| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Bernard S. Bloom, University of Pennsylvania, Department of Medicine, Division of Geriatrics, 3615 Chestnut Street, Philadelphia, PA 19104-2676. E-mail: bsbloom{at}mail.med.upenn.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Alzheimer's disease • Dementia • Cost • Economics
Cost-of-illness (COI) studies are useful to understand the economic burden of illness on patients, families, health care systems, payers, and society. They are the economic equivalent of epidemiological surveys of prevalence, incidence, and natural history of a disease, and they are typically the initial step in an economic evaluation.
The burdens of Alzheimer's disease (AD) on health care systems, families, and social services are widely recognized in the scientific and popular literature. Since AD is a disease of senescence, population aging ensures increasing prevalence and intensifying challenges for policy makers (Brookmeyer, Gray, & Kawas, 1998; Jorm, Korten, & Henderson, 1987; Lobo et al., 2000; Ritchie, Kildea, & Robine, 1992; Trabucchi, 1999). Current U.S. estimates of prevalence range from 1.09 to 4.58 million people, affecting between 2.7% and 11.2% of the population aged older than 65 years (Brookmeyer et al., 1998; Trabucchi, 1999). Prevalence is estimated to increase nearly four-fold to between 4.2 and 15.4 million whereas annual incidence is expected to rise from 360,000 new cases in 1997 to 1.14 million in 2047 (Brookmeyer et al.). As prevalence increases, so will expenditures for this chronic and progressive disease for multiple family, medical, social, and community resources (Ernst & Hay, 1997; Meek, McKeithan, & Schumock, 1998). However, reported annual U.S. estimates of AD cost vary five-fold, from $20 billion to $100 billion (Ernst & Hay, 1994; Leon, Cheng, & Neumann, 1998; Menzin, Lang, Friedman, Neumann, & Cummings, 1999; Rice et al., 1993; Schumock, 1998). A majority of costs are for long-term care, reduced function, and family care giving (Ernst & Hay, 1994, 1997; Max, Webber, & Fox, 1995; Rice et al., 1993).
Most people with AD are elderly and often have multiple high-cost diagnoses. Thus, allocating precisely direct, indirect, and nonmedical costs of AD across all diagnoses for the same person is particularly difficult. This is the problem of joint product, wherein it is difficult to attribute specific medical services and their costs, in part or in whole, to each diagnosis. Accurate attribution is especially problematic in estimating AD costs as there are few, only modestly efficacious medical treatments with mainly short-term effects (Winblad & Wimo, 1999).
The goal of this study was to examine the usefulness of published cost-of-illness estimates of AD. Accurate cost data are fundamental for informed policy and clinical decisions by patients, physicians, and payers to meet clinical, quality-of-life, and economic objectives.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
We compared results and conclusions of all studies, examining similarities and differences in study methods, data collected, cost-calculation processes, sources used, and cost estimates. We collected epidemiological estimates of AD prevalence and incidence. We collected or calculated from each study annual total and costs per person with AD (per case costs); we calculated incremental costs of AD for controlled studies. Costs attributable to AD, or any diagnosis, are difficult to estimate in the face of multiple diagnosis. For example, it is nearly impossible to separate the components of an ambulatory visit or an inpatient stay into those costs specifically attributable to each diagnosis for the individual patient. The best alternative is to estimate incremental AD costs by examining in two comparable cohorts, one with and one without AD, all costs of care. The differences (incremental) in costs can be defined as those of AD. We found two types of controlled studies—(a) those that measured costs in populations with and without AD, and (b) those that compared costs of care for people with AD living in an institution to people with AD living in the community.
For all studies, the total number of people with AD was the denominator. For U.S. studies, we adjusted costs by sector specific (e.g., hospital, pharmaceuticals) medical care price indices, 1982–1984 = 100 (U.S. Bureau of the Census, 1999). Non-U.S. study costs were expressed in that country's currency, adjusted for their inflation rates (1982–1984 = 100). Further, we adjusted costs across countries to U.S. dollars by purchasing power parity (PPP) for the year of each country's data (World Bank Group, 2000). PPP is a technique that allows for accurate comparisons across countries of income and cost of purchasing goods and services. For example, a metric can be the number of hours a person must work to purchase a kilowatt of electricity or an ambulatory physician visit. PPP is accepted as the most accurate way to measure the ability of the currency of one country to purchase goods and services in other countries.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Definition of Alzheimer's Disease
We used each study's authors' definition of AD. Fourteen studies included all types of dementia in their analysis. Four studies used secondary sources like paid claims data and relied on ICD recorded coding (Gutterman, Markowitcz, Lewis, & Fillit 1999; Menzin et al., 1999; Taylor & Sloan, 2000; Weiner, Powe, Weller, Shaffer, & Anderson, 1998). Four studies relied on the Mini-Mental State Examination (Hux et al., 1998; Østbye & Crosse, 1994; Rice et al., 1993; Souêtre, Thwaites, & Yeardley, 1999), two used National Institute of Neurological and Communicative Diseases and Stroke criteria (Kronborg-Andersen et al., 1999; Leon et al., 1998), and one used results of a randomized control trial (O'Brien et al., 1999). Five studies gave no specific source of disease definition. Adjustment for AD severity was made in 10 studies (Ernst & Hay, 1997; Kinosian et al., 2000; Kronborg-Andersen et al., 1999; Leon et al., 1998; Max et al., 1995; Rice et al., 1993; O'Brien et al., 1999; Schneider, Kavanaugh, Knapp, Beecham, & Netten, 1993; Sôuetre et al., 1999; Taylor & Sloan, 2000). Five studies adjusted for comorbidity (Gutterman et al., 1999; Kavanaugh & Knapp, 1999; Kinosian et al., 2000; Leon et al., 1998; Menzin et al., 1999).
