| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Maximiliane E. Szinovacz, Department of Gerontology, University of Massachusetts Boston, 100 Morrissey Boulevard, Boston, MA 02125-3393. E-mail: maxi.szinovacz{at}umb.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Family caregiving • Care networks • Care transitions • Intergenerational transfers • Intergenerational supports
Literature Review
Research has assessed elder care networks in several different ways. One set of studies has portrayed the composition of elders' care networks, showing that adult children's care for their elderly parents is often shared with several family members, including other adult children, spouses, or even the adult children's children (Dilworth-Anderson, Williams, & Gibson, 2002; Ingersoll-Dayton, Neal, Ha, & Hammer, 2003; National Alliance for Caregiving, 1997; Stommel, Given, & Given, 1998; Szinovacz, 2003; Wolf, Freedman, & Soldo, 1997). Thus, a focus on a single or primary caregiver may be too narrow to capture the reality of informal care.
A second set of studies has addressed relatives' (and especially adult children's) participation in care and their care decisions, showing how specific care contexts and characteristics of parents and adult children influence each child's participation in care (Caron, Griffith, & Arcand, 2005; Henretta, Hill, Li, Soldo, & Wolf, 1997; Marks, 1996; McGarry & Schoeni, 1995; Wolf et al., 1997). These studies confound characteristics that explain parent care per se with predictors related to each specific child's entry into or exit from the caregiver network because they capture caregivers at different stages of their careers. Adult children who provide care at the beginning of the parent's care trajectory may differ from those who assume care later on. Also, children who assist parents over a longer time period are overrepresented in cross-sectional designs and may differ from those involved in short care spells.
Using a similar approach, a related set of qualitative studies has explored care decisions and negotiations among adult children (Caron et al., 2005; Hequembourg & Brallier, 2005; Matthews, 1992). Both sets of studies have revealed that adult children's participation in care evolves from negotiations among adult child siblings that reflect normative factors, children's ability and propensity to provide care, characteristics of the entire sibling network, as well as characteristics of the parents (Finch & Mason, 1993; Henretta et al., 1997; Marks, 1996; McGarry & Schoeni, 1995; Wolf et al., 1997). Studies have documented gender differences in care, with daughters being more prone to help than their brothers (Henretta et al., 1997; Wolf et al., 1997). They have further indicated that characteristics of the sibling network influence each specific child's participation in care. For example, children with sisters provided fewer hours of care than did those with brothers or no siblings, but parents still received a net gain in support when more siblings were available to provide assistance (Wolf et al., 1997). Although these studies have offered insights into the complexity of care decisions, they have failed to address the dynamic character of care negotiations.
Longitudinal studies that assess circumstances leading to transitions between informal and formal care (Kelman, Thomas, & Tanaka, 1994; Lyons, Zarit, & Townsend, 2000; Miller & McFall, 1991; Stoller, 1990) have demonstrated considerable stability in type of care (formal vs informal), although changes within the informal network itself have been observed. However, these studies have provided only limited information on changes within informal care networks. Stoller, for example, noted a shift toward female helpers over time. Research on caregiving careers has portrayed caregivers' career trajectories (e.g., in terms of duration or intensity; Seltzer & Li, 1996, 2000) and has shown effects of career trajectories and care contexts on caregivers' well-being (Aneshensel, Pearlin, Mullan, Zarit, & Whitlatch, 1995; Lawton, Moss, Hoffman, & Perkinson, 2000; Seltzer & Li, 2000). Although these studies have captured change in care arrangements over time, they have only rarely attended to the influence of care network characteristics on these changes.
Fewer studies have explored changes of individual caregivers over time. Jette and colleagues (1992) reported that 23% of their Massachusetts sample changed primary caregivers over a 4-year period. These changes typically involved shifts among adult child caregivers or from spouses to adult children and prevailed among care recipients who did not coreside with the original primary caregiver. Dwyer, Henretta, Coward, and Barton (1992) investigated the assumption and relinquishment of care over time by adult child caregivers. Their analyses indicated that adult children who commenced care were more likely to be female, unmarried, younger, not employed full time, and to live closer to the parent. Adult children who relinquished care were more likely to be male, to have married between waves, and to live farther away from parents. In contrast to most other researchers, Dwyer and colleagues further considered characteristics of the adult child sibling networks as predictors of care transitions. Their analyses (based on the National Long-Term-Care Survey) revealed a positive relationship between individual adult children's care transitions (starting or stopping) and those of their siblings. This suggests some confounding of adult children's provision of care per se and shifts in care responsibilities among adult children. Thus, researchers should consider care networks in their entirety to identify the extent and predictors of change in caregivers over time. They also need to differentiate between adult children's provision of care per se and changes in adult children's care involvement over time. Our analyses addressed these issues by focusing on change in adult child care networks over a 2-year period.
