The Stress Process of Family Caregiving in Institutional Settings

Purpose: This study adapts the Stress Process Model (SPM) of family caregiving to examine the predictors of depression in a sample of caregivers ( n (cid:2) 133) with demented relatives residing in suburban skilled nursing facilities. Design and Methods: We interviewed family caregivers of family members residing in skilled nursing homes using a variety of measures to assess primary stressors, secondary strain, nursing home stressors, and caregiver depression. We used block-wise regression analyses to determine the predictors of caregiver depression. Results: Results indicated that positive resident adjustment to placement was best predicted by the closeness of the resident–caregiver relationship and nursing home stressors. Caregiver strain, resident adjustment, and nursing home stressors best predicted caregiver adjustment. In turn, the best predictors of caregiver depression included caregiver age, caregiver adjustment to the nursing home, and nursing home stressors. Implications: These results suggest that caregiver depression is closely linked to how well both the resident and caregiver adjust to the nursing home environment. Results also indicate that by broadening the SPM to include stressors common to the nursing home experience, researchers will be able to understand more clearly the specific components of the stress process that may lead to depression in family caregivers of persons living in nursing homes.

Over the years, research efforts have led to greater specification of mental health indicators for in-home caregivers, yet few studies have examined these indicators for caregivers who have relatives living in long-term care environments. Early research suggests that most caregivers do not abandon their relatives after placement (Pratt, Schmall, Wright, & Hare, 1987;Smith & Bengtson, 1979), but remain involved with their relatives in a variety of ways (Chenoweth & Spencer, 1986;Rosenthal, Sulman, & Marshall, 1993;Townsend, 1990). Family caregivers continue to provide a great deal of hands-on care for and remain emotionally involved with their relatives after placement (Zarit & Whitlatch, 1992). They visit often, help with feeding and personal care, and work with nursing home staff to ensure their relative obtains the best care possible. Only a handful of studies have examined how the experiences of placing a relative affect the mental health of family caregivers, such as increased feelings of guilt and depres-sion. Family caregivers who have recently placed a demented relative report relatively high levels of depression, anger, and anxiety and many continue to experience these negative consequences up to three years after placing their relative (Zarit & Whitlatch, 1992). These same caregivers, however, also report feeling less overwhelmed and tired (Zarit & Whitlatch, 1992) than their in-home care peers. Placement seems to have a selective effect on caregiver outcomes suggesting that caregiver stress and emotional well-being must be thoroughly evaluated within the context of the various transitions experienced by family caregivers.
Research by Pearlin and colleagues (Aneshensel, Pearlin, Mullan, Zarit, & Whitlatch, 1995;Pearlin, Mullan, Semple, & Skaff, 1990;Zarit & Whitlatch, 1992) indicates that the Stress Process Model (SPM) of family caregiving is useful in the conceptualization of stress proliferation and containment for both inhome caregivers and those who continue to provide care after placing their relative. Briefly, this model proposes that there are a host of stressors specifically associated with family caregiving. As described by Aneshensel and associates (1995), caregiving stressors are "the problematic conditions and difficult circumstances experienced by caregivers" which "exceed or push to the limit" the caregiver's ability to adapt (p. 34). These stressors are influenced by a variety of background and contextual characteristics of the caregiver and impaired adult (i.e., sociodemographic factors), primary objective and subjective stressors (i.e., impaired adult functional abilities and caregiver overload), and secondary role strains (e.g., work and ր or economic strain, etc.). Together, stressors and contextual factors can proliferate to such an extent that they produce negative caregiver outcomes that are the consequences of caregiving stress.
The SPM is broad in scope and dynamic in nature making it easily adapted to test the course of stress for persons involved in a variety of caregiving situations. In the original presentation of the SPM, Pearlin and colleagues (1990) emphasized the stress associated with providing in-home care to persons with dementia. More recent and longitudinal work has since expanded the model to explain stress proliferation and containment for "placement" caregivers, those facing the death of a demented elder, and caregivers of persons with AIDS (Aneshensel et al., 1995;Mullan, 1992Mullan, , 1998Pearlin, Aneshensel, & LeBlanc, 1997;Zarit & Whitlatch, 1992). The present study further expands the SPM by introducing stressors associated with the changing nature of the relationship between the family caregiver and impaired adult in institutional settings.

The SPM for Family Caregivers of Nursing Home Residents
The process of placing a family member into a long-term care setting brings additional stressors and strains to the lives of caregivers. Stressors common to placement involve the new relationships and interac-tions that must be negotiated with staff, residents, and families of other residents. A wife who finds it necessary to place her demented husband must interact with staff and administrators to insure her husband receives the best care possible. She develops relationships with other residents and their families. These interactions have the potential to be positive and ր or negative. As well, placement leads to role changes for the family caregiver. The caregiver must relinquish her role as the primary provider of her husband's personal care. She may continue to provide limited hands-on care for her husband, but may find that there are new roles she is expected to fulfill as well. If her husband has difficulty adjusting to his new environment (e.g., becomes more disoriented, refuses to eat), she may experience increased stress and compromised emotional well-being. These caveats illustrate the many changes and resulting stress that a caregiver may experience once a relative moves to an institutional setting. We have incorporated these additional stressors into an expanded SPM that broadens the conceptualization of primary subjective stressors and secondary role strains and adds the dimension of nursing home stressors in order to reflect the experience of "placement" caregivers.
