| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
1 Departments of Health Policy and Management, Bloomberg School of Public Health, Johns Hopkins University, Baltimore, Maryland.
2 Departments of Psychiatry and Behavioral Science and Medicine, School of Medicine, Johns Hopkins University, Baltimore, Maryland.
 |
Abstract |
|---|
 Top Abstract MethodsResultsDiscussionReferences |
|---|
Methods. A community sample of 449 cognitively impaired elders and their knowledgeable informants (KIs) was followed for 1 year. Subjects were classified as having dementia (
) or mild cognitive impairment (MCI) (
) using a neuropsychiatric battery of 4 tests. Subject characteristics (behavioral and psychological symptoms, demographic, health related, and health services use) and KI characteristics were entered into the Cox proportional hazard regression analysis.
Results. The incidence rate of NHP was 8.9%. At baseline, 57.7% of subjects had at least 1 behavioral and psychological symptom. Presence of these symptoms was associated with shorter time to NHP only in a univariate analysis. Other factors significant in the multivariate Cox modeling were being white (hazard ratio 
), having fair or poor physical health rating (
), having greater numbers of difficulties with activities of daily living (
), and having a physician's diagnosis of dementia (
). An interaction was found between the last 2 variables (
), indicating that among those with a diagnosis of dementia, a greater number of difficulties with activities of daily living delayed time to NHP. KI characteristics were not associated with NHP.
Conclusions. Behavioral and psychological symptoms were common, but having these symptoms was not an independent risk factor of NHP. Caregiver characteristics may not play as important a role in determining NHP as subject characteristics. Interventions aimed at improving or maintaining physical or cognitive functioning may have better chances of delaying NHP.
CARING for dementia patients is a stressful process for family members (1). Dementia care also poses substantial economic burdens on families and society, in large part because of the need for nursing home care (2). Estimates are that between half and three quarters of the nursing home population suffers from dementias (3).
The term "behavioral and psychological symptoms of dementia" (BPSD) has been used to describe "a heterogeneous range of psychological reactions, psychiatric symptoms, and behaviors resulting from the presence of dementia" (4, p. 3). These symptoms are common, with community population-based studies reporting a point prevalence of greater than 60% (5,6). The relationship between BPSD and nursing home placement (NHP) has been controversial. Some studies have found a direct association (7–12), others an indirect association (13,14), and still others no association at all (15,16).
Many other factors have been identified as determinants of NHP in elderly people with cognitive impairment. Presence and severity of cognitive or functional impairment are consistently found to be associated with higher risk of nursing home admission (9,10,17,18). Patient's and caregiver's sociodemographic characteristics also predict NHP in some studies (8,9,16,17), as does caregiver well-being (such as health status, burden, and stress) (7,8,13,14,16). Several studies have identified more use of patient or caregiver supportive services as factors related to placement (7,9,14,19).
Previous studies on determinants of NHP among persons with cognitive impairment have shared several limitations. Most used samples from clinical settings and thus lack generalizability. Furthermore, some studies included only patient characteristics or only caregiver characteristics, not both. The purpose of this study was to determine (a) whether the presence of BPSD is an independent risk factor for NHP, and (b) other patient and caregiver characteristics associated with risk of NHP in a community-dwelling cohort of elders with cognitive impairment.
|
Methods |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
) or mild cognitive impairment (MCI) (
) using a neuropsychological battery of 4 tests (described in the paragraphs that follow) were included in this study. The term mild cognitive impairment describes individuals with cognitive impairment that does not meet diagnostic criteria for dementia (25,26), but who may be at a subclinical stage of dementia (27). [Another term for this subset of persons, cognitive impairment no dementia (CIND), has been used, e.g., in the Canadian Study of Health and Aging (25).] The Johns Hopkins Bloomberg School of Public Health Institutional Review Board approved the study. A knowledgeable informant (KI) was identified for each subject. These KIs completed the baseline and first annual follow-up interviews of the MMCS. At the first annual follow-up, 32 subjects were lost to follow-up. Among the 450 with follow-up information, 41 subjects were admitted to nursing homes (14 of them died after placement), 30 died in the community, and 379 still lived in the community. Most subject and KI demographic characteristics did not vary between those lost to follow-up and those with follow-up information. However, subjects lost to follow-up were younger and their KIs were older, had fewer years of education, and were less likely to have a paying job (Table 1).
|
). Subjects who died in the community and subjects who lived in the community at the first annual follow-up were censored. One subject with the date of NHP coded later than the date of death was excluded from further analysis. The final sample size was 449.
Independent Variables
BPSD determination.--
The depression subscale of the Neuropsychiatric Inventory (NPI) (28) and the Behavior Symptom Rating Scale (BSRS) (29) were used to measure the behavioral and psychological symptoms. Dichotomized indicators of presence of specific BPSD—psychotic symptoms, depressive symptoms, and agitated symptoms—and any BPSD were created for subsequent analyses (5).