Theoretical Framework and Hypotheses
Research on intergenerational supports has relied on theoretical models from various disciplines, including altruism, self-interest exchange, social convoys, intergenerational solidarity, and the so-called behavioral model of health care use (Andersen, 1995; Silverstein, Conroy, Wang, Giarrusso, & Bengtson, 2002; Soldo & Hill, 1993). Most of these models identify predictors of whether support is provided but fail to address the circumstances under which support structures change. Except for the convoy model (Antonucci, 1990), they also focus on individual support contributions rather than on the distribution of support among network members. The theoretical framework informing our analyses integrated findings from previous research into a life course and family systems perspective. The life course perspective draws attention to three concepts that are crucial to our theoretical approach: contextual embeddedness of life transitions, the notion of linked lives, and life course development (Bengtson & Allen, 1993; Moen, 1996). Similarly, family system theorists argue that changes experienced by one family member have consequences for the entire family system. They also stress feedback processes, that is, tendencies to correct system imbalances or disruptions that may be caused by internal or external stressors (Broderick & Smith, 1979). Based on the premise of contextual embeddedness, we argue that family support systems are influenced by contexts (cultural, socioeconomic, familial) that define individuals' propensity to provide supports, their ability to do so, and their access to family-external care. The notions of linked lives and family systems suggest that the allocation of support responsibilities among family members derives from negotiations within the family system as a whole and is responsive to each family member's abilities and obligations at any point in time (Davey & Norris, 1998; Finch & Mason, 1993). The emphasis on life course development means that family members' needs, obligations, and resources change over time in response to specific life course events. Thus, family support systems must be understood dynamically, reflecting the changing life circumstances of each family member.
In line with this general framework, we considered three broad groups of predictors of changes in support networks, namely contexts (cultural, socioeconomic, familial), caregiver abilities and resources, and care burden. In discussing these potential influences on changes in care networks, we focus on predictors that are available in the Health and Retirement Study (HRS).
Cultural Contexts
The assumption of contextual embeddedness implies that selected contexts shape care decisions. Cultural contexts reflect the normative underpinnings of care that are tied to adult children's sense of filial obligation and thus their willingness to maintain stressful parent care over time. Because the HRS lacks direct indicators of filial responsibility, we relied on cultural contexts with known ties to filial obligation. Specifically, past research has suggested that both gender and race/ethnicity may function in this manner. People often see care work as women's work (Campbell & Martin-Matthews, 2003; Shuey & Hardy, 2003), and women tend to carry the majority of care work unless only male relatives are available (Shuey & Hardy, 2003). These gender norms may influence network change in a dual way. On the one hand, norms that favor care work by women should also enhance the stability of women's care (i.e., women may feel more obliged than men to maintain stressful caregiving activities over prolonged time periods). On the other hand, the availability of alternative female caregivers (sisters) may promote network change. We addressed this ambiguity by including both gender composition and the number of siblings (see "Familial Contexts," below) in our models. Controlling for number of brothers and sisters, we expected that care networks that included men would have more changes in caregivers than would caregiver networks that included only women (Hypothesis 1).
A second gender-specific norm pertains to cross-gender care. Both caregivers and care recipients prefer that personal care (which is the focus of our analyses) be provided by same-gender caregivers (Campbell & Martin-Matthews, 2003; Pillemer & Suitor, 2006). To the extent that more problematic care situations promote change among caregivers, we thus hypothesized that care networks involving cross-gender care would be less stable than would care networks involving only same-gender care (Hypothesis 2).
There is also some evidence that some minority groups (and especially African Americans) feel stronger obligations to care for ailing parents and are more prone than are Whites to reject parents' institutionalization (Sudha & Mutran, 1999). These attitudes would promote maintenance of stressful care over long time periods. Thus, we expected that there would be less change in the sibling networks of African Americans and Hispanics than in those of Whites (Hypothesis 3).
Socioeconomic Contexts
Socioeconomic contexts can impinge on care networks in a dual fashion. First, parents in higher socioeconomic status (SES) groups have greater access to formal care, which may reduce stress among caregiving adult children. Second, some researchers have argued that provision of care is grounded on self-interest exchange (Cox & Rank, 1992), either concurrent (inter-vivos transfers) or in the future (bequests). Because parents in higher SES groups control more economic resources than do their poorer counterparts, their children may be more motivated to maintain long-term care arrangements based on the receipt of current, or the expectation of future, financial contributions from parents. Empirical evidence for these perspectives is mixed. Some researchers have found little evidence of bequest or self-interest exchange (Altonji, Hayashi, & Kotlikoff, 1996), whereas others have shown that supports to parents are contingent on past receipt of supports by parents (Henretta et al., 1997; Silverstein et al., 2002) and that the presence of siblings induces more competition among siblings and thus more giving to parents (White-Means & Hong, 2001). Nevertheless, both perspectives (access to formal care, self-interest exchange) imply a tendency toward longer maintenance of care for higher status parents. We thus expected there to be less change in caregiver networks among higher status parents. Lacking information on parents' income or current SES, we selected two of the few available indicators of their status (viz., SES when the respondent was a child and education). We thus hypothesized that there would be more network change among parents with lower education and with lower SES when the respondent was growing up than among parents with higher education and with higher SES (Hypothesis 4).