First, as seen in Figure 1, in addition to caregiver strain, we have broadened the conceptualization of primary subjective stress to include a dimension of the emotional closeness of the resident and family caregiver. In addition, we have included a measure of the caregivers' subjective appraisal of their own difficulty in dealing with the relative's mental and emotional state. These two additional stressors reflect the "restructured relationship between caregiver and patient" (Pearlin et al., 1990, p. 588) that may occur as a result of nursing home placement.
Next, we have responded to anecdotal evidence suggesting that caregivers are keenly aware of their relative's reaction and adjustment to living in a nursing home, regardless of the relative's level of cognitive impairment. We have conceptualized this awareness of resident adjustment as a secondary role strain for a number of reasons. First, caregivers often feel responsible for their relative's comfort, happiness, and satisfaction with the new living arrangement. Feeling responsible for their relative's well-being and happiness is especially significant if the caregivers believe that the move to a nursing home was somehow against the wishes or preferences of the relative. Often, we see caregivers assuming new roles, which require them to do their best to ensure their relative's well-being and happiness. Therefore, we have added a measure of the caregivers' perception of "how well" their relative has adjusted to the nursing home environment. If caregivers believe their relative is not adjusting well, caregivers are likely to feel great pressure to remedy the situation thereby increasing the potential for additional role strain in the caregivers' lives. In addition, caregivers' perceptions of their own adjustment to the nursing home and the new roles they assume are indicators of the stress they feel. Thus, we have added two secondary role strains to the SPM to account for the stress associated with the adjustment of both residents and caregivers to the nursing home.
Third, consistent with the original SPM's emphasis on the context of caregiving, we have added the domain of nursing home stressors. As noted, the potential stress of placement takes many forms for caregivers including visiting the facility, interactions with others (e.g., the impaired relative, staff, and families of other residents), and activities while visiting (e.g., helping to feed the relative). These stressors are likely to influence how well both the impaired resident and family caregiver adjust to the nursing home environment and, in turn, contribute to caregiver outcomes, specifically, caregiver depression.

Research Questions
The present study broadens the SPM of family caregiving for use with a sample of caregivers of demented family members residing in nursing homes. By examining "placement" caregivers, a group whose well-being has the potential to be compromised, we are able to explore the particular mechanisms of the stress process in order to understand the various contexts within which caregiving stress occurs. Specifically, the present study focuses on the influence of resident and caregiver adjustment to placement and how this adjustment predicts depression in family caregivers. Our work addresses three important re-search questions: (a) What are the stressors associated with family caregiving in a long-term care environment (e.g., problems or dissatisfaction with staff, services, etc.)? (b) How are these nursing home stressors related to resident and caregiver adjustment to placement? (c) How do resident and caregiver adjustment and nursing home stressors contribute to depression for caregivers of institutionalized family members?

Study Sites
We obtained cross-sectional data from in-person interviews with family caregivers of residents with diagnosed dementia. The five study sites were skilled nursing facilities located in suburban areas of Cleveland, Ohio, that ranged in size from 82 to 203 beds. We selected these facilities because they primarily serve residents with Alzheimer's disease or other types of dementia. Two of the facilities were philanthropic, and three were proprietary; four of the five included special dementia care units. The data presented here are from a larger study that also included in-person interviews with 114 nursing assistants of the residents to examine factors affecting their job satisfaction and the quality of their relationship with families (Looman, Noelker, Schur, Whitlatch, & Ejaz, 1997; Noelker, Schur, Looman, Ejaz, & Whit- Figure 1. Expanded stress process model. ADLs ϭ activities of daily living; NA ϭ nursing assistant; NH ϭ nursing home. latch, 1996;Schur, Noelker, Looman, Whitlatch, & Ejaz, 1998).

