International Journal of Technology Assessment in Health Care, 7:2 (1991), 209-219. Copyright © 1991 Cambridge University Press. Printed in the U.S.A.
ECONOMIC EVALUATION OF A SUPPORT PROGRAM FOR CAREGIVERS OF DEMENTED ELDERLY Michael F. Drummond University of York
E. Ann Mohide Michelle Tew David L. Streiner McMaster University
Dorothy M. Pringle University of Toronto
J. Raymond Gilbert McMaster University
Abstract An economic evaluation was undertaken concurrently with a randomized trial comparing a Caregiver Support Program (CSP) with existing conventional community nursing care for those caring for elderly relatives at home. The differences in resource consumption were compared with changes in caregiver quality of life, as measured by the Caregiver Quality of Life Instrument (CQLI). A 20% difference from baseline in the CQLI favored the experimental (CSP) group, although this did not reach conventional levels of statistical significance. A comparison of improvement in quality of life with costs implies an incremental cost per quality-adjusted life year gained of Canadian $20,000 for the CSP, which compares favorably with other health care interventions. Further, larger studies are required to confirm this result.
In many countries, there is increased emphasis on community care for the elderly. While this can be beneficial to those elderly wishing to remain in their own homes, there This work was supported by a grant from the Ontario Ministry of Health, Research and Planning Division, 1986-1989. During this study, E. A. Mohide was a Career Scholar with The R. Samuel McLaughlin Centre for Gerontological Health Research; at the time of funding, D. M. Pringle was Research Director for VON Canada. Address reprint requests to: E. Ann Mohide, McMaster University, Faculty of Health Sciences, 1200 Main St. W., Room 2J40f, Hamilton, Ontario L8N 3Z5, Canada. We are grateful for the cooperation and encouragement of the Hamilton-Wentworth and Halton Branches of the VON, the Hamilton-Wentworth Visiting Homemakers Association and the Nursing Divisions of the Hamilton-Wentworth and Halton Departments of Health Services, and the Alzheimer Society for HaltonWentworth, Ontario, where this research was conducted.
Drummond et al.
are concerns that community care places an additional burden on the informal caregiver (3). Consequently, there has been recent interest in developing caregiver support programs, where additional services are provided to assist those family members or friends caring for the elderly in the community. However, given the pressures on health sector budgets, it is increasingly important to demonstrate that the benefits of new services justify their costs. This can be accomplished by an economic evaluation, a method by which the relative costs of alternative health care programs are compared with their relative consequences in terms of savings in other health care resources and improvements in health (6). One form of economic evaluation compares the relative monetary value and benefits from health care interventions in terms of cost per quality-adjusted life year (QALY) gained (24;28). A QALY is calculated by multiplying the changes in quality of life or utility by the number of years of life for which the change is experienced. Quality of life is measured on an interval scale from (0) dead to (1) complete health or- well-being. Therefore, an improvement of 0.1 on the quality of life scale for 1 year is equal to 0.1 QALY gained. Economic evaluation has been widely applied to health care, including geriatrics (25). For example, it has been shown that community care for elderly persons of cer-. tain levels of dependency is more cost-effective than alternative forms of institutional care (15;29). However, Weissert et al. (27) showed that not all community care programs are cost-effective, and Weissert (26) outlined seven reasons why this may often be the case. Very few economic evaluations have been undertaken of community care programs for the demented elderly. Kyle et al. (12) found that a comprehensive care package, comprising psychiatric nurse visiting, daycare, and inpatient respite care was of lower cost to the statutory agencies than institutional care. However, they did not address the issue of the burden on caregivers or the evaluation of programs to support caregivers. This article reports the first economic evaluation of a support program for caregivers of the demented elderly. Because it is based on a randomized trial (13), it provides the first evidence of whether such programs represent a good use of health service resources. METHODS Study Design An economic evaluation was carried out comparing the costs and consequences of the Caregiver Support Program (CSP) with those of conventional community nursing care. The main additional costs of the CSP were for nursing visits and the provision of respite care in the home. For the purpose of the economic evaluation, it was hypothesized that the main consequences would be improvement in the quality of life of the caregiver and reduction in the use of other services. The economic evaluation was conducted alongside a controlled clinical trial. Caregiver-patient pairs were recruited from an urban center in southern Ontario and were block-randomized using a computer-generated random numbers table to receive the experimental set of interventions, either the CSP or conventional community nursing care, for a 6-month period. The decision to offer a form of treatment to the control group was made because subjects probably would not have agreed to be assigned to a "no-care" alternative, especially those who were receiving some form of community care at the time of recruitment. The care provided to the control group was similar to that currently available in the locality studied. The length of the intervention period 210
INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
Economic evaluation of a caregiver support program
was based on the judgments of five geriatric clinicians that an overall effect should be demonstrated by this time. Subjects and Eligibility In order to be eligible for enrollment in the trial, both the caregiver and the cognitively impaired relative had to satisfy a number of criteria. Caregivers were eligible if they (a) lived with the demented relative, (b) identified themselves as the principal family member providing day-to-day care to the relative, (c) spoke English, (d) scored 8 or more on a short mental status questionnaire (19), (e) did not suffer from a life-threatening illness that would prevent them from continuing in the caregiver role over the length of the trial, (0 agreed to have nursing services provided by a specific visiting nursing agency, and (g) gave written consent to participate. Relatives were eligible if they (a) had a medical diagnosis of primary degenerative, multiinfarct, or mixed dementia, (b) were assessed as moderately to severely impaired (18), (c) did not suffer from a serious concomitant illness(es) that would be likely to result in acute or long-term hospitalization or death during the intervention period, and (d) were not likely to be placed in a long-term care setting during the course of the study. The relatives were rated using the Dementia Rating Scale (DRS) (2) and the MiniMental State Examination (MMSE) (9). Following this testing, a geriatric nurse clinician, who was blind to the DRS and MMSE test scores, rated the relative using the Reisberg Global Deterioration Scale (18). Those with eligible scores on at least two of the three scales (8 or more on the DRS, 19 or less on the MMSE, stage 5 or greater on the Reisberg Scale) had their medical records reviewed to confirm the diagnosis of progressive irreversible dementia. Caregivers were stratified on the basis of gender and whether or not the demented relative attended a day program to ensure that the groups were balanced with respect to these two variables. In the latter case, severely burdened caregivers might be more likely to have their relative referred to a day program than those who are less burdened. Description of the Interventions The experimental set of supportive interventions was directed at helping the caregivers to enhance their competence at giving care and achieve a sense of control in their roles as caregivers. Caregiver support nurses (CSNs) were assigned to caregivers, and made regularly scheduled home visits at a time that was convenient to the caregivers. Initially these visits were weekly but were adjusted upward or downward depending on the needs of the caregiver. Health assessments were completed on all caregivers using forms designed to encourage the assessment of common caregiver problems. Caregivers were encouraged to seek medical attention for neglected health problems, such as hypertension. The CSNs consulted with the caregivers' family physicians on a regular basis whenever the caregivers' health was judged to be unstable. The caregivers received education about dementia and caregiving using content and teaching methods tailored to their knowledge level, caregiving situation, and learning style. The CSP intervention included a 4-hour block of scheduled weekly in-home respite, with additional respite on demand. The respite workers were assigned to specific families and were encouraged to discuss the demented patients' plans of care with the assigned CSN. The weekly respite permitted the caregivers to anticipate relief from the rigors of caregiving on a consistent basis, and respite on demand provided relief between periods of scheduled respite. INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
211
Drummond et al.
