| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
a Department of Community Health Sciences, School of Public Health, University of California, Los Angeles
b School of Medicine, University of California, Los Angeles
Carol S. Aneshensel, Department of Community Health Science, School of Public Health, Box 951772, University of California, Los Angeles, CA 90095-1772; E-mail: anshnsl@ucla.edu
Decision Editor: Fredric D. Wolinsky, PhD
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. The data were obtained from a 5-year longitudinal study of family caregiving. This analysis used a subsample of caregivers who moved their relatives from home to a nursing home (n = 210). The frequency (times per week) and length (hours per week) of visits for up to 5 years after placement were examined in relation to select caregiver and care-recipient characteristics. A semiparametric, group-based method using a latent class mixture model was used to identify distinctive trajectories over time.
Results. The majority of caregivers maintained the frequency and length of their visits for extended periods of time. Five distinct patterns for frequency and three patterns for length were identified. Being a spouse, lower education, a close past relationship, a strong sentiment against placement, and living close to the facility predicted membership in groups visiting frequently and for longer times. Caregiver characteristics were more strongly associated with frequency, whereas care-recipient characteristics were related to length.
Discussion. To a large extent, family visits to nursing home residents with dementia become established in the period immediately after relocation, which means that efforts to enhance continued family involvement should commence at the time of admission.
THE contemporary increase in the number of elderly persons has been accompanied by an increase in the number of persons afflicted by Alzheimer's disease (AD) or other forms of dementia, a segment of the population with substantial need for long-term care. In 1997, an estimated 2.3 million persons in the United States were afflicted with AD, and this figure is projected to almost quadruple in the next 50 years, largely because of the demographic increase of the "old-old" population (Brookmeyer, Gray, and Kawas 1998
). The total cost in the United States for Alzheimer care was estimated at $67.3 billion in 1991 (Ernst and Hay 1994). The care of elderly persons with dementia has become an important social issue in terms of health care costs and the well-being of older persons and their family members.
Family members play a significant role in caring for elderly persons, including those who suffer from dementia. Family constitutes 72% of paid and unpaid caregivers for elderly persons with activity limitations, with adult children accounting for the largest segment (National Academy on an Aging Society 2000). The recognition that this care is not a "cost-free" alternative to institutional care has generated a substantial accumulation of research, but most studies focus on caregiving in the home, limiting our knowledge about caregiving to those living in institutions. This oversight has far-reaching consequences because many elderly people are eventually institutionalized because of their progressive degenerative illness. In particular, those who suffer from dementia have a relatively high rate of institutionalization and having dementia increases the risk of institutionalization (Lee and Tussing 1998). However, only limited information exists about the continuing role families play in the care of relatives with dementia after these persons have been relocated to a nursing home. This study seeks to provide this information by describing the extent to which caregivers continue to visit their institutionalized relatives after admission to a nursing home.
Despite the prevailing myth that caregiving ends with placement in a nursing facility, family members typically continue their caregiving activities after institutionalization (Aneshensel, Pearlin, Mullan, Zarit, and Whitlatch 1995; Bowers 1988; Dempsey and Pruchno 1993; Fink and Picot 1995; Fleming 1998; Kellet 1998; Minichiello 1987, Minichiello 1989; Smith and Bengtson 1979; Stull, Cosbey, Bowman, and McNutt 1997; York and Calsyn 1977; Zarit and Whitlatch 1992). For example, in a comparative study of family caregivers for a relative with dementia at home and those who had placed their relative in a nursing home, Stephens, Kinney, and Ogrocki 1991 found that caregivers visited their institutionalized relatives, on average, 2.9 days per week. Other research demonstrates that a majority of family caregivers visit relatives living in nursing homes at least weekly (Bitzan and Kruzich 1990; Fleming 1998; Hook, Sobal, and Oak 1982). Some studies find length of residence is negatively correlated with the frequency of visits (e.g., Bitzan and Kruzich 1990; Green and Monahan 1982; Hook et al. 1982), but other studies find no association (e.g., Minichiello 1989). Minichiello 1989 suggests that recently admitted patients may receive frequent visits from relatives who are concerned about the person's adjustment to a new environment, but that these concerns may abate over time and some relatives may forget the older person. The extent to which caregivers maintain their visits over the often lengthy duration of institutional life, however, has not been established. In this study, we describe patterns of visitation after institutionalization for up to 5 years, much longer than previous investigations.
Caregiver strain does not cease after institutionalization because families continue to be involved in elder care. In an earlier analysis of the data used in the present study, Aneshensel and colleagues 1995 compared those who institutionalized their relatives and those who continued in-home care. There were no significant changes over time in strain related to family, finance, work, loss of self, or depression. However, a sense of guilt about the elderly relative was intensified after institutionalization. An Australian study similarly showed a continued sense of burden among informal carers who institutionalized relatives with dementia (Waltrowicz, Ames, McKenzie, and Flicker 1996). These findings suggest that continuing commitment to caregiving after institutionalization maintains or exacerbates at least some of the burdens experienced during in-home care, contrary to the implicit expectation that caregiver burden ends when the older person is placed.
