Effectiveness of Case Management for Homeless Persons: A Systematic Review

Abstract

We reviewed the literature on standard case management (SCM), intensive case management (ICM), assertive community treatment (ACT), and critical time intervention (CTI) for homeless adults. We searched databases for peer-reviewed English articles published from 1985 to 2011 and found 21 randomized controlled trials or quasi-experimental studies comparing case management to other services.

We found little evidence for the effectiveness of ICM. SCM improved housing stability, reduced substance use, and removed employment barriers for substance users. ACT improved housing stability and was cost-effective for mentally ill and dually diagnosed persons. CTI showed promise for housing, psychopathology, and substance use and was cost-effective for mentally ill persons.

More research is needed on how case management can most effectively support rapid-rehousing approaches to homelessness.

Homelessness is a serious and widespread public health problem. In the United States and Europe, estimates for the lifetime prevalence of homelessness range between 5.6% and 13.9%.¹ The global financial crisis has negatively affected the prevalence of homelessness. In the United States, certain groups, such as families and people living in suburban and rural areas, have become more vulnerable to homelessness.² In Europe, austerity measures implemented after the start of the crisis have increased poverty and homelessness, with possibly the worst to come because of a strong time lag effect.³

Homelessness is often accompanied by other problems. People who are homeless experience a lower quality of life than those who are domiciled.^4,5 Several longitudinal studies have found that quality of life improves as independent housing is obtained.^5–7 Societal participation is limited; many homeless persons are unemployed, have few sources of income, and have a limited social network. They often experience extreme poverty and a lack of social support.^8,9 Although few are felony offenders, homeless persons are at risk of arrest for transgressions resulting from their lifestyle (e.g., panhandling, public intoxication, squatting, and failing to pay fines).^8,10 Moreover, estimates suggest that almost 40% of homeless people are dependent on alcohol and 25% on drugs. Many suffer from a mental disorder, such as a psychotic illness (13%), major depression (11%), or personality disorder (23%).¹¹ Physical health problems are more prevalent among this group than in the general population.^12,13 Recent studies found that up to 73% of homeless individuals have unmet health needs.^14,15 Consequently, homelessness should be regarded as a significant and increasing threat to public health, which should be addressed.

In recent years, the focus of policy measures to reduce homelessness has changed. The Homeless Emergency and Rapid Transition to Housing Act, an amendment to the McKinney–Vento Homeless Assistance Act, was enacted in 2009 to modernize the US Department of Housing and Urban Development’s homelessness assistance programs.¹⁶ In 2010, the jury recommendations of the European Consensus Conference on Homelessness laid out a road map for ending homelessness in the European Union.³ Both proposals called for a shift away from the "staircase" approach, which requires homeless persons to prove housing readiness while transferring through shelters and transitional housing situations before they become eligible for independent housing. The proposed alternative is a rapid-rehousing,¹⁶ or housing-led,³ approach, which focuses on providing access to permanent independent housing as the initial response to resolving situations of homelessness, in conjunction with flexible support services as required by the service needs of those who are rehoused to prevent recurrent homelessness.¹⁷ Case management has been identified as a strategy to support rapid rehousing, especially for those with complex needs.³ Little is known, however, about what patterns of services are most suitable to accompany housing for different subgroups of homeless people.^16,18

Since the 1980s, several models of case management have been developed that provide the same basic functions: outreach, assessment, planning, linkage, monitoring, and advocacy.^19,20 Services delivered by case managers often include practical support, help with developing independent living skills, acute care in crisis situations, support with medical and psychiatric treatment, and assistance with contacts between clients and people in their social and professional support systems.²⁰

We focused on 4 models of case management that have been recommended and widely implemented for homeless persons¹⁹: standard case management (SCM), intensive case management (ICM), assertive community treatment (ACT), and critical time intervention (CTI). The models are distinguished by the functions they emphasize (Table 1). SCM is a coordinated and integrated approach to service delivery, with the goal to provide ongoing supportive care.²¹ ICM is typically targeted to individuals with the greatest service needs and prescribes more intensive services, more frequent client contact, and smaller individual caseloads than does SCM.²² ACT is closely related to ICM; however, in ACT the responsibility for providing services to clients is shared by a multidisciplinary team that is accessible 24 hours a day, 7 days a week.²³ CTI is an intensive time-limited case management approach to enhance continuity of care by bridging the gap between services and strengthening clients’ social and professional networks. CTI is designed to be deployed at critical moments in the lives of clients, for instance, when a person is about to make a transition from a shelter to independent housing.²⁴

