|
Cost-effectiveness of supported housing for homeless persons with mental illness |
Rosenheck R, Kasprow W, Frisman L, Liu-Mares W |
|
|
Record Status This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn. Health technology Three interventions were evaluated:
a joint programme of the US Department of Housing and Urban Development (HUD) and the US Department of Veterans Affairs (VA): the HUD-VA Supported Housing (HUD-VASH) programme, in which HUD Section 8 housing vouchers (rent subsidies which authorize payment of a standardised local fair-market rate less 30% of the individual beneficiary's income) were paired with intensive case management services provided by VA clinicians;
intensive case management without access to Section 8 vouchers; and
standard VA homeless services (comparator), which consisted of short-term broker case management as provided by the VA's Health Care for Homeless Veterans program (HCHV) outreach workers.
In the intensive case management strategies, trained nurses or social workers using an approach based on the Assertive Community Treatment model helped clients locate an apartment, negotiate the lease, furnish and move, and encouraged at least once weekly contacts and community based service delivery. They also provided substance abuse and employment counselling and facilitated linkage with other VA services.
Economic study type Cost-effectiveness analysis.
Study population The study population comprised homeless veterans (living in a homeless shelter or on the streets) with a major psychiatric disorder (schizophrenia, major affective disorder, bipolar disorder, or posttraumatic stress disorder) and/or an alcohol or drug abuse disorder.
Setting The setting was community. The economic study was conducted in the USA.
Dates to which data relate Recruitment took place between June 1992 and December 1995, and the study had a 3 year follow-up. The price year was 1996.
Source of effectiveness data Effectiveness data were derived from a single study.
Link between effectiveness and cost data Prospective costing was carried out on the same sample of study participants as that used in the effectiveness analysis.
Study sample Power calculations were not reported by the authors. Veterans who met minimal criteria and who were contacted through the outreach programme of the four sites (Cleveland, La Jolla, San Francisco and San Diego) were eligible for the study. Study participants appeared to have been self-selected and gave written informed consent. Out of 3,489 eligible veterans, 460 gave written, informed consent. Consequently, the study sample may not have been representative of the study population. Participants were more likely to be younger, female, or African American and had a greater likelihood of past hospitalisation due to drug abuse. Following weighted randomisation 182 participants entered the HUD-VASH arm, 90 entered the case management only arm, and 188 entered the standard care arm.
Study design This was a multi-centred, randomised, controlled trial with three year follow-up. Randomisation was conducted by means of a telephone call to the central evaluation staff, who identified the next assignment from a deck of cards specific to each site. Randomisation was weighted to assign half as many participants to the intensive case management arm. The unit of randomisation was the patient, and neither veterans nor staff could be masked to group assignment. There were higher follow-up rates at 3 year assessment in the HUD-VASH arm with 70% compared to 48% in the case management only arm and 40% in the standard care arm. There were consistent differences during that time period. There were also partial cross-overs as 21.4% of the HUD-VASH arm did not receive vouchers and 5.6% of the case management only arm did receive vouchers.
Analysis of effectiveness Although the authors did not state it explicitly, they appear to have conducted the analysis on the basis of intention to treat. The primary outcome measures were the number of days housed in the previous 90, and the number of days homeless in the previous 90. Secondary outcomes included psychiatric, alcohol and drug problems; quality of life; quality of residency; social support; and five kinds of indicators of treatment process (computerised records, therapeutic alliance, use of vouchers, case managers summaries, and use of regular VA mental health outpatient services). The groups were comparable at baseline.
As the follow-up rate was higher in the HUD-VASH arm, followed by the intensive case management and then standard care, two strategies were used to address the potential bias arising from data loss, which gave similar results (adjustment for baseline measures that differed among interviewed groups at any point in time; and marginal structural modelling to inversely weight observations from patients on the basis of likelihood of being followed-up).
Effectiveness results There was greater participation and visits in the HUD-VASH group than in the case management group and in the case management group than in standard care group across all years.
Averaging across all 3 years, veterans assigned to the HUD-VASH group had 25% more days in an apartment, room or house than the standard care group (59.4 versus 47.6 days; p<0.001) and 16.9% more days housed than the case management only group (59.4 versus 50.8 days; p<0.004).
The case management only group had only 7% more days housed than the standard care group, (p=0.29).
Differences across time were significant for the first two years but attenuated in year three.