Familial Contexts
The main indicator of familial contexts available in the HRS is sibship size (the number of parents' children). Past studies have suggested that care provision by specific adult children declines as their number of siblings (and especially sisters) increases (McGarry & Schoeni, 1995; Soldo & Hill, 1995; Wolf et al., 1997), although the availability of more siblings tends to increase the total amount of help provided by the entire sibship group. This suggests that individual adult children adapt their care provision to the availability of care by other adult children, and that the potential for alternative care arrangements will be greater the larger the sibship. Controlling for the gender composition of the care network (see Hypothesis 1), we thus expected that there would be more network change with increasing numbers of adult child siblings, although, given the gender mandate in caregiving, the effect of the number of sisters on changes in care networks would be stronger than the effect of the number of brothers (Hypothesis 5). This hypothesis does not contradict Hypothesis 1. For instance, we expected to see less change in only-daughter networks if there were only two daughters than if there were three or more daughters.
Caregiver Abilities and Resources
Both the assumption of linked lives and the assumption of life course development imply that siblings' relative involvement in care as well as changes in their care participation evolve from considerations that include each sibling's abilities and resources. Indeed, earlier research has shown that competing family obligations can preclude or reduce care involvement. For example, some studies have indicated that married individuals are less prone to provide care than are single people (Engers & Stern, 2002; McGarry & Schoeni, 1995) and that childless caregivers or those without dependent children are more likely than caregivers with children in the home to assume care (Pezzin & Schone, 1999; Wolf et al., 1997). Because the HRS either did not obtain specific adult child characteristics such as presence of dependent children or only measured these for selected siblings, we operationalized competing family obligations with siblings' age and marital status. Younger siblings are more likely than older siblings to still have dependent children and to be in the labor force. Thus, younger sibling networks may be more prone to change than are networks composed of older siblings. Although marital obligations may compete with caregiving and spouses' conflict over parent care involvement can contribute to caregiver stress (Stephens & Franks, 1995; Suitor & Pillemer, 1994), spouses also can provide essential supports to caregivers (Globerman, 1996). If these two influences balance each other, the marital status composition of the sibling network may have little effect on network change. Nevertheless, based on the overarching assumption that competing family obligations will promote network change, we hypothesized that care networks composed of younger adult children and those that included a greater proportion of married adult children would be less stable than would networks with older adult children and with a lower proportion of married children (Hypothesis 6).
Caregiver Burden
The concepts of linked lives and feedback processes suggest that siblings who are not involved in care will help or take over care if the caregiving siblings experience undue burden, although some studies have failed to find a link between caregiver stress and caregiver turnover (Caron et al., 2005; Jette et al., 1992; Miller & McFall, 1991). Unfortunately, the HRS has no measures of caregiver burden or stress. We thus addressed this issue in terms of characteristics of the care situation that may be linked to caregiver burden. Such conditions include additional financial support to parents, care for multiple parents, as well as parent attributes (e.g., age, can be left alone, end-of-life care) that may serve as indicators for care requirements and care burden. We thus hypothesized that care networks would be less stable if the other parent also required care, if parents also required financial assistance, if parents were older, if parents could not be left alone, or if parents had died between waves (Hypothesis 7). One other condition that researchers have linked to caregiver change is coresidence (Jette et al., 1992). Coresidence seems to reduce the probability of changes in care arrangements, most likely because moving the parent may be difficult and undesirable from the perspective of the care recipient. We thus proposed that parental coresidence would reduce the likelihood of change in care networks (Hypothesis 8).
In summary, our conceptual framework implied that change in care networks would be influenced by norms pertaining to filial obligation and reciprocity (cultural and socioeconomic contexts; Hypotheses 1–4) and would evolve from negotiations among siblings that reflected access to alternative care arrangements (socioeconomic and familial contexts; Hypothesis 5), competing caregiver responsibilities (Hypothesis 6), and caregiver burden (Hypotheses 7 and 8). Due to data limitations in the HRS, several of our hypotheses constitute only indirect assessments of the main theoretical arguments.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Figure 1 shows the selection of the analytic sample. Because we focused on network changes, the analyses relied on those cases in which a parent (a) received care from any adult child for two consecutive waves and (b) had more than one child. This subsample represented 12.3% of all living parents and 31.6% of parents receiving care (n = 1,017 parents). Care occasions were the unit of analysis, pooling over waves and parents to capture all care occasions. Thus, if a parent received care from any adult child between Wave 1 and Wave 2, this constituted one care occasion. If the same parent also received care between Waves 2 and 3 or the other parent received care between Waves 1 and 2, we treated these as additional care occasions. Thus, the same parents or adult child networks were represented several times in the sample if adult children either cared for more than one parent or cared for the same parent over several consecutive waves; we adjusted for the resulting nonindependence of observations in our analyses. There were a total of 1,577 care occasions for analyses pertaining to changes in any adult child caregiver (we lost 2 cases due to missing data that we could not impute) and 1,457 care occasions for change in the primary caregiver. The lower number of cases for primary caregiver resulted from missing cases for the adult child "most involved" in parent care that we could not impute because data were based on person identification numbers.
