A Prospective Study of the Associations Among Housing Status and Costs of Services in a Homeless Population
Abstract
Objective:
The complex needs of homeless populations result in use of a wide range of services and high costs for housing programs and psychiatric and general medical care. Allocation of resources often is not congruent with assessed needs. A series of cost-congruence hypotheses was developed to test assumptions that needs are associated with resources provided for appropriate services in homeless populations.
Methods:
Individuals (N=255) who were homeless were followed for two years and were categorized by housing status over time (consistently housed, housed late, lost housing, or consistently homeless). Detailed information about the individuals was obtained at baseline, and follow-up data were collected one and two years later. Extensive data about the costs of services provided by type (medical, psychiatric, substance abuse, and homeless maintenance and amelioration) were derived from 23 agencies, and service use information was collected from the agencies and by self-report. Multiple regression models were used to test the hypotheses.
Results:
Medical, psychiatric, and homeless maintenance costs varied by housing status. Serious mental illness predicted costs for psychiatric services, as expected, but also costs for substance abuse services and acute behavioral health care and total costs. Alcohol use disorders predicted substance abuse service costs.
Conclusions:
This study followed a homeless cohort prospectively and provided estimates of costs of service use derived from a large number of agencies. This research increases the understanding of patterns of service use in a homeless population and informs the provision of services appropriate to the complex needs of this difficult-to-serve population.
Providing for the complex needs of homeless populations involves a variety of services, yielding high costs for housing programs, psychiatric and substance abuse treatment, and general medical care. Estimated costs of these services may top $50,000 per person annually (1–3). Research demonstrates that many individuals move in and out of housing, and some do so repeatedly. Sixty percent of stays in emergency shelters last less than one month, 60% of stays in transitional housing last less than six months, and permanent supportive housing typically lasts less than five years (4).
Our research group has shown how changes in housing status relate to substance abuse (5). Because service needs of persons who are homeless do not necessarily remain constant over time, costs related to these services are likely to also vary. For example, attaining stable housing should eliminate further costs for sheltering.
For homeless populations, greater service costs may not necessarily reflect better outcomes. One possible reason is that satisfaction with services is not necessarily included as an outcome measure. In a recent study (6), persons who were homeless perceived benefit from service use if they sensed that it is appropriate to accept social support from others and the service is considered helpful. In this case, receiving a specific service is less about helping meet a particular need and more about improving general satisfaction. An individual’s satisfaction is not typically considered and measured as an effectiveness outcome in research. Consequently, costs of services accrued would not be congruent with effectiveness. In addition, there is no consensus on how attaining housing affects service use costs. In contrast to our research (7,8), Larimer and colleagues (3) found that attaining housing was associated with reduced medical care costs.
Providing services that are not sufficiently matched to the assessed problems of a population may generate costs but not necessarily contribute to improved outcomes. This issue is not about whether outcomes align with costs but about whether the allocation of resources is congruent with assessed needs. Given that needs for treatment of substance use and psychiatric disorders are disproportionately high among persons who are homeless compared with the general population, it is logical to examine whether service costs are congruent with these specific treatment needs.
We developed a series of cost-congruence hypotheses to test assumptions that needs are associated with resources provided for appropriate services in homeless populations. These hypotheses are as follows: attainment of stable housing is associated with lower costs for homeless maintenance and amelioration, attainment of stable housing is associated with higher medical care costs, alcohol and cocaine use disorders are associated with higher costs for substance abuse treatment services, and serious mental illness is associated with higher costs for acute psychiatric and behavioral care.
The purpose of this study was to test the cost-congruence hypotheses in the context of two related issues—whether changes in housing status were associated with changes in costs of specific services used and whether specific service needs were associated with costs. Little is known about how costs for services change with the temporal evolution of housing status in homeless populations. This study empirically tested whether these hypotheses were affected by housing status by using prospective two-year longitudinal data collected from a systematically selected homeless sample in St. Louis, Missouri.