Veterans assigned to the HUD-VASH group had 36.2% fewer days homeless than the standard care group (13.1 versus 20.5 days; p<0.001) and 35.8% fewer days homeless than the case management only group (13.1 versus 20.3 days; p<0.004).
There was no significant difference between the case management only group and the standard care group, (p=0.96).
There were no significant differences on any measures of psychiatric or substance abuse status or community adjustment, although HUD-VASH clients had a larger social network.
Subgroup analysis showed results to be similar among subgroups defined by severity of illness, dual diagnosis, substance abuse diagnosis, chronic homelessness, racial/ethnic minority, level of social support.
Clinical conclusions Veterans assigned to the HUD-VASH intervention had more days housed and fewer days homeless than the other two control strategies: intensive case management only and standard care. There were no significant differences on any measures of psychiatric or substance abuse status or community adjustment, although HUD-VASH clients had larger social networks.
Estimates of effectiveness and key assumptions The measure of benefit used in the economic analysis was the increase in days of housing.
Direct costs Discounting was not mentioned in the paper. VA healthcare costs were estimated by multiplying the number of units of service consumed by each patient by the unit cost of the type of service using published VA cost data. Unit costs of the HUD-VASH case management were estimated separately using more detailed data on programme expenditures and service delivery during a sample year. The VA service utilisation data were derived from the VA national workload data systems for inpatient, outpatient, and residential programme data. Non-VA services (medical and mental health inpatient, residential, nursing home, medical-surgical outpatient, and mental health outpatient), were documented by quarterly interviews and costs used were published non-VA sources. Non-health care costs included days in shelter beds, jail or prison, cash transfer payments, earnings, and the cost of Section 8 vouchers. Although cash transfer payments (including housing subsidies) were included in the governmental agencies perspective, only the administrative costs of these payments were included in the societal perspective. Resources were measured during the study period (1992-1995) and 1996 prices were used. Resource use and unit costs are not reported separately.
Statistical analysis of costs Costs were treated in a stochastic way. Multiple regression models were used to impute missing cost data of all cost categories other than VA health costs. To examine longitudinal time trends on costs, analyses of variance were conducted using VA health care costs among groups at 6 month intervals from the year before the trial to 3 years after randomisation.
Indirect Costs Discounting was not mentioned by the authors. Earnings documented by interview data during the study period were estimated and included in order to evaluate the societal perspective.
Sensitivity analysis Variability in the data was explored through the 95% confidence intervals of the results. A subgroup analysis was carried out to evaluate outcomes according to severity of illness, dual diagnosis, substance abuse diagnosis, chronic homelessness, racial/ethnic minority, level of social support.
Estimated benefits used in the economic analysis Average incremental days housed in the past 90 days, measured during a three year follow-up as compared to standard care, were: HUD-VASH 11.79, (p<0.001); intensive case management 3.21, (p=0.29).
Cost results Total 3-year VA health costs for HUD-VASH clients were $36,524, for case management $35,095 and for standard care $28,515.
The incremental 3-year VA health costs for the case management only group as compared with standard care was $6,580 and the incremental costs for HUD-VASH clients compared with case management only was $1,429.
The greater HUD-VASH costs were almost entirely attributable to the greater homeless programme costs.
Three year non-VA health costs were $1,047 lower for HUD-VASH than for standard care clients and case management only was $2,421 less than for HUD-VASH. So, from the perspective of the health system as a whole, costs of HUD-VASH were $6,962 (18%) greater than standard care.
From the governmental perspective HUD-VASH was $10,295 more expensive than standard care.
Combining health and non-health care costs to estimate the societal perspective, HUD-VASH clients consumed $6,200 (15%) more resources than standard care clients.
Synthesis of costs and benefits Incremental cost-effectiveness ratios (ICERs) with their 95% confidence intervals were used to compare each of the experimental conditions with standard care from the 4 different perspectives. Cost-effectiveness acceptability curves (CEACs) were also constructed for analysing net benefits. ICERs show that each additional day housed among HUD-VASH clients cost $58 (95% CI: $4 - $111) from the VA perspective; $50 (95% CI: -$17 - $117) from the total health system perspective; $74 (95% CI: $5 - $143) from the governmental agencies perspective, and $45 (95% CI: -$19 - $108) from the perspective of society as a whole.
CEACs show that, from the societal perspective, benefits are likely to outweigh the costs with a probability of 56% if a day of housing were valued at $50; 80% if valued at $75; 92% valued at $100 and 97% above $125.
These probabilities were modestly greater than from the health care system or VA perspectives, and modestly smaller than from the governmental perspective.