Subject demographic characteristics.-- Subject's age, sex, race, marital status, years of education, and income were derived from the baseline KI interview.
Subject health-related characteristics.-- The neuropsychiatric battery of 4 tests consisted of 3 measures selected from a battery used by the Consortium to Establish a Registry for Alzheimer's Disease (30)—the Boston Naming Test, the Word List Memory Test, and the Verbal Fluency Test—and the Digit Symbol subscale of the Wechsler Adult Intelligence Scale–Revised (31). A determination of dementia was made if the subject scored at or below 2 SD from the mean for normal subjects of comparable age and education on at least 2 of 4 tests. A classification of MCI was made if the subject scored at or below 1.5 SD from the mean on at least 1 test or scored at or below 1 SD from the mean on 2 tests, and did not meet the criteria for dementia. Tests were administered in the same order to all subjects by trained interviewers.
The physician's diagnosis of dementia was based on the KI's answer to the following 2 questions: "Has a doctor ever said that the subject had Alzheimer's disease?" and "Has a doctor ever said that the subject had any other type of dementia or senility?"
Both the study classification and the physician's diagnosis of dementia variables were highly correlated with baseline MMSE scores (
,.5307, respectively; 
for both). These 2 variables were used as indicators of severity of cognitive impairment.
Difficulties with 5 basic activities of daily living (ADL) (32), a 5-point rating of general physical health (from excellent to poor), and the number of comorbid conditions from a lists of 9 possible conditions (heart disease, hypertension, diabetes, arthritis, stroke, cancer, chronic obstructive pulmonary disease, vision, and hearing problems) as reported by KIs were used as health status measures.
Subject health service use characteristics.-- Pharmacy records from approximately 6 months before the baseline interview were obtained. Subjects were identified as using a psychotropic drug if they were prescribed cognitive enhancers, antidepressants, neuroleptics, anxiolytics–hypnotics, or certain anticonvulsants (gabapentin and divalproex sodium).
The KI was asked about the subject's use of 5 paid in-home services (home-delivered meals, paid housekeepers, shopping services, visiting nurses or home health aides, and paid sitters) and 5 paid community services (senior centers–social events, adult daycare, transportation services, eating together programs, and social clubs) at a frequency greater than once per month.
KI characteristics.-- KI characteristics included age, sex, marital status, years of education, income, relationship to subject, living with subject or not, working status, general health, use of caregiver supportive services, and attitudes toward NHP.
Statistical Analysis
Data were analyzed using Stata (33) statistical packages. Survival analyses used Cox proportional hazard regression (34) to identify risk factors for time to NHP. Univariate analyses were performed first. Presence of any BPSD and other variables with 
were included in multivariate analysis. A stepwise backward selection procedure based on a likelihood ratio test was performed to delete variables that ceased to be significant at the 
level. Presence of any BPSD was included in all models as the main independent variable of interest. Nonlinear and interaction terms were examined. Graphic displays and diagnostic model checking were applied for identifying violations of proportionality assumptions and influential points. All reported p values were two sided. A hazard ratio greater than 1 indicates shorter time to NHP in a survival analysis.
|
Results |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
) days, ranging from 5 to 680 days. Forty subjects (8.9%) were admitted into nursing homes between baseline and the first annual follow-up. The mean time to NHP for those admitted into nursing homes was (
) days, ranging from 5 to 458 days. The median time to NHP was not calculated because of the small number of subjects admitted. Nearly three fifths (57.5%) of subjects had at least 1 behavioral and psychological symptom in the 4-week period prior to baseline (Table 2): 19.6% had depressive symptoms, 14.3% had psychotic symptoms, and 37.6% had symptoms of agitation.
|
Three quarters of the subjects (73.5%) were classified as having dementia, and 26.5% were classified as having MCI. Only 16.3% of cognitively impaired subjects had a physician's diagnosis of dementia. Thirty-seven percent had 1 or more ADL difficulties, and 56.4% had more than 2 comorbid conditions. Almost half (49.2%) had fair or poor physical health.
Nearly one quarter (24.7%) of the subjects were prescribed at least 1 psychotropic drug. However, 17.4% had missing information on psychotropic drug use. Approximately one fourth (24.5%) used at least 1 paid in-home service, and 27.6% used at least 1 community service.
Thirty-seven percent of KIs were aged 65 or older, the majority (74.6%) were female, and 51.9% were married. Only 12.9% of KIs completed fewer than 9 years of education. One third had incomes below 125% of the poverty line, 38.1% had incomes above 125% of the poverty line, and 23.6% had missing income information. Nearly half (48.8%) of the KIs were adult children, 57.9% lived with subjects, 42.8% had paying jobs, and 24% had fair or poor general health. More than one seventh (15.4%) of the KIs used at least 1 caregiver supportive service, and 35.6% had a positive attitude toward NHP for persons with dementia.