Medical and Cost Data Collected
There was wide variability of cost data items across studies. Additionally, service definitions were not uniform. For example, hospital costs may have included, individually or in varying combinations, acute, short-term, long-term, mental, and/or rehabilitation inpatient care. Indirect costs for activities like caregiver time were estimated by opportunity cost or replacement cost methods, and different average wages were used for each cost estimation process. Finally, primary and secondary data sources often were intermingled.
Data Sources
Eight studies were of national or large populations (Coon & Edgerly, 1999; Gray & Fenn, 1993; Hux et al., 1998; Kavanaugh & Knapp, 1999; O'Brien et al., 1999; O'Shea & Reilly, 2000; Taylor & Sloan, 2000; Wimo, Karlsson, Sandman, Corder, & Winblad, 1997), and the other 13 were of selected or random population samples. Thirteen (63.8%) studies used secondary data sources. Some used national data, like U.S. Medicare, or from all of Sweden, and some regional data like those from northern California Medicaid.
Among the 11 control studies, 8 used matched cohort designs, and 1 was based on a previously published randomized control trial. Eight of the control studies compared costs of persons with AD with costs of those without AD and 3 estimated incremental cost of people with AD living in the community compared with those in nursing home. Age and gender were the most common matching criteria in all control studies.
Prevalence of Alzheimer's Disease
Wide ranges of AD prevalence estimates characterized all studies; the greatest was observed in U.S. studies, 1.5 million to 4.3 million people (Ernst & Hay, 1994; Rice et al., 1993). In England, there were two-fold prevalence variations, 370,000 and 730,000 (Schneider et al., 1993). Some differences were clearly related to individual study definition of dementia. U.S. studies with narrowly defined AD led to lowest prevalence of 1.5 million (Ernest & Hay, 1994), and more inclusive diagnoses of all types of dementia to 4.3 million (Rice et al., 1993). But as about 70% of all types of dementia are estimated as Alzheimer's type, the large variation is not well accounted for by multiple dementia diagnoses.
Cost of Alzheimer's Disease—Adjusted for Inflation and Purchasing Power Parity
Controlled Studies
Four of the eight control studies comparing cost of care for people with and without AD were of U.S. populations (Gutterman et al., 1999; Menzin et al., 1999; Taylor & Sloan, 2000; Weiner et al., 1998), two were from Canada (O'Brien et al., 1999; Østbye & Crosse, 1994), and one each from the United Kingdom (Souêtre et al., 1999) and Denmark (Kronborg-Andersen, 1999; Table 1). After adjusting for inflation among U.S. studies, and inflation and PPP among non-U.S. studies, annual incremental AD costs across control studies varied from about $1,500 (Weiner et al.) to $79,000 per case (O'Brien et al.).
|
Much of the cost variance across studies was due to differences in types of care cost collected. For example, four studies did not collect nursing home and/or indirect and/or family costs (Gutterman et al., 1999; Kronborg-Andersen et al., 1999; Menzin et al., 1999; Weiner et al., 1998; Table 1). In one U.S. study, nursing home care was 87.4% of per capita costs but no indirect and/or family costs were collected (Menzin et al., 1999). In one Canadian study, nursing home care was 76.3% of annual costs and indirect and/or family expenses were 19.8% of costs (Østbye & Crosse, 1994). One high-cost British study found an annual cost of PPP $27,675 per case; it collected direct medical, nursing home, and indirect expenses (family out-of-pocket cost, caregiver time away from employment plus reduced work productivity); the latter expenditures represented 71.3% of estimated costs (Souêtre et al., 1999).
