| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
a Stigler Center for the Study of the Economy and the State, University of Chicago, Illinois
b Fuqua School of Business, Duke University, Durham, North Carolina
c National Bureau of Economic Research, Cambridge, Massachusetts
d Center for the Study of Aging and Human Development, Duke University Medical Center, Durham, North Carolina
e Durham VA Geriatric Research, Education and Clinical Center (GRECC), North Carolina
f Department of Medicine, Division of Geriatrics, Duke University Medical Center, Durham, North Carolina
Michael J. Moore, Darden School of Business, Faculty 121, University of Virginia, Charlottesville, VA 22902 E-mail: MooreMi{at}darden.virginia.edu.
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. Female primary caregivers were drawn from the first wave (N = 2043) of the National Longitudinal Caregiver Study, a survey of informal caregivers of elderly male veterans diagnosed with probable Alzheimer's disease or vascular dementia. Cost measures include the following four items: value of caregiving time, caregiver's lost income, out-of-pocket expenditures for formal caregiving services, and caregiver's excess health costs. Disease severity was indexed by the number of impairments in activities of daily living. Patient's problem behaviors were measured using the Behavior Rating Scale–Dementia.
Results. The annual cost of providing informal care to elderly community-dwelling veterans with dementia was estimated to be $18,385 per patient in 1998. The larger components of this cost are caregiving time ($6,295) and caregiver's lost earnings ($10,709). All aspects of costs increase with disease severity and problem behavior. Most of this cost increase derives from the increased caregiving time required for the provision of physical care.
Discussion. This study provides a comprehensive estimate of the excess costs that result from providing informal dementia care in the community. Unlike previous studies, our estimates excluded costs that caregivers would have incurred if they had not been caregivers. Therefore, results reported here reflect only costs due to informal dementia care.
PROGRESSIVE dementing disorders such as Alzheimer's disease (AD) and multi-infarct/vascular dementia (VAD) are debilitating conditions characterized by loss of cognitive function, with subsequent losses in motor and physical functioning. AD is the most common cause of dementia in older adults, accounting for about 50% to 60% of all dementia cases; VAD accounts for 10% to 20% of such cases (
Katzman 1995
;
Small et al. 1997;
Zarit and Zarit 1998). The incidence of AD increases with age, doubling every 5 to 10 years. One study (
Katzman 1995) estimated that about 2.8% of persons between age 65 and 74 have probable AD, 9% among those 75 and 84, and 28% among those 85 or older. Age-specific incidence rates derived from five major epidemiological studies estimated the prevalence of AD in the United States to be 2.32 million as of 1997 (
Brookmeyer, Gray, and Kawas 1998;
Evans et al. 1990;
General Accounting Office 1998; National Advisory Council on Aging, 1995;
Small et al. 1997). Furthermore, the General Accounting Office (GAO) predicted the prevalence to nearly quadruple, to 8.64 million in the next 50 years, by which time approximately 1 in 45 Americans will be afflicted.
Although onset is mild, individuals with dementia become less capable of self-care over time and must increasingly rely on others to manage and supervise the most basic mental and physical tasks. Eventually, dementia patients reach a level of disability that requires constant care and supervision. Because the majority of dementia patients are cared for in the community by family and friends, the cost of providing informal care is very high. The National Advisory Council on Aging (1995) estimates that the annual costs of care equal $56,000 per patient for community-dwelling and institutionalized AD patients alike. However, for community-dwelling AD patients, the majority of care costs are due to unpaid informal caregiving ($40,668 in 1998 dollars), almost three times as high as the cost of formal care in the community ($14,185). Recent studies (
Brookmeyer et al. 1998;
Leon, Cheng, and Neumann 1998) indicate that relatively small delays in the onset and progression of dementia could substantially reduce disease costs.
Several studies have estimated the cost of time family members spend on informal care provision to AD patients in the community (a more extensive literature review is available from authors upon request). Table 1 summarizes the results from these studies. The estimates vary considerably: For AD patients in the community, the hours of care reported in these studies ranged from about 11 to more than 70 hours per week, and annual costs of informal caregiving per patient ranged from $12,730 to $57,937.
|
Other areas largely overlooked in previous studies include how informal caregiver time and its related costs are affected by patient variables such as disease severity and dementia problem behaviors. This information is crucial for health policy analysis. In the absence of an understanding of the marginal relationship between cost and disease severity, interventions aimed at alleviating the economic problems associated with the disease will be misguided.