Sample
To obtain the sample, each nursing home supplied research staff with a list of residents who met the following criteria: (a) a diagnosis of irreversible dementia, (b) a resident in the current unit for at least 2 months, and (c) a family member listed as the resident's primary responsible party. Participating nursing homes supplied contact information for eligible family members. We sent each family member a letter signed by the facility administrator on nursing home letterhead. The letter briefly described the study and indicated that someone from the research staff would contact families within a few weeks unless they preferred not to be contacted. No family members dropped out of the study at this point. Next, field operation staff screened family members via telephone to determine whether the family members visited the resident at least twice a month and had contact with the nursing assistants. Of the initial 331 caregivers screened, the staff found 101 (30.5%) ineligible (e.g., primary responsible party was not a family member, resident admitted less than two months previously, or responsible party was unable to participate because of the death of their relative), 72 (21.8%) refused to participate, and 25 (7.6%) could not be contacted. The final sample ( n ϭ 133) represents 57.8% of the 230 caregivers eligible to participate in the study. Field operation staff forwarded contact information for family members who met all inclusion criteria and agreed to participate to interviewers for scheduling. Interviewers conducted interviews face-to-face in the caregiver's home or the relative's nursing home from July 1994 through August 1995.

Measures
Structured interviews consisted of close-and openended questions which examined five domains of the SPM: (a) caregiver and resident background characteristics (i.e., sociodemographics), (b) primary objective and subjective stressors related to the resident's functional status, problem behaviors, caregiver strain, and the closeness of the resident-caregiver relationship, (c) nursing home related stressors, (d) secondary role strains in the form of resident and caregiver adjustment to placement, and (e) caregiver depression.
Caregiver and resident background characteristics included age, gender, education, marital status, employment status, caregiver and resident living arrangement preplacement, length of resident's occupancy in the current facility, and caregiver kin group (e.g., wife, husband, and adult child caregiver).
We measured primary objective stressors related to the resident's functional status in two ways. First, we asked caregivers whether their impaired relative was currently able to perform seven activities of daily living (ADLs) including positioning self in bed, getting up from bed and ր or chair, moving about room, dressing self, feeding self, managing own personal care (combing hair, shaving, etc.), and using the toilet. Response options ranged from "none of the time," and "some of the time," to "all or a good part of the time." Possible summary scores ranged from 0 to 14 with higher scores indicating greater physical impairment ( ␣ ϭ .92). Second, we asked caregivers if their family member showed signs of problematic behaviors (e.g., moaning, shouting or yelling, biting or spitting, hitting or slapping or other physically aggressive behaviors). Response options included "none of the time," "some of the time," "a good part of the time" or "all of the time." A summary measure of these items ranged from 4 to 16 with higher scores indicating greater frequency of occurrence ( ␣ ϭ .71).
We measured primary subjective stress in three ways. First, a seven-item indicator of care-related strain assessed the stress a caregiver feels as a result of having a relative in a nursing home. Examples of questions include: (a) "I don't have enough time for myself," (b) "I have more things to do than I can handle," (c) "I feel that my social life is limited," and (d) "I have trouble managing all the demands on my life." Caregivers chose between four response options: (a) strongly agree, (b) agree, (c) disagree, and (d) strongly disagree. Possible scores ranged from 7 to 28, with higher scores indicating greater carerelated strain ( ␣ ϭ .90).
Second, we added a measure of the caregiver's perception of the "closeness of the relationship" between the caregiver and resident. This measure of closeness ( ␣ ϭ .90) included six items: (a) "My relative always understood what I value in life," (b) "My relationship with my relative has always been close," (c) "My relative always made me feel that whatever I did for him ր her was not enough," (d) "My relative always made me feel like a special person," (e) "My relative was often critical of me," and (f) "My relative and I could always discuss things together." Four response options were possible from "strongly agree" to "strongly disagree" with higher scores reflecting a more close relationship between caregiver and impaired relative (range from 6 to 24).
Third, we added a single-item measure of the caregiver's difficulty in dealing with the relative's mental state. This question asked caregivers "How much difficulty do you have with your relative's mental or emotional state?" (no difficulty ϭ 0, some difficulty ϭ 1, a great deal of difficulty ϭ 2). Responses were coded so that higher scores indicated more stressful reactions (e.g., caregiver had a great deal of difficulty with their relative's mental and emotional state).
We expanded secondary role strains to include the adjustment of both the resident and caregiver to the experiences of institutionalization. We added these stressors to our model to reflect the role stress experienced by caregivers as they adjust to the new and shifting responsibilities associated with nursing home care. In addition, we propose that caregivers' perception of how well the relative adjusts to the nurs-ing home will have a profound influence on the role stress experienced by caregivers. The placement of these indicators within the domain of secondary role strains is based on the original SPM, which defines secondary stress as derived from the primary stressors driving the stress process. Thus, the caregiver's adjustment to placement is a product of primary conditions related to the resident's physical abilities and adjustment. Caregivers rated both their relative's level of adjustment to the nursing home as well as their own adjustment. These single item questions reflected adaptation ranging from very poor ր poor, fair, good, and very good ր excellent with higher scores indicating better perceived adjustment (range 0 to 3).