Finally, caregivers and other interested family members were encouraged to attend a monthly 2-hour self-help support group specifically for families with demented relatives. In the control group, conventional (existing) community nursing focused on care of the demented patient, rather than care of the family caregiver. This care addressed such physical needs of the demented patient as bathing, grooming, managing incontinence, and monitoring medication. Although some time might be spent discussing the patient's behavioral problems with the caregiver, this was not the main focus on the visit. The nurses in the control group were assigned specific patients on the basis of geographical location and the frequency of visits was left to the nurses' discretion. The nurses in the control group did not have concentrated training in this area and did not use the consultative services of psychogeriatric clinicians. Caregivers were not recommended to receive other health and social services in a standardized or systematic manner. Outcome Measures
Ideally, the primary outcome variable for a trial such as this should be a direct measure of the burden on caregivers. Accordingly, the psychometric properties of several measures of burden (10;20;30) were appraised, but none of the measures had been shown in previous studies to be responsive to change. Consequently, it was necessary to select instruments specific to symptoms. Because negative emotional effects, in particular depressive and anxious symptomatology, were cited frequently in the literature (1;7;8;11); these were used as the major outcomes of interest. Measures with sound psychometric properties were selected: the Center for Epidemiologic Studies Depression Scale (CES-D) (17) and the state-anxiety portion of the State-Trait Anxiety Inventory (STAI) (22). When responding to the scale, subjects were asked to consider the past week in order that the scale be responsive to changes in circumstances and yet not be overly dependent on daily fluctuations in mood. This modification is consistent with the CES-D Scale response interval of 1 week. In addition to these two major measures of outcome, the caregiver's quality of life was assessed using the Caregiver Quality of Life Instrument (CQLI) (14), an application of the time trade-off technique for measuring health status (23). Using this approach, the subjects decided how much length of life (expressed in future years of life) to trade off for improved quality of life (expressed in terms of dimensions of health, e.g., physical, psychological, or social function). To achieve this, subjects were asked to choose between being in two alternative health states for different periods of time. This approach gives valuations of health states, or quality of life, on an interval scale from 0-1. The CQLI was particularly relevant to the economic evaluation as it enabled the calculation of the QALYs gained by the caregivers from the intervention. Cost Items
A wide range of resources was recorded. These included nursing visits and respite care provided by the study, as well as those services sought by the caregivers, such as physician visits, homemaker services, day programs, overnight respite (both in the home and in institutions), visits by social workers, and other paid help including some companion services. In addition, days of acute hospital care were recorded, although it was recognized at the outset that these events were too infrequent to draw any strong inferences about the impact on them of the CSP. There were a number of sources of financial data reflecting the wide range of agencies involved in the program. These included the records of relevant health and social 212
INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
Economic evaluation of a caregiver support program
agencies and interviews with the caregivers themselves. The economic evaluation concentrated on the resources that could be unambiguously expressed in monetary terms and compared them with changes in caregiver quality of life as measured by the CQLI. A further article will consider the nonfinancial costs of caring, such as forgoing leisure activities in order to give care. Data Collection and Analysis
The research assistant interviewed the subjects at entry, at midpoint (3 months), and upon completion of the intervention at 6 months. If placement in long-term care was anticipated, an end-of-study interview was conducted prior to the withdrawal wherever possible. The study was completed by 22 subjects in the experimental group, and 20 in the control group completed the study; 10 demented relatives were admitted to long-term care facilities (5 in each group), 6 were admitted to a hospital for longer than 1 month, one caregiver died, and one couple moved out of the study catchment area. Our primary analysis is based on those who completed the study. This group reflected the impact of the program over a fixed period of time and provided results for the major outcome measures without the added variables of several major life events (such as long-term institutionalization, major illness, and unstable health of the demented relative). Exclusion of the subjects from the primary analysis was conservative since, although the number of relatives admitted to long-term facilities was equal between the groups, those in the control group were admitted for longer periods of time. The demographic data were analyzed using a t test and chi-square. The major effectiveness analyses were conducted using multivariate analysis of covariance (MANCOVA) with between-groups and within-group independent variables (intervention and time, respectively), and three dependent variables (CES-D, STAI, CQLI). Analysis of covariance (ANCOVA) was used to examine individual outcome measures. Pearson's r was used to examine correlations and stepwise regression to examine the variance accounted for by specific variables. The data on cost were analyzed by the Mann-Whitney U test (20); a nonparametric statistic for the comparison of two independent means was used. Differences in overall cost can result from differences in the physical qualities of resources consumed, or differences in the unit prices of those resources. Since some unit prices, such as those for daycare programs, could potentially vary by location, rather than as a result of the CSP, statistical tests were applied to the physical quantities of resources as well as the overall financial amounts. The incremental cost of the CSP was then compared with the improvement in quality of life (as measured by the CQLI), and the incremental cost per QALY calculated.
RESULTS
Sixty caregivers were enrolled in the trial. Table 1 provides the baseline characteristics of the 30 subjects allocated to each group. The groups did not differ on most of the variables. There was a statistically significant difference between the two groups in the caregivers' perception of the length of caregiving. Three control subjects had extremely long periods of caregiving in comparison to the other subjects; only two of the three completed the trial. Overall, the experimental group had a higher annual family income than the control group. In addition, there was a marginally significant INTL. J. OF TECHNOLOGY ASSESSMENT IN.HEALTH CARE
213
Drummond et al. Table 1 . Characteristics of the Caregivers and Their Relatives at Baseline Control group (» = 30)
Experimental group (» =: 30)
Characteristics Caregiver Number of females Number of spouses Wife Husband Mean age Mean months caregiving Number with some or all of high school Mean annual income Number of moderate to extremely healthy (self-rated) Number using psychotropic drugs Demented Relative Number of females Mean age Mean DRS score Number at stage 5 or 6 on the Reisburg Global Deterioration Scale Mean MMSE score Number with Alzheimertype dementia Mean Barthel score
No.
No.
(% or SD)
(% or SD)
21
(70%)
22
(73%)
14 7 66.10 39.80
(47%) (23%) (13.47) (30.23)
17 8 69.40 72.40
(57%) (27%) (8.61) (61.93)
23 $27,607
(77%) (19,708)
18 $8,155
(60%) (11,930)
22
(73%)
20
(67%)
3
(10%)
3
(10%)
15 77.80 8.75
(50%) (9.19) (3.01)
14 75.90 10.75
(47%) (7.70) (4.75)
27 13.21
(90%) (6.83)
26 11.04
(87%) (6.17)
19 68.67
(63%) (18.1)
21 62.50
(70%) (24.52)
p value
.013 .035
.057
difference in the mean DRS score, although this does not appear to have been clinically important. Effectiveness analyses were conducted on those who completed the trial. The mean scores of the major outcome variables used in this analysis are provided in Table 2. The covariates used in the MANCOVA included the baseline values for the major outcomes and the variables that differed between groups (annual family income and perceived length of caregiving). The three outcomes (CES-D, STAI, and CQLI) were examined individually using ANCOVAs with the same covariates. There were no statistically significant differences for any of the analyses. For purposes of comparison, we conducted an analysis of intent to treat that included those who completed at least 18 weeks of the intervention. This did not alter the values to a clinically important extent, and there were no significant differences between the groups. Despite the constancy of the depression and the anxiety scores, the quality of the caregivers' lives, as measured by the CQLI, increased in the experimental group, while it decreased in the control group. By the end of the trial, there was a difference of 20% from the baseline scores favoring the experimental group. We judged this 20% difference to be clinically important. However, a sample size of 178 subjects per group would be required to show that this difference is statistically significant (4). 214
INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
Economic evaluation of a caregiver support program Table 2. Means (and SD) for the Major Outcomes Experimental group (n = 22)
Control group (n = 20)
Baseline
3 mos.