These gaps in the literature provide the motivation for the following series of analyses on the informal costs related to dementia caregiving. The contributions of this study include the use of a national sample of informal dementia caregivers, examination of multiple aspects of informal caregiving costs, and examination of the extent to which patients' disease severity and problem behaviors are associated with informal cost estimates.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
More formally, let y denote the outcome we are interested in (e.g., caregivers' lost income), X a set of socioeconomic variables that affect the outcome, Z a dummy variable that takes on the value of 1 if the individual has caregiving responsibilities, and 
the unobservables. The outcome equation is
 | (1) |
We assumed that 1 holds for both caregivers and noncaregivers. Because only dementia caregivers are included in the NLCS, where Z = 1, we estimate 1 for a comparison sample of noncaregivers, where Z = 0 and obtain the estimated coefficients 
0, and 1. We use these estimated coefficients to obtain the predicted outcomes for the sample of caregivers in NLCS, 
c,
 |
c therefore is the outcome that would have been observed for the caregivers if they did not have caregiving responsibilities. The difference between the actual outcome and the predicted outcome, y - c, estimates the effects of caregiving responsibilities on that outcome. We first computed the value of time that the caregivers allocated to caregiving tasks. The value of caregiving time was computed as the product of the reported hours spent on each task and its corresponding hourly wage. We estimated the value of caregivers' time using two approaches. First, we followed the approach employed in previous studies and used state average wage rates to compute one estimate of the cost. Second, as market wages may not accurately reflect caregivers' opportunity cost of time, we also predicted the caregivers' wage rate to compute time cost. The estimation of predicted wages is described in the following section, where we compare cost estimates from these two approaches.
Second, we estimated caregivers' lost earnings attributable to caregiving activities as in 1. We can only observe wage rates and hours of work for those individuals who are employed. Because the decision to work is unlikely to be made randomly, ordinary regression models estimating wage rates and hours of work may lead to biased estimates of caregiver wages and lost income. Therefore, the Heckman selection model (
Heckman 1979) was used to obtain unconditional parameter estimates of the wage and hours of work equations. (Estimation results from the selection models can be obtained from the authors upon request.) Using the parameter estimates from the comparison sample, we predicted the wage rates (
c) and hours of work (
c) of the caregivers in our sample. The predicted earning were then computed as the product of the predicted wages and the predicted annual hours of work, c = cc. These predicted earnings estimated the earnings of individuals with the same socioeconomic characteristics as the NLCS caregivers in the absence of caregiving responsibilities. Finally, lost earnings for each caregiver were computed as the difference between actual and expected earnings.
Wage rates are not observed for the majority of caregivers, either because they are not working or because of missing values on wage rates. Lost earnings for this group of caregivers are computed by subtracting actual hours of work from predicted hours of work and multiplying the difference by the predicted wage, that is,
 |
A small group of employed caregivers (n = 160) reported wage rates. Lost earnings for this group of caregivers are computed by subtracting actual earnings from predicted earnings, that is,
 |
We then estimated excess health care utilization attributable to caregiving. Because the health care utilization outcomes are count variables, negative binomial regression models were used. Similar to the lost earnings predictions, this allowed us to estimate the level of health care utilization of those individuals with the same socioeconomic characteristics as the NLCS caregivers but without caregiving responsibilities. Using the estimated coefficients for health care use from the comparison sample, total annual hospital days and doctor visits were predicted for the NLCS caregivers. Excess hospital days and doctor visits were then estimated for each caregiver by subtracting actual use from predicted use. Excess use was then converted to cost measures by multiplying it by state average costs of hospital stays and physician visits.
Lastly, as data on formal care purchases were not available in the comparison surveys, we could not estimate out-of-pocket costs of purchasing formal services for noncaregivers. We assumed that the noncaregivers did not utilize any formal care and computed caregivers' out-of-pocket cost of hiring paid helpers directly as the product of the reported hourly wage rates for these helpers and their hours of caregiving.