|
The dependent variables also relied exclusively on the respondents' reports of who provided help and thus neglected potentially different perceptions on the part of other adult children. Several studies have documented that family members' reports on caregiving often differ (Horowitz, Goodman, & Reinhardt, 2004; Matthews, 1995; Pruchno, Peters, & Burant, 1995), but it is not clear to what extent siblings disagree on the mere fact of who was involved in care to their parents. Both of these data limitations may have introduced some bias into our analyses.
To assess network change, we constructed a matrix of adult child helpers for each wave (by person number) and compared these matrices to define the change variables. We derived three dependent variables from these matrix comparisons: change in any adult child caregiver, type of change in caregivers, and change in the primary adult child caregiver. Change in any adult child caregiver was a dichotomous variable (1 = yes, 0 = no) that indicated whether care involvement of any of the adult children for a specific parent had changed across waves (i.e., whether any adult child either started or dropped out of care at Time 2). Elaborating upon this variable, type of change in caregiver identified what kind of change had occurred. It was a nominal variable that partitioned the group reporting any change (see above) into three categories: one or more adult children were added but none were dropped, one or more adult children were dropped but none were added, and exchange (some adult children were dropped and some were added during the same time period). Those who reported no change served as reference category. The third dependent variable referred to changes in the adult child who helped most with care. It was a dichotomous variable indicating whether the primary adult child caregiver had been replaced between waves. We should note that this variable did not necessarily refer to the primary caregiver. In some cases the primary caregiver could have been the parent's spouse or another relative. The HRS provides no information on who the primary caregiver is. Parent's marital status may provide some indication of who the primary caregiver is, but some spouses may be unable to help and thus marital status alone could only partially adjust for this data limitation. Furthermore, inclusion of both parent's marital status and whether adult children provided care to another parent led to multicollinearity problems. Because care for another parent is conceptually and empirically (we conducted sensitivity analyses including care for the other parent and parent's marital status in alternative models) more important, we used only this variable in the final models.
Some respondents indicated that "all siblings helped" (in response to which of their siblings provided care) or that "all siblings helped equally" (in response to who provided the most care). Because these answers may have inflated the extent of change, we conducted sensitivity analyses to assess whether this variable influenced results. These analyses revealed that although the overall extent of change was altered slightly (see Results section), the regression results remained essentially the same.
The five major dimensions of independent variables were contexts (cultural, socioeconomic, and familial), caregiver abilities and resources, and caregiver burden. Selected variables differed slightly for the analyses that addressed any change and type of change in caregivers and the analyses pertaining to change in the primary adult child caregiver. Specifically, we used network-level variables such as the gender composition of caregivers for the former analyses but characteristics of the primary caregiver (e.g., gender of the primary adult child caregiver) for the latter analyses.
Cultural contexts consisted of caregivers' and parent's gender and race. For the gender composition of the adult child care network, we used two dummy variables (female only, mixed gender), with only male caregivers serving as the reference. We coded the gender of both the primary adult child caregiver and the parent as 1 = female, 0 = male. Because Hypothesis 2 addressed cross-gender care, we created interaction terms between the adult child gender and the parent gender variables. Race was only available for respondents. However, because all questions pertained to biological siblings, siblings' race should have been the same as respondents'. We coded race into three dummy variables, namely Black, Hispanic, and other (White was the reference).
Socioeconomic context variables included parents' SES and education. The only SES variable for parents pertained to the time when the adult child was growing up; we originally coded this into three categories. We created two dummy variables (low SES, medium SES), with high SES as the reference. In addition, we included current education (in years) of the parent receiving care.
We represented familial contexts by the number of brothers and sisters in the adult child network as well as the number of adult child caregivers at Time 1.