On the other hand, continued family relationships often have significant meaning to family members (Fleming 1998; Kellet 1998; Tilse 1997). Tilse 1997 conducted a qualitative content analysis of interview data from 18 Australian spousal caregivers, finding that visiting was embedded in the long-term relationship between the caregiver and their husbands or wives. The purpose and meaning of visiting included "maintaining continuity of self and the pattern of daily life, demonstrating commitment, expressing attachment, coping with grief and loss, and avoiding loneliness" (p. 201). Fleming 1998 linked meaning to the construction of self and to a "lifeline of special care." Smith and Bengtson 1979 suggested that some family-elder dyads experienced "renewed closeness" after institutionalization because relief from some burdens facilitated socioemotional interactions.
Similarly, continuing family involvement seems to benefit the quality of life of institutionalized elderly persons, either by improving psychological functioning (Green and Monahan 1982), or by assuring the quality of institutional care (Fleming 1998). For example, Green and Monahan 1982 found that elderly persons with frequent family visits had lower levels of psychosocial impairment, suggesting that these visits had a significant therapeutic influence. Psychological impairment was expected to impede family visits because it makes visiting stressful, but only impairment in activities of daily living (ADLs) was associated with visiting, and this association was positive. The researchers suggest that the most seriously impaired patients were maintained at home the longest by highly motivated caregivers who then tend to visit frequently. Dempsey and Pruchno 1993 report that frequency of visits is positively associated with the amount of supplemental care provided by family members, improving the overall quality of care, although lack of coordination between staff and families can result in duplication of effort or neglect of key tasks (Bowers 1988). Family caregivers are able to personalize care because of their biographical expertise and intimate knowledge about the care-recipient (Bowers 1988), including serving as an advocate to ensure quality of care and quality of life (Stull et al. 1997).
These results suggest that family members' continued involvement with institutionalized elderly persons has implications for the quality of life of both the elderly person and his/her family. However, little is known about the factors that influence family visitation. In a review article, Naleppa 1996 summarized the following as significant predictors: financial conditions—the use of family resources to fund nursing home care and patient affluence are associated with more visits (Kosberg 1972); physical distance to the facility—families living close to the facility tend to visit more often (Bitzan and Kruzich 1990; Green and Monahan 1982; Hook et al. 1982; Minichiello 1989); kinship distance—closer kin tend to visit more often (Hook et al. 1982); and nursing home regulations—nursing homes that include families in activities and those with more flexible schedules tend to receive more visits (Montgomery 1982). Quality of relationship has also been implicated as a predictor of visits (Minichiello 1989). The impact of these factors on caregiving as it evolves over the course of the care-recipient's institutional life, however, is unclear.
The current analysis expands our understanding of the family's relationship with institutionalized relatives over time, specifically family members who have dementia. The existing literature involves cross-sectional designs or limited follow-up observation, usually 1-year comparisons of only two time points (e.g., at institutionalization and 1 year later). These design features hinder the examination of processes of change in the family's behavior over time. Longitudinal analysis is necessary because the health and cognitive function of the care-recipient decline as residential life extends over long periods of time, generating new caregiving demands. This study encompasses enough observation points to describe a true trajectory of change for several years after nursing home admission, providing much longer follow-up than existing studies. Simultaneously, caregivers are influenced by factors outside of this realm, developing life-course trajectories in work, in other family life, and so forth. The patterns of visitation that evolve over time, therefore, are not necessarily static, but may be dynamic. This process emerges as a life-course trajectory—caregiving in the postinstitutional phase. This study describes for the first time several trajectories for the frequency and duration of visiting one's relative in a nursing home, documenting the fact that caregivers do not all follow a single course.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
). Data were collected by means of a multiwave panel survey, with six face-to-face interviews. All care-recipients resided in the community at baseline (1988). Caregivers were subsequently reinterviewed annually irrespective of whether they continued to provide care at home, found it necessary to place their relatives in institutions, or became bereaved. The results reported here are limited to data collected during the in-home and institutional phases of caregiving.
A total of 273 patients were placed in long-term care institutions at varying times after Wave 1 and before Wave 6: 156 (57.1%) subsequently died, 92 lived through Wave 6, and 25 caregivers were lost to follow-up. Of these persons, 63 experienced the death of their relatives immediately after their admission and before the next scheduled interview, resulting in an analytic sample size of 210. The length of postadmission follow-up varies from 1 year (admissions between Waves 5 and 6) through 5 years (admissions between Waves 1 and 2). Follow-up data are stacked to create a synthetic cohort, with time counted from the interview immediately after institutionalization. Thus, data are available for a first postadmission interview for 210 persons irrespective of whether these data were obtained at Wave 2 (N = 100), Wave 3 (N = 52), Wave 4 (N = 33), Wave 5 (N = 20), or Wave 6 (N = 5); data for a second postadmission interview were obtained from Wave 3 through Wave 6 for the first five groups, and are not available for the last group because data collection had ended. This pattern is repeated for the 3-, 4-, and 5-year follow-up intervals. The sample size necessarily declines as the length of follow-up increases (N1 = 210, N2 = 146, N3 = 109, N4 = 70, N5 = 38).