We categorized nursing home stressors into five domains. The first two domains assessed caregivers' perceptions of the nursing assistants' (NAs') attitudes toward them as family members of residents in their care. We asked caregivers to respond to 20 questions by answering either "most of the time," "some of the time," or "hardly ever" with higher scores reflecting that NAs had more positive attitudes toward the family caregivers. Results of factor analysis revealed two distinct measures. The first measure, NAs' communication with residents' families ( ␣ ϭ .73), assessed the caregivers' opinions of how well they were treated by the NAs when they visited the nursing home. This six-item measure included questions such as do NAs (a) "greet you when you visit," (b) "seem glad to see you," (c) "seem interested in learning more about your relative by talking with you," (d) "have all the information they need to care for your relative properly," (e) "respond to your questions promptly," and (f) "act rude or unpleasant." Scores ranged from 2 to 12 with higher scores indicating more positive communication by NAs with residents' family members. The second measure, NAs' support for families ( ␣ ϭ .76), was composed of five items that rated the caregivers' perceptions that NAs provided support and reassurance to family caregivers. Items included statements such as the NA (a) "reassures me that my family member's behavior is not unusual," (b) "helps me know who to contact at the home regarding a problem," (c) "encourages me to talk about my fears and concerns," (d) "keeps me informed about changes in my relative's condition," and (e) "understands that having my family member in a home is stressful for me." Scores ranged from 0 to 10 with higher scores indicating more supportive and reassuring NA relationships.
The third domain included measures that assessed dimensions associated with personal interactions and experiences within the nursing home environment. We asked family caregivers if there were persons at the facility that they confided in, felt frustrated by, made them laugh, etc. Questions included the number of positive and negative interactions family caregivers had with (a) their institutionalized relatives, (b) the staff at the facility, (c) other residents, and (d) the families of other residents. We worded five positive items to determine (a) if caregivers felt they were "able to confide in relative ր staff ր families," (b) if they felt "relative ր staff ր families were capable of reassuring them when they were upset," (c) if relative ր staff ր families "showed respect to families," (d) if relative ր staff ր families "made them laugh," and (e) if relative ր staff ր families were "affectionate." We worded negative items to determine if caregivers felt the relative, staff, or families of other residents made the caregiver feel "frustrated," "angry or provoked," "guilty," or if these persons made them "cry" or "got on their nerves." We coded items dichotomously to reflect whether an interaction had occurred with each of the four possible categories of persons (i.e., interactions with placed relative, staff, other residents, or families of other residents). To illustrate, a caregiver with a score of five positive interactions with staff would signify that the caregiver had experienced all five types of positive interactions with staff (i.e., confiding, reassuring, respect, laugh, and affectionate). Thus, we were able to assess the positive and negative interactions caregivers might have had with a variety of persons involved in the nursing home.
Alphas for the measures of positive and negative interactions are as follows. Interactions with other residents: positive interactions variable discarded because of too few cases, negative interactions ␣ ϭ .79.
Interactions with the families of other residents: positive interactions ␣ ϭ .78; negative interactions variable was discarded because of too few cases. Interactions with facility staff: positive interactions ␣ ϭ .65; negative interactions ␣ ϭ .79. Interactions with institutionalized relative: positive interactions variable discarded because of too few cases; negative interactions ␣ ϭ .80.
The fourth domain of nursing home stress included three measures to assess whether caregivers performed any of 18 specific activities while visiting their family members (e.g., bring food, talk with relative, hold hands with relative, tidy relative's room; ␣ ϭ .66), the caregivers' visiting patterns (e.g., length of time visiting), and frequency of visits (e.g., daily, weekly).
The fifth and final domain assessed how upset the caregivers were that their relative was in a nursing home. This single-item measure questioned caregivers about "how upsetting is it for you to see your relative in a nursing home" (very upsetting, somewhat upsetting, or not upsetting). Responses were coded so that higher scores indicated more stressful reactions (e.g., caregiver felt very upset seeing their relative in a nursing home).
We measured caregiver depression using the Center for Epidemiological Studies-Depression Scale (CES-D; Radloff, 1977). This 20-item, 4-point scale covers the areas of depressed mood, feelings of guilt and worthlessness, feelings of helplessness and hopelessness, loss of energy, and disturbances of sleep and appetite ( ␣ ϭ .88). The CES-D is designed to measure current state and has been used in numerous caregiving studies (Deimling, 1991;Pruchno, Kleban, Michaels, & Dempsey, 1990;Schulz & Williamson, 1991). Scores of 16 or higher indicate significant depressive symptoms (Radloff & Teri, 1986

Analytic Design and Procedures
Given the large number of independent variables we wished to examine, we used a step-wise analysis strategy that entered variables by blocks consistent with the four conceptual domains. We carried over variables that were significant and included them in the next block of variables we entered. To determine the final regression predicting depression, we first considered the effects of secondary role strains (i.e., resident and caregiver adjustment to placement). As seen in Figure 1, secondary role strains are viewed as being influenced by primary and nursing home stressors. Additionally, they have the potential to have an effect on each other and on caregiver depression. In designing our analyses, we chose to examine resident adjustment to placement first because of anecdotal evidence suggesting that caregivers' perception of how well their relative has adjusted to placement may influence how well caregivers react to the placement experience. In other words, we conceptualized resident adjustment as a predictor of caregiver adjustment rather than the reverse.