Methods
This study was funded by the National Institute on Drug Abuse and was approved in advance by the Institutional Review Board of Washington University in St. Louis. Participants provided written informed consent.
Sample
For this study, homelessness was defined as having no current fixed address and having spent the previous 14 nights in a public shelter, in some other unsheltered location, or on the streets. Individuals who had resided during this period in flophouses or cheap motels were included if they had been otherwise unhoused for less than 30 days. This definition included individuals who had stayed in a public shelter or unsheltered location for most of the past 14 days only if they had stayed for less than one half of those days temporarily with friends or relatives or in temporary single room occupancy facilities. Participants who met this definition of homelessness were systematically recruited from shelter beds and street routes from 1999–2001 with procedures previously described (8–10) and were tracked for two years (11). Participants received $20 for each interview. A total of 435 were eligible for the study, 400 (92%) were included in the baseline sample, and of those 288 (72%) were successfully tracked over two years. Of the 400 participants, 29 were classified as ineligible for reassessment at one or both follow-up points because of death (N=5), severe illness (N=6), or incarceration (N=18); of the 371 who were eligible for follow-up reassessments, 255 (69%) were reassessed in both follow-up years with complete data. No significant differences in demographic characteristics at baseline were detected between those who completed follow-up interviews versus those who did not over the two-year period.
Measures
At baseline, participants completed structured interviews including sociodemographic sections of the National Comorbidity Survey interview (12), the substance abuse sections of the Composite International Diagnostic Interview substance abuse module for DSM-III-R (13), the Diagnostic Interview Schedule for DSM-IV (14), and the homeless supplement to the Diagnostic Interview Schedule (15). These interviews elicited detailed lifetime history and recency of symptoms and disorders, substance use patterns, and amount and types of services used during the previous year.
At both the one- and two-year interviews, participants provided detailed information on their housing status and the amount and types of services they used during the previous year. Participants were considered to have stable housing at the end of each year if they had been housed in their own places for the previous 30 days and for most of the past year.
Agencies that delivered services to this homeless sample provided service use and cost data. The 23 participating agencies represented four primary types of service agency: medical centers, mental health treatment facilities, substance abuse treatment facilities, and homeless shelters. Using structured instruments that were standardized across agencies, we collected study participants’ service use and cost data from the agencies for the two follow-up years. Service use data collected from the agencies were combined with the self-report service use data from interviews of the homeless individuals (8). This method of data collection contributes to the literature in two main ways. First, the cost estimates reflect the social value of resources used for each service (16) rather than the gross average cost per participant or the reimbursement per service from billing records. Second, because service use data from two sources are combined, cost estimates are more likely to be complete.
Creation of Variables
Details of the development and definition of service use and cost variables for this study are provided elsewhere (8). The main service use categories for this analysis are medical, psychiatric, substance abuse, homeless maintenance and amelioration, and total services. Medical services included Medicaid, Medicare, other insurance, and utilization data from service providers. These data were not limited to the specific collaborating agencies. Homeless amelioration services are vocational programs and housing programs directly focused on helping people get out of homelessness and attain stable housing. Homeless maintenance is provision of shelter, food, and things to maintain people’s survival while they are homeless. In addition, for certain analyses, psychiatric and substance abuse services were divided into subcategories of acute and outpatient behavioral health services.
Housing status was determined separately for both follow-up years. Four main housing patterns over the course of the study were possible: consistently housed (stable housing obtained in the first year and maintained in the second year), housed late (stable housing obtained in the second year), lost housing (stable housing obtained in the first year but not in the second year), and consistently homeless (stable housing not achieved in either year). Additional variables were created to represent lifetime diagnoses of alcohol use disorder, cocaine use disorder, and serious mental illness (representing lifetime diagnoses of schizophrenia, bipolar disorder, or major depression), as determined through baseline interviews.
