Several studies [6-14] have followed caregivers of elderly persons with dementia for periods of 2 to 30 months. The changes in caregiver health found in these studies were generally small. The absence of a non-caregiving control group in all but one [14] of the earlier studies makes it difficult to compare changes observed among caregivers with changes in comparable persons who were not providing care. Furthermore, many of the earlier studies have been criticized for including caregivers of patients with low levels of impairment [15]. Such caregivers may be in situations that are not particularly stressful, and thus these studies may not constitute a valid test of the hypothesis relating the stress of caregiving to changes in mental and physical health.

The principal objective of our study was to estimate the change in depression and physical health during a 1-year period in a group of caregivers for elderly persons with dementia and in a group of comparison participants. We also aimed to examine the relation between changes in caregiver health and certain characteristics of the patient with dementia.

Methods

Top Methods Results Discussion Author & Article Info References

Study Design and Participant Recruitment

Details of the study design have been reported elsewhere [16]. Briefly, identification of caregivers and comparison participants required two steps: identification of two groups of elderly index patients (one with dementia and one without) and then identification of these elderly patients' close family members. Patients with dementia were recruited at the outpatient Geriatric Assessment Unit of the Jewish General Hospital in Montreal, Canada. The Geriatric Assessment Unit's multidisciplinary team assesses elderly patients with complicated medical, social, and functional problems, many of whom have dementia. Patients are referred to the Geriatric Assessment Unit by their primary care physicians or by a social worker; the referral process is often initiated by family members.

Patients were included in the study if they were 55 years or older; had received a clinical diagnosis of dementia, according to the criteria of the Diagnostic and Statistical Manual of Mental Disorders (third edition) [17]; and had a history of cognitive decline of at least 6 months' duration. Patients were excluded if they had been referred to the geriatric unit for preplacement assessment or if they had a major, acute medical problem; 24.1% of potential patients with dementia were excluded for these reasons.

Patients having cataract surgery at the same hospital as the Geriatric Assessment Unit were chosen as the source of elderly persons without dementia. Patients with cataracts were excluded if they were younger than 55 years, if they were cognitively impaired (defined as a score of less than 25 on the Mini-Mental State Examination [18]), or if they had a severe or total functional impairment according to criteria of the Older Americans Research and Services questionnaire [19]; 40.8% of potential patients were considered ineligible for these reasons.

The demented patient's primary caregiver was defined as the person who had the most responsibility for and provided the most care to the patient. In the case of the comparison group, each patient with cataracts was asked to name the person who currently provided the most daily assistance. Both caregivers and comparison participants were excluded if the person who was identified was not the patient's spouse or child. Structured interviews were conducted in the participant's home by trained interviewers. Only participants who could speak English or French were interviewed. The investigators did not discuss the specific hypotheses of the study with the interviewers. The first interview (time 1) was conducted soon after referral of the patient to the study. The second interview (time 2) was conducted approximately 1 year later (median interval between interviews, 12.4 months; range, 11 to 20 months). The research project was approved by the ethics committee of the Jewish General Hospital. Signed consent was obtained from study participants and a close family member after the nature of their participation was explained. Recruitment and time 1 interviewing took place from 1987 to 1989. Time 2 interviews were conducted from 1988 to 1991.

Measurements

Depression was measured using the Center for Epidemiologic Studies depression (CESD) scale [20, 21], which includes 20 items, graded on a scale corresponding to the frequency, in the preceding week, of each symptom (ranging from 0, rarely or none of the time to 3, most or all of the time). Possible scores range from 0 to 60. In keeping with previous research [22-26], a score of 16 or more was used to indicate the presence of depression. Aday and Andersen's 24-item checklist, which has been used in studies of health care access [27, 28], was used to quantify physical symptoms occurring in the month preceding the interview. The checklist includes various items, such as coughing, fatigue, headache, weight loss, diarrhea, and chest pain.

Consumption of prescription medications by caregivers and comparison participants was determined using the checklist from the Older Americans Research and Services questionnaire [19]. Three questions from this questionnaire [19] were used to assess the caregivers' and comparison participants' perceptions of their overall health: Self-rated health scores were computed as the sums of the scores for the three questions. Participants were asked to indicate whether they had any of the 21 chronic diseases included in the Canada Health Survey questionnaire [29]: the total number of diseases for which the participant answered "yes" was used as a measure of his or her chronic disease status.

The coping efficacy of both groups of study participants was assessed using Olson and colleagues' [30] inventory of family coping strategies. Social support was measured using Cutrona and Russell's 20-item Social Provisions Scale [31, 32]. A three-item index of social support satisfaction, similar to one designed by Stokes [33], was used to quantify the caregivers' assessment of the quality of their social contacts. The caregivers' experience of stressful life events was evaluated using the 34 undesirable events from the Louisville Older Person Events Scale [34]. To determine their lifetime exposure to cigarette smoking, the number of years during which the caregivers smoked was multiplied by the average number of packs of cigarettes smoked per day. The procedure developed for the Established Populations for Epidemiologic Studies of the Elderly surveys was used to measure total alcohol consumption in the month preceding the interviews [35].