As shown in the equation BC ϩ PS ϩ NHS ϭ RA, for the first series of block-wise regressions, resident adjustment (RA) was regressed onto three blocks of variables consistent with the conceptual domains of background characteristics (BC), primary stressors (PS), and nursing home stressors (NHS). We carried over the significant predictors of resident adjustment from the first block (i.e., background characteristics) and included them in the next block of variables we entered (i.e., primary stressors). For example, because no background variables were significant predictors of resident adjustment, we did not include any in the next regression that added primary stressors. After we entered all blocks, we determined the model predicting resident adjustment.
Next, we regressed caregiver adjustment (CgA) to placement onto three blocks of variables consistent with the conceptual domains of background characteristics (BC), primary stressors (PS), resident adjustment (RA), and nursing home stressors (NHS; BC ϩ PS ϩ RA ϩ NHS ϭ CgA). Similar to the regression predicting resident adjustment, no background characteristics were significant predictors of caregiver adjustment so that we did not carry over any to the next block that added primary stressors. We carried over significant primary stressors to the next regression that added resident adjustment and nursing home stressors as predictors of caregiver adjustment.
The last blockwise analysis regressed caregiver depression (CgD) onto background characteristics (BC), primary stressors (PS), caregiver adjustment (CgA), and nursing home stressors (NHS; BC ϩ PS ϩ CgA ϩ NHS ϭ CgD). As with the previous models, we included any variables that were significant in the next block of variables to be added to the model. Using this method, we were able to develop systematically a model predicting caregiver depression that started with the influence of background characteristics on primary stressors, the effect of primary stressors and nursing home stressors on secondary role strains (i.e., resident and caregiver adjustment), and the final effects on caregiver depression.

Primary Subjective Stress
On average, caregivers reported moderately low levels of care-related strain (mean ϭ 13.8, SD ϭ 4.4, range 0-23). Turning to the first of two measures of primary subjective stress that we added to the SPM, we found that caregivers reported fairly high levels of closeness (mean ϭ 18.92, SD ϭ 4.36, range 6-24). When asked if they had any difficulty with their relative's mental or emotional state, 33% of caregivers reported they had a great deal of difficulty, 44% had some difficulty, and 23% reported they had no difficulty.

Secondary Role Strains
Caregivers generally felt their relative's adjustment to the nursing home was positive. Specifically, one third responded that the relative's adjustment had been very good or excellent (34%), while the remainder rated the relative's adjustment as good (34%), fair (20%), or poorրvery poor (11%). Similarly, the caregivers' perceptions of their own adjustment to the placement of their relative was found to be relatively positive. One third (34%) reported their adjustment had been very good or excellent, while 32% felt their adjustment had been good, 20% felt their adjustment had been fair, and 14% felt they had adjusted poorly or very poorly.

Nursing Home Stressors
Results for the first and second domains of nursing home stress (i.e., NAs' communication with and level of support of the caregivers), indicated that caregivers felt NAs were generally quite positive with them (mean ϭ 9.9, SD ϭ 2.2, range 2-12). Caregivers felt that NAs greeted them most of the time (86%), hardly ever acted rude or unpleasant (92%), and hardly ever ignored the caregiver (89%). Similarly, caregivers felt NAs were fairly supportive of them (mean ϭ 6.3, SD ϭ 2.8, range 0-10). For example, caregivers felt that most of the time NAs helped solve a questionրconcern the caregiver had (73%), helped the caregiver to know who to contact when a problem arose (69%), understood that it is stressful to have a relative in a nursing home (66%), and reassured the caregiver that the relative's behavior was not unusual (63%).
The third and fourth domains assessed dimensions associated with personal interactions and experiences common to caregivers involved in placement settings (i.e., activities while visiting and positive and negative interactions with others at the facility). Results indicated that caregivers remained very involved in the care of their relatives after placement. Caregivers visited often, with most (45%) visiting two to five times per week while others visited three to four times per month (29%). During each visit to the facility, many caregivers spent 30 min to 1 hr (48%) while others stayed for a few hours (35%). When at the facility, nearly all caregivers (over 90%) talked with their relative, held hands, talked to RNs or LPNs, andրor talked to NAs. Over half (50%-89%) brought others to visit their relative, brought food, helped with personal care, took their relative for an outing, arranged for services, talked to social workers, tidied their relative's room, inventoried their clothing, andրor talked to other residents and their families.