The magnitude of the variance of the forecasting errors is of concern. Aside from the variance due to random disturbances, the variance of the forecasting errors depends on three factors: size of the sample used for estimation, dispersion of explanatory variables of the sample, and the distance between given values of explanatory variables and the sample means. Variance of the forecasting errors is smaller the larger the sample size, the greater the dispersion of the explanatory variables, and the smaller the distances between the population parameters and the sample means. If large differences exist between the characteristics of the noncaregiver comparison samples from which the outcomes equations were estimated and the characteristics of the NLCS caregivers, the estimated variance of prediction errors will be large. We followed
Kmenta 1986 and computed estimates of the variance of the forecasting errors that take these factors into account.
Sample
The sample is drawn from the first wave of the NLCS, a longitudinal investigation of informal caregivers of elderly male U.S. veterans with dementia living in communities throughout the 48 contiguous states and Puerto Rico. The NLCS identified all patients in the 1997 Outpatient Diagnosis File at the Austin Automation Center (
12,000) who met the following criteria: age 60 or above and residing in the community; with next-of-kin or emergency contact; and had a formal diagnosis of AD (ICD-9 331.0) or VAD (ICD-9 290.4). Closer inspection found 24% of the patients had either incorrect diagnosis, no next-of-kin, unusable addresses, or were already institutionalized. Caregiver identification letters targeted the remaining 9,124 households, and 5,773 (63%) responded. Upon further inspection, 2,108 respondents were deemed to be ineligible (299 patients had died, 1,268 were in nursing homes, 391 were not caregivers, 15 spoke only Spanish, and 12 refused). Finally, 3,665 surveys were sent in 1998 to eligible primary caregivers; 62% responded, providing a baseline sample of 2,278. For the empirical analysis, we excluded the following 235 observations: 73 observations from Puerto Rico; 43 caregivers who reported that the care recipients were currently working; 88 caregivers who reported not living with the patient; 30 male caregivers; and one patient with a missing diagnosis code. Thus, our analysis sample consists of 2,043 female primary informal caregivers coresiding with their elderly, demented, nonworking care recipients.
Comparison Surveys
Two nationally representative surveys were used as comparison samples to our study: the Panel Study of Income Dynamics (PSID) and the National Health Interview Survey (NHIS). Because the PSID focuses on income dynamics of families, it does not include variables on health and health care utilization of the respondents. On the other hand, the NHIS focuses on health and thus does not include detailed information on respondents' wage rates and employment behaviors. Both data sets were used in our analysis to maximize variable correspondence between these studies and our caregiver sample, the NLCS.
PSID.
The PSID is a longitudinal survey of a nationally representative sample of 5,000 families in the United States. The study began in 1968 and has followed all individuals living in or born to those families every year. The PSID emphasizes the dynamic aspects of economic behaviors, including labor market work, income, and wages. We used the 1992 wave of the PSID, the latest publicly available wave of the study, as a population-based comparison to predict the hours of market work and earnings that the caregivers in our sample would have had in the absence of caregiving responsibilities. The original sample for the PSID consisted of two independent samples: a national sample of civilian noninstitutional families and a sample of low-income families. To maintain the representativeness of our sample, we excluded the low-income subsample from our analysis. To make the PSID sample comparable to the NLCS sample, male respondents were also excluded. All dollar values were updated to 1998 dollars.
NHIS.
The NHIS is a large ongoing survey of health and health care expenditures conducted every year by the National Center for Health Statistics. The data contain, along with the major socioeconomic variables, data on a series of chronic conditions, and measures of health care use in terms of number of doctor visits and hospital days. We used the pooled 1991 and 1993 NHIS data to predict health care utilization that the caregivers in our sample would have experienced in the absence of caregiving responsibilities. To make the NHIS sample comparable to the NLCS sample, male respondents were excluded.
Aggregate data.
We matched several state level data sets to our individual data by the respondents' state of residence. Data on average hospital cost per day were obtained from the
American Hospital Association 1999. We used average cost per day per patient in community hospitals to calculate the costs of hospital stays. Data on physician fees were obtained from the
American Medical Association 1998. We used mean fees for an office visit with an established patient for internal medicine to calculate doctor visit costs. Data on hourly wages for caregiving tasks were obtained from the U.S. Department of Labor (
Bureau of Labor Statistics 1999a).