Data Analysis
Service use data were aligned and cleaned in a Microsoft Excel spreadsheet; imported into SAS, version 9.3; and merged with individual interview data in SAS for analysis. Missing data were imputed by using previously described methods (8). Service use costs among pairs of housing groups were compared in bivariate analyses by using nonparametric Wilcoxon-Mann-Whitney U tests (PROC NPAR1WAY in SAS) to accommodate instances of nonnormal data. Because comparisons among six pairs of housing groups were made for each variable, the alpha value used to define statistical significance was equal to .05 divided by 6, or .008. These comparisons were repeated by using linear regression models, with overall results consistent with the results of nonparametric testing. Because nonparametric testing of multivariate models is not possible, the consistent results of these two different methods support the use of multiple regression models for multivariate analysis. Generalized linear models were estimated in SAS by using PROC GLM to predict service use costs (the dependent variable in separate models for each service type) by housing group (independent variable). The models were controlled for need variables and demographic variables (sex, age, racial-ethnic minority group, years of education, and marital status) by entering the independent covariates simultaneously with other variables in the model (Table 1). In these multivariate models, square-root transformation was applied to the cost data to address positive skew.
| Service cost and characteristic | F | df | p |
|---|---|---|---|
| Medical costs and housing status | 4.10 | 3 | .007 |
| Psychiatric costs and housing status | 3.24 | 3 | .023 |
| Psychiatric costs and serious mental illness | 10.57 | 1 | .001 |
| Substance abuse costs and alcohol use disorder | 5.36 | 1 | .022 |
| Substance abuse costs and serious mental illness | 3.84 | 1 | .051 |
| Homeless maintenance costs and housing status | 3.98 | 3 | .009 |
| Homeless amelioration costs and female sex | 17.37 | 1 | <.001 |
| Total costs and serious mental illness | 4.75 | 1 | .030 |
| Total costs and female sex | 4.07 | 1 | .045 |
| Acute behavioral health care costs and serious mental illness | 7.48 | 1 | .007 |
| Outpatient behavioral health care costs and female sex | 19.96 | 1 | <.001 |
| Outpatient behavioral health care costs and age | 5.36 | 1 | .022 |
| Outpatient behavioral health care costs and alcohol use disorder | 4.84 | 1 | .029 |
a
Separate multivariate models were conducted for each type of service, and only variables that significantly predicted costs are listed: medical, F=2.25, p=.013; psychiatric, F=3.02, p<.001; substance abuse, F=1.84, p=.048; homeless maintenance, F=1.67, p=.083; homeless amelioration, F=3.38, p<.001; total services, F=2.80, p=.002; acute behavioral health care, F=2.39, p=.008; and outpatient behavioral health care, F=3.37, p<.001 (df=11). Acute behavioral health care and outpatient behavioral health care are subcategories of psychiatric and substance abuse services.
Results
Details about the baseline characteristics of this sample of 255 homeless individuals have been previously reported (5,8). To summarize, the sample was predominantly African American (N=192, 75%) and male (N=187, 73%), with a high school education on average (N=149 of 249 participants; 60%). Many had a lifetime history of “legal trouble” (N=158 of 247 participants; 64%), alcohol use disorder (N=151 of 254 participants; 59%), cocaine use disorder (N=113 of 253 participants; 45%), or serious mental illness (N=82, 32%).
Figure 1 shows costs over the two study years by service use category among the four housing groups. The group that was housed late had the highest total service use costs (mean±SD=$16,545±$17,595), followed closely by the group that was consistently housed ($15,634±$23,279). The medical services category was the most costly overall ($6,029±$13,394). The cost of the next most expensive overall category, psychiatric services ($3,305±$6,973), was significantly lower (Wilcoxon S=75,636, p<.001). Costs of psychiatric services were marginally greater than costs of substance abuse services ($2,004±$4,961; Wilcoxon S=68,238, p=.051).