Given the amount of time spent at the nursing home, it is not surprising that caregivers were involved in both positive and negative interactions with others. Positive items reflected interactions with their relative, the staff, other residents, and the families of other residents that allowed the caregiver to confide in someone, to feel reassured or respected, to laugh, or to receive affectionate gestures. Negative interactions included those which made the caregiver cry, feel frustrated, guilty, or angry, or "got on their nerves." Caregivers had an average of 4.79 positive and 1.50 negative interactions with their relative, 6.41 positive and 1.42 negative interactions with facility staff, 1.91 positive and 2.5 negative interactions with other residents, and 2.34 positive and 2.25 negative interactions with other residents' family members. Nearly all caregivers (99%) had at least one positive interaction whereas 69% had at least one negative interaction. Overall, results indicated that caregivers experienced more positive (10.82) than negative (4.51) interactions with their relatives, staff, and others in the nursing home.
The fifth and final dimension of nursing home stress examined how upset caregivers were at seeing their relatives in a nursing home. Nearly one third (30%) of the caregivers reported that they were not upset seeing their relative in a nursing home, nearly half (47%) reported feeling somewhat upset, and 23% were very upset at seeing their relative in a nursing home.

Caregiver Depression
We found levels of caregiver depression to be relatively low for this sample of "placement" caregivers (mean ϭ 9.69, SD ϭ 7.31, range 0-34). Over three quarters (79.8%) were under the CES-D cutoff of 16, reflecting that 20.2% reported symptoms of clinical depression.

Multiple Regressions Predicting Resident Adjustment, Caregiver Adjustment, and Caregiver Depression
Using multivariate ordinary least squares regression, we examined the predictors of resident adjustment to the nursing home, caregiver adjustment to the nursing home, and caregiver depression. As seen in Table 3, resident adjustment to the nursing home was predicted by the closeness of the relationship between the caregiver and resident, and nursing home stressors (R 2 ϭ .28, p Ͻ .0001). This regression analysis indicated that caregivers viewed their relatives as better adjusted if caregivers felt greater closeness to the relative (␤ ϭ .23, p ϭ .008), if caregivers and their relative had few negative interactions (␤ ϭ Ϫ.28, p ϭ .004), if caregivers and other residents had more negative interactions (␤ ϭ .37, p ϭ .000), and if caregivers felt less upset by seeing the relative in a nursing home (␤ ϭ Ϫ.17, p ϭ .099). Better resident adjustment was predicted by nursing home stressors as well. For example, residents were viewed as better adjusted by caregivers who felt the NAs offered more support to families (␤ ϭ .21, p ϭ .018).
Predictors of family caregivers' adjustment (R 2 ϭ .28, p Ͻ .0001) included the resident's adjustment (␤ ϭ .15, p Ͻ .10). In addition, caregivers who had little difficulty with their relative's mental and emotional state were more likely to adjust successfully (␤ ϭ Ϫ.34, p ϭ .001), as were caregivers who felt that NAs displayed higher levels of communication toward them (␤ ϭ .30, p ϭ .007).
Finally, predictors of caregiver depression (R 2 ϭ .20, p Ͻ .001) included caregiver age (␤ ϭ .19, p ϭ .03) and adjustment to placement (i.e., caregivers who had adjusted poorly to the relative's placement were more depressed; ␤ ϭ Ϫ.18, p ϭ .05). Caregivers who reported a higher frequency of negative interactions with their relatives were more likely to be depressed (␤ ϭ .17, p ϭ .04), as were caregivers who reported having difficulty with their relatives' mental and emotional state (␤ ϭ .21, p ϭ .04). In addition, caregiver depression was predicted by negative interactions with the families of other residents (␤ ϭ .22, p ϭ .007; see Figure 2).

Discussion
The present study examined the factors leading to caregiver depression in a sample of families with demented relatives residing in skilled nursing facilities. Our analytic approach demonstrated that the predictors of caregiver depression are connected to various components of the SPM of family caregiving. In the first step of this pathway, variables associated with the interpersonal relationship between the caregiver and resident (i.e., closeness, negative interactions, caregiver being upset seeing the relative in a nursing home) along with stressful characteristics of the nursing home (i.e., negative interactions with others) were predictive of resident adjustment to placement (i.e., secondary role strain). Next, we found that resident adjustment in combination with the caregiver's difficulty in dealing with the relative's mental state and the NAs' communication with the family (nursing home stressor) were predictive of caregiver adjustment to the relative's placement (i.e., secondary role strain). Lastly, the final outcome or consequence of caregiving, caregiver depression, was predicted by caregiver age, negative interactions between the caregiver and resident, the caregiver's difficulty with the relative's mental state, caregiver adjustment, and the caregiver's negative interactions with the families of other residents.