Three occupations were chosen from the description of job duties from the U.S. Department of Labor (
Bureau of Labor Statistics 1999b) to match the caregiving tasks reported by NLCS respondents: home health aides, housekeepers, and bookkeepers. Wage rates of home health aides were used for physical care tasks (e.g., bathing). Since both home health aides and housekeepers may perform housekeeping duties, an average wage rate of the two was used for housekeeping tasks (e.g., cleaning house). Wage rates of housekeepers were used for all personal management tasks (e.g., taking medicine) except managing money and dealing with agencies, for which wage rates for bookkeepers were used.
Data reduction.
The NLCS study examined a total of 19 caregiving tasks. We grouped these tasks into three types: (1) personal care tasks, which include eight items (eating, bathing, dressing, using the toilet, grooming, getting in and out of bed, going up stairs, and getting around the house); (2) housekeeping tasks, which include four items (preparing meals, cleaning the house, doing laundry, and grocery shopping); and (3) personal management tasks, which include seven items (driving, shopping for clothes for the patient, taking medicine, repairs inside the house, yard work, managing money, and dealing with agencies).
The NLCS questions explicitly asked for hours spent on each task since the respondent had become a caregiver. In particular, for personal care tasks, the survey asked whether the caregiver had performed the task since she became a caregiver; for those who answered yes, the survey asked how many times per day or week she performed the task and the average hours and minutes it took each time. For housekeeping and personal management tasks, the survey asked whether the caregiver performed the task; for those who answered yes, the survey asked whether she had performed that task only since becoming a caregiver or whether she had always done so. The survey then asked how frequently the caregiver performed each task and the hours and minutes spent on the task at each time. In order to determine caregiving hours as a result of dementia care, only hours spent on each task since the respondent became a caregiver are included in the analysis. Caregiving hours by those who reported having always performed the tasks were excluded.
We categorized the data in two ways to capture the effects of patients' disease severity and problem behaviors on informal costs. First, for disease severity, functional limitations were divided into four categories: no difficulty with any activities of daily living (ADLs); difficulty with 1–2 ADLs; difficulty with 3–6 ADLs; and difficulty with all 7 ADLs. Second, we assessed patients' behavior problems using the Behavior Rating Scale–Dementia (BRS–D), which is designed to be completed by the primary caregivers and measures the manifestation of behavioral pathology exhibited in the previous month by persons with dementia (
Mack and Patterson 1996). Because our survey used the BRS–D Short Form, we transformed the scores to the standard BRS–D scores according to the CERAD BRS–D manual to be comparable to the results reported in the literature (
Mack and Patterson 1996).
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
The NLCS categorized self-reported health status with a 4-item scale (excellent, good, fair and poor), compared with the 5-item scale (excellent, very good, good, fair and poor) reported in both the PSID and NHIS. To make this variable comparable between surveys, we combined the excellent and good categories in the NLCS, and the excellent, very good, and good categories in both the PSID and NHIS. The NLCS sample is more likely to report own health as fair (38.5%) than the PSID (10.9%) and NHIS (21.9%).
We matched the caregivers' chronic conditions reported in NHIS to those included in the NLCS as closely as possible. Although the list of conditions reported in the NHIS is more comprehensive than in the NLCS, 19 conditions match in both surveys and are used for comparison: anemia, arthritis/rheumatism, asthma, circulation problems, diabetes, emphysema/chronic bronchitis, epilepsy, glaucoma, heart disease, high blood pressure, kidney disease, liver disease, multiple sclerosis, stomach/intestinal disorder/gall bladder, Parkinson's disease, skin disorder, thyroid/other glandular disorder, tuberculosis, and ulcer of the digestive system. Fewer comorbid conditions (0.8) were reported in the NLCS than in the NHIS (1.3). While the percentage reporting overnight hospital stays and doctor visits were both lower in the NLCS than the NHIS, the number of overnight hospital stays and doctor visits were similar between the two samples.