6 mos.
Baseline
3 mos.
6 mos.
CES-D
21.55 (11.57)
18.36 (11.78)
21.50 (12.98)
15.40 (9.21)
15.35 (7.45)
18.20 (10.05)
STAI a
50.41 (16.40)
46.68 (11.56)
49.64 (14.48)
46.25 (12.62)
46.85 (9.35)
48.50 (14.38)
CQL1
.55 (.33)
.58 (.37)
.64 (.33)
.56b (.41)
.52" (.42)
.53" (.41)
Measure
0 b
Raw scores for the STAI state-anxiety scale. n = 19; one subject did not complete the CQLI at baseline.
Table 3. Resource Costs for Individual Services Item (unit)
Unit cost (1988 Canadian $)
Payment sources'1
Nursing6 (per visit) Respite workerb (per hour) Homemaker (per hour) Day program (per day)
29.19 8.13 8.13-8.50 13.26-69.56
Physician (per visit) Overnight institutional respite (per night) Other paid help (per visit)
14.90-110.65
MCSS or MOH MCSS or MOH Caregiver or MOH MOH, MCSS, or municipality MOH
45.00-130.00 5.OO-55.OO
Caregiver or MOH Caregiver
MCSS, Ministry of Community and Social Services; MOH, Ministry of Health Since this trial was conducted, variations on the CSP have been funded in Ontario. Payment sources have been categorized according to the current funding trends. b During the period of the trial, payment arrangements were slightly different, as nursing and respite worker services were provided as part of the study.
0
The unit costs of the main resource inputs to care are given in Table 3. (All costs are expressed in January 1988 Canadian dollars, Canadian $1 = US $.83.) The ranges for some of the items, such as day programs, reflected the variety of methods of provision. In addition, Table 3 gives the usual sources of payment during regular operation of the CSP. The sources of payment are important in assessing costs and benefits by viewpoint. Our primary viewpoint considered all costs regardless of source of payment, but other viewpoints may be of interest to those considering particular governmental budgets. Table 4 gives the total costs of services received by the experimental and control groups during the 6-month intervention. As expected, those resource inputs that pertained to the CSP intervention, nursing visits to caregiver and respite worker visits, were higher for the experimental group. On the other hand, costs for private nursing, day programs, and overnight institutional respite care were higher for the control group. However, with the exception of the respite worker and private nursing, these differences were not significant, either in physical quantities of resources consumed or financially. These data can be interpreted in two ways. First, the overall differences in costs INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
215
Drummond et al. Table 4. Summary of Total Costs During the 6-Month Intervention, in January 1988 Canadian $ (Range) Service
Experimental group (n = 22)
Control group (n = 20)
Nursing visits To caregiver3 To relative
14,507(379-1,022) 12,143 (0-1,430)
1,314(0-759) 12,318 (146-1,547)
Attendant care Respite worker 3 Homemaker Private nursing Other
9,747 (0-943) 18,710(0-7,072) 0 48 (0-48)
0 11,679(0-1,691) 2,473 (0-2,006) 858 (0-858)
2,699 (0-550) 2,904 (0-757)
2,535 (0-484) 2,244 (0-406)
MD visits To caregiver To relative Other paid help
1,910(0-528)
1,198(0-660)
Day programs Overnight institutional respite Total (per caregiver)
13,169(0-4,243) 2,535 (0-1,540) 3,562
16,853(0-4,243) 6,447 (0-2,735) 2,897
Incremental cost per QALY (over control group)
20,036
-
Note: Range is given inside parentheses. Services delivered as part of the Caregiver Support Program (CSP). Two caregivers in the control group were receiving nursing visits for their own health problems prior to their participation in the trial.
a
could be taken as the net cost of adding the intervention. This amounts to $665 ($3,562 less $2,897) for 6 months, or $1,330 per annum. Alternatively, one could take the view that since the differences in individual cost items did not reach conventional levels of statistical significance, the only financial impact of the CSP is the additional annual cost of the intervention. This was $1,102 ($14,507 plus $9,747 divided by 22) over 6 months, or $2,204 per annum. (The costs of the monthly 2-hour support group meetings were trivial and have been omitted.) The more conservative approach is to take the higher incremental cost estimate of $2,204 per annum for the calculation of cost per QALY gained. The improvement in the quality of life of caregivers, as measured by the CQLI, was 0.11 over the 6-month intervention period. Over 1 year, this would be equivalent to 0.11 QALYs gained. Therefore, the implied incremental cost per QALY gained is $2,204 divided by 0.11, or $20,036. DISCUSSION
This article reports the first economic evaluation of a support program for caregivers of the elderly conducted concurrently with a randomized trial. Taking the more conservative approach, the annual incremental cost of providing the support program was $2,204 per caregiver. The CSP had no impact of any consequence on the other resources from statutory sources. Economic evaluation seeks to compare the additional costs of health care interventions with their consequences in terms of health effects or QALYs. In this case, the economic evaluation depends greatly on the interpretation placed on the results of the clinical trial. The magnitude of the differences between the groups in terms of the principal outcome measures was not clinically important. 216
INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
Economic evaluation of a caregiver support program
Therefore, statistical power was not an issue in this instance. However, a clinically important difference of 20% above baseline was observed in the CQLI, but the study did not have sufficient sample size to confirm this difference at conventional levels of statistical significance. Mohide et al. (13) pointed out that there are a number of possible explanations for finding no differences in the CES-D and STAI measures. The burdens of caring for a relative with dementia may be intractable, or the stage of the progression of the disease may have been too late for any intervention to make an impact on the emotional stress of caregivers. Alternatively, the intervention may not have been potent enough, or its effect may have been minimized because some experimental subjects did not accept all aspects of the intervention. Finally, it could be that measures of emotional symptomatology are not the optimal constructs to be used as major outcome variables; indeed, a measure that incorporates more domains while at the same time being specific to caregivers, namely, the CQLI, may be a more sensitive and relevant measure of program effectiveness. Whatever the reason for the lack of statistically significant results, the conclusions of the concurrent economic evaluation are necessarily tentative. This study alone cannot demonstrate that caregiver support programs represent good value for the money. However, it does show that the implied cost per QALY (of approximately Canadian $20,000) compares favorably with other health care interventions. Torrance and Zipursky (24) have calculated the incremental cost per QALY of a range of health care interventions. These range from $4,200 (in 1983 US $) for coronary artery bypass surgery for left main artery disease, to $54,000 for hospital hemodialysis. Allowing for inflation and converting from U.S. to Canadian dollars, the cost per QALY of the CSP falls in the middle of the range, being marginally higher than the estimate for treatment of severe hypertension (diastolic > 105 mm Hg) in 40-year-old males, which Torrance and Zipursky reported at $9,400 (1983 US $) per QALY. Further work is therefore required to confirm our results and to demonstrate, through larger studies, that caregiver support programs give good value for the money. However, this will not be an easy task. A multicenter trial is likely to be required, since in a population of 500,000 in southern Ontario, only 146 caregiver-relative pairs could be located, of which only 60 were eligible for our study. Multicenter trials of communitybased interventions like the CSP raise considerable methodological problems of comparability, coordination, and standardization of the intervention. However, our study has shown that it is feasible to design a support program for caregivers of the demented elderly and that such programs could represent good value for the money when compared with other health care interventions. A number of interesting methodological issues are raised by this study. First, the variance in quality of life measures like the CQLI is such that much larger sample sizes are required to detect a difference between means at the conventional levels of statistical significance. In this case, the CQLI was not the primary outcome measure, and the difference observed, while clinically important, was not statistically significant. Since measures like the CQLI are more relevant to economic evaluations than are more specific clinical outcome measures, it is possible that clinical trials with concurrent economic evaluations will require larger sample sizes, with consequent resource implications for clinical research. Second, considerable effort was devoted in this study to recording a wide range of resources, many of which were neither different between the two groups nor significant in terms of cost. Drummond and Stoddart (5) stressed the need for a "phasing policy" for undertaking economic evaluation alongside clinical trials, in order to minimize unnecessary work. However, with an evaluation of a new intervention, it is diffiINTL J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
217
Drummond et al.