Figure 1 Costs of services over two years among 255 participants grouped by housing status, in adjusted U.S. dollarsa
a*p<.01, **p<.001, corresponding to bar of same shade
Compared with mean spending for all other groups, mean costs for medical services were highest for the group that was consistently housed ($10,922±$22,130 versus $4,835±$9,921; Wilcoxon S=7,585, p=.012) and lowest among persons who lost housing ($1,295±$2,007 versus $6,566±$14,020; Wilcoxon S=2,377, p=.008). Among all the groups, psychiatric service costs were lowest for the group that was consistently housed ($1,181±$2,055 versus $3,823±$7,626; Wilcoxon S=5,598, p=.074) and highest among persons who lost housing ($5,090±$8,515 versus $2,707±$6,288; Wilcoxon S=9,515, p=.007). Substance abuse service costs did not vary among the housing groups. Most psychiatric and substance abuse service costs were for acute behavioral care ($3,114±$7,693), and these costs were greater for persons who remained unhoused in one or both years ($3,575±$8,217) compared with those who were consistently housed ($1,224±$4,598; Wilcoxon S=5,133, p=.001). Persons who were housed late had higher homeless maintenance costs than those who were consistently housed. Figure 1 also shows significant differences in costs in all pairwise comparisons of the four housing groups by each type of service use.
Table 1 summarizes significant results from models testing whether the costs of each service type were predicted by need variables, housing group, and demographic control variables. There was one model for each service type, and the dependent variables were costs. Consistent with the bivariate models, medical and psychiatric service use costs and homeless maintenance costs varied significantly among housing groups, although substance abuse service costs and homeless amelioration services did not. As expected, serious mental illness predicted psychiatric service costs, and alcohol use disorders—but not cocaine use disorders—predicted substance abuse service costs. Serious mental illness also predicted substance abuse, acute behavioral health, and total costs. Women had higher costs relative to men for homeless amelioration ($965±$3,259 versus $2±$20), outpatient behavioral health ($890±$1,634 versus $290±$774), and total services ($15,928±$21,204 versus $12,383±$15,484).
Discussion
To our knowledge, this study is the first to follow a homeless cohort prospectively and use data from the agencies providing services to calculate cost estimates reflecting resource use per service. Follow-up over a two-year period allowed examination of how obtaining, retaining, and losing housing from year to year affected service use costs during the natural progression of homelessness.
This study found that medical service costs were highest among persons who were consistently housed and lowest among persons who lost housing. One possible explanation is that persons with chronic and severe general medical problems who require costly services may be able to obtain housing through additional resources available for people with disabling general medical conditions. It also may be that persons with housing tend to have greater access to general medical care.
Psychiatric and acute behavioral health service costs showed an opposite relationship with stable housing. Psychiatric service costs were lowest among persons who were consistently housed and highest among persons who were housed late. Similarly, acute behavioral health costs were lower for persons who were consistently housed than for those with any other pattern of housing status. Perhaps services that address psychiatric morbidity do not facilitate the acquisition of housing. Perhaps psychiatric illness, by its nature, does not necessarily lead to service utilization, even when stable housing is obtained. This study examined data on various services directly addressing homeless maintenance. Not unexpectedly, persons who did not obtain stable housing until the second year had higher homeless maintenance costs than those with stable housing in both years.
As discussed further below, the cost-congruence hypotheses predicted that the group that never attains stable housing will have the highest overall service use costs and that the group that attains stable housing will have the lowest overall service use costs. The findings, however, demonstrated a much different pattern, largely attributed to higher medical costs among persons who were consistently housed. This team has previously postulated that obvious acute needs, such as psychiatric illness and literal homelessness, may create barriers to meeting chronic needs (8).
In contrast, findings of this study supported the cost-congruence hypotheses that costs of medical services, homelessness services, psychiatric and mental health services, and substance abuse services are associated with service needs. Results of multivariate analyses further demonstrated that the costs of services for persons who are homeless are sometimes predicted by factors that are not logically congruent with the persons’ needs. Examples are the associations of housing group with psychiatric services (as noted above), serious mental illness with substance abuse services, alcohol use disorder with outpatient behavioral health services, and serious mental illness with overall costs. In addition, women had greater costs than men for homeless amelioration, total services, and outpatient behavioral health services. Although the cost-congruence hypotheses were generally supported by the findings, understanding the larger cost picture requires application of more nuanced and complex behavioral models to this study’s specific patterns of findings.