Uncontrolled Studies
Studies without control populations exhibited the widest range of annual per case AD costs, and with even less consistency among specific sectors costs than for controlled studies (Table 2; Cavallo & Fattore, 1997; Ernst & Hay, 1994; Gray & Fenn, 1993; Huang, Cartwright, & Teh-Wei, 1998; Hux et al., 1998; Kinosian et al., 2000; Leon et al., 1998; O'Shea & O'Reilly, 2000; Weinberger et al., 1993; Wimo, Karlsson, et al., 1997). Annual inflation-adjusted costs per case among U.S. studies varied from about $12,000 (Leon et al., 1998) to $91,000 (Huang et al., 1998; Table 2). Among non-U.S. studies, inflation- and PPP-adjusted annual per case costs varied from PPP $2,300 (Weinberger et al., 1993) to PPP $30,000 (Wimo, Karlsson, et al., 1998).
|
The study with the highest AD cost estimated inflation-adjusted U.S. costs for 4.28 million people at $41 billion, or $91,000 per case (Huang et al., 1998). Direct medical costs were 47.4% of the total and included short- and long-stay hospitalizations, physician and other provider services, and prescription pharmaceuticals; nursing home care was 43.2% and indirect and/or family care was 9.4% of total costs.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The singular finding of this study is the widely divergent annual per case cost estimates of AD ($1,500–$91,000 per case), with inflation-adjusted U.S. estimates ranging from $5.6 billion to $88 billion per year. Controlled and uncontrolled studies have similar cost variability irrespective of country.
Similar wide cost variations were found in an extensive review of 110 U.S. inflation adjusted cost-of-illness studies across 80 diagnoses (Bloom, Bruno, Maman, & Jayadevappa, 2001) and of a review of diabetes mellitus cost-of-illness studies specifically (Pagano, Brunetti, Tediosi, & Garattini, 1997). Both studies concluded that low quality of methods and study design were the most important factors for cost variations and low quality scores. The present intensive case study of AD confirms the findings of the earlier studies. This AD study provides additional opportunities to examine causes of variation in costs because there are a relatively large number of published studies, care for people with AD is expensive, and direct and indirect costs are frequently estimated.
Diversity of cost components collected and methods of monetarizing cost also have important effects. For example, for studies that measured indirect and/or family costs, some used actual or mean wages to value time lost from work and others used opportunity cost. In addition, variable estimates of Medicare cost may be due to the fact that Medicare data appear to underreport diagnoses and overreport payments for AD (Newcomer, Clay, Luxenberg, & Miller, 1999). Such inconsistencies are likely problems related to joint product—the difficulty of distributing costs accurately across diagnoses during an episode of illness, especially for studies without control populations.
Multiple study enrollment processes (i.e., precision of AD diagnosis) confounded results. There is frequently inadequate differentiation between cost of AD by controlled studies and cost of caring for people with AD from uncontrolled designs even after estimating costs per case and adjusting all costs to a common index. As a consequence of variations in disease definition and case ascertainment, the distributions of disease severity for AD and other diseases may vary substantially among population-based studies, and between population studies and those using selected samples.
Despite expectations of decreased costs under managed care, their per case expenditures are nearly two-fold greater than those of nonmanaged care, without appreciable differences in type of services included or methods used (Taylor & Sloan, 2000; Weiner et al., 1998). But to the contrary, Rice and colleagues (1993), found that annual hospitalization costs for AD added $504 per person for an additional 1.34 inpatient days per annum and $233 for 7.8 more physician visits, both paid mainly by Medicare (Rice and colleagues). Prescription pharmaceuticals added $185 per year.
The challenge of AD cost variability is compounded by variability of epidemiological estimates of AD incidence and prevalence. Lack of agreement on the definition of AD and absence of consensus on a gold standard diagnostic test necessarily leads to inconsistent criteria for diagnosing and coding AD and likely contributes to four-fold differences in U.S. AD prevalence estimates.
Despite the wide cost variations we are able to discern common patterns across studies. Long-term care and family costs are the most important determinants of total cost. AD patients have higher direct and indirect costs than people without AD. Costs increase with disease severity and decline of any activity of daily living. When those with mild or moderate AD severity are cared for at home, indirect costs are higher than in nursing home. Although some uncontrolled U.S. and non-U.S. studies find care at home more costly or similar to institutionalized care, all controlled studies find institutional care more costly than home care. Comorbid conditions may explain some of the higher hospital and physician costs of AD patients compared with those without AD. For example, incremental medication costs between those with and without AD, however, were less than $300 per year in any country, supporting the findings on incremental medication costs of Rice and colleagues (1993). Finally, when similar methods and populations are used, irrespective of country and controlled and uncontrolled design, similar results are produced (Kinosian et al., 2000; Max et al., 1995; O'Brien et al., 1999; Rice et al., 1993). A simulation analysis from the Netherlands (McDonnell et al., 2001) that was published after our analysis was complete and that used similar methods to those of O'Brien and colleagues (1999) and Kinosian and colleagues (2000), found 10-year results comparable to both these studies.