Data about visiting behavior are from each of the postadmission interviews. Data for independent variables are obtained primarily from the baseline interview and from the continuing care interview immediately preceding institutionalization.
Measurement
The outcome variable is the trajectory over time of visiting behavior assessed along two dimensions. The first dimension is frequency, assessed with the following question: "During a typical week, including weekends, on how many days do you visit your ... (mother/father/husband/wife)?" The second dimension is duration, obtained by summing answers to a two-part question, "During the week, that is Monday through Friday (and "On Saturday and Sunday"), about how many hours in total do you spend visiting with your ... ?" These dimensions were assessed at each postadmission interview and arrayed into a trajectory spanning the duration of the postadmission interval.
There are several sets of independent variables. Demographic characteristics were assessed with standard questions asked at the baseline interview and updated annually as appropriate. Quality of relationship was assessed at baseline by asking, "Thinking of your relationship with your ... before (his/her) illness ever began, how close or distant were the two of you?" Given that responses were highly skewed, answers were collapsed into a dichotomy of (1) for "very close" versus (0) for "somewhat close" to "very distant." Preadmission sentiment regarding institutionalization was asked with, "Do thoughts of the possibility of placing your relative in a nursing home or other care facility make you feel ... ?" with responses collapsed into (1) for "very upset" versus (0) for "fairly upset" to "not at all upset."
Aspects of the care-recipient's dementia are included in the analysis because social isolation may be especially likely among persons who are insensible or who have been ill for very long periods of time. Duration was assessed by incrementing perceived onset at baseline—"How long ago did you realize that something was wrong with your ... ?"—with the interval from the baseline interview to nursing home admission. This assessment is subjective because illness recognition occurs well after the true onset of the underlying disease. However, this report was obtained independently of subsequent institutionalization, meaning that inaccuracies are unlikely to systematically bias results. Cognitive deficit at the interview immediately preceding admission was measured by asking, "How difficult is it for your ... to" perform seven tasks, such as understanding simple instructions (Pearlin, Mullan, Semple, and Skaff 1990). Response categories ranged from (0) for "not at all difficult" to (5) for "very difficult" or "can't do at all." The summated scale is divided by the number of items and has excellent reliability (
= .82). Although reported by caregivers, this measure is based on standard clinical tests (Folstein, Folstein, and McHugh 1975) and was validated with independent clinical evaluations using the Mini-Mental State Examination (r = .65) for a subsample of 75 care-recipients.
Reason for admission was measured at the first postadmission interview by asking a list of 12 common reasons, such as "you became too ill to carry on" or "your ... had lost bowel and/or bladder control." Multiple reasons could be given, and each was coded (0) for "no" and (1) for "yes." Similar reasons were collapsed into combined categories (see Table 3 ). Relocation from one care facility to another was traced for the entire follow-up period. In addition, time to travel from home to the care facility was coded in minutes through 90. The difficulty of making this trip was rated by the caregiver from (1) for "not at all difficult" through (4) for "very difficult."
|
Three dimensions of caregiving during the in-home period were assessed at the interview immediately preceding admission. Each of these measures is a composite of existing measures and study-specific measures developed during qualitative pilot work; each was pretested before data collection (Aneshensel et al. 1995). Problem behaviors were measured by asking, "In the past week, on how many days did you personally have to deal with the following behavior of your (relative)?" There were 14 items (e.g., "keep you up at night"), and response categories ranged from (1) for "no days" through (4) for "5 days or more." The average summated scale has excellent reliability ( = .84). In addition, assistance with ADLs and instrumental ADLs (IADLs) was assessed by asking, "I'm going to read from a list and ask you how much your ... depends upon you personally for help." Altogether, 15 items were asked for ADLs (e.g., "eating") and IADLs (e.g., "taking medication"). Response categories ranged from (1) for "not at all" through (4) for "completely." The average summated scales have excellent reliability ( = .89 for ADLs and = .90 for IADLs).
Analytic Strategy
The analytic objectives are to describe how patterns of nursing home visitation vary over time and to link these trajectories to attributes of the care-recipient and the caregiver. Data are analyzed with a latent class mixture model, a semiparametric group-based method for modeling unobserved heterogeneity in a population by identifying distinctive clusters of relatively homogeneous individual trajectories of behavior over time (Jones, Nagin, and Roeder 2001; Nagin 1999). A polynomial relationship is used to link visiting behavior with the number of interview years since admission to a nursing home, specifically a quadratic censored normal model:
 | (1) |
|
|
|
|
|
|
). The likelihood ratio test is not used because alternative models are not nested (Nagin 1999).