These findings support the application of the SPM in this sample of "placement" caregivers. By adding conceptually relevant measures that reflected primary subjective stress, secondary role strain, and nursing home stress, we were able to demonstrate the utility of the SPM within the context of institutional care. For caregivers of a demented relative living in a nursing home, level of depression is closely linked to how well their relative adjusts to the nursing home environment. In turn, the relative's perceived adjustment is associated with the caregivers' reported adjustment to the nursing home. Moreover, adjustment of the relatives and caregivers is associated with specific nursing home stressors, which, in turn, play a role in the caregivers' level of depression. Thus, the relationship between caregivers and institutionalized relatives in combination with specific nursing home stressors are paramount to the adjustment of both residents and caregivers, which, in turn, are related to the caregivers' level of depression. It is interesting to note that the caregivers in this study had fairly low levels of depression (i.e., mean CES-D scores less than 10). This mean rate of depression is nearly half the level of previously reported CES-D rates for in-home caregivers (Whitlatch, Feinberg, & Sebesta, 1997), yet comparable to other studies of the general adult population (e.g., mean CES-D scores ranging from 8.6 to 9.9; Linn, Dean, & Ensel, 1986). Our sample's CES-D scores were also lower than scores reported in comparable stud- ies of in-home caregivers with reported mean CES-D scores ranging from 14 to 18 (Cattanach & Tebes, 1991;Deimling, 1991;Pruchno et al., 1990). To date, no comparable studies of caregivers that have placed a demented relative have reported mean levels of depression using the CES-D.
Surprisingly, the residents' physical functioning and problem behaviors (primary objective stressors) and the caregivers' visiting behavior (i.e., frequency of visits, activities performed while visiting) had no bearing on the caregivers' adjustment or depression. This finding supports previous research documenting that the change from in-home to institutional caregiving is marked by a decrease in primary stressors related to the impaired adult's functioning and hands-on care provided by the caregiver. Placement may alter the stressors caregivers experience but may also lead to the generation of new and potent caregiving stressors (Aneshensel et al., 1995;Zarit & Whitlatch, 1992).
These results point to the importance of interpersonal relationships and interactions within the nursing home environment. Recall that caregivers reported greater number of positive rather than negative interactions with staff and others in the facilities. Yet, negative interactions were more powerful predictors than positive interactions of resident and caregiver outcomes (i.e., adjustment and caregiver depression). The measures of positive and negative interactions were fairly simple suggesting that there is much to learn about the complex nature and potential stressfulness of interactions within nursing homes. The results indicate that compared to positive interactions, negative interactions have a greater impact on the well-being of both resident and caregiver. While this result may appear obvious, this study is the first to document its occurrence, suggesting that it may be more important to "eliminate the negative" rather than "accentuate the positive" within the context of family caregiving in a nursing home environment.
An interesting and unexpected finding concerns the fact that caregivers who reported more negative feelings from interactions with other residents (e.g., frustration, annoyance, sadness) also reported that their impaired relative had adjusted better to the facility. It is possible that, within the context of having negative feelings from interactions with other residents, caregivers may form a more positive appraisal of their own relative's adjustment. A phenomenon similar to this has been reported in studies examining the influence of social comparisons as important predictors of well-being (Heidrich & Ryff, 1993a, 1993b. This body of work draws upon Rosenberg's (1979) description of the mechanisms by which individuals give meaning to their experiences. Most appropriate to the current research is the principle of social comparisons whereby persons evaluate themselves through selected comparisons with others. In our work, we extend this idea to caregivers of nursing home residents who may evaluate their relative's well-being by comparing him or her to certain resi-dents seen as "worse off" (downward comparison) than their relative. In turn, caregivers perceive their institutionalized family member as having adjusted better as compared to other residents in the facility.
The relative strength of our single-item measures of resident and caregiver adjustment warrant discussion and further inquiry. One explanation for the strength of these preliminary constructs draws upon studies of older women that examine the reasons for and outcomes of relocation. Ryff & Essex (1992) suggest that specific aspects of well-being (e.g., purposefulness, continued development, positive selfregard, etc.) can be compromised when women relocate to settings that are discrepant with their reasons for moving; that is, "push-pull discrepancy." For example, a woman may consider relocating because of health problems, death of her significant other, loss of friends, or other "push factors." If she wishes, but is unable, to relocate to a place with health facilities and opportunities for social activities (i.e., "pull factors"), she will likely experience compromised wellbeing. For the present study, it may be that our exploratory dimension of resident andրor caregiver adjustment may touch upon this notion of "pushpull discrepancy." Caregivers may perceive themselves andրor their relative as poorly adjusted when they believe that their relative's relocation is discrepant with their reasons for the placement: "I wanted him to get three meals a day because he wasn't eating at home. Now he refuses to eat because the nursing home food doesn't taste like home cooking." Future research could expand upon these findings by exploring the specific components underlying perceived resident and caregiver adjustment.
