| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
RESEARCH ARTICLE |
School of Social Work, University of Michigan, Ann Arbor.
Address correspondence to Lydia Li, School of Social Work, University of Michigan, 1080 S. University, Ann Arbor, MI 48109-1106. E-mail: lydiali{at}umich.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Caregiving and bereavement are often interrelated, as most deaths of older persons in America occur after a period of chronic illness, disability, and family caregiving (Bass, Bowman, & Noelker, 1991
; Schulz, Newsom, Fleissner, Decamp, & Nieboer, 1997). The literature, however, tends to treat these events separately. Studies of family caregiving have largely focused on the time when caregiving is active, whereas studies of bereavement usually focus on adjustment after the death (Schulz et al., 2001). Rarely is the same sample of individuals followed from caregiving through bereavement. In this study, I view caregiving and bereavement as an ongoing process that exists in a chronic context (Pearlin, 1989). The study has two aims: to examine how depressive symptoms of wife and daughter caregivers change in the transition from caregiving to bereavement, and to investigate whether the depressive symptom trajectories of caregivers during this transition vary by their predeath caregiving stress and social support, and their kinship relationship with care recipients and income.
Bereavement Models
The death of a close relative is one of the most stressful life events (Holmes & Rahe, 1967). Studies of the general bereaved population, however, tend to concur that the mental health effect of bereavement attenuates over time; few bereaved persons experience prolonged adverse effects (Norris & Murrell, 1990). For those who have cared for people before their death, two models regarding the adjustment process have been proposed, both based on the stress and coping paradigm (Bass & Bowman, 1990; Schulz et al., 1997). The depletion model suggests that caregiving is a chronic stressor that depletes personal and social resources. Therefore, in facing another stressor (bereavement), caregivers are not equipped to cope and are vulnerable to negative consequences. The relief model also asserts that caregiving is stressful; however, it proposes that the death of care recipients provides stress relief to caregivers and allows them to reengage in other roles and activities, which ease their adjustment to bereavement.
Empirical studies on caregivers' change in mental health following their care recipients' death have been inconclusive. Some found that caregivers do not become less depressed after bereavement; rather, their predeath levels of depression continue well into the postloss period, with a trajectory similar to that of continuing caregivers (Bodnar & Kiecolt-Glaser, 1994). Others found that caregivers' depressive symptoms increase in the months immediately after the death and then decline at varying rates over time (Chentsova-Dutton et al., 2002; Mullan, 1992; Schulz et al., 2003).
The different trajectories of depressive symptoms found in caregivers after the death of their care recipients might in part be due to study differences in timing, interval, duration, and number of observations. Caregivers' experience of distress is likely to vary during the time before and after care recipients die. The direction and extent of change may therefore depend on when preloss and postloss assessments are conducted. The interval between assessments and the duration of observation matter, because too long an interval may cause the observer to miss changes that are transient, whereas too short an observation period leaves the observer unable to detect long-term changes. Finally, using two time points to analyze change assumes linearity, which may not capture the dynamics of caregiver depression. Thus, my primary aim in this study is to examine changes in caregivers' depressive symptoms during the bereavement transition by using a design and methodology that overcomes limitations of prior studies.
Diversity Within Caregivers
Caregivers' response to bereavement is likely to vary, given their variation in responding to caregiving (Pinquart & Sorensen, 2003). On the basis of the caregiving stress process model proposed by Pearlin, Mullan, Semple, and Skaff (1990), in this study I examine the effects of three sets of factors—caregiving stress, social support, and caregiver background characteristics—on caregivers' trajectories of depressive symptoms during the transition from caregiving to bereavement.
Caregiving Stress
The term caregiving stress refers to conditions, experiences, and activities arising from caregiving that are problematic (Pearlin et al., 1990). According to the model by Pearlin and colleagues, caregiving stress has both objective and subjective aspects. Objective caregiving stress refers to stressors that stem from care recipients' needs and the associated care demand, such as care recipients' problematic behavior and limitations in basic and instrumental activities of daily living (ADLs and IADLs, respectively; these include, e.g., bathing, dressing, and housekeeping). Subjective caregiving stress refers to caregivers' subjective experience of hardships in the caregiving role, such as feelings of overload and burden.
From the perspective of the aforementioned relief model, caregivers who experience high stress during caregiving should feel great relief when they exit the caregiving role; hence, their mental health recovery should be faster than low-stress caregivers. In contrast, the depletion model suggests that more intense stress during caregiving depletes more coping resources, so caregivers have more difficulties adapting to bereavement. Neither model specifies whether subjective and objective stress operate the same way.