Value of Caregivers' Time
We first reported each type of task, the rates of care provision and hours of care provision since become a caregiver, and then computed the value of caregiving time (Table 3 ). Almost all of the caregivers (92.9%) reported that they began performing some caregiving tasks that they had not performed before becoming a caregiver. On average, 72.4% of the caregivers helped with physical care tasks, 23.5% provided help with housekeeping, and 82.6% provided care with personal management tasks. The average caregiver in the sample spent about 16 hours per week helping with a total of seven tasks providing dementia care. About half of the time (7.8 hours) was spent on providing physical care. Using market prices for various types of formal care services, the value of the caregivers' time is estimated to be $121 per week, with about half on physical care ($60), a third on personal management ($37), and the rest on housekeeping ($25). Estimates of the value of the caregivers' time using predicted wage rates of the caregivers were similar.
|
Caregivers' Lost Earnings
By far the largest component of caregiving cost is the caregivers' lost earnings (Table 4 ). Among the NLCS caregivers, labor force participation rates for all age groups are lower than that for the general population. The
U.S. Department of Commerce 1998 reported that in the general population, 80% of women aged 35–54, 56% of those aged 55–64, and 9% of those aged 65 and older were employed in 1997. Among the female caregivers in our sample, however, only 54% aged 35–54, 25% of those aged 55–64, and 6% of those aged 65 and older were employed. Many caregivers reported reducing hours of work or retiring because of caregiving responsibilities. Among the 54% of the caregivers who had retired, half (27%) reported that they retired because of caregiving responsibilities. Among the 16% of the caregivers who were still working, 60% reported reducing hours of work due to caregiving responsibilities, 42% reported being late for work because of caregiving, and 39% reported taking sick leave related to caregiving duties.
|
Caregivers' lost earnings were then examined by disease severity (i.e., level of ADL impairment). Actual hours worked by the caregivers decreased as disease severity increased. Therefore, caregivers' loss in earnings rose from just about $10,000 to over $12,000 annually as dementia progressed.
Formal Care Costs
Not only did caregivers spend substantial amounts of time providing care; some also paid out of pocket for formal care services. Table 5 describes the types and weekly hours of formal care purchases and related out-of-pocket costs. Just over half of the caregivers reported paying for some type of formal care (53%), most of which (48%) was for work that patients could no longer do. Because of the relatively low utilization rates of a particular type of formal care purchase, Table 5 reports per user hours of formal care purchases and out-of-pocket costs for each type of task. On average, caregivers paid $86 for 10 hours of formal services per week.
|
Excess Health Care Utilization
Using the NHIS as a population-based comparison survey, we estimated the number of hospital days and doctor visits expected for individuals with the same socioeconomic characteristics as those of the NLCS caregivers, but without caregiving responsibilities (Table 6 ). (Estimation results from the NHIS can be obtained from the authors upon request.) We then estimated excess health care costs due to caregiving for our sample of caregivers. Overall, this sample of caregivers had $114 annual excess cost in terms of hospital and physician care. Although saving $31 per caregiver annually from 1.2 fewer doctor visits, the NLCS caregivers spent $162 more than predicted on the 0.2 excess hospital days. Thus, these data suggest that excess health care utilization is a minor component of total informal cost.
|
Studies have shown that the BRS–D score, an indicator of dementia problem behaviors, is closely correlated with dementia severity (
Mack, Patterson, and Tariot 1999;
Patterson et al. 1997). We examined BRS–D scores in relation to informal costs of dementia care. Because our survey used the BRS–D Short Form questionnaire, we transformed the scores to the standard BRS–D scores according to the CERAD BRS–D manual to be comparable to the results reported in the literature (
Mack and Patterson 1996). We calculated an average total cost at each BRS–D score. The predicted caregiving costs ranged from $14,377 for patients with low BRS–D scores to more than $30,854 for patients with high BRS–D scores, indicating rising costs with more severe dementia problem behaviors.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The largest component of this cost is the caregivers' lost earnings, which equals $10,709 annually. The differences in lost earnings for caregivers younger than 65 and those 65 and older are of note. For both age groups, labor force participation rates among the caregivers in this sample are lower than those for the general population. While predicted hours of work are lower for older caregivers than for younger caregivers, reductions in actual hours of work are greater for older caregivers. Therefore, the difference between the predicted hours of work and the actual hours of work is much greater among the older NLCS caregivers than among younger caregivers, resulting in higher lost earnings among the older caregivers.