We captured caregiver abilities and resources by caregivers' age and marital status. For the analyses pertaining to any change and type of change, we assessed these variables at the network level. For marital status we used two dummy variables (all married, some married), with all caregivers not married as the reference. For adult children's age we used the average age of the caregiver network. In the analyses concerning change in the primary caregiver, we again used the characteristics of the primary caregiver at Time 1, namely, married (1 = yes, 0 = no) and age in years. We operationalized caregiver burden with other support obligations, parent's age and condition, and coresidence. Other support obligations included whether adult children provided care to the other parent (1 = yes, 0 = no) and whether they provided financial support to any parent (1 = yes, 0 = no). Because the HRS lacks direct information on parents' health, such as their health condition or respondent-rated health status, we used other available indicators of parents' condition, namely age (in years), whether they could be left alone (1 = yes, 0 = no), and whether they had died between waves (1 = yes, 0 = no). We coded parental coresidence with any adult child as 1 = yes, 0 = no.
All models further controlled for the waves from which the care information was drawn (Waves 4 and 5 = reference). Table 1 shows the means and standard deviations of all variables for the full sample (any change in caregivers).
|
We also assessed potential multicollinearity problems. Both the correlation matrix (see Table 2) and variance inflation factors under 2.0 indicated that there were no multicollinearity problems.
|
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Extent of Change
Change in adult child caregivers occurred in 54.3% of the care occasions. When change did occur, roughly equal numbers of networks (more than one third) either added (36.6%) or dropped (38.8%) one or more adult children. Exchange of caregivers occurred somewhat less frequently (24.7%), and more than one fourth of primary adult child caregivers changed. Sensitivity analyses including and excluding those cases in which respondents indicated that "all siblings helped" (n = 100) revealed no major influence on these outcomes for change in any sibling. Change in primary adult child caregivers deceased from 29.4% to 26.6% when we excluded responses indicating "all siblings equally" at either Time 1 or time 2 (n = 63). Taken together, these data suggest considerable fluctuation among adult child caregivers over time.
Tables 3, 4, and 5 present results for the regressions, one for each dependent variable. However, because the discussion of results focuses on the hypotheses, we organize the presentation of results around the five sets of predictors (indicators for cultural contexts, socioeconomic contexts, familial contexts, caregiver obligations and resources, and caregiver burden).
|
|
|
Hypothesis 2 pertained to the impact of cross- versus same-gender care. We found that the gender of the care recipient per se had no influence on change in caregivers, type of change, or change in primary caregivers (Tables 3–5). Furthermore, the data did not support the predicted positive effect of cross-gender care (based on interactions between the gender composition of the network and parent's gender, data not shown) on change. For change in any caregiver and type of change in caregivers, the interactions were not significant. Contrary to our hypothesis, we found a significant positive effect of the Female Caregiver x Female Care Recipient interaction on change in primary caregivers. Thus, female primary caregivers involved in care for mothers seemed more prone to change than did female caregivers caring for fathers.
Hypothesis 3 addressed racial/ethnic variations in network change. Contrary to the hypothesis, race/ethnicity did not exert a significant influence on any change in caregivers (Table 3) or in the primary caregiver (Table 5). However, type of change was significantly affected by race/ethnicity (Table 4). Specifically, both Black and Hispanic caregivers were more likely to engage in an exchange of caregivers than were Whites.
Socioeconomic Contexts
The second set of predictors of network change was socioeconomic contexts, measured by the parent's education and the family's SES when the adult child was growing up. Based on higher status parents' better access to formal care and their ability to reciprocate, we expected less network change for parents with higher SES or education (Hypothesis 4). This hypothesis received no support for any change in caregivers (Table 3). However, exchange of caregivers and change in the primary caregiver were less common among higher educated parents than among less educated parents (Tables 4 and 5).
Familial Contexts
Indicators for familial contexts were the number of adult children's brothers and sisters and the number of caregivers at Time 1. We expected more change in care networks with increasing sibship size and in larger care networks. Furthermore, we assumed that the number of sisters would have a greater influence on network change than would the number of brothers (Hypothesis 5). The data supported this hypothesis. As far as change in any caregiver was concerned (Table 3), both number of living brothers and number of living sisters were independently positively related to change, as was the number of adult child caregivers at Time 1. Furthermore, number of brothers had a less pronounced effect than did number of sisters (odds 1.10 vs 1.24, respectively). However, the effects of number of siblings and caregivers differed in terms of type of network change (Table 4). Addition and exchange of caregivers were primarily determined by availability of alternative caregivers (number of brothers or sisters), whereas number of caregivers enhanced the likelihood of dropping caregivers from the network. Both number of sisters and number of caregivers (but not number of brothers) also enhanced the probability of change in the primary adult child caregiver (Table 5), demonstrating again the greater importance of sisters as alternative caregivers. Thus, larger sibships and especially larger female sibships seemed to take advantage of their opportunity for caregiver additions and exchanges, whereas large caregiver networks seemed to shrink over time.