Previous studies investigating the effects of caregiving stress on caregivers' bereavement adjustment have mostly focused on subjective stress, and the results are mixed. Schulz and colleagues (2001), for example, found that spouse caregivers reporting more strain associated with caregiving are more likely to improve in health risk behaviors and have no further increase of depressive symptoms following their spouse's death, compared with spouse caregivers with low strain and with spouse noncaregivers. Skaff, Pearlin, and Mullan (1996) found that bereaved caregivers with higher levels of role captivity during caregiving have greater increase of mastery after their care recipients die. Bass and Bowman (1990), however, found that caregivers reporting more difficulties in caregiving also report more difficulties in bereavement. Bernard and Guarnaccia (2003) found that for husband (but not daughter) caregivers, greater prebereavement caregiver strain predicts worse bereavement adjustment. The latter two studies, however, did not control for caregivers' predeath distress, which may confound the findings. Few studies have examined the effect of objective caregiving stress on bereavement adjustment. One exception is the research by Mullan (1992), who found that ADL assistance provided by caregivers predicts increased depressive symptoms of caregivers after their recipients' death.
Social Support
Social support has been conceptualized as a resource that has direct and indirect benefits to caregivers (Pearlin et al., 1990). Various dimensions of social support have been used to predict bereavement outcomes and well-being in the literature. Research consistently shows that perceived support, such as satisfaction with support, is the strongest predictor (Bass & Bowman, 1990; Krause, 2001). A prior study found that more satisfaction with the support received during caregiving predicts less difficulty experienced in bereavement by caregivers, and that satisfaction with caregiving support is a better predictor than satisfaction with bereavement support for caregivers' bereavement adaptation (Bass et al., 1991).
Caregivers' Background Characteristics
It has been suggested that the influence of caregivers' key characteristics be threaded through the entire stress process (Pearlin et al., 1990). In this study, I examine two caregiver characteristics—kinship and income. Kinship matters because different role relationships involve different expectations, obligations, and identities. Wife and daughter caregivers, for instance, vary greatly in their obligation to care and the meaning of loss when their care recipients die. Wives should be more psychologically reactive to caregiving and bereavement than daughters because of the relative salience of the role of wife to women (Seltzer & Li, 2000).
Caregivers' income is an indicator of their socioeconomic status (SES). Prior research has shown that individuals with low SES are more susceptible to negative effects of caregiving (Roth, Haly, Owen, Clay, & Goode, 2001) and bereavement (Arbuckle & de Vries, 1995; Norris & Murrell, 1990), and caregivers with limited income are likely to decrease in mastery over time (Skaff et al., 1996).
Research Questions and Hypotheses
I ask two questions in this study.
First, how do depressive symptoms of wife and daughter caregivers change during the course of the transition from caregiving to bereavement? The relief and depletion models predict different patterns of change in the depression of caregivers following their care recipients' death, with the former suggesting a decline from high levels of depressive symptoms and the latter a continuation or elevation of previous levels. Overall, the literature provides more support for the relief hypothesis (e.g., Mullan, 1992; Schulz et al., 2003). Changes in caregiver depression before the death are less clear (George, 2002). An increase in caregivers' depressive symptoms from about 4 months before to the time of the care recipients' death has been reported (Schulz et al.). Hence, I expect the trajectory of caregivers' depressive symptoms during the course of caregiving and bereavement to be curvilinear, with symptoms increasing as care recipients are closer to death and decreasing thereafter.
Second, how do objective and subjective caregiving stress, social support, kinship, and caregiver income affect caregivers' trajectories of depressive symptoms? On the basis of the relief model and prior work (Schulz et al., 2001; Skaff et al., 1996), I hypothesize that caregivers who experience high levels of subjective caregiving stress have a greater sense of relief on exiting the caregiving role, which leads to faster recovery, indicated by steeper decline in depressive symptoms. The literature does not provide enough information to specify a hypothesis regarding the effects of objective caregiving stress on caregivers' bereavement adjustment; therefore, this issue is an exploratory one in this study.
Social support has been found to ease the bereavement transition (e.g., Bass et al., 1991; Skaff et al., 1996). I hypothesize that more satisfaction with the support received during caregiving predicts faster decline in caregiver depression following the care recipient's death. Wife caregivers are expected to experience more difficulties adjusting to bereavement (i.e., slower decline in depressive symptoms after the death) than daughter caregivers, because of the greater impact of bereavement on the identity and structure of life for wives. On the basis of previous research findings (Arbuckle & de Vries, 1995; Skaff et al., 1996), I also expect low-income caregivers to have more difficulties in this transition.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
).
|
Among the 336 caregivers at Wave 1, 149 lost their care recipients as a result of death by Wave 4 and 60 were lost to follow-up (Figure 1). In addition, 8 noncaregivers at Wave 1 became caregivers at a later wave, prior to being bereaved. Hence, a total of 157 women experienced the transition from caregiving to bereavement during the four-wave study period. These women constituted the sample for the present analysis.