cult to decide which items require detailed assessment and measurement. Based on the experience of this study, future evaluations of caregiver support programs may be able to screen resources to assess which should be estimated precisely. REFERENCES
1. Anthony-Bergstone, C. R., Zarit, S. H., & Gatz, M. Symptoms of psychological distress among caregivers of dementia patients. Psychology and Aging, 1988, 3, 245-48. 2. Blessed, G., Tomlinson, B. E., & Roth, M. The association between quantitative measures of dementia and of senile change in the cerebral grey matter of elderly subjects. British Journal of Psychiatry, 1968, 114, 797-811. 3. Caserta, M. S., Lund, D. A., Wright, S. D., et al. Caregivers to dementia patients: The utilization of community services. The Gerontologist, 1987, 27, 209-14. 4. Cohen, J. Statistical power analysis for the behavioral sciences. New York: Academic, 1969. 5. Drummond, M. F., & Stoddart, G. L. Economic analysis and clinical trials. Controlled Clinical Trials, 1984, 5, 115-28. 6. Drummond, M. F., Stoddart, G. L., & Torrance, G. W. Methods for the economic evaluation of health care programmes. Oxford and New York: Oxford University Press, 1987. 7. Eagles, J. M., Beattie, J. A. G., Blackwood, G. W., et al. The mental health of elderly couples. I. The effects of a cognitively impaired spouse. British Journal of Psychiatry, 1987, 150, 299-303. 8. Fitting, M , Rabins, P., Lucas, M. J., et al. Caregivers for dementia patients: A comparison of husbands and wives. The Gerontologist, 1986, 26, 248-52. 9. Folstein, M. R, Folstein, S. E., & McHugh, P. R. Mini-Mental State: A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 1975, 12, 189-98. 10. Greene, J. G., Smith, R., Gardiner, M., et al. Measuring behavioral disturbance of elderly demented patients in the community and its effect on relatives: A factor analytic study. Age and Aging, 1982, 11, 121-26. 11. Haley, W. E., Levine, E. G., Brown, S. L., et al. Psychological, social, and health consequences of caring for a relative with senile dementia. Journal of the American Geriatrics Society, 1987, 35, 405-11. 12. Kyle, D. R., Drummond, M. F, & White, D. M. D. The Hereford District Department of Mental Health of the Elderly: A preliminary evaluation. Community Medicine, 1987, 9, 35-46. 13. Mohide, E. A., Pringle, D. M., Streiner, D. L., Gilbert, J. R., et al. A randomized trial of family caregiver support in the home management of dementia. Journal of the American Geriatrics Society, 1990, 38, 446-54. 14. Mohide, E. A., Torrance, G. W., Streiner, D. L., et al. Measuring the wellbeing of family caregivers using the time trade-off technique. Journal of Clinical Epidemiology, 1988, 41, 475-82. 15. Mooney, G. H., Russell, E. M., & Weir R. D. Marginal analysis: A case study in care of the elderly. In Choices in Health Care, London: Macmillan, 1980, 77. 16. Rabins, P., Mace, N., & Lucas, M. The impact of dementia on the family. Journal of the American Medical Association, 1982, 248, 333-35. 17. Radloff, L. S. The CES-D Scale: A new self-report depression scale for research in the general population. Applied Psychological Measurement, 1977, 1, 385-401. 18. Reisberg, B., Ferris, S., DeLeon, M., et al. The Global Deterioration Scale for assessment of primary degenerative dementia. American Journal of Psychiatry, 1982, 139, 1136-39. 19. Robertson, D., Rockwood, K., & Stolee, P. A short mental status questionnaire. Canadian Journal on Aging, 1982, 1, 16-20. 20. Robinson, B. C. Validation of a caregiver strain index. Journal of Gerontology, 1983, 38, 344-48. 21. Siegel, S., & Castellan, N. J. Nonparametric statistics for the behavioral sciences, 2nd ed. New York: McGraw-Hill Book Company, 1988.
218
INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
Economic evaluation of a caregiver support program 22. Spielberger, G., Gorsuch, R., & Lushene, R. STAI manual for the State-Trait Anxiety Inventory. California: Consulting Psychologists Press, 1968. 23. Torrance, G. W., Thomas, W. H., & Sackett, D. L. A utility maximization model for evaluation of health care programs. Health Service Research, 1972, 7, 118-33. 24. Torrance, G. W., & Zipursky, A. Cost-effectiveness of antepartum prevention of Rh immunization. Clinics in Perinatology, 1984, 11, 267-81. 25. Warner, K. E., & Luce, B. R. Cost-benefit and cost-effectiveness analysis in health care. Ann Arbor, MI: Health Administration Press, 1982. 26. Weissert, W. G. Seven reasons why it is so difficult to make community-based long-term care cost-effective. Health Services Research, 1985, 20, 423-33. 27. Weissert, W. G., Wan, T. T. H., Livieratos, B. B., et al. Cost-effectiveness of homemaker services for the chronically ill. Inquiry, 1980, 17, 230-43. 28. Williams, A. H. Economics of coronary artery bypass grafting. British Medical Journal, 1985, 291, 326-29. 29. Wright, K. G., Cairns, J. A., & Snell, M. C. Costing Care. Social Services Monographs, Sheffield, UK: University of Sheffield Joint Unit for Social Services Research, 1981. 30. Zarit, S., Reever, K. E., & Bach-Peterson, J. Relatives of the impaired elderly: Correlates of feelings of burden. The Gerontologist, 1980, 20, 649-55.
INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
219
ECONOMIC EVALUATION OF A SUPPORT PROGRAM FOR CAREGIVERS OF DEMENTED ELDERLY Michael F. Drummond University of York
E. Ann Mohide Michelle Tew David L. Streiner McMaster University
Dorothy M. Pringle University of Toronto
J. Raymond Gilbert McMaster University
Abstract An economic evaluation was undertaken concurrently with a randomized trial comparing a Caregiver Support Program (CSP) with existing conventional community nursing care for those caring for elderly relatives at home. The differences in resource consumption were compared with changes in caregiver quality of life, as measured by the Caregiver Quality of Life Instrument (CQLI). A 20% difference from baseline in the CQLI favored the experimental (CSP) group, although this did not reach conventional levels of statistical significance. A comparison of improvement in quality of life with costs implies an incremental cost per quality-adjusted life year gained of Canadian $20,000 for the CSP, which compares favorably with other health care interventions. Further, larger studies are required to confirm this result.
In many countries, there is increased emphasis on community care for the elderly. While this can be beneficial to those elderly wishing to remain in their own homes, there This work was supported by a grant from the Ontario Ministry of Health, Research and Planning Division, 1986-1989. During this study, E. A. Mohide was a Career Scholar with The R. Samuel McLaughlin Centre for Gerontological Health Research; at the time of funding, D. M. Pringle was Research Director for VON Canada. Address reprint requests to: E. Ann Mohide, McMaster University, Faculty of Health Sciences, 1200 Main St. W., Room 2J40f, Hamilton, Ontario L8N 3Z5, Canada. We are grateful for the cooperation and encouragement of the Hamilton-Wentworth and Halton Branches of the VON, the Hamilton-Wentworth Visiting Homemakers Association and the Nursing Divisions of the Hamilton-Wentworth and Halton Departments of Health Services, and the Alzheimer Society for HaltonWentworth, Ontario, where this research was conducted.
Drummond et al.
are concerns that community care places an additional burden on the informal caregiver (3). Consequently, there has been recent interest in developing caregiver support programs, where additional services are provided to assist those family members or friends caring for the elderly in the community. However, given the pressures on health sector budgets, it is increasingly important to demonstrate that the benefits of new services justify their costs. This can be accomplished by an economic evaluation, a method by which the relative costs of alternative health care programs are compared with their relative consequences in terms of savings in other health care resources and improvements in health (6). One form of economic evaluation compares the relative monetary value and benefits from health care interventions in terms of cost per quality-adjusted life year (QALY) gained (24;28). A QALY is calculated by multiplying the changes in quality of life or utility by the number of years of life for which the change is experienced. Quality of life is measured on an interval scale from (0) dead to (1) complete health or- well-being. Therefore, an improvement of 0.1 on the quality of life scale for 1 year is equal to 0.1 QALY gained. Economic evaluation has been widely applied to health care, including geriatrics (25). For example, it has been shown that community care for elderly persons of cer-. tain levels of dependency is more cost-effective than alternative forms of institutional care (15;29). However, Weissert et al. (27) showed that not all community care programs are cost-effective, and Weissert (26) outlined seven reasons why this may often be the case. Very few economic evaluations have been undertaken of community care programs for the demented elderly. Kyle et al. (12) found that a comprehensive care package, comprising psychiatric nurse visiting, daycare, and inpatient respite care was of lower cost to the statutory agencies than institutional care. However, they did not address the issue of the burden on caregivers or the evaluation of programs to support caregivers. This article reports the first economic evaluation of a support program for caregivers of the demented elderly. Because it is based on a randomized trial (13), it provides the first evidence of whether such programs represent a good use of health service resources. METHODS Study Design An economic evaluation was carried out comparing the costs and consequences of the Caregiver Support Program (CSP) with those of conventional community nursing care. The main additional costs of the CSP were for nursing visits and the provision of respite care in the home. For the purpose of the economic evaluation, it was hypothesized that the main consequences would be improvement in the quality of life of the caregiver and reduction in the use of other services. The economic evaluation was conducted alongside a controlled clinical trial. Caregiver-patient pairs were recruited from an urban center in southern Ontario and were block-randomized using a computer-generated random numbers table to receive the experimental set of interventions, either the CSP or conventional community nursing care, for a 6-month period. The decision to offer a form of treatment to the control group was made because subjects probably would not have agreed to be assigned to a "no-care" alternative, especially those who were receiving some form of community care at the time of recruitment. The care provided to the control group was similar to that currently available in the locality studied. The length of the intervention period 210
INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
Economic evaluation of a caregiver support program
was based on the judgments of five geriatric clinicians that an overall effect should be demonstrated by this time. Subjects and Eligibility In order to be eligible for enrollment in the trial, both the caregiver and the cognitively impaired relative had to satisfy a number of criteria. Caregivers were eligible if they (a) lived with the demented relative, (b) identified themselves as the principal family member providing day-to-day care to the relative, (c) spoke English, (d) scored 8 or more on a short mental status questionnaire (19), (e) did not suffer from a life-threatening illness that would prevent them from continuing in the caregiver role over the length of the trial, (0 agreed to have nursing services provided by a specific visiting nursing agency, and (g) gave written consent to participate. Relatives were eligible if they (a) had a medical diagnosis of primary degenerative, multiinfarct, or mixed dementia, (b) were assessed as moderately to severely impaired (18), (c) did not suffer from a serious concomitant illness(es) that would be likely to result in acute or long-term hospitalization or death during the intervention period, and (d) were not likely to be placed in a long-term care setting during the course of the study. The relatives were rated using the Dementia Rating Scale (DRS) (2) and the MiniMental State Examination (MMSE) (9). Following this testing, a geriatric nurse clinician, who was blind to the DRS and MMSE test scores, rated the relative using the Reisberg Global Deterioration Scale (18). Those with eligible scores on at least two of the three scales (8 or more on the DRS, 19 or less on the MMSE, stage 5 or greater on the Reisberg Scale) had their medical records reviewed to confirm the diagnosis of progressive irreversible dementia. Caregivers were stratified on the basis of gender and whether or not the demented relative attended a day program to ensure that the groups were balanced with respect to these two variables. In the latter case, severely burdened caregivers might be more likely to have their relative referred to a day program than those who are less burdened. Description of the Interventions The experimental set of supportive interventions was directed at helping the caregivers to enhance their competence at giving care and achieve a sense of control in their roles as caregivers. Caregiver support nurses (CSNs) were assigned to caregivers, and made regularly scheduled home visits at a time that was convenient to the caregivers. Initially these visits were weekly but were adjusted upward or downward depending on the needs of the caregiver. Health assessments were completed on all caregivers using forms designed to encourage the assessment of common caregiver problems. Caregivers were encouraged to seek medical attention for neglected health problems, such as hypertension. The CSNs consulted with the caregivers' family physicians on a regular basis whenever the caregivers' health was judged to be unstable. The caregivers received education about dementia and caregiving using content and teaching methods tailored to their knowledge level, caregiving situation, and learning style. The CSP intervention included a 4-hour block of scheduled weekly in-home respite, with additional respite on demand. The respite workers were assigned to specific families and were encouraged to discuss the demented patients' plans of care with the assigned CSN. The weekly respite permitted the caregivers to anticipate relief from the rigors of caregiving on a consistent basis, and respite on demand provided relief between periods of scheduled respite. INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
211
Drummond et al.