The caregivers' perceptions of the support provided by and communication with NAs were also important predictors of the adjustment of the resident and caregiver, respectively. Our research suggests that caregivers may depend on the information provided by NAs about the impaired relative in order to formulate their appraisal of the resident's adjustment to the facility. Comments by NAs such as "your relative's behavior is not uncommon" or "there's nothing to worry about," may help caregivers feel that their relative is doing well. Similarly, caregivers' perception of an "open line" of communication between themselves and the NA may help caregivers to adjust better to the stressful experiences of nursing home care. In either case, nursing home stressors are important and added components of the SPM that can have differential effects on the caregiver and resident.
Although we were successful in adapting the SPM to this sample of placement caregivers, it is important to note three limitations of the study. First, many stress indicators described in the original model were not included in the interviews and analyses since the stress model was not used to guide the development of the original research. Key concepts, such as caregiver overload, were not included in the present study although a similar measure of "carerelated strain" was found to be unrelated to adjustment and depression. We also had no exact measure of the loss of the relationship between caregiver and resident, although a similar measure of the current level of closeness was measured and found to be nonsignificant. Moreover, although single-item measures of adjustment were central to the results, future work will be needed to clarify the exact components and mechanisms underlying and defining resident and caregiver adjustment to placement. For example, is the better adjusted resident one who does not cry, is fairly active, andրor does not ask to go home? Is the well-adjusted caregiver one who is not consumed by their visits to the facility or who has become reintegrated into their larger social network in the community? These limitations mark a critical dilemma facing researchers who wish to apply the SPM to secondary analyses of existing data. Using secondary data to test a theoretical model that was not used to ground the initial study can lead to an inaccurate representation of the model's components. Although there may be missing pieces in our adapted SPM, the concepts we added fit well and our results proved promising.
Second, the present study was cross-sectional, unlike the original methodology testing the SPM, which included five waves of data. Without longitudinal data it is difficult to say that the present results confirm that resident adjustment at Time 1 will affect caregiver adjustment at Time 2, which will affect caregiver depression at Time 3. As well, the specific nature of a causal relationship between adjustment and depression remains unclear. Depression in family caregivers prior to placing their relatives may have hindered the adjustment of both the caregiver and resident. We hope that these findings prompt further and more detailed inquiry into the longer term effects of "placement" caregiving.
Third, it is important to recognize the potential limitations of the study's generalizability to the larger population of family caregivers. This sample consisted of caregivers who remained involved in the care of their placed relatives by visiting often and providing hands-on care. Residents were predominantly living in suburban facilities whose clients were predominantly white elderly. Compared to many previous studies, where reported samples consist predominantly of spouse caregivers, our sample is unusual in that most caregivers were adult daughters. It may be that sample recruitment from within a nursing home leads to samples of adult daughters rather than spouses. This is not to say that adult daughters are more likely to place a demented relative, although it is possible that the competing demands of work, family, and caregiving may lead daughters to place relatives earlier than spouses. The group may also be atypical of the larger population of caregivers of demented relatives who are unable to place their relatives because of inadequate financial resources or lack of adequate facilities. Our sample may be more representative of families where the placed relative is quite impaired as indicated by the inability to perform ADLs. Further research on placement caregivers will undoubtedly shed light on the mechanisms and timing of placement, which will increase the generalizability of findings.
Recently, there has been increased interest in developing interventions for individuals caring for persons with dementia. While the effectiveness of these interventions has been criticized, improvements in design and methods of evaluation indicate the usefulness of focused caregiver interventions. It may prove helpful to caregivers if these interventions were further developed to address the issues of communitydwelling caregivers who continue to care for persons residing in nursing homes (e.g., support groups or counseling for caregivers throughout the various stages of the placement process). Past research indicates that the effects of placement are both shortand long-term suggesting that interventions should be designed to meet the changing needs of placement caregivers. A limited number of service agencies and nursing homes provide interventions to promote the adjustment and well-being of caregivers who have placed a relative, yet no empirical investigations have examined the effectiveness of these interventions.
The results of the current study suggest that reducing caregiver depression is a multistage phenomenon influenced chiefly by the caregiver's own adjustment and the relative's adjustment to placement. The importance of negative relationships and interactions among caregivers, residents, and others at the nursing home are also paramount to the caregiver's adjustment and depression. Interventions that help eliminate or contain caregiving distress by focusing attention on caregivers' perceptions of the residents' adjustment may improve the well-being of placement caregivers.
A final note concerns the importance of using secondary data to test current or emerging theories of caregiving stress. Rather than spend precious time and resources collecting new caregiving samples, researchers should consider adapting models such as the SPM to existing data sets. Our results suggest that the SPM is broad enough in its application that it can be used successfully with caregivers who have made the transition to institutional care. By adapting and expanding the model, we were able to confirm that nursing home stressors are critical to the adjustment of both the resident and caregiver and to the caregiver's level of depression. Research that builds upon and generates more detailed specificity of process and outcome holds great promise for advancing caregiving research and developing effective caregiver interventions.