The sample of individuals consisted of 53 wife and 104 daughter caregivers. Virtually all were White (96.2%). At baseline (Wave 1), they averaged 63.6 years old (SD = 9.0). Most had a high school or more education (83.4%). Their care recipients averaged 83.1 years old (SD = 8.7) at baseline; the recipients were quite diverse in their reason for needing care, including dementia (n = 31), arthritis (n = 21), heart trouble (n = 17), asthma or emphysema (n = 16), stroke (n =15), diabetes (n = 7), cancer (n = 6), blindness (n = 6), and other disease categories (n < 6 for each). Approximately 20.2% of the daughter caregivers cared for their fathers. The bereaved caregivers were comparable with the continuing caregivers and those who were lost to follow-up (no difference in Wave 1 measures of depressive symptoms, kinship, income, education, satisfaction with support, overload, burden, duration of care, dementia diagnosis, and problematic behavior of care recipients). However, relative to continuing caregivers and those lost to follow-up, the bereaved caregivers cared for elders with more functional limitations and who were older.
Measures
Outcome variable
The outcome variable was depressive symptoms, which I measured by use of the 20-item Center for Epidemiological Studies–Depression scale (CES-D; Radloff, 1977). The CES-D asks respondents to report symptoms of depression in the previous week on a 4-point scale ranging from rarely to most of the time. I computed a scale score, with higher scores indicating more depressive symptoms. Cronbach's alpha reliabilities of CES-D ranged from 
=.85 to =.88 throughout the four waves. Because of a skewed distribution, I transformed the variable (natural log of 1 + raw CES-D score; I added 1 because some respondents scored 0 on CES-D).
Predictor variables
The predictor variables included three sets—caregiving stress, social support, and caregiver background characteristics. I measured most predictors (except kinship) at the wave before and closest to the death, which occurred.82 years (SD =.42) before the care recipients died, on average. As I subsequently explain in the Data Analysis section, the sample varied in the number of measurement points before and after the death. Each respondent, however, had at least one assessment before the death, which made it possible for me to examine the effects of predeath measures on changes in caregiver depression during the bereavement transition. Table 1 displays the ranges, means, standard deviations, Cronbach's alpha reliabilities, and correlations of all predictor variables.
|
). Caregivers rated each item from 0 (never) to 2 (usually). I measured functional limitations by use of a modification of the Barthel Index (Mahoney & Barthel, 1965), which assesses care recipients' performance in 14 ADL and IADL areas. Caregivers rated each from 0 (independent) to 2 (not at all). The scale scores of problematic behavior and functional limitations were the sum of their respective items. Dementia diagnosis referred to whether the care recipient had Alzheimer's disease or other types of dementia (coded 1). Research shows that caring for older adults with dementia is more challenging than caring for those without it (Ory, Hoffman, Yee, Tennstedt, & Schulz, 1999). Duration of care indicates the length of exposure to care-related stressors, measured by number of years that the caregiver had provided care to her care recipient before his or her death.
Subjective stress was indicated by role overload and burden. I assessed overload by use of a four-item scale that asks about the caregivers' experience of being overwhelmed by care-related responsibilities, such as feeling exhausted at the end of the day and having more things to do than one can handle (Pearlin et al., 1990). Each item was rated from 1 (not at all) to 4 (completely). I measured burden by use of the Zarit Burden Interview (Zarit, Reever, & Bach-Peterson, 1980), which assesses problems frequently faced by caregivers, including health, finances, and social life. I computed scale scores for overload and burden, with higher scores indicating higher stress levels.
Social support
Support was indicated by caregivers' satisfaction with the support received from their personal networks, assessed by a single item with four response categories (0 = completely dissatisfied to 3 = completely satisfied).
Caregiver background characteristics
Background characteristics included kinship and income. Kinship was a dummy coded variable (wife = 1, daughter = 0). Income referred to total household income of the caregiver, measured in dollars. I transformed the income variable by using the natural log to reduce skew.
Data Analysis
I conducted hierarchical linear modeling (HLM) based on maximum likelihood estimation, using the software developed by Raudenbush, Bryk, Cheong, and Congdon (2000). The repeated measures of depressive symptoms were conceived as being nested within individuals; thus the analysis of change had two levels: within persons (Level 1) and between persons (Level 2).