The probability of an individual's membership in the various groups that make up the model is estimated as:
 | (2) |
). Group assignment is based on the highest posterior probability of group membership calculated from the estimated values of trajectory parameters. Some misclassification of cases into latent classes is almost inevitable, especially when there are many classes (Clogg 1995; Hagenaars 1990). To evaluate this possibility, the assignment probability that determined group membership was examined. Misclassification is of concern for analysis based on group assignment (see Table 3 and Table 5 ), but not for the latent class analysis (see Table 4 and Table 6 ).
|
|
|
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
Of the caregivers who placed their relatives in nursing homes during the course of the study, only 38 (18.1%) completed five postplacement interviews. More (25.7%) continued to provide care until their last interview, but had less than five postplacement interviews because data collection ended. The largest group (44.3%) left the cohort because the care-recipient died. Few (11.9%) were lost to follow-up. Interview status is significantly associated with only two characteristics, income and ethnicity of the caregiver, as also shown in Table 1 . Caregivers in the highest income group were disproportionately high in the long-term institutional care group, and low in the lost-to-follow group. Non-Hispanic Whites were disproportionately high in the bereavement group and disproportionately low in the lost-to-follow group.
Frequency of Visits to the Nursing Home
The average number of days per week that the caregiver visits declines over time, but only slightly from 3.41 (SD = 2.44) in postplacement follow-up Year 1, to 3.13 (SD = 2.38), 2.73 (SD = 2.41), 2.67 (SD = 2.53), and 2.90 (SD = 2.44) in Years 2 through 5, respectively. Very few caregivers did not visit the care-recipient at all: 5.3%, 2.1%, 4.7%, 1.4%, and 0%, at Years 1 through 5, respectively. Thus, most caregivers continue to visit their relatives frequently, and there is little evidence for the notion that caregivers abandon their relatives after nursing home admission.
Table 2 shows BIC for different numbers of groups in the latent class analysis. For comparison, all groups are calculated using linear and quadratic terms. Although the BIC suggested that a 6-group model was statistically most appropriate, it yielded an extremely small group, making the 5-group model preferable on the basis of stability and parsimony. The linear and quadratic terms in 3 of the 5 groups were not statistically significant and are omitted in the final model, yielding the final BIC of -1020.86 for frequency.
|
|
The other group showing change, "biweekly increasing," is very small. These caregivers initially visited about 3 days per week and then gradually increased over time to daily visits. Visits increased most during the early period of the nursing home stay and gradually leveled off over time. This pattern is defined by a positive linear coefficient (b1 = 1.08, SE = .31; p < .001) and a negative quadratic coefficient (b2 = -0.62, SE = .21; p < .01), along with the intercept (b0 = 7.70, SE = 0.46, p < .001). Given the small number of subjects, especially late in the study time (n = 5 for the 4th year and n = 1 for the 5th year), considerable caution is needed in the interpretation of this group.
The posterior probabilities for group assignment range from .37 to 1.00, with two groups, "weekly stable" and "biweekly increasing," having minimum probabilities of less than .50. However, mean probabilities for each group range from .82 to .93 ("weekly stable": .88; "biweekly stable": .82; "daily stable": .83; "daily declining": .93; and "biweekly increasing": .89) for the five groups. These mean values suggest a generally high probability of correct assignment, but the ranges suggest the presence of some misclassification, particularly for the two groups with low minimum probabilities. Therefore, some caution is required for the next analysis, which is based on group assignment.
As can be seen in Table 3 , several caregiver characteristics showed statistically significant associations with group membership. The "daily stable" group was disproportionately male and spousal caregivers, whereas the "weekly stable" group was disproportionately female and adult children. The "daily declining" group contained the highest proportions of employed and well-educated caregivers, whereas the "daily stable" group was lowest on these characteristics. The percentage who had a "very close relationship" before the illness was highest for the "daily stable" group and lowest for the "weekly stable" group. Length of care before admission and preadmission sentiment about the prospect of placement, however, were not significantly associated with the subsequent pattern of visit frequency.
As also shown in Table 3 , none of the reasons for admission that were attributable to care-recipient characteristics were significant, but two caregiver-related reasons were significant. The "biweekly increasing" and "daily stable" groups contained an especially large proportion of admissions attributed to caregiver illness or exhaustion, whereas these reasons were uncommon in the "daily declining" group. These results suggest that the small "biweekly increasing" group consists of spousal caregivers who relocated their husbands and wives because their own health and stamina were depleted, and they were unable to continue caregiving at home. These caregivers appear to increase visits with their spouses after a period of recovery or rest.