The Mini-Mental State Examination [18] was used to assess the severity of the cognitive impairment in the patients with dementia. Possible scores on the 11-item examination range from 0 to 30, with a lower score indicating more cognitive impairment. The functional status of index patients was assessed using the functional subscale of the Older Americans Research and Services questionnaire [19, 36]. The Dementia Behavior Disturbance scale [37] was used to quantify the frequency of behavioral problems in the patients with dementia.

Analysis

Change scores were computed for both depressive symptoms and physical symptoms by subtracting the value at time 1 from the value at time 2. Thus, a difference with a positive sign indicated worsening of the caregivers' level of depression or physical symptoms, whereas a difference with a negative sign indicated an improvement. Analysis of variance was used to test the difference between caregivers and comparison participants with respect to the change in the dependent variables, after adjusting for potential confounding variables. In these analyses, the change score was the dependent variable, caregiving status (1, caregiver; 0, comparison participant) was the independent variable, and the time 1 value of the dependent variable was included as a covariate. For depressive symptoms, the following covariates were also included: sex (1, female; 0, male), ethnic group (1, Jewish; 0, other), education (1, high school completed; 0, high school not completed), age, number of chronic diseases, number of stressful life events in the preceding 6 months, coping score, social provisions score, and support satisfaction score. For physical symptoms, two covariates were included in addition to those listed above: lifetime exposure to cigarette smoking (in pack-years) and number of ounces of alcohol consumed in the previous month.

Multiple Classification Analysis, available in the Statistical Package for the Social Sciences [38], was used to estimate the mean change among caregivers and comparison participants after adjusting for covariates. Multiple Classification Analysis uses the results of the analysis of variance procedure to compute adjusted mean values of the dependent variable in subgroups defined by the categories of the independent variable(s).

Changes in caregiver health were considered in relation to four patient characteristics: cognitive status, behavioral disturbances, functional impairment, and the occurrence of institutionalization in the period between time 1 and time 2. Stepwise linear regression, with the change score as the dependent variable, was used to select statistically significant caregiver characteristics: Age, sex, and the time 1 value of CESD score were selected for change in depression, whereas sex and the time 1 number of physical symptoms were selected for change in physical symptoms. Analysis of variance was carried out, with the change score as the dependent variable, the four patient characteristics as independent variables, and the selected care-giver characteristics as covariates. Multiple Classification Analysis was used to estimate the adjusted change in care-giver depression and physical symptoms in subgroups defined by the categories of the four patient variables.

Results

Top Methods Results Discussion Author & Article Info References

Of the 218 study participants interviewed at time 1 (103 caregivers and 115 comparison participants), 193 were interviewed again at time 2. Of those who were not reinterviewed, 18 refused and 7 could not be contacted. Two participants were excluded because of poor data quality, for an overall follow-up rate of 88%. Among patients with dementia, 7.5% died and 32.3% were admitted to an institution in the interval between time 1 and time 2; among patients with cataracts, 3.1% died and 2.0% were admitted to an institution. Among patients with dementia who were admitted to an institution, the median interval between the time 1 interview and institutionalization was 8.7 months (range, 0.9 to 13 months). Family members of the 10 patients who died in the study interval were eliminated from analysis, yielding a total sample of 181 (86 caregivers and 95 comparison participants).

Characteristics of the elderly patients who were receiving care are shown in Table 1. Patients with dementia were, on average, older than patients with cataracts and included a slightly lower proportion of women. As expected, they also received a much larger amount of help with activities of daily living and more hours of home care services than did patients in the comparison group. The clinical condition in the patients with dementia worsened substantially during the 1-year study period: On average, their cognitive and functional status decreased and the degree of behavioral disturbance increased. Although the amount of help with activities of daily living provided by the caregiver to the patient with dementia was slightly lower at time 2 than at time 1, the volume of home care services more than doubled. The results shown in Table 1 probably represent an underestimation of the patients' deterioration because the time 2 measurements were only available for those patients still living at home.

The average age of caregivers was higher than that of comparison participants (Table 2). Caregivers were slightly more likely to live with the patient than were comparison participants. The proportion of Jewish participants was higher among caregivers than among comparison participants, and caregivers were less likely to have graduated from high school. More than 60% of participants were the patient's spouse and nearly 60% were women. Only small differences were found between the two groups with respect to mean coping and social support scores and the mean number of chronic conditions.

The mean depression score increased slightly during the 1-year study period among caregivers and comparison participants (Table 3). The increase was largest among caregivers of patients with dementia who were admitted to an institution during the study interval. A similar pattern of associations was observed for all the other health measurements.