Caregiver Abilities and Resources
Assuming that competing responsibilities of caregivers would encourage network change, we expected networks with younger caregivers and a greater proportion of married caregivers to be less stable than networks with older caregivers or a smaller proportion of married caregivers (Hypothesis 6). This hypothesis received mixed support. Networks that included both married and nonmarried children were more likely to experience some change in caregivers than were networks composed of all-married or all nonmarried caregivers (Table 3). Such networks tended toward dropping and exchanging caregivers (Table 4). In addition, networks composed solely of married siblings were more likely to drop caregivers than were networks containing only nonmarried adult children. However, marital status of the primary adult child caregiver was not significantly related to change in primary care (Table 5). Age of caregivers did not exert a significant influence on any change in caregivers or in type of change (Tables 3 and 4), but older adult children were more likely to maintain their role as primary caregiver than were younger adult children (Table 5).
Caregiver Burden
The fifth set of predictors consisted of indicators of caregiver burden as available in the HRS. In line with our general expectation that higher caregiver burden would elicit more change in care networks, we expected that care networks would be less stable if there were care commitments to both parents; if parents also received financial assistance; or if parents were older, could not be left alone, or had died between waves (Hypothesis 7). The data showed that change in any caregiver prevailed if adult children also provided financial help to a parent, if they cared for another parent, and if they were involved in end-of-life care (Table 3). Provision of financial support as well as end-of-life care were tied to a greater likelihood of dropping caregivers from the network, whereas caring for another parent led to caregiver additions (Table 4). In contrast, care for parents who could not be left alone reduced the probability of change in any caregiver (Table 3) and especially the likelihood of adding siblings to the network (Table 4). However, none of the caregiver burden indicators were significantly associated with change in the primary caregiver (Table 5).
In addition, we also examined whether parental coresidence reduced the likelihood of network change (Hypothesis 8). Our data showed no effect of this predictor on any of the change variables.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Our findings provide some important insights that can hopefully guide future research. We found considerable change in care networks. More than one fourth of primary adult child caregivers were replaced by one of their siblings during any two-wave interval, similar to findings reported in other research (Jette et al., 1992); 33.3% of all networks added at least one adult child during the same time period, and 34.5% dropped at least one adult child. As evidenced by the considerable extent of change in adult child care networks, we clearly captured phenomena that apply to a considerable number of caregiving families.
Our theoretical framework addressed five broad categories of influence factors on network change (i.e., cultural, social, and familial contexts; caregiver abilities and resources; and care burden), which we tested in terms of eight specific hypotheses. In terms of cultural contexts, we assessed the influences of caregivers' and parents' gender as well as of race/ethnicity (Hyotheses1–4). As far as caregivers' gender is concerned, the data suggest greater stability of female primary caregivers as well as of female-only networks than of male networks. In addition, our finding that mixed-gender networks are more prone to drop caregivers over time is in line with earlier studies indicating a feminization of care networks over time (Stoller, 1990). The data did not support our expectations pertaining to cross-gender care, suggesting that the gender matching of caregivers and care recipients may be more complex than previously assumed. In regard to race/ethnicity, we found higher probabilities of exchange of one caregiver for another among African Americans and Hispanics than among Whites. This result may be suggestive of greater coordination in the networks of minorities, but we do not have information about care from other relatives, which some research has shown to be important (e.g., Silverstein & Waite, 1993). Overall, then, the gendered mandate for participation in care seems to be the strongest cultural factor influencing change in care networks.
Assessing influences of socioeconomic contexts, we found that exchange of one caregiver for another and changes in the primary caregiver were less likely for parents with better education, but we found no other effects of parents' SES. Thus, Hypothesis 4, which suggested that control of greater resources would predict more stable care networks, was not supported beyond parents' education. This finding may reflect greater access to alternative (paid) sources of support that render long-term care more feasible via knowledge or other resources. Although it is conceivable that an exchange motive (Cox & Rank, 1992) also comes into play, previous research has found little support for this (Altonji et al., 1996).
In terms of familial contexts, we expected (Hypothesis 5) that change would be more common in larger sibling groups, with the number of daughters having a stronger effect than the number of sons. We found strong evidence that number of siblings in general and number of daughters in particular increased the probability of additions and exchanges among network members and of changes in the primary caregiver, whereas number of caregivers increased the potential of dropping caregivers from the network. This is perhaps the strongest evidence for both linked lives and family systems feedback processes. These findings suggest that siblings may join the care network when needed either by providing additional help or by replacing a too burdened or otherwise no-longer-able sibling. Thus, decisions made by one sibling have implications for the involvement of the other siblings as well as for the composition of the parent's support network.
The shrinking of large care networks over time suggests, in addition, that the benefits involved in distributing care obligations among many adult children sometimes are outweighed by the costs of coordinating and managing such complex care arrangements. Taken together, these findings support the life course theory notion that contexts influence network change. We found strong support for the cultural (gender) and family embeddedness of such change; however, we found only weak support for the influence of socioeconomic contexts, perhaps due to measurement deficiencies.