We estimate that the value of caregivers' time equals $6,295. This estimate includes only the hours spent performing tasks that the caregiver started to perform after becoming a caregiver. The hours spent performing tasks that the caregiver always performed (e.g., preparing meals for her husband) were excluded from the analysis. It is most likely that hours spent on many "always performed" tasks increased as a result of dementia care. However, data limitations preclude us from determining the magnitude of such increases. Unless hours spent on tasks that the caregiver has always performed remained the same after she became a caregiver, excluding the time spent on them will underestimate the total caregiving time allocation. Correspondingly, the value of caregiving time also will be underestimated. Based on an assumption of no increase in caregiving time on "always performed" tasks after one becomes a caregiver, our result represents a lower bound estimate of the value of caregiving time. We calculated an upper bound value of caregiving time to be $15,896 by including all caregiving hours reported by the caregivers. While the lower bound estimate represents 16 hours of care due to dementia, the upper bound estimate reflects more than 42 hours of care provision, more than the time spent on a full-time job. The magnitude of the cost is more alarming when we consider that AD and VAD are dementing disorders of later life, and that the majority of caregivers are spouses of the care recipients who are themselves older adults of retirement age.
In an influential time use study conducted about a decade ago (
Juster and Strafford 1985), adult time use was found to fall into one of the following 10 major categories: market work, house/yard work, child care, services/shopping, personal care, education, organizations, social entertainment, active leisure, and passive leisure. Aside from child care, there was no mention of caregiving activities. According to projections from the U.S. Census Bureau, 82 million Americans will be 65 or over by 2050, more than double the 34.6 million in that age group when the NLCS caregivers provided data for this study. As the population ages, so too will the incidence and prevalence of dementing disorders. Caring for impaired family or friends will become a significant occupation of many adults and likely will change previously identified time use patterns.
In addition to providing informal care, the caregivers paid almost $4,500 out of pocket annually for formal care. Also, about one fifth of the caregivers reported using their own money to provide regular financial support to and purchase other services (e.g., laundry, barber) for their care recipients. The average out-of-pocket cost to the caregiver for these services is $360 per month ($4,320 annually), almost as high as that for formal care purchases. In the baseline survey, however, it appeared that some caregivers may have misunderstood the time frame in the questions concerning other-services use; therefore, these costs were not included in the total caregiving costs reported in this article. In the follow-up surveys these questions were clarified. Our baseline estimates of out-of-pocket cost of formal care purchases are therefore conservative and may underestimate the true cost of formal care purchases. Because many NLCS families are elderly couples who live on fixed incomes, this may further compound the financial burden associated with informal caregiving.
The very minor excess health cost of the NLCS caregivers is worth noting. The NLCS respondents did not report worse health or higher health care utilization than the comparison group of noncaregivers. Lower rates of health care utilization by caregivers have been reported previously (
Pruchno and Potashnik 1989), including fewer hospital days and physician visits than community-matched controls. Because of the tremendous amount of time devoted to informal care provision and the high cost of care, many caregivers may ignore their own health and health care; their health care utilization may therefore be lower than the comparison group of noncaregivers. Previous research suggests that the time demands of caregivers may prevent them from seeking needed medical care until they believe it is absolutely necessary, and that caregivers neglect their own self-care needs whenever caregiving demands increase (
Snyder and Keefe 1985). This may mean that excess health care costs of the caregivers might not materialize until much later, perhaps following death or institutionalization of the care recipients.
Results of this study suggest that all aspects of informal costs rise with disease progression. Most of this increase comes from the increased cost of caregiver time, especially from providing physical care. This is not surprising, as the housekeeping and personal management tasks correspond closely to the instrumental activities of daily living (IADLs). By the time the care recipients begin to develop basic ADL limitations, they are likely to also have IADL limitations and problems with all aspects of personal management.
The dementia patients in our sample are almost evenly distributed between having Alzheimer's disease (43.4%) and vascular dementia (56.6%). To evaluate whether informal caregiving costs differ by type of dementia, we reestimated caregiving costs for AD caregivers (n = 898) and VAD caregivers (n = 1145) separately. We tested for differences by dementia type in all cost components estimated in the study. No significant differences in informal care costs were found for AD and VAD caregivers.
The magnitude of the variances of the forecasting errors is of concern. We report consistent estimates of the variance of the forecasting errors, correcting for the "distance" between the characteristics of the noncaregiver comparison samples and the characteristics of the NLCS caregivers. Not correcting the bias that results from out-of-sample prediction did not lead to large reductions in the variance of the forecasting errors. This indicates that most of the variances in our cost estimates come from sizable variations in the data.