Authors' conclusions The authors concluded that supported housing for homeless people with mental illness resulted in superior housing outcomes than intensive case management alone or standard care and modestly increased societal costs.
CRD COMMENTARY - Selection of comparators Although the choice of comparator (standard care) was not explicitly justified by the authors, it would appear to represent current practice in the author's setting. You should decide if the comparator represents current practice in your setting.
Validity of estimate of measure of effectiveness The study design, a randomised controlled trial, was appropriate for the study question. There appears to have been a degree of selection bias in recruiting participants as they were self-selected. This makes generalisation from the study sample to all of the study population less straightforward. The randomisation method was centralised, which strengthens the validity of the study, and, although the system used (a deck of cards by site), could potentially be manipulated, this does not seem to have occurred, as all groups were similar at baseline. The principal limitation of the study, as acknowledged by the authors, was the substantial and differential follow-up attrition across treatment groups after the first year, with participants in the two experimental conditions more likely to be re-interviewed. Since veterans more severely affected were more likely to be lost to follow-up, the standard group would be biased in the favourable direction, which may explain the negative findings for clinical outcomes. Appropriate statistical analyses were undertaken to take this into account. It should also be emphasized, as the authors' did, that the comparator was standard care in a full service health care system with an outreach clinician to link them with a full range of health-care services. The comparator results were impressive in this setting, and care should be taken not to conclude that case management is ineffective.
Validity of estimate of measure of benefit The estimate of benefit was obtained directly from the effectiveness analysis, and it was one of the primary outcomes of the study. This choice of presenting the results as the increase in days housed rather than the decrease in homeless days (the other primary outcome) was not explicitly justified by the authors, but it seems to be reasonable choice as it would lead readers to think about the monetary value of a day of housing.
Validity of estimate of costs All categories of cost relevant to the four perspectives evaluated were included in the analysis, and, although details were not exhaustive within each cost category, the authors cited all the sources used to obtain unit costs. Unit costs and quantities were not reported separately, which limits transferability exercises to other settings. Resource use was assessed through a single study, and a statistical analysis of quantities was performed. A sensitivity analysis of prices was not conducted. Although the date to which prices relate was reported, a significant omission in the paper was the discounting practice applied both to benefits and costs, which should have been mentioned as the study time horizon was three years.
Other issues The authors make appropriate comparisons of their results with findings from other studies. The experimental interventions were neither highly standardised nor monitored. However, as the authors point out, this increases the relevance and generalisability to "real world" practice. As the authors also pointed out, generalisability cannot be guaranteed to a more homogeneous population, to women, or to health care systems other than the VA. Other potential limitations stated by the authors were that the same case managers implemented both experimental conditions, which could have blurred their differences; and an intrinsic limitation of any cost-effectiveness analysis is that, in the absence of a monetary valuation of the outcome, it is not possible to decide whether added costs were justified by the benefits. Other topics to which the authors' referred were that any large scale implementation of HUD-VASH would require increased taxation, and that some authorities regard housing as a basic right which should be guaranteed to all citizens regardless of costs.
Implications of the study The agency-specific approach was successful at integrating clinical and housing services and in improving housing outcomes. This study demonstrates the potential benefits of housing vouchers for this population, although the associated clinical costs are not inconsiderable. The finding that case management and standard care were similar suggests further research could be carried out to evaluate whether housing vouchers are effective but without some special support service. Further research is also needed on the relationship of service intensity and outcomes.
Bibliographic details Rosenheck R, Kasprow W, Frisman L, Liu-Mares W. Cost-effectiveness of supported housing for homeless persons with mental illness. Archives of General Psychiatry 2003; 60(9): 940-951 Indexing Status Subject indexing assigned by NLM MeSH Adult; Case Management /economics /standards; Community-Institutional Relations /economics; Cost of Illness; Cost-Benefit Analysis; Female; Financing, Government /economics /standards; Government Agencies /economics; Health Care Costs; Health Status; Homeless Persons /statistics & Humans; Male; Mental Disorders /economics /rehabilitation /therapy; Program Evaluation; Prospective Studies; Public Housing /statistics & Social Adjustment; Substance-Related Disorders /economics /rehabilitation; United States; United States Department of Veterans Affairs; Veterans /statistics & numerical data; numerical data; numerical data AccessionNumber 22003001232 Date bibliographic record published 31/03/2004 Date abstract record published 31/03/2004 |
|
|
|