We also found some support for Hypothesis 6, regarding caregiver abilities and resources, that the age and marital composition of the sibling group would have implications for the stability of care networks. Although the assessment of these influences was necessarily indirect, it did confirm the potential influence of other familial roles on individual family members' ability to provide assistance to older parents (Stephens & Franks, 1995). In line with the life course assumption of linked lives and the system theory concept of feedback processes, these findings also demonstrate that the distribution of responsibilities among network members shifts to accommodate conflicting obligations of some network members over time.
One must assume that parents' need for assistance (or, in life course theory language, the care context itself) ultimately drives the structure and stability of care networks. Unfortunately, many of our indicators of need were indirect. We expected in Hypothesis 7 that situations associated with greater need for assistance would be associated with less stability. Except in the case in which the parent could not be left alone, which was associated with a lower probability of any change in the network (but also a lower likelihood of adding a new caregiver), this was consistent with our findings. Interestingly, however, not a single care context variable was associated with changes in the adult child providing the most care. Thus, adjustments evolving from the burden of care seem to occur more among peripheral members of the care network than among primary caregivers. Thus, earlier results that indicate that greater caregiver burden does not predict relinquishment of care (Caron et al., 2005; Jette et al., 1992) may only apply to primary caregivers and not to network members in other stages of the care trajectory. Such changes may also be tied to the level of skill required for care. For example, addition of caregivers was less likely for parents who could not be left alone. Hypothesis 8, that coresidence with the adult child would decrease the likelihood of change in care networks, received no support. Perhaps our data were not finely grained enough to capture the dynamics that led up to the selection of this care arrangement before care began. Unfortunately, data limitations prevented us from pursuing this further.
In assessing our results, it is important to consider both sample restrictions and data limitations. Our findings refer only to adult child care and are restricted to families with two or more adult children. Respondents in the HRS are also somewhat older than adult child caregivers in nationally representative studies of caregivers (National Alliance for Caregiving, 1997; Stone, Cafferata, & Sangl, 1987; Wolff & Kasper, 2006), and many entered the study when at least one of their parents was already deceased (fewer than one fifth had a surviving father at Wave 1, and fewer than one half a surviving mother). Also, our results pertain only to care situations that lasted at least 2 years. Based on the HRS sample, this means that close to one third of adult child parent care has the potential for change in adult child caregivers. However, other studies show that changes in care networks may occur even within shorter time spans than the 2-year period covered by the HRS (Lyons et al., 2000).
The HRS remains one of the most important and comprehensive nationally representative data resources for studying issues related to aging. However, in part because of its breadth, we could not assess many of the key constructs for this study in an ideal fashion. Specifically, although we could identify the adult child providing the most assistance to an adult parent, it was not possible to identify the primary caregiver or to identify all individuals providing assistance to the older parent. This study could thus not evaluate important differences, such as differences in the probability of changes among primary versus secondary caregivers.
The retrospective nature of the caregiving items was also an important limitation. Because the questions referred to the time between consecutive interviews, we had to select all predictor variables from the previous wave in order to ensure an appropriate temporal ordering. This issue also limited our ability to track key changes between waves, such as changes in parents' health status, changes in the lives of adult children, and so forth. We also do not know precisely how siblings allocated their assistance to parents. Given these limitations in temporal ordering, the longitudinal results we report should be seen as conservative estimates of longitudinal effects.
In addition, the design of the HRS requires that information about assistance be provided by a single adult child for all of his or her siblings. Although likely to introduce some bias into the measures, this limitation is likely more than offset by the benefit provided by the large and representative nature of the sample.
Not possible with the current data set, future research should shift from the emphasis on individual caregivers to a focus on the composition, characteristics, and dynamics of entire care networks. Some previous research, adopting a qualitative approach, has provided guidance in this regard (e.g., Finch & Mason, 1993). However, current national data sets are poorly equipped to address questions regarding the structure and evolution of care networks over time. Such a shift could provide needed information on underexplored territory such as the specific supports that caregiving family members provide for one another. Research on the intersection between caregiving and marital contexts, for example, has focused more on these roles than on how couples or siblings support one another's activities, or the specific ways in which adult children provide assistance to support a caregiving parent. Such an analytic shift will also require some theoretical reorientation, specifically from rational choice models (Silverstein et al., 2002) toward models based on life course and system theory concepts.