Although the findings agree with the services literature documenting the complexity of homelessness, they do not necessarily agree with the conclusion that housing leads to lower societal costs (3). This difference in findings may in part reflect methodological differences. Prior studies intervened by providing housing and then estimating costs, a method that may not have adequately accounted for the two-way relationship between housing and service use. Further, these intervention studies compared housed groups with groups that remained homeless after the intervention, not accounting for the well-documented phenomenon of movement in and out of homelessness over time—a finding that this team has demonstrated to be significantly associated with overall service use (5,7).
Although this study had significant methodological strengths—including a probability sample, longitudinal data collection, and agency-based cost derivation—it also had limitations. The relatively modest analytic sample size (N=255) may have limited power to detect effect sizes. A larger sample might have revealed more subtle interactions among individual needs and appropriate costs. As noted elsewhere (8), this study’s collection of cost data did not include criminal justice costs—a significant cost driver in other studies (17). In addition, had follow-up continued for longer than two years, further significant findings would likely have emerged. Perhaps housing needs to be in place for more than one to two years before true cost increases or decreases are fully realized. It may also be the case that cost increases are associated primarily with a small number of high-cost service utilizers. Important information may be gained by focused analysis of high-cost outliers. Further, the cross-sectional sample provided a relative overrepresentation of homelessness chronicity, and thus the findings are especially pertinent to population groups with a history of chronic homelessness. Finally, a proportion of the self-report service use data for each year of the study was collected after collection of the housing-status data.
The findings provided important information to help direct future research. To understand costs over time, studies may need larger samples and examination of costs over longer periods. Studies collecting data over shorter periods may risk inclusion of costs incurred for acute conditions and fail to capture costs associated with periodic acute exacerbations of chronic conditions. This point is of particular relevance for studies examining immediate cost consequences of interventions for specific acute conditions.
Future research is needed to address how societal costs change with housing status, both in the aggregate and across agencies. The findings suggest that aggregate societal costs associated with services for persons who had been homeless may increase after the individuals find housing, and there may be cost shifting among agencies. Future research may indicate that providing services to improve outcomes among people in need does not always save money. However, detailed and accurate estimates of costs are critical to understanding how to distribute scarce service resources.
The findings also suggest the need to consider the potential benefits of including preventive health services as part of the care provided to homeless persons. As we suggested in another publication (8), medical costs may increase if providers prefer addressing acute needs, ignoring asymptomatic chronic conditions. As indicated in the broader homelessness literature, the relative risk of death among persons who are homeless is considerably higher than in the general population (18), consistent with this study’s findings of five known deaths over the course of the study (11). Including medical services as part of routine homelessness care may decrease medical service costs and improve outcomes related to both mortality and morbidity.
Conclusions
This study followed a homeless cohort prospectively and derived information about the costs of services from the agencies that provided the services. This research contributes to understanding of the pattern of service use in homeless populations and can help to provide services that are more in line with this population’s complex needs.
Acknowledgments
This work was funded by grant R01 DA10713 from the National Institute on Drug Abuse.
The authors report no competing interests.
References
1.
Gilmer TP, Stefancic A, Ettner SL, et al: Effect of full-service partnerships on homelessness, use and costs of mental health services, and quality of life among adults with serious mental illness. Archives of General Psychiatry 67:645–652, 2010
2.
Basu A, Kee R, Buchanan D, et al: Comparative cost analysis of housing and case management program for chronically ill homeless adults compared to usual care. Health Services Research 47:523–543, 2012
3.
Larimer ME, Malone DK, Garner MD, et al: Health care and public service use and costs before and after provision of housing for chronically homeless persons with severe alcohol problems. JAMA 301:1349–1357, 2009
4.
The 2010 Annual Homeless Assessment Report to Congress. Washington, DC, US Department of Housing and Urban Development, 2010
5.