Many care-recipients (42.8%) were moved from one nursing home to another. Whether they moved, however, was not significantly associated with the pattern of visiting frequency. Because many of the care-recipients were moved, a composite variable was created to examine the influence of short distance (below average) among those who did not move (compared with long distance didn't move and moved). The "daily stable" and "daily declining" groups had an especially high percentage of caregivers who continued to live close to the institution.
Both the distance from home to the nursing home and the difficulty of making this trip were associated with group membership. The average commuting time for those who visited daily was only a quarter of the time for those who visited weekly, and those who visited biweekly had about half the commute of those who visited weekly. A similar association is seen for the difficulty of the trip. Thus, those who are consistently infrequent visitors have by far the greatest distance to travel and experience the greatest difficulty in making the trip.
Whether the caregiver identified problems in the facility soon after admission was not significantly associated with the frequency trajectory. Similarly, aspects of the functional status of the care-recipient before admission were not significant.
Multivariate Analyses of Frequency Data
The statistically significant bivariate correlates of the frequency trajectories were included in a multivariate multinomial logit model. The "biweekly increasing" group was dropped because its small size made multivariate analysis infeasible. Some covariates were omitted because of multicollinearity and covariates that were not significant (p > .10) were trimmed from the final model, shown in Table 4 .
Being a spouse resulted in significantly higher log odds of belonging to both the "biweekly stable" and "daily stable" groups, as contrasted to the "weekly stable" group net of other variables in the model. Education was significant for only one contrast: having more than a high school education reduces the chances of belonging to the "daily stable" group. Having a "very close" past relationship with the care-recipient independently yields higher chances of belonging to two groups, "biweekly stable" and "daily stable." Lastly, short distance and no relocation are associated with higher log odds of belonging to the "daily stable" and "daily declining" groups net of other variables in the model. Examined by group, "biweekly stable" was predicted by being a spouse and having had a very close relationship. All of the examined characteristics were associated with being a member of the "daily stable" group. "Daily declining" was predicted only by living close and not moving the care-recipient after admission.
The overall probabilities of group membership for these characteristics were calculated using the logit coefficients. When none of the conditions listed in Table 4 are present, the probabilities of each group membership are 79% "weekly stable," 16% "biweekly stable," 1% "daily stable," and 4% "daily declining." If instead the caregiver is a spouse, the probability of belonging to the "weekly stable" group declines to 42%. If the caregiver also had a "very close" past relationship with the care-recipient, the chances of belonging to the "weekly stable" decline to only 26%. When all the factors are present, the probabilities of group membership are 15% "weekly stable," 28% "biweekly stable," 33% "daily stable," and 24% "daily declining."
Duration of Visits to the Nursing Home
Relatively few caregivers visited the care-recipient less than 1 hour per week: 6.3%, 4.2%, 5.7%, 2.9%, and 0% for Year 1 through Year 5, respectively. The average number of hours per week at each follow-up interview was 7.45 (SD = 8.29), 6.58 (SD = 7.83), 6.43 (SD = 9.26), 5.75 (SD = 7.53), and 6.25 (SD = 6.92) for Year 1 through Year 5, respectively. Repeated-measures analysis of variance using 35 participants who completed all five follow-up interviews revealed a significant change over time (F(4,136) = 3.92, adjusted p [Greenhouse-Geisser] < .05). Examined by paired t tests, there were significant differences between Year 1 and Year 4 (n = 68; t = 3.52, p < .001) and Year 1 and Year 5 (n = 36; t = 2.40, p < .05). Thus, there was a general decrease in the number of hours spent visiting over time, but the average remains relatively high even after several years have passed. This pattern testifies again to the continuing commitment of family members to their institutionalized relatives.
BICs for different numbers of groups are calculated also for the duration data (Table 2 ). Although the BIC suggested that a 6-group model was statistically most appropriate, the models with more than three groups yielded extremely small groups (1.0% or less), making the 3-group model preferable on the basis of stability and parsimony. The linear and quadratic terms in 2 of the 3 groups were not statistically significant and are omitted in the final model, yielding the final BIC of -1691.70.
Fig. 2 shows the observed and expected trajectories of three groups of duration pattern identified using the latent class mixture model. The largest group was "brief & stable" caregivers who initially visited their relatives about 3.6 hr in the average week and maintained this pattern over time. The second largest group, "moderate & stable," was similarly consistent, but visits were more lengthy, about 13.7 hr a week. Finally, the smallest group, "lengthy & declining," started visiting an average of 32.2 hr per week (more than 4 hr per day, assuming they visited every day) and gradually shortened their visits. Despite the decline, this group had the longest visits, on average, even in the 5th year of the nursing home stay. An intercept-only model best describes both the "brief & stable" group (b0 = 3.63, SE = .23, p < .001) and the "moderate & stable" group (b0 = 13.66, SE = .69, p < .001), but the "lengthy & declining" group has, in addition to the intercept (b0 = 30.03, SE = 1.12, p < .001), statistically significant negative linear (b1 = -2.97, SE = .55, p < .001) and quadratic (b2 = -0.94, SE = .45, p < .05) terms. These coefficients mean that the rate of decline increased over time.