The results of this study have several implications for both practice and policy. Health care providers should be sensitive to the broad and dynamic nature of support networks. In addition to identification of a primary caregiver, it is important to identify the full extent of involvement by all adult children, spouses, and other relatives. This will enable practitioners to assess the vulnerability of care networks if one or more caregivers drop out or if some caregivers are overburdened. Likewise, the practitioner should be sensitive to the possibility that information will need to be disseminated to multiple individuals and especially to those caregivers who require the information for their care tasks. All too often, the caregiver at hand is the one to whom health professionals direct instructions and recommendations, but these caregivers may not be those who will implement the recommended care regiments (e.g., an adult child providing transportation to the doctor's office is not necessarily the one in charge of administering medications). Furthermore, we have demonstrated the dynamic nature of care networks. Network changes are limited not simply to secondary or auxiliary caregivers, but also to the adult child providing the most assistance. Thus, health care providers will need to reassess care networks periodically to ensure that relevant information reaches all concerned caregivers. Our findings also point to the importance of latent networks: Even if care is provided by only one adult child at one point in time, other adult children may be available and willing to enter the care network at a later time point. For this reason, experts should solicit information about the care situation from all relevant family members (e.g., those who potentially could provide care) whenever possible, as each family member's perspective may be different. Similarly, care decisions should be made with input from all current and potential caregivers, and care instructions should facilitate the sharing of information among caregivers (e.g., written instructions or even multiple copies of instructions). Another area where practitioners can make an important contribution is in helping family members identify and activate all potential sources of support, when needed, and in mediating potential conflicts among all (current and potential) caregivers. Indeed, previous research has long recognized the potential value of such a family-system-level intervention with caregivers and their families (Zarit, Anthony, & Boutselis, 1987).
From a policy perspective, our findings have relevance in three areas: the availability of family care and, implicitly, the need for formal care in the future; needed programmatic changes; and health care costs associated with caregiving. Demographers have documented that declines in fertility, the expanding age gap between generations due to postponement of parenthood into midlife, and continuing high divorce rates will lead to a significant decline in the number of potential caregivers for older individuals (Uhlenberg & Cheuk, in press). In view of our findings, this means that many caregivers who can now share care work with other family members may not be able to do so in the future, and they may also be no longer able to find a replacement if they cannot continue care. Consequently, the rising demand for formal care due to the greater number of the baby boomers entering old age in the next two decades will be acerbated by simultaneously shrinking informal caregiver networks.
In order to contain the dramatic increase in formal care costs that this scenario implies, policy and programmatic changes may be necessary to lighten the burden of family caregivers and to broaden informal networks through participation of nonfamily. Compared to those of other Western nations, U.S. policies (unpaid family leave, state-level caregiver supports) provide comparatively little relief to caregivers (Wisensale, in press). Such policies and programs will have to be expanded to address the potential future shortage in informal care. New policies could include direct financial supports to caregivers, paid leave, expansion of workplace programs such as leave or flexible work time to a wider range of potential caregivers than just close kin, or financial incentives such as tax reductions or subsidies to nonfamily.
A related policy issue concerns increases in costs associated with caregiving itself. Caregiving can have long-term detrimental health outcomes (Schultz & Beach, 1999), and these outcomes are likely to increase as care networks—and thus the potential of sharing or alternating care among network members—shrink. Because delayed fertility will likely lead to caregiving at younger ages, this will mean an increase not only in direct health-care-related costs but also in costs to employers. Again, expansion of policies and programs that support caregivers and broaden care networks will be essential to avert such outcomes.
Our research underlines the need to view caregiving from a dynamic life course and family systems perspective. Adult child caregivers enter and leave the care network in response to cultural mandates, socioeconomic and family structural contexts, their own and their siblings' obligations and resources, and the changing demands of the care situation. These dynamics will not be captured unless the focus of caregiving research shifts from an emphasis on current, primary caregivers to consideration of long-term fluctuations among all caregivers. Understanding these dynamics is essential to meeting the care needs of the large cohorts of older adults in the coming decades.
|
|
Footnotes |
|---|
The Health and Retirement Study data are collected and managed by the University of Michigan's Institute of Social Research and funded by National Institutes of Health. Angela T. Rouse and Jill E. Neagle participated in running the data analyses, developing the tables, and compiling the references.
1 Gerontology Institute, University of Massachusetts Boston.
2 College of Health Professions, Temple University, Philadelphia, PA.
Decision Editor: Linda S. Noelker, PhD
Received for publication March 22, 2006. Accepted for publication February 5, 2007.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  R. A. Ward Multiple Parent-Adult Child Relations and Well-Being in Middle and Later Life J. Gerontol. B. Psychol. Sci. Soc. Sci., July 1, 2008; 63(4): S239 - S247. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Savla, D. M. Almeida, A. Davey, and S. H. Zarit Routine Assistance to Parents: Effects on Daily Mood and Other Stressors J. Gerontol. B. Psychol. Sci. Soc. Sci., May 1, 2008; 63(3): S154 - S161. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. C. Talley and J. E. Crews TALLEY AND CREWS RESPOND Am J Public Health, November 1, 2007; 97(11): 1931 - 1932. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