Finally, caregivers and other interested family members were encouraged to attend a monthly 2-hour self-help support group specifically for families with demented relatives. In the control group, conventional (existing) community nursing focused on care of the demented patient, rather than care of the family caregiver. This care addressed such physical needs of the demented patient as bathing, grooming, managing incontinence, and monitoring medication. Although some time might be spent discussing the patient's behavioral problems with the caregiver, this was not the main focus on the visit. The nurses in the control group were assigned specific patients on the basis of geographical location and the frequency of visits was left to the nurses' discretion. The nurses in the control group did not have concentrated training in this area and did not use the consultative services of psychogeriatric clinicians. Caregivers were not recommended to receive other health and social services in a standardized or systematic manner. Outcome Measures
Ideally, the primary outcome variable for a trial such as this should be a direct measure of the burden on caregivers. Accordingly, the psychometric properties of several measures of burden (10;20;30) were appraised, but none of the measures had been shown in previous studies to be responsive to change. Consequently, it was necessary to select instruments specific to symptoms. Because negative emotional effects, in particular depressive and anxious symptomatology, were cited frequently in the literature (1;7;8;11); these were used as the major outcomes of interest. Measures with sound psychometric properties were selected: the Center for Epidemiologic Studies Depression Scale (CES-D) (17) and the state-anxiety portion of the State-Trait Anxiety Inventory (STAI) (22). When responding to the scale, subjects were asked to consider the past week in order that the scale be responsive to changes in circumstances and yet not be overly dependent on daily fluctuations in mood. This modification is consistent with the CES-D Scale response interval of 1 week. In addition to these two major measures of outcome, the caregiver's quality of life was assessed using the Caregiver Quality of Life Instrument (CQLI) (14), an application of the time trade-off technique for measuring health status (23). Using this approach, the subjects decided how much length of life (expressed in future years of life) to trade off for improved quality of life (expressed in terms of dimensions of health, e.g., physical, psychological, or social function). To achieve this, subjects were asked to choose between being in two alternative health states for different periods of time. This approach gives valuations of health states, or quality of life, on an interval scale from 0-1. The CQLI was particularly relevant to the economic evaluation as it enabled the calculation of the QALYs gained by the caregivers from the intervention. Cost Items
A wide range of resources was recorded. These included nursing visits and respite care provided by the study, as well as those services sought by the caregivers, such as physician visits, homemaker services, day programs, overnight respite (both in the home and in institutions), visits by social workers, and other paid help including some companion services. In addition, days of acute hospital care were recorded, although it was recognized at the outset that these events were too infrequent to draw any strong inferences about the impact on them of the CSP. There were a number of sources of financial data reflecting the wide range of agencies involved in the program. These included the records of relevant health and social 212
INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
Economic evaluation of a caregiver support program
agencies and interviews with the caregivers themselves. The economic evaluation concentrated on the resources that could be unambiguously expressed in monetary terms and compared them with changes in caregiver quality of life as measured by the CQLI. A further article will consider the nonfinancial costs of caring, such as forgoing leisure activities in order to give care. Data Collection and Analysis
The research assistant interviewed the subjects at entry, at midpoint (3 months), and upon completion of the intervention at 6 months. If placement in long-term care was anticipated, an end-of-study interview was conducted prior to the withdrawal wherever possible. The study was completed by 22 subjects in the experimental group, and 20 in the control group completed the study; 10 demented relatives were admitted to long-term care facilities (5 in each group), 6 were admitted to a hospital for longer than 1 month, one caregiver died, and one couple moved out of the study catchment area. Our primary analysis is based on those who completed the study. This group reflected the impact of the program over a fixed period of time and provided results for the major outcome measures without the added variables of several major life events (such as long-term institutionalization, major illness, and unstable health of the demented relative). Exclusion of the subjects from the primary analysis was conservative since, although the number of relatives admitted to long-term facilities was equal between the groups, those in the control group were admitted for longer periods of time. The demographic data were analyzed using a t test and chi-square. The major effectiveness analyses were conducted using multivariate analysis of covariance (MANCOVA) with between-groups and within-group independent variables (intervention and time, respectively), and three dependent variables (CES-D, STAI, CQLI). Analysis of covariance (ANCOVA) was used to examine individual outcome measures. Pearson's r was used to examine correlations and stepwise regression to examine the variance accounted for by specific variables. The data on cost were analyzed by the Mann-Whitney U test (20); a nonparametric statistic for the comparison of two independent means was used. Differences in overall cost can result from differences in the physical qualities of resources consumed, or differences in the unit prices of those resources. Since some unit prices, such as those for daycare programs, could potentially vary by location, rather than as a result of the CSP, statistical tests were applied to the physical quantities of resources as well as the overall financial amounts. The incremental cost of the CSP was then compared with the improvement in quality of life (as measured by the CQLI), and the incremental cost per QALY calculated.
RESULTS
Sixty caregivers were enrolled in the trial. Table 1 provides the baseline characteristics of the 30 subjects allocated to each group. The groups did not differ on most of the variables. There was a statistically significant difference between the two groups in the caregivers' perception of the length of caregiving. Three control subjects had extremely long periods of caregiving in comparison to the other subjects; only two of the three completed the trial. Overall, the experimental group had a higher annual family income than the control group. In addition, there was a marginally significant INTL. J. OF TECHNOLOGY ASSESSMENT IN.HEALTH CARE
213
Drummond et al. Table 1 . Characteristics of the Caregivers and Their Relatives at Baseline Control group (» = 30)
Experimental group (» =: 30)
Characteristics Caregiver Number of females Number of spouses Wife Husband Mean age Mean months caregiving Number with some or all of high school Mean annual income Number of moderate to extremely healthy (self-rated) Number using psychotropic drugs Demented Relative Number of females Mean age Mean DRS score Number at stage 5 or 6 on the Reisburg Global Deterioration Scale Mean MMSE score Number with Alzheimertype dementia Mean Barthel score
No.
No.
(% or SD)
(% or SD)
21
(70%)
22
(73%)
14 7 66.10 39.80
(47%) (23%) (13.47) (30.23)
17 8 69.40 72.40
(57%) (27%) (8.61) (61.93)
23 $27,607
(77%) (19,708)
18 $8,155
(60%) (11,930)
22
(73%)
20
(67%)
3
(10%)
3
(10%)
15 77.80 8.75
(50%) (9.19) (3.01)
14 75.90 10.75
(47%) (7.70) (4.75)
27 13.21
(90%) (6.83)
26 11.04
(87%) (6.17)
19 68.67
(63%) (18.1)
21 62.50
(70%) (24.52)
p value
.013 .035
.057
difference in the mean DRS score, although this does not appear to have been clinically important. Effectiveness analyses were conducted on those who completed the trial. The mean scores of the major outcome variables used in this analysis are provided in Table 2. The covariates used in the MANCOVA included the baseline values for the major outcomes and the variables that differed between groups (annual family income and perceived length of caregiving). The three outcomes (CES-D, STAI, and CQLI) were examined individually using ANCOVAs with the same covariates. There were no statistically significant differences for any of the analyses. For purposes of comparison, we conducted an analysis of intent to treat that included those who completed at least 18 weeks of the intervention. This did not alter the values to a clinically important extent, and there were no significant differences between the groups. Despite the constancy of the depression and the anxiety scores, the quality of the caregivers' lives, as measured by the CQLI, increased in the experimental group, while it decreased in the control group. By the end of the trial, there was a difference of 20% from the baseline scores favoring the experimental group. We judged this 20% difference to be clinically important. However, a sample size of 178 subjects per group would be required to show that this difference is statistically significant (4). 214
INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
Economic evaluation of a caregiver support program Table 2. Means (and SD) for the Major Outcomes Experimental group (n = 22)
Control group (n = 20)
Baseline
3 mos.