Some researchers contend cost of illness studies offer little to decision makers and may lead to inadequate policies because of poor data, lack of outcomes results, or no incremental analysis (Drummond, O'Brien, Stoddart, & Torrance, 1997). However, there are criteria for executing COI studies (Drummond et al., 1997; Jefferson, Demichelli, & Mugford, 2000), many of which are fulfilled in a few of the studies we reviewed (e.g., Kinosian et al., 2000; O'Brien et al., 1999). Unfortunately, adherence with these guidelines is inconsistent within studies. Methodological standards, though, tend to be less frequently discussed or debated for COI than for cost-effectiveness studies (Gold, Siegel, Russell, & Weinstein, 1996). Inadequate studies thus are due mainly to lack of implementation of existing standards. This burden is added to the central AD problem: Lack of a uniformly accepted definition of AD and of a gold standard diagnostic test contribute to highly variable incidence, prevalence, and cost estimates. Thus, future studies of AD COI should include available performance methodological standards for epidemiological, clinical, and economic evaluations. Further, they should be long-term cohort studies samples starting perhaps at age 55. Such studies would likely require 10–15 years to complete. Some parts of such a study are already underway, for example, the AHEAD study that measures indirect and family costs of AD.
Limitations
The most important limitation of this study is the inability to estimate clearly effects of individual factors that contribute to variability of AD cost results. An inability to disentangle joint product among all services compounds this problem (Drummed et al., 1997). Few of the studies adjusted for comorbidity, which appears to lead to overestimating incremental AD cost (Lee, Meyer, & Clouse, 2000). In addition, we could not test accurately for factors associated with cost variation because of small cell sizes relative to the number of independent variables. Third, limiting reviews to English-language studies may introduce bias even in the face of our including English language studies from non-English speaking countries. Finally, as with any literature review we could only review the publication, not the actual study. A poor evaluation may be due to inadequate method or analysis, incomplete reporting by the authors, or space limits by the journal.
Conclusion
COI studies offer estimates of current status of a disease, measure changes over time (e.g., following introduction of new treatments) and generate new research hypotheses. Like clinical research, we find that controlled studies lead generally to more consistent estimates than uncontrolled ones.
One problem that bedevils uncontrolled studies is the difficulty of differentiating between cost of AD and cost of all care for people with AD even when authors attempt to separate services and costs into specific diagnoses. Second, controlled studies have a tighter distribution of total and individual sector costs than uncontrolled studies, but an equally wide per case cost distribution. Similar wide variation is found for epidemiological estimates of AD prevalence and incidence. Next, AD cost variation narrows across time and countries once we control for inflation and PPP. Fourth, AD entails relatively low direct medical costs because there are no highly effective medical treatments that importantly alter symptoms or long-term natural history of disease. However, social, family, and daily function changes, work loss, and early mortality make AD economically burdensome. Of one thing we can be certain, most if not all of future medical and nonmedical AD costs will be higher than current ones in the absence of any effective prophylaxis or treatment.
|
Footnotes |
|---|
1 Department of Medicine, Division of Geriatrics, University of Pennsylvania, Philadelphia.
2 Université de Paris V, Hôpital Cochin, Paris, France.
3 Outcomes Research and Management, Merck & Company, West Point, PA.
Received for publication February 27, 2002. Accepted for publication September 6, 2002.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  S. C. Burgener and B. Berger Measuring perceived stigma in persons with progressive neurological disease: Alzheimer's dementia and Parkinson's disease Dementia, February 1, 2008; 7(1): 31 - 53. [Abstract] [PDF]
|
|
|
|
 |
|
 J. Lopez-Bastida, P. Serrano-Aguilar, L. Perestelo-Perez, and J. Oliva-Moreno Social-economic costs and quality of life of Alzheimer disease in the Canary Islands, Spain Neurology, December 26, 2006; 67(12): 2186 - 2191. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
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]
|
|
|
|
 |
|
 D. M. Fick, A. M. Kolanowski, J. L. Waller, and S. K. Inouye Delirium Superimposed on Dementia in a Community-Dwelling Managed Care Population: A 3-Year Retrospective Study of Occurrence, Costs, and Utilization J. Gerontol. A Biol. Sci. Med. Sci., June 1, 2005; 60(6): 748 - 753. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. Feldman, S. Gauthier, J. Hecker, B. Vellas, M. Hux, Y. Xu, E.M. Schwam, S. Shah, and V. Mastey Economic evaluation of donepezil in moderate to severe Alzheimer disease Neurology, August 24, 2004; 63(4): 644 - 650. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. S. Bloom, S. Chhatre, and R. Jayadevappa Cost effects of a specialized care center for people with Alzheimer's disease American Journal of Alzheimer's Disease and Other Dementias, July 1, 2004; 19(4): 226 - 232. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