Table 4 shows the results of the multivariate analysis for the two major outcome measurements. The CESD score increased by 1.9 among caregivers and decreased by 0.2 among comparison participants between time 1 and time 2. As a point of reference, an increase of 2 in CESD score is equivalent to the appearance of one symptom that occurs "a moderate amount of time" or two symptoms that occur "a little of the time". The number of physical symptoms increased by 0.5 among caregivers and by 0.1 among comparison participants. The difference between the groups with respect to the change over time was not statistically significant for either variable.

Table 5 shows changes in caregiver health with respect to patient characteristics. For both depression and physical symptoms, the deterioration was greater among caregivers of patients who were admitted to an institution during the study period; for physical symptoms, the association with institutionalization was statistically significant. Caregivers of patients with more behavioral disturbances had larger decrements in both depression and physical symptoms; the association with behavioral disturbance was statistically significant for both variables. The associations between the patient's cognitive status and the caregiver's depressive symptoms and physical symptoms were not significant.

Table 6 presents the distribution of caregivers and comparison participants with respect to changes between time 1 and time 2. For depression, comparison participants were more likely than caregivers to have experienced little or no change (difference in CESD score of ±2 points). Caregivers and comparison participants were almost equally likely to have a higher level of depressive symptoms at time 2 than at time 1 (36.0% compared with 33.7%, respectively). Among caregivers, the proportion who deteriorated was equal to the proportion who improved. For physical symptoms, there was more stability among comparison participants than among caregivers: During the study period, 64.2% of comparison participants and 38.4% of caregivers experienced little or no change in the number of symptoms (difference of ±1 symptom). Caregivers were considerably more likely than comparison participants to have an increase in physical symptoms. Still, almost as many caregivers improved (26.7%) as deteriorated (34.9%).

Discussion

Top Methods Results Discussion Author & Article Info References

Caregivers were somewhat more likely than comparison participants to experience an increase in depression and physical symptoms during the 1-year study period, although the differences were small and not statistically significant. Although the level of depression among caregivers in our study was higher than in previous studies, the magnitude of the changes is consistent with earlier research. For example, in the study by Pruchno and coworkers [8], spouse caregivers had mean CESD scores of 14.9, 15.5, and 14.5 at 0, 6, and 12 months, respectively. In studies by Williamson and Schulz [11, 15], CESD scores among 79 caregivers whose relative with dementia was still living at home after 24 months were 11.9, 13.5, 12.6, and 13.7 at 0, 6, 18, and 24 months, respectively.

Kiecolt-Glaser and associates [14], in the only previous longitudinal study in which a comparison group was included, studied 69 spouse caregivers of patients with dementia and 69 volunteer control participants. At time 1, the prevalence of a psychiatric diagnosis of a depressive disorder was 24.6% among the caregivers and 0% among control participants. At time 2 (approximately 13 months later), the prevalence was 31.9% and 5.8%, respectively. Because these authors used a more stringent definition of depression, the prevalence was considerably lower than in our study. However, the magnitude of the change in the prevalence of depression during a period of 1 year was remarkably similar to that observed in our study.

The study by Haley and colleagues [13] is the only previous study of caregivers in which changes in physical symptoms over time were considered. These investigators found that the score on a self-rated health problems checklist increased during a 30-month period from 1.3 to 1.8 among 18 caregivers of elderly patients with dementia who were still living at home at follow-up, and from 1.9 to 2.1 among 15 caregivers of patients who had entered an institution during the follow-up period.

A higher level of behavioral disturbance in patients with dementia at time 1 was significantly associated with deterioration in both caregiver depression and physical symptoms during the study period. Because these findings were based on exploratory analysis and relatively small numbers, they must be viewed with caution. However, this corroborates those results obtained in almost every cross-sectional study in which the relation between behavioral disturbance and care-giver health was examined [39-46]. Ours is the first study to observe an association between patient behavioral disturbance and changes in caregiver health over time. Together these results suggest that families in which patients manifest high levels of behavioral disturbance require close monitoring.

Caregivers of patients with dementia who were admitted to an institution during the study period were more likely to experience an increase both in depression and physical symptoms; for physical symptoms, the association was statistically significant. There are several possible explanations for the observed association between institutional placement of the patient and deterioration of the caregiver's health. First, the mean interval between institutionalization and the time 2 interview was only a few months. Thus, the increase in depression and physical symptoms that was observed for caregivers of patients who were admitted to an institution may represent simply a short-term reaction to the loneliness and guilt associated with placement of a close family member. More research will be necessary to determine whether the increased risk is maintained during longer follow-up periods. Second, caregivers may become more aware of their own psychological state after the placement of their impaired relative: This may make them more willing to discuss their depressed mood and physical health. Finally, increased health problems in the caregiver may be the cause rather than the consequence of the decision to admit the patient to an institution.