Several data limitations are of note. First, although the sample used in this study is drawn from a large, national survey, it is not a nationally representative sample of caregivers. All patients in this sample are elderly male veterans, a subgroup that makes up about 60% of the entire male population of the same age. If these veterans and their caregivers have unique characteristics that are associated with the aspects of caregiving costs we examined, then our estimates will be biased. Our sample may also have selected slightly more spouse caregivers than the female dementia caregivers in the general population. If spouse caregivers spend more time than other caregivers providing care to dementia patients, then our estimate of the cost of caregivers' time may be overstated. On the other hand, if the spouse caregivers are less likely to purchase services than other caregivers, then our estimate of the cost of formal care may be understated. The net effect of the potential bias in our estimate is unclear. Second, due to their veterans health care eligibility, coverage by public health insurance is much higher in our sample than among members of a nationally representative sample such as the NHIS. Caution is needed in generalizing the results of this study to patients and caregivers with lower rates of public health insurance coverage. Third, the majority of caregiving research has relied on self-selected caregiver samples that tend to have higher levels of education and income. The NLCS sample includes a broader spectrum of caregivers that have relatively lower levels of income and education. Conversely, the NLCS provides an opportunity to analyze the burden of dementia care among a large segment of U.S. caregivers with relatively few personal and financial resources. Fourth, reductions in work hours due to caregiving and hours spent on caregiving may overlap and lead to possible double counting of caregiving costs. However, determination of this overlap is beyond the scope of this study. Our future work will exploit the longitudinal design of the NLCS survey to examine the magnitude of the overlap and obtain more accurate caregiving cost estimates.
This study is the first in a sequence of studies of the informal costs of dementia care. The findings come from a cross-sectional accounting of the various components of informal care costs and indicate that these costs are positively associated with disease severity and the presence of dementia-related behaviors. An understanding of the marginal relationship between informal cost and dementia severity is crucial to accurately timing interventions aimed at alleviating economic burdens associated with dementia care. However, because these data are cross-sectional, little can be said about informal cost–disease progression trajectories. For example, when and under what conditions in the course of dementia do the various types of informal costs estimated in this study lead certain caregivers to relinquish care to institutions? Can we identify such key thresholds? Controlling for disease severity, are such decisions made earlier in the disease course if informal costs rise? Our future investigations will examine these questions by exploiting the longitudinal design of this ongoing study.
|
Acknowledgments |
|---|
Received for publication June 21, 2000. Accepted for publication December 20, 2000.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  C. W. Zhu, N. Scarmeas, R. Torgan, M. Albert, J. Brandt, D. Blacker, M. Sano, and Y. Stern Longitudinal study of effects of patient characteristics on direct costs in Alzheimer disease Neurology, September 26, 2006; 67(6): 998 - 1005. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. W. Zhu, N. Scarmeas, R. Torgan, M. Albert, J. Brandt, D. Blacker, M. Sano, and Y. Stern Clinical features associated with costs in early AD: Baseline data from the Predictors Study Neurology, April 11, 2006; 66(7): 1021 - 1028. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. L. Douglas and B. J. Daly Caregivers of Long-term Ventilator Patients: Physical and Psychological Outcomes Chest, April 1, 2003; 123(4): 1073 - 1081. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Sabir, K. Pillemer, J. Suitor, and M. Patterson Predictors of successful relationships in a peer support program for Alzheimer's caregivers American Journal of Alzheimer's Disease and Other Dementias, March 1, 2003; 18(2): 115 - 122. [Abstract] [PDF]
|
|
|
|
|
|
D.L. Murman, Q. Chen, M.C. Powell, S.B. Kuo, C.J. Bradley, and C.C. Colenda The incremental direct costs associated with behavioral symptoms in AD Neurology, December 10, 2002; 59(11): 1721 - 1729. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. M. Langa, S. Vijan, R. A. Hayward, M. E. Chernew, C. S. Blaum, M. U. Kabeto, D. R. Weir, S. J. Katz, R. J. Willis, and A. M. Fendrick Informal Caregiving for Diabetes and Diabetic Complications Among Elderly Americans J. Gerontol. B. Psychol. Sci. Soc. Sci., May 1, 2002; 57(3): S177 - 186. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