|
Table 5 shows the association between caregiver and care-recipient variables and the duration pattern of visiting. Some variables that were associated with frequency trajectories (see Table 3 ) were also associated with the duration trajectories, but others differed. Gender of the caregiver was not significantly associated with the duration trajectories, although it was associated with the frequency trajectories. In contrast, type of relationship showed an association similar to frequency: spouses comprised almost all of the "lengthy & declining" group, whereas they were less than half of the "brief & stable" group. Employment was also a significant factor in visitation hours: only 1 in 10 of the "lengthy & declining" group worked, whereas 4 in 10 of the "brief & stable" group were employed. Similarly, education was a significant factor in group membership, with a disproportionate number of caregivers with more than a high school education in the "brief & stable" group, 3 in 4, compared with the "lengthy & declining" group, only 1 in 4.
The quality of relationship before admission was significantly associated with duration, as it was for frequency. Those who had "very close" relationships were especially unlikely to be found in the "brief & stable" group. Length of care before admission was not significant.
Some variables were only associated with duration trajectories. Caregiver income was one of these variables: the "brief & stable" group had the highest proportion of high income caregivers. Negative sentiments about placement before admission were especially low among this group. In addition, the only significant reason for admission was care-recipient (and not caregiver) specific: those who cited a decline in care-recipient's condition were most commonly found in the "moderate & stable" group and were least likely to be found in the "brief & stable" group.
Of the relocation and traveling variables, only distance to the nursing home was significant. The average commuting time was longest for the "brief & stable" group. The "lengthy & declining" group had the highest proportion of caregivers who lived a short distance from the facility and did not move their relatives.
Unlike frequency, problems the caregiver experienced with the staff of the facility were significant for duration: those who belonged to the "lengthy & declining" group experienced the most problems. The care-recipient's condition before admission was also significant. Both cognitive deficit and ADL assistance were highest among the "moderate & stable" group. For cognitive deficit, the "lengthy & declining" group showed the lowest score, whereas the "brief & stable" group showed the lowest score for ADL assistance.
Multivariate Analyses of Duration Data
Table 6 shows the results of the multivariate multinomial logit analysis for duration. Being a spouse yielded a higher chance of belonging to the "moderate & stable" and the "lengthy & declining" groups. Having more than a high school education increased the probability of belonging to the "brief & stable" group, compared with the "lengthy & declining" group, although education was not significant for the "moderate & stable" group. Those who previously felt "very upset" at the thought of placing the care-recipient in a nursing home were more likely to belong to the "moderate & stable" group. Lastly, those who continuously lived a short distance from the nursing home were more likely to be in the "lengthy & declining" group.
Examined by each group, belonging to the "moderate & stable" group was predicted by being a spouse and having had intense anticipatory feelings about placement. Belonging to the "lengthy & declining" group was predicted by being a spouse, having no more than a high school education, and continuously living a short distance from the nursing home.
The overall probabilities of group membership for the characteristics listed in Table 6 were calculated using the logit coefficients. When none of the factors are present, the probabilities of group membership are: 95% "brief & stable," 4% "moderate & stable," and 1% "lengthy & declining." Being a spouse increases the probability of belonging to the "lengthy & declining" group (13%) and decreases the probability for the "brief & stable" group (71%). Simply having had more than a high school education yields the highest possibility of belonging to the "brief & stable" group (98%). When all conditions are present, the probabilities of group membership are: 9% "brief & stable," 35% "moderate & stable," and 56% "lengthy & declining."
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Differences in findings may also be the result of the population studied. Our sample consists of primary caregivers who have demonstrated their commitment to the person with dementia by providing in-home care before institutionalization. These caregivers may be less likely to "forget" the nursing home resident than visitors in general, some of whom have only weak ties to the resident. In addition, the care-recipients in our study are spouses and parents, people who often inspire especially intense and long-lasting commitments. Along these lines, it is important to recall that all of our care-recipients had primary caregivers, which is not the case for many nursing home residents. Indeed, the absence of a close family relation substantially increases the risk of institutionalization. Thus, our results apply to a select subgroup and should not be generalized to the entire nursing home population.
This caveat calls attention to another feature of the study that merits comment: all of our care-recipients were ill with dementia. Although this selection criterion limits generalizability, it also provides a test case. The underlying disease makes interpersonal interactions especially difficult because it robs the individual of his or her persona, personal history, and ability to recognize and interact with others. Family visits may become especially challenging and thankless in conventional ways, although some families find rising above these obstacles is its own reward. In many ways then, the case of dementia serves as a conservative estimate of visiting by close family members because severe cognitive impairment makes these persons difficult to visit.