North CS, Eyrich-Garg KM, Pollio DE, et al: A prospective study of substance use and housing stability in a homeless population. Social Psychiatry and Psychiatric Epidemiology 45:1055–1062, 2010
6.
Christian J, Clapham D, Thomas S, et al: The relationship between well-being, future planning and intentions to utilise intervention programmes: what can be learned from homeless service users. International Journal of Housing Policy 12:159–182, 2012
7.
Pollio DE, Spitznagel EL, North CS, et al: Service use over time and achievement of stable housing in a mentally ill homeless population. Psychiatric Services 51:1536–1543, 2000
8.
Fuehrlein BS, Cowell AJ, Pollio DE, et al: Deriving costs of service use among an urban homeless population. Journal of Substance Abuse Treatment 46:491–497, 2014
9.
Smith EM, North CS, Spitznagel EL: A systematic study of mental illness, substance abuse, and treatment in 600 homeless men. Annals of Clinical Psychiatry 4:111–120, 1992
10.
Smith EM, North CS, Spitznagel EL: Alcohol, drugs, and psychiatric comorbidity among homeless women: an epidemiologic study. Journal of Clinical Psychiatry 54:82–87, 1993
11.
North CS, Black M, Pollio DE: Predictors of successful tracking over time in a homeless population. Social Work Research 36:153–159, 2012
12.
Kessler RC, McGonagle KA, Zhao S, et al: Lifetime and 12-month prevalence of DSM-III-R psychiatric disorders in the United States: results from the National Comorbidity Survey. Archives of General Psychiatry 51:8–19, 1994
13.
Cottler L, Compton W: Advantages of the CIDI family of instruments in epidemiological research of substance use disorders. International Journal of Methods in Psychiatric Research 3:109–119, 1993
14.
Robbins L, Cottler L, Bucholz K, et al: Diagnostic Interview Schedule for the DSM-IV (DIS-IV). St Louis, Mo, Washington University, 1995
15.
North CS, Eyrich KM, Pollio DE, et al: The homeless supplement to the Diagnostic Interview Schedule: test-retest analyses. International Journal of Methods in Psychiatric Research 13:184–191, 2004
16.
Cowell AJ, Pollio DE, North CS, et al: Deriving service costs for a clubhouse psychosocial rehabilitation program. Administration and Policy in Mental Health and Mental Health Services Research 30:323–340, 2003
17.
Poulin SR, Maguire M, Metraux S, et al: Service use and costs for persons experiencing chronic homelessness in Philadelphia: a population-based study. Psychiatric Services 61:1093–1098, 2010
18.
Beijer U, Andreasson S, Agren G, et al: Mortality and causes of death among homeless women and men in Stockholm. Scandinavian Journal of Public Health 39:121–127, 2011
Information & Authors
Information
Published In
Cover: In the Loge, by Mary Cassatt, circa 1879. Pastel and metallic paint on canvas prepared with a pastel ground. Gift of Mrs. Sargent McKean, 1950 (1950-52-1), the Philadelphia Museum of Art. Photo credit: the Philadelphia Museum of Art/Art Resources, New York.
History
Published ahead of print: 31 October 2014
Published in print: January 01, 2015
Published online: 2 January 2015
Authors
Funding Information
National Institute on Drug Abuse10.13039/100000026: NIH R01 DA10713
Metrics & Citations
Metrics
Citations
Export Citations
If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.
For more information or tips please see 'Downloading to a citation manager' in the Help menu.
View Options
View options
PDF/EPUB
View PDF/EPUBGet Access
Login options
Already a subscriber? Access your subscription through your login credentials or your institution for full access to this article.
Personal login Institutional Login Open Athens loginNot a subscriber?
PsychiatryOnline subscription options offer access to the DSM-5-TR® library, books, journals, CME, and patient resources. This all-in-one virtual library provides psychiatrists and mental health professionals with key resources for diagnosis, treatment, research, and professional development.
Need more help? PsychiatryOnline Customer Service may be reached by emailing [email protected] or by calling 800-368-5777 (in the U.S.) or 703-907-7322 (outside the U.S.).