6 mos.
Baseline
3 mos.
6 mos.
CES-D
21.55 (11.57)
18.36 (11.78)
21.50 (12.98)
15.40 (9.21)
15.35 (7.45)
18.20 (10.05)
STAI a
50.41 (16.40)
46.68 (11.56)
49.64 (14.48)
46.25 (12.62)
46.85 (9.35)
48.50 (14.38)
CQL1
.55 (.33)
.58 (.37)
.64 (.33)
.56b (.41)
.52" (.42)
.53" (.41)
Measure
0 b
Raw scores for the STAI state-anxiety scale. n = 19; one subject did not complete the CQLI at baseline.
Table 3. Resource Costs for Individual Services Item (unit)
Unit cost (1988 Canadian $)
Payment sources'1
Nursing6 (per visit) Respite workerb (per hour) Homemaker (per hour) Day program (per day)
29.19 8.13 8.13-8.50 13.26-69.56
Physician (per visit) Overnight institutional respite (per night) Other paid help (per visit)
14.90-110.65
MCSS or MOH MCSS or MOH Caregiver or MOH MOH, MCSS, or municipality MOH
45.00-130.00 5.OO-55.OO
Caregiver or MOH Caregiver
MCSS, Ministry of Community and Social Services; MOH, Ministry of Health Since this trial was conducted, variations on the CSP have been funded in Ontario. Payment sources have been categorized according to the current funding trends. b During the period of the trial, payment arrangements were slightly different, as nursing and respite worker services were provided as part of the study.
0
The unit costs of the main resource inputs to care are given in Table 3. (All costs are expressed in January 1988 Canadian dollars, Canadian $1 = US $.83.) The ranges for some of the items, such as day programs, reflected the variety of methods of provision. In addition, Table 3 gives the usual sources of payment during regular operation of the CSP. The sources of payment are important in assessing costs and benefits by viewpoint. Our primary viewpoint considered all costs regardless of source of payment, but other viewpoints may be of interest to those considering particular governmental budgets. Table 4 gives the total costs of services received by the experimental and control groups during the 6-month intervention. As expected, those resource inputs that pertained to the CSP intervention, nursing visits to caregiver and respite worker visits, were higher for the experimental group. On the other hand, costs for private nursing, day programs, and overnight institutional respite care were higher for the control group. However, with the exception of the respite worker and private nursing, these differences were not significant, either in physical quantities of resources consumed or financially. These data can be interpreted in two ways. First, the overall differences in costs INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
215
Drummond et al. Table 4. Summary of Total Costs During the 6-Month Intervention, in January 1988 Canadian $ (Range) Service
Experimental group (n = 22)
Control group (n = 20)
Nursing visits To caregiver3 To relative
14,507(379-1,022) 12,143 (0-1,430)
1,314(0-759) 12,318 (146-1,547)
Attendant care Respite worker 3 Homemaker Private nursing Other
9,747 (0-943) 18,710(0-7,072) 0 48 (0-48)
0 11,679(0-1,691) 2,473 (0-2,006) 858 (0-858)
2,699 (0-550) 2,904 (0-757)
2,535 (0-484) 2,244 (0-406)
MD visits To caregiver To relative Other paid help
1,910(0-528)
1,198(0-660)
Day programs Overnight institutional respite Total (per caregiver)
13,169(0-4,243) 2,535 (0-1,540) 3,562
16,853(0-4,243) 6,447 (0-2,735) 2,897
Incremental cost per QALY (over control group)
20,036
-
Note: Range is given inside parentheses. Services delivered as part of the Caregiver Support Program (CSP). Two caregivers in the control group were receiving nursing visits for their own health problems prior to their participation in the trial.
a
could be taken as the net cost of adding the intervention. This amounts to $665 ($3,562 less $2,897) for 6 months, or $1,330 per annum. Alternatively, one could take the view that since the differences in individual cost items did not reach conventional levels of statistical significance, the only financial impact of the CSP is the additional annual cost of the intervention. This was $1,102 ($14,507 plus $9,747 divided by 22) over 6 months, or $2,204 per annum. (The costs of the monthly 2-hour support group meetings were trivial and have been omitted.) The more conservative approach is to take the higher incremental cost estimate of $2,204 per annum for the calculation of cost per QALY gained. The improvement in the quality of life of caregivers, as measured by the CQLI, was 0.11 over the 6-month intervention period. Over 1 year, this would be equivalent to 0.11 QALYs gained. Therefore, the implied incremental cost per QALY gained is $2,204 divided by 0.11, or $20,036. DISCUSSION
This article reports the first economic evaluation of a support program for caregivers of the elderly conducted concurrently with a randomized trial. Taking the more conservative approach, the annual incremental cost of providing the support program was $2,204 per caregiver. The CSP had no impact of any consequence on the other resources from statutory sources. Economic evaluation seeks to compare the additional costs of health care interventions with their consequences in terms of health effects or QALYs. In this case, the economic evaluation depends greatly on the interpretation placed on the results of the clinical trial. The magnitude of the differences between the groups in terms of the principal outcome measures was not clinically important. 216
INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
Economic evaluation of a caregiver support program
Therefore, statistical power was not an issue in this instance. However, a clinically important difference of 20% above baseline was observed in the CQLI, but the study did not have sufficient sample size to confirm this difference at conventional levels of statistical significance. Mohide et al. (13) pointed out that there are a number of possible explanations for finding no differences in the CES-D and STAI measures. The burdens of caring for a relative with dementia may be intractable, or the stage of the progression of the disease may have been too late for any intervention to make an impact on the emotional stress of caregivers. Alternatively, the intervention may not have been potent enough, or its effect may have been minimized because some experimental subjects did not accept all aspects of the intervention. Finally, it could be that measures of emotional symptomatology are not the optimal constructs to be used as major outcome variables; indeed, a measure that incorporates more domains while at the same time being specific to caregivers, namely, the CQLI, may be a more sensitive and relevant measure of program effectiveness. Whatever the reason for the lack of statistically significant results, the conclusions of the concurrent economic evaluation are necessarily tentative. This study alone cannot demonstrate that caregiver support programs represent good value for the money. However, it does show that the implied cost per QALY (of approximately Canadian $20,000) compares favorably with other health care interventions. Torrance and Zipursky (24) have calculated the incremental cost per QALY of a range of health care interventions. These range from $4,200 (in 1983 US $) for coronary artery bypass surgery for left main artery disease, to $54,000 for hospital hemodialysis. Allowing for inflation and converting from U.S. to Canadian dollars, the cost per QALY of the CSP falls in the middle of the range, being marginally higher than the estimate for treatment of severe hypertension (diastolic > 105 mm Hg) in 40-year-old males, which Torrance and Zipursky reported at $9,400 (1983 US $) per QALY. Further work is therefore required to confirm our results and to demonstrate, through larger studies, that caregiver support programs give good value for the money. However, this will not be an easy task. A multicenter trial is likely to be required, since in a population of 500,000 in southern Ontario, only 146 caregiver-relative pairs could be located, of which only 60 were eligible for our study. Multicenter trials of communitybased interventions like the CSP raise considerable methodological problems of comparability, coordination, and standardization of the intervention. However, our study has shown that it is feasible to design a support program for caregivers of the demented elderly and that such programs could represent good value for the money when compared with other health care interventions. A number of interesting methodological issues are raised by this study. First, the variance in quality of life measures like the CQLI is such that much larger sample sizes are required to detect a difference between means at the conventional levels of statistical significance. In this case, the CQLI was not the primary outcome measure, and the difference observed, while clinically important, was not statistically significant. Since measures like the CQLI are more relevant to economic evaluations than are more specific clinical outcome measures, it is possible that clinical trials with concurrent economic evaluations will require larger sample sizes, with consequent resource implications for clinical research. Second, considerable effort was devoted in this study to recording a wide range of resources, many of which were neither different between the two groups nor significant in terms of cost. Drummond and Stoddart (5) stressed the need for a "phasing policy" for undertaking economic evaluation alongside clinical trials, in order to minimize unnecessary work. However, with an evaluation of a new intervention, it is diffiINTL J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
217
Drummond et al.