Although our study tracks visits for up to 5 years after relocation to a nursing home, most persons were not tracked for the full 5 years, raising the possibility of bias associated with sample attrition. However, few persons were actually lost to follow-up (i.e., could not be located or refused participation). Instead, losses were primarily caused by either the study design or mortality. Specifically, the length of follow-up was limited by the time of admission: Admissions that occurred late in the study could only be followed for short intervals. However, at baseline, the sample was diverse in terms of illness duration, which means that study time is not associated with the severity or progression of the illness. In other words, the timing of our observations, including the follow-up observations, is unrelated to the progression of the dementia. This feature reduces concern about censoring because of the end of data collection. Finally, the tracking of visits ended for many persons because the care-recipient died. These persons were necessarily dropped from analysis, but are not losses to follow-up because this stage of the caregiving career had come to its natural end. Although we could increase sample size by shortening the duration of the trajectories, providing the long-term data seems to outweigh the advantages that would be gained.
With few exceptions, the pattern of visiting that is established in the first year after nursing home admission is maintained thereafter. Caregivers who follow trajectories that do change markedly over time appear to be either rebounding from their own state of depletion, or moderately reducing what started out as extremely intense supplemental care. Many caregivers in the "biweekly increasing" group, for example, cited their own illness and exhaustion as reasons for institutionalization, suggesting that subsequent increases in the frequency of their visits may reflect their own recuperation. This very small group is the only one that follows a trajectory of increasing involvement over time. The two groups that follow trajectories of decreasing involvement over time, "daily declining" and "lengthy declining," start at the maximum level of effort, which is undoubtedly difficult to maintain even among the most committed and vigorous. In other words, the intensity of the exertion may portend its own erosion over time. In most instances, however, the initial frequency and length of visits seem to be self-perpetuating. Programmatic efforts to maintain family involvement over the duration of the residential life of the person with dementia, therefore, should generally start early. Caregivers who are physically unable to visit, however, should be encouraged to reconnect with the care-recipient if this subsequently becomes feasible.
Consistent with previous research, we find that a short commute to the nursing home is associated with frequent visits, a finding with a straightforward commonsense interpretation. The impact of distance appears to be established during the initial phase of residential life, insofar as distance does not differentiate stable and declining patterns of visiting among those who start out as daily visitors. Although caregivers who consistently commute short distances decrease the amount of time they spend at the nursing home, they initially visit for a great many hours and therefore remain the most extensive visitors. Distance does not seem to influence changes in the length of visits because the other two duration groups remain stable over time. Thus, the impact of distance on visiting seems to be established early and to be sustained over time by the stability of visiting patterns, unless, of course, the patient is moved to a different care facility.
When distance is taken into consideration, two caregiver characteristics remain associated with both the frequency and duration trajectories: being a spouse or adult child, and education. In general, spouses visit more often and spend more time visiting than adult children, patterns that persist over the course of the residential life of the care-recipient. One contrast is especially noteworthy: the extremely high log odds of being in the daily stable group (relative to the weekly stable group) for husbands and wives. Married persons share a unique relationship that may account for the intense involvement with their husbands and wives. In addition, these married persons are mostly older themselves and, therefore, may have greater latitude in devoting large amounts of time to caregiving than is the case for adult children, who are generally midlife with extensive commitments to other activities, such as work and family.
Although spouses have overall less education than adult children, the impact of education on visiting is not entirely caused by this generational difference. Spouses have less education, and they visit often and for a considerable amount of time. The multivariate model controls for being a spouse, and for this reason also controls for generational differences on a variety of other characteristics, including education. In this model, the impact of education is quite specific: having more than a high school education decreases the chances of being in the "daily stable" group and the "lengthy & declining" group (relative to the "brief & stable" group). Stated differently, other things being equal, those with higher educations are especially likely to visit once a week for a brief period and to maintain this pattern over time. The impact of education, net of being a spouse, does not appear to be caused by employment or income because these variables did not enter into the multivariate model. It may reflect a constellation of social status and role obligations that compete with visiting, or an education-based set of attitudes and norms concerning nursing homes and visiting, but the validity of such interpretations requires further investigation.
Two psychosocial variables are also implicated in visiting trajectories: closeness with the care-recipient before caregiving for frequency and preplacement sentiments about nursing home admission for duration. Caregivers who were close to their spouses or parents tend to continue visiting twice a week or more (compared with continued weekly visits). Equally important, closeness appears to differentiate the subsequent trajectories of those who initially visit daily insofar as it increases the chances of being in the "daily stable" group (relative to the "weekly stable" group), but not the "daily declining" group. Thus, closeness appears to result not only in frequent visiting, but also in the maintenance of frequent visiting over time.