Univariate Analysis
Presence of any BPSD and symptoms of agitation were significantly associated with shorter time to NHP at univariate analysis (Table 3, unadjusted). Only presence of any BPSD variable was used in subsequent analysis.
|
Multivariate Analysis
Presence of BPSD was no longer significant in the multivariate model controlling for other potential risk factors (Table 3, adjusted). Being white (adjusted hazard 
), having fair or poor physical health rating (2.12), having a physician's diagnosis of dementia (6.76), and having a greater number of ADL difficulties (1.46) were significantly associated with shorter time to NHP in the final model. There was a significant interaction between number of ADL difficulties and physician's diagnosis of dementia (0.59). Because of this interaction, the effects of these 2 variables could not be interpreted separately. For subjects without a physician's diagnosis of dementia, a greater number of ADL difficulties was associated with shorter time to NHP. For subjects with a physician's diagnosis of dementia, a lower number of ADL difficulties was associated with less time to NHP (Figure 1). Reanalysis with multivariate logistic regression was performed to check the impacts of imputation of some NHP dates on the survival analysis results (data not shown). The statistical significances of variables were the same with 2 statistical techniques.
|
|
Discussion |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
Presence of behavioral and psychological symptoms, especially agitation, was associated with shorter time to NHP at the univariate level. Several other previous studies have found that BPSD were predictors of NHP in univariate but not multivariate analyses (13,14). These researchers argued that BPSD were only indirectly associated with NHP. These symptoms may serve as primary stressors for caregivers. They may increase caregiver burden or the caregiver's desire for NHP, and these may be the direct risk factors for NHP.
The only significant subject demographic characteristic in the multivariate model was race. Consistent with the research by Kasper and Shore (9), white subjects were 2 times more likely to have shorter time to NHP than were black subjects. The observed racial differences were not explained by socioeconomic status (education or income).
Subjects with fair or poor physical health were twice as likely to have shorter time to NHP compared with subjects with excellent, very good, or good physical health. This finding was consistent with that from Montgomery and Kosloski (19). Also consistent with other studies (9,10,15,17–19,35), we found both severity of functional and cognitive impairment were predictive of time to NHP. We used 2 variables (study classification as MCI or dementia and physician's diagnosis of dementia) but not MMSE score as indicators of severity of cognitive impairment. A physician's diagnosis of dementia was a stronger predictor of time to nursing home entry than classification as dementia or MCI. Osterweil and colleagues (35) also found that when a dementia diagnosis was in the model, MMSE score was no longer predictive of NHP.
We found an interaction between the number of ADL difficulties and a physician's diagnosis of dementia. This finding suggests that caregivers may have lower expectations of subjects with a clinical diagnosis of dementia, because fewer ADL difficulties were associated with shorter time to entry. A diagnosis may trigger consideration of NHP. Persons with dementia who are functionally intact may also be more likely to show physically aggressive behaviors or wander and thereby receive a diagnosis. They may be more difficult to care for in some respects than those who are bed-bound.
Our findings that use of any psychotropic drug and use of supportive services were associated with shorter time to NHP at the univariate level only is consistent with other studies (11,14). Nursing home placement has increasingly been added as an outcome measure in dementia drug trials (15), and the relationship between supportive services use and NHP has been inconsistent (8,9,14). More research on the effects of health services use on NHP is needed.
Consistent with several studies (9,10), we did not find any caregiver (KI) characteristics to be associated with time to NHP in the multivariate model. Our findings should be interpreted with caution because we did not have information on several often-cited caregiver risk factors such as caregiver burden, stress, and informal care network characteristics (7,13,16,19).
There were several limitations to this study. First, with only 1-year of follow-up data and relatively few (
) cases admitted into nursing homes, we may not have enough power to detect several risk factors of interest. Second, with relatively few cases admitted into nursing homes, model overadjustment might be a concern. Third, although it was a population-based study, all subjects were from 1 state, Maryland, and our findings may not be generalizable to other regions. Finally, we have imputed dates on NHP for several cases with missing data. This might affect the precision of the statistical model. However, an analysis with logistic regression models with binary outcomes produced the same sets of risk factors, so the estimation error could be negligible.
Conclusion
In conclusion, the presence of behavioral and psychological symptoms was not an independent risk factor for shorter time to nursing home admission in elderly persons with cognitive impairment. Other characteristics that were predictive to NHP included race, general physical health, severity of functional impairment, and severity of cognitive impairment. The effect of severity of functional impairment differed for those who were and were not diagnosed with dementia. Caregiver characteristics may not play as important a role in determining NHP as subject characteristics. Interventions aimed at improving or maintaining physical or cognitive functioning may have a better chance of delaying NHP, but the role of a dementia diagnosis requires more investigation as well.
|
Acknowledgments |
|---|
Address correspondence to Ding-Cheng (Derrick) Chan, MD, PhD, Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, #157A Versailles Circle, Towson, MD 21204. E-mail: dcchan{at}jhsph.edu
Received August 5, 2002
Accepted October 21, 2002
|
References |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
).