cult to decide which items require detailed assessment and measurement. Based on the experience of this study, future evaluations of caregiver support programs may be able to screen resources to assess which should be estimated precisely. REFERENCES
1. Anthony-Bergstone, C. R., Zarit, S. H., & Gatz, M. Symptoms of psychological distress among caregivers of dementia patients. Psychology and Aging, 1988, 3, 245-48. 2. Blessed, G., Tomlinson, B. E., & Roth, M. The association between quantitative measures of dementia and of senile change in the cerebral grey matter of elderly subjects. British Journal of Psychiatry, 1968, 114, 797-811. 3. Caserta, M. S., Lund, D. A., Wright, S. D., et al. Caregivers to dementia patients: The utilization of community services. The Gerontologist, 1987, 27, 209-14. 4. Cohen, J. Statistical power analysis for the behavioral sciences. New York: Academic, 1969. 5. Drummond, M. F., & Stoddart, G. L. Economic analysis and clinical trials. Controlled Clinical Trials, 1984, 5, 115-28. 6. Drummond, M. F., Stoddart, G. L., & Torrance, G. W. Methods for the economic evaluation of health care programmes. Oxford and New York: Oxford University Press, 1987. 7. Eagles, J. M., Beattie, J. A. G., Blackwood, G. W., et al. The mental health of elderly couples. I. The effects of a cognitively impaired spouse. British Journal of Psychiatry, 1987, 150, 299-303. 8. Fitting, M , Rabins, P., Lucas, M. J., et al. Caregivers for dementia patients: A comparison of husbands and wives. The Gerontologist, 1986, 26, 248-52. 9. Folstein, M. R, Folstein, S. E., & McHugh, P. R. Mini-Mental State: A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 1975, 12, 189-98. 10. Greene, J. G., Smith, R., Gardiner, M., et al. Measuring behavioral disturbance of elderly demented patients in the community and its effect on relatives: A factor analytic study. Age and Aging, 1982, 11, 121-26. 11. Haley, W. E., Levine, E. G., Brown, S. L., et al. Psychological, social, and health consequences of caring for a relative with senile dementia. Journal of the American Geriatrics Society, 1987, 35, 405-11. 12. Kyle, D. R., Drummond, M. F, & White, D. M. D. The Hereford District Department of Mental Health of the Elderly: A preliminary evaluation. Community Medicine, 1987, 9, 35-46. 13. Mohide, E. A., Pringle, D. M., Streiner, D. L., Gilbert, J. R., et al. A randomized trial of family caregiver support in the home management of dementia. Journal of the American Geriatrics Society, 1990, 38, 446-54. 14. Mohide, E. A., Torrance, G. W., Streiner, D. L., et al. Measuring the wellbeing of family caregivers using the time trade-off technique. Journal of Clinical Epidemiology, 1988, 41, 475-82. 15. Mooney, G. H., Russell, E. M., & Weir R. D. Marginal analysis: A case study in care of the elderly. In Choices in Health Care, London: Macmillan, 1980, 77. 16. Rabins, P., Mace, N., & Lucas, M. The impact of dementia on the family. Journal of the American Medical Association, 1982, 248, 333-35. 17. Radloff, L. S. The CES-D Scale: A new self-report depression scale for research in the general population. Applied Psychological Measurement, 1977, 1, 385-401. 18. Reisberg, B., Ferris, S., DeLeon, M., et al. The Global Deterioration Scale for assessment of primary degenerative dementia. American Journal of Psychiatry, 1982, 139, 1136-39. 19. Robertson, D., Rockwood, K., & Stolee, P. A short mental status questionnaire. Canadian Journal on Aging, 1982, 1, 16-20. 20. Robinson, B. C. Validation of a caregiver strain index. Journal of Gerontology, 1983, 38, 344-48. 21. Siegel, S., & Castellan, N. J. Nonparametric statistics for the behavioral sciences, 2nd ed. New York: McGraw-Hill Book Company, 1988.
218
INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
Economic evaluation of a caregiver support program 22. Spielberger, G., Gorsuch, R., & Lushene, R. STAI manual for the State-Trait Anxiety Inventory. California: Consulting Psychologists Press, 1968. 23. Torrance, G. W., Thomas, W. H., & Sackett, D. L. A utility maximization model for evaluation of health care programs. Health Service Research, 1972, 7, 118-33. 24. Torrance, G. W., & Zipursky, A. Cost-effectiveness of antepartum prevention of Rh immunization. Clinics in Perinatology, 1984, 11, 267-81. 25. Warner, K. E., & Luce, B. R. Cost-benefit and cost-effectiveness analysis in health care. Ann Arbor, MI: Health Administration Press, 1982. 26. Weissert, W. G. Seven reasons why it is so difficult to make community-based long-term care cost-effective. Health Services Research, 1985, 20, 423-33. 27. Weissert, W. G., Wan, T. T. H., Livieratos, B. B., et al. Cost-effectiveness of homemaker services for the chronically ill. Inquiry, 1980, 17, 230-43. 28. Williams, A. H. Economics of coronary artery bypass grafting. British Medical Journal, 1985, 291, 326-29. 29. Wright, K. G., Cairns, J. A., & Snell, M. C. Costing Care. Social Services Monographs, Sheffield, UK: University of Sheffield Joint Unit for Social Services Research, 1981. 30. Zarit, S., Reever, K. E., & Bach-Peterson, J. Relatives of the impaired elderly: Correlates of feelings of burden. The Gerontologist, 1980, 20, 649-55.
INTL. J. OF TECHNOLOGY ASSESSMENT IN HEALTH CARE
219