On average, the degree of change among caregivers over time was fairly small, both in this study and in most earlier studies. Using a score of 16 or more on the CESD scale to classify individuals as symptomatic, Schulz and Williamson [15] found that 59% of caregivers were stable with respect to depression at four measurement points during a period of 2 years. In our study, using the same definitions, almost 70% were stable during 1 year (data not tabulated). Methodologic factors may contribute to the finding of stability in care-giver health over time. Thus, it is possible that caregivers with rapidly deteriorating health status are more likely than others to abandon the caregiving role; these caregivers would be under-represented in research studies. The scales and questionnaires commonly used as indicators of caregiver health may not be sufficiently sensitive to change or quantitative research methods may not adequately capture the negative effect of caregiving. Another possible explanation of the observed stability is that most caregivers, although distressed, can meet the challenges of caregiving without serious adverse consequences to their own health [15].

The absence of large changes in group means does not preclude the possibility of substantial changes in certain subgroups. In this study, for example, the aggregate change scores masked a considerable amount of individual variability. Thus, for both depression and physical symptoms, the proportion of caregivers who worsened was almost the same as the proportion of caregivers who improved. This suggests that prognosis may differ in different subgroups of caregivers. Future research could usefully focus on predictors of improvement. Such knowledge could be applied in interventions with caregivers, whereas knowledge about predictors of deterioration would help identify high-risk persons.

Given the relative lack of change seen in care-giver health over time, it may be useful to revise some of our assumptions concerning the temporal pattern of the effects of caregiving. According to the "wear and tear" hypothesis, prolonged exposure to stress affects well-being by depleting the caregiver's physical and psychological resources [47]. According to this hypothesis, well-being declines linearly over time with continuing exposure to the stressful situation. Despite the plausibility of this hypothesis, studies of caregiving for patients with dementia have provided little evidence to support it [13]. Other models of change have been proposed that incorporate the idea that the caregiver's preexisting psychological and social resources are the most salient correlates of change or that focus on the possibility of adaptation to stressful experiences over time [15]. More sophisticated studies, with multiple measurement points, are necessary to test the validity of these models adequately.

In our earlier paper describing the cross-sectional component of this study, we discussed methodologic issues that could influence the validity of the results, such as the ethnic composition of the study sample and the comparability of the study groups [16]. An additional methodologic point must be considered in relation to the results reported in the present paper. Although use of a longitudinal design avoids many of the problems related to cross-sectional research, our study was not based on an inception cohort: Thus, at time 1, the caregivers included in the study had already been in the caregiving situation for an average of 3 years. Because important differences may exist between those who continue in the caregiving role and those who abandon it, our results can only legitimately be generalized to a group of caregivers who have persisted through approximately 3 years of caregiving.

As in all clinic-based studies, various selection factors may limit the generalizability of the results. An outpatient geriatric assessment unit, such as the one in which this study was done, may attract patients and families who are particularly distressed. However, the bias is probably smaller than in some earlier studies, in which caregivers were recruited through media publicity and Alzheimer disease support groups. Another design feature that might influence generalizability is that, to ensure homogeneity of the study sample, only spouses and children of patients were included. Thus, the results cannot be generalized to caregivers who have some other relation with the patient. Because spouses and children account for most of the caregivers of elderly persons with dementia in the community [48], this feature of the study does not severely limit its generalizability.

The outcome measures used in our study were related to the frequency of depressive and physical symptoms and did not necessarily reflect the presence of a particular medical or psychiatric diagnosis. The CESD, for example, is nonspecific because it measures symptoms that are characteristic of major, minor, and secondary depression, situational depressive reaction, and transient mood variations [20]. We felt that use of a scale such as the CESD would be more appropriate for the assessment of the effect of caregiving than would a psychiatric diagnosis. The severe stresses of caregiving are likely to result in feelings of grief, despair, helplessness, and hopelessness, but these reactions do not necessarily lead to psychiatric illness [4, 49]. Thus, Becker and Morrissey [49] contend that the stress of caregiving is unlikely to lead to a major depressive disorder except in predisposed persons. A similar argument can be made for the physical symptoms checklist, which is composed of a diverse series of items and does not purport to identify the presence of specific disease states. Rather, it assesses the frequency of somatic symptoms that may arise in response to the stress of caregiving.

Overall mean changes in depression and physical symptoms during 1 year were small, both among caregivers and comparison participants. However, a substantial amount of variability exists in individual response to the caregiving situation. Thus, certain subgroups of caregivers (that is, those caring for a patient with a high level of behavioral disturbance at time 1 and those caring for a patient who was admitted to an institution during the study period) experienced greater deterioration in their health status. Future research could profitably focus on identifying other important prognostic factors.