The HLM analysis proceeded in two stages. First, I examined the intraindividual variability in depressive symptoms over time. In this study, time refers to the length of time before and since the death of the care recipient. I created this by subtracting the date of interview from the date of death, and I measured it in years. Hence, a minus sign of time indicated time before the death, a zero indicated the time at death, and a plus sign indicated time since the death.
In the data set, each caregiver had four measurement points covering about 4.5 years (Waves 1 to 4), from before to after the care recipient's death. However, because care recipients died at different times during the study, the period of time covered by the sample can be 9 years, ranging from –4.5 years (care recipients died right before Wave 4) to +4.5 years (care recipients died right after Wave 1).
At the first stage of the analysis, I compared three Level 1 models, which represented that caregivers' depressive symptoms (a) do not change over time (no time effect), (b) change at a constant rate (linear time effect), and (c) change at a rate that accelerates (or decelerates) over time (quadratic time effect); I constrained the Level 2 model to be unconditional (i.e., no predictors). I used likelihood ratio tests to determine the best model. This analysis would answer the first research question and identify the change parameters that best describe individual change in depressive symptoms over time.
Then, I examined interindividual differences in the trajectory of depressive symptoms by modeling the individual change parameters as a function of the proposed predictors at Level 2. To improve estimation efficiency and construct a model that was parsimonious, I first examined each predictor variable separately; I dropped those with no significant correlation with any change parameters. I then simultaneously entered all significant predictors from the separate analyses to predict each individual change parameter; I retained only significant predictors in the final model.
Except kinship, all other study variables had missing data, with burden having the highest proportion (12.7% missing). I used multiple imputation that involved a regression approach and data-augmentation algorithm to impute these missing values, using the software developed by Schafer (1999). I analyzed three imputed data sets. I obtained a single point estimate by averaging across the estimates from the three imputed data sets, and I used a formula to calculate the standard errors (Schafer & Olsen, 1998). I used a level of <.05 for all statistical tests in this study.
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
Individual and Mean Change of Depressive Symptoms
The first HLM analysis examined how the depressive symptoms of the individuals in the study sample changed during the course of caregiving and bereavement. Using likelihood ratio tests to compare three nested models (no time effect, and linear and quadratic effects of time, respectively), the analysis suggests that quadratic change functions best describe intraindividual variation in depressive symptoms over time.
Model A in Table 3 presents the estimated mean in quadratic change functions, and Figure 2 displays the mean depression trajectory of the sample. As shown, caregivers' depressive symptoms exhibited a curvilinear shape of change, increasing as their care recipients were closer to death and decreasing thereafter, which is consistent with the hypothesis.
|
|
Model B shows that, at the time of the care recipients' death, caregivers who felt overloaded, burdened, dissatisfied with support received, and were wives of the care recipients had higher levels of depression. In addition, changes in depressive symptoms during the transition from caregiving to bereavement varied by care recipients' problematic behavior and by caregivers' overload, kinship, and income. To illustrate their effects on caregivers' trajectories of depressive symptoms, Figures 2 and 3 display the adjusted change curves of depressive symptoms according to differences in problematic behavior, overload, kinship, and income.
|
Caregivers who experienced high overload had a steeper increase of depressive symptoms before their care recipients died, and a more rapid decline of symptoms following the death, compared with caregivers with low overload (Figure 2). This change pattern suggests that caregivers who feel overloaded during caregiving are more likely to improve in mental health following their care recipients' death.
Wife and daughter caregivers had very different trajectories of depressive symptoms (Figure 3). The depressive symptoms of wife caregivers accelerated as their husbands were closer to death, and they decelerated after the death. Daughters had lower levels of depression at all times and were relatively stable in depressive symptoms during the course of caregiving and bereavement. I explored the interaction effects of kinship and other predictor variables and found them to be statistically insignificant.
Caregivers with high income were more likely to have a downward trajectory of depressive symptoms that began before their care recipients died and extended to after the death, whereas low-income caregivers were likely to increase in depressive symptoms during caregiving and their high levels of depression continued into bereavement (Figure 3). The longer the time since bereavement, the bigger the depression gap was between high- and low-income caregivers.
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
and suggest that the transition from caregiving to bereavement provides relief rather than poses mental health risks for caregivers.
