-recipient disengagement). Interrater reliability (n = 13, from the entire sample in the parent study) was less satisfactory, however. The expressed anger (
= 0.143) and expressed positive affect ( = 0.408) subscales had unacceptable interrater reliability; in addition, interrater reliability could not be calculated on the care-recipient disengagement subscale because it had insufficient variability in one set of ratings. The other three subscales had kappa values ranging from = 0.773 (enmeshment or cohesion) to = 0.824 (conflict resolution). Thus, some preliminary information suggests that the SFSR-DC may be somewhat less reliable in two-person families, but further investigation is warranted.
Finally, because of the relatively small number of families and missing SFSR-DC data at later time points, the present study could only test a very simple stress-process model within a cross-sectional framework. In addition, the SFSR-DC has yet to be cross-validated, and the current study uses a subset of the sample upon which the measure was developed (Mitrani et al., 2005).
Directions for Future Research
Further research is needed to elucidate the family mechanisms of caregiver stress processes. An important area for further study is the impact of caregiving stress on the family system, and how the family responds to the ongoing changes in the caregiving needs of the care recipient. Structural theory would predict that the family will either adapt to changing needs by reorganizing itself, and thus protect individual family members; or that it will fail to adapt and therefore lead to problems in individual family members. The latter represents deterioration of family functioning as a secondary stressor that has a cascading effect on individual family members. Beyond that, we need to identify the factors that predict whether a family will successfully adjust or fail to respond in an adaptive manner.
The program of research outlined herein requires longitudinal designs that follow the family and individual caregiver's progress through the milestones of the dementia caregiving process. Such research could ultimately contribute to the development of precisely focused and theory-driven family interventions to prevent and alleviate caregiver distress. Furthermore, adding family functioning into the stress-process picture can provide knowledge regarding how family functioning moderates the effects of individual, demographic, cultural, and environmental factors that influence caregiver outcomes, thus guiding the development of interventions customized for particular segments of the dementia caregiving population.
|
Footnotes |
|---|
1 Department of Psychiatry and Behavioral Sciences, University of Miami, FL.
2 University Center for Social and Urban Research, University of Pittsburgh, PA.
Decision Editor: Linda S. Noelker, PhD
Received for publication April 15, 2005. Accepted for publication July 29, 2005.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  S. W. Williams, C. S. Williams, S. Zimmerman, J. Munn, D. Dobbs, and P. D. Sloane Emotional and Physical Health of Informal Caregivers of Residents at the End of Life: The Role of Social Support J. Gerontol. B. Psychol. Sci. Soc. Sci., May 1, 2008; 63(3): S171 - S183. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Pillemer and J. J. Suitor Making choices: a within-family study of caregiver selection. Gerontologist, August 1, 2006; 46(4): 439 - 448. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