Extreme negative sentiments about nursing home placement before the admission of the care-recipient are associated with the amount of time spent visiting, but not the frequency of visits, distinguishing the "moderate & stable" group from the "brief & stable" group, but not the "lengthy & declining" group. In other words, these caregivers seem to visit about as often as those who had less intense feelings; but, when they visit, they tend to stay longer, a pattern that persists over time. Many caregivers who are averse to institutionalization make promises to their spouses and parents that they will not be put in a nursing home, promises these caregivers are eventually unable to keep. They may be trying to keep the spirit of these promises by spending as much time as possible with the care-recipient, a commitment that extends over time.
We expected visiting to be influenced by the functional status of the care-recipient and by the caregiver's assessment of problems with the nursing home and its staff, but these factors were associated only with duration, not frequency, and only in bivariate, not multivariate, analysis. Where there are bivariate differences, impairment is greatest in the "moderate & stable" group, but problems are greatest in the "lengthy & declining" group. Dissatisfaction with care, then, is not simply an indicator of perceptions that are tied to the special needs of the most severely impaired. Initial dissatisfaction is associated with lengthy initial visits that become shorter over time. In the early stages of institutional life, these caregivers may be staying longer to provide supplemental care because institutional care is seen as inadequate, or caregivers may use this time to monitor and advocate for the care of their relatives. The subsequent decline in the length of visits may mean that these issues have been resolved. However, the care-recipients in this study tend to be severely impaired, and their caregivers tend to be highly satisfied. The weak associations with visiting trajectories, then, may reflect limited variation on these attributes.
Significant limitations of this study include that the participants are self-selected at the time of the baseline interview and therefore probably biased toward those who are more committed to caregiving. Selection may have resulted in higher levels or less diversity in the frequency and duration of visiting after institutionalization. It should also be recognized that the latent class analysis usually requires a large sample. The limited number of participants in this analysis, especially in the later period, may have reduced the stability of results. In this regard, some trajectories are based on very few participants at the longest follow-up observation, which also detracts from reliability. The group assignment based on the maximum posterior probability is subject to the possibility of misclassification, especially when the number of groups is as large as five, as in the frequency analysis. The number of groups is determined by the relative comparison of BIC rather than statistical model testing. Taken together, these comments mean that the findings presented here should be understood as suggestive rather than definitive. Nevertheless, these results provide a unique source of information for the further development of hypothesis/theory building, information that has not been available in the past literature, and should also aid in the design of further longitudinal studies.
Although our results demonstrate considerable uniformity among caregivers toward maintaining visits over time, this tendency subsumes several distinct trajectories. It is not possible, then, to describe a single trajectory that characterizes most caregivers, aside from the tendency to continue to do what they start out to do. Instead, some visit every day and stay for a good part of the day, whereas others visit for a short time once a week, and still others fit somewhere in between. In other words, the institutional phase of family caregiving is as diverse as the in-home care phase, and as diverse as the family life that preceded the need for care. Moreover, these differences are linked to both structural characteristics, such as generation, and social psychological factors, like closeness to the care-recipient. That is, visiting reflects, at least in part, the constraints and possibilities encountered by the caregiver, and his or her relationship with the care-recipient. A reasonable goal for nursing home staff, then, would be helping families discover the type of connection that is best for that particular family and assisting families in coming to terms with this schedule even if it is restricted, compared with that of other families, or with regard to some societal ideal.
|
Acknowledgments |
|---|
We thank the San Francisco and Los Angeles chapters of the Alzheimer's Disease and Related Disorders Association for their assistance with this research. We gratefully acknowledge Bobby L. Jones for his statistical consultation. This article was prepared while Noriko Yamamoto-Mitani was supported by a research leave from the University of Tokyo, Japan.
Received for publication April 24, 2001. Accepted for publication August 23, 2001.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  J. E. Gaugler, A. M. Pot, and S. H. Zarit Long-Term Adaptation to Institutionalization in Dementia Caregivers Gerontologist, December 1, 2007; 47(6): 730 - 740. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. E. Gaugler and R. L. Kane Families and Assisted Living Gerontologist, December 1, 2007; 47(suppl_1): 83 - 99. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. L. Port Informal caregiver involvement and illness detection among cognitively impaired nursing home residents. J. Gerontol. A Biol. Sci. Med. Sci., September 1, 2006; 61(9): 970 - 974. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. L. Port, S. Zimmerman, C. S. Williams, D. Dobbs, J. S. Preisser, and S. W. Williams Families Filling the Gap: Comparing Family Involvement for Assisted Living and Nursing Home Residents With Dementia Gerontologist, October 1, 2005; 45(suppl_1): 87 - 95. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Schulz, S. H. Belle, S. J. Czaja, K. A. McGinnis, A. Stevens, and S. Zhang Long-term Care Placement of Dementia Patients and Caregiver Health and Well-being JAMA, August 25, 2004; 292(8): 961 - 967. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