Deviation from the average trajectory is evident, however. I found care recipients' problematic behavior and caregivers' feeling of overload, kinship, and income to moderate caregivers' change of depressive symptoms during the bereavement transition. The finding related to problematic behavior seems to support the depletion hypothesis. Prior research has shown that problematic behaviors of care recipients have adverse effects on caregiver mental health, in part because they lead to an erosion of psychological and social resources (e.g., mastery, self-concept, and social support; see Aneshensel, Pearlin, Mullan, Zarit, & Whitlatch., 1995; Li, Seltzer, & Greenberg, 1999; Skaff & Pearlin, 1992). The diminishment of these resources may be the reason for the finding that caregivers of elders with more problematic behaviors recover slower during bereavement. Such findings suggest that caregiving to elders with problematic behaviors might have long-term negative effects on caregivers' psychological well-being.
The finding related to overload, however, supports the relief model. Although overload has been found to predict caregiver depression during active caregiving (Alspaugh, Stephens, Townsend, Zarit, & Greene, 1999); its adverse effects may be more role specific and limited to the time when one is an occupant of that role. Research shows that overload remains stable during caregiving but decreases dramatically following care recipients' death (Aneshensel et al., 1995). The relief felt by caregivers might have therapeutic effects; for instance, it might evoke a new sense of control over their lives (Skaff et al., 1996). Caregivers who have felt more overloaded during caregiving may free up more time and energy to attend to neglected needs and roles after caregiving ends, which facilitates adaptation.
Following the death of their care recipients, wife caregivers had a steeper decline of depressive symptoms than did daughters. Although this finding is unexpected, it is understandable considering the high depression level of wives before the death. Relief, however, does not mean freedom from grief (Skaff et al., 1996). When their husbands die, wife caregivers may experience competing feelings of grief and relief. Low-income caregivers are more likely than high-income caregivers to endure high levels of depressive symptoms from before to after the loss of their care recipients. The vulnerability of low-income caregivers may be related to the fact that poverty itself is a chronic stressor that takes a toll on mental health, in addition to diminishing caregivers' coping capacity.
This study contributes to the literature in four significant ways. First, in this study I conceptualized caregiving and bereavement as an ongoing process and examined changes in caregivers' depressive symptoms in this process. I utilized multiple measurement points spanning caregiving to bereavement, and the analytic approach incorporated their information simultaneously in the same model rather than simply capturing changes between pairs of preloss and postloss measurements. Second, using the HLM approach, I maximized the use of available data and examined the patterns of change in caregivers' depressive symptoms over a relatively long period of time. To my knowledge, this is the first study that has taken such a long view to understand the dynamics of caregivers' depressive symptoms during the bereavement transition. Third, the findings regarding problematic behavior and role overload have not been reported before. Although future studies are needed to validate whether and understand why problematic behavior and overload have differential effects on caregivers' bereavement adjustment, research in this direction may shed light on our understanding of the long-term consequences of objective and subjective stress. Fourth, this study shows that the social standing of the caregiver, including income and kinship, is relevant to the caregiver's emotional adaptation during the bereavement transition. These findings illustrate the importance of social structural factors in psychological processes.
The findings have practice implications. First, bereavement support services to caregivers should target vulnerable groups, including those with low incomes and those who have cared for recipients with high levels of problematic behavior. Second, the year before care recipients die seems to be particularly distressing for caregivers. Services to help caregivers cope with the dying process may be needed. Third, some caregivers, including wife caregivers and caregivers who feel overloaded, have a great need for supportive services when they are active caregivers.
Limitations of this study include a virtually all-White sample of individuals, a relatively small sample size, and the lack of a comparison group. The sample composition limits generalization of the findings to non-White populations. The limited sample size, coupled with the approach used to select predictors into the final model, warrants caution when the findings are interpreted. Cross-validation of the findings on independent samples is needed. The lack of a comparison group makes it less clear whether the patterns and predictors of change found in this study are unique to caregivers and the bereavement transition (Amirkhanyan & Wolf, 2003). Also note that the care recipients in this sample had diverse diseases, which could affect caregivers' depression trajectory. The small number of care recipients in each disease category made it difficult for me to examine the effects of disease type in this study.
To conclude, a particularly important contribution of this study is that it describes a relatively long-term trajectory of caregivers' depressive symptoms during the transition from caregiving to bereavement, using prospective data and multiple measurement points. It shows that caregivers respond differently during the final stage of caregiving. Further research on individual differences in responding to the bereavement transition may illuminate the stress process and help in the design of interventions to support caregivers.
|
Acknowledgments |
|---|
I am grateful to Dr. Seltzer for sharing the data, Drs. Seltzer and Jan Greenberg for comments on an earlier version of this article, Ms. Natalya Verbitsky for statistical consultations, and Ms. Terri Torkko for editorial assistance.
|
Footnotes |
|---|
Received for publication September 10, 2004. Accepted for publication February 2, 2005.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