TABLE 1—

Characteristics of Case Management Models for Homeless Adults

	Standard Case Management21	Intensive Case Management21,22	Assertive Community Treatment21,23	Critical Time Intervention21,24
Focus of services	Coordination of services	Comprehensive approach	Comprehensive approach	Targeted to continuity of care
Target population	Homeless persons	Homeless persons with the greatest service needs	Homeless persons with the greatest service needs	Homeless persons at critical transitions in their lives
Duration of services	Time limited	Ongoing	Ongoing	Time limited
Average caseload, no.	35	15	15	25
Outreach	No	Yes	Yes	Yes
Coordination or service provision	Coordination	Service provision	Service provision	Service provision and coordination
Responsibility for clients’ care	Case manager	Case manager	Multidisciplinary team	Case manager
Importance of client–case manager relationship	Somewhat important	Important	Important	Important

To our knowledge, 4 reviews on the effectiveness of case management for homeless adults have been published.^19,25–27 All 4 reviews underscore the effectiveness of ACT in producing positive outcomes for homeless people. Nevertheless, whether ACT is effective for all homeless subgroups in achieving more positive outcomes than other services, including other case management models, remains to be seen. These reviews have limitations: (1) they focus solely on homeless individuals with severe mental illness,^25–27 (2) they examine only 1 or 2 of the 4 models in use^26,27 or do not distinguish between different models and their individual effects,²⁵ and (3) they consider only certain outcomes.^25,26 Morse provides a more complete overview; however, he did not conduct a systematic literature search and failed to describe inclusion criteria for studies. Furthermore, this review is dated and was not published in a peer-reviewed journal.¹⁹

Our primary goal was to examine the consistency of findings across various models of case management and their applicability in a variety of homeless subgroups and settings through a complete overview of the existing literature on the effectiveness of the 4 case management models. We categorized and evaluated all outcome measures that were included in randomized controlled trials and quasi-experimental studies comparing these models to other services for the general homeless population or specific homeless subgroups.

METHODS

We conducted an electronic systematic literature search for peer-reviewed articles published in English between January 1985 and June 2011 in the PsycINFO, MEDLINE, Cochrane Library, Embase, and CINAHL databases. To identify study populations that were predominantly homeless, we used the following keywords: “homeless,” “homelessness,” and “homeless people.” We combined these keywords with the following terms to search for the 4 case management models: “strengths-based,” “strengths perspective,” “case management,” “intensive case management,” “assertive community treatment,” “critical time intervention,” “outreach,” “outreach programs,” “mental health,” “mental illness,” “psychiatric,” and “substance abuse” (Appendix A, Table A, available as a supplement to this article at http://www.ajph.org). We used Web of Science for a cited reference search.

After we conducted the search, we removed duplicates, and 2 reviewers independently screened titles and abstracts of the retrieved publications. We excluded reports that did not match our inclusion criteria, and 2 other reviewers independently evaluated the remaining publications. We resolved disagreements through discussion among at least 3 reviewers to achieve consensus.

Selection Criteria

Participants in eligible study samples were aged 18 years or older. The recruitment strategy of the study had to target a predominantly homeless population, as evidenced by the description of the target population, recruitment setting, or selection criteria. For the purpose of our review, we defined homeless persons as

1. persons who lacked a fixed, regular, and adequate nighttime residence or resided at night in a place not meant for human habitation;

2. persons who were living in a shelter;

3. persons who were exiting an institution and resided in a shelter or place not meant for human habitation before institution entry;

4. persons who would imminently lose their housing and lacked the resources to obtain other permanent housing;

5. unaccompanied youths or homeless families with children who experienced unstable housing; and

6. persons who were fleeing dangerous conditions in their current housing situation and lacked the resources to obtain other permanent housing.²⁸

We imposed no restrictions regarding other participant characteristics, such as being elderly, suffering from a mental illness, or having a military service history.

The title or abstract had to indicate that the study included an intervention. In the full-text article, at least 1 of the included interventions had to be identified as adhering to, or being based on, 1 of the 4 models of case management that we selected. Furthermore, the study had to be designed as a randomized controlled trial or a before-and-after study, incorporating a baseline and at least 1 follow-up assessment of outcome variables, comparing 2 or more groups that received different interventions. The article had to include participant-level outcomes. Because our aim was to provide a complete overview of all previously reported effects, we did not limit our selection to preselected outcomes of interest or impose restrictions regarding the services received by participants in control groups or length of follow-up.

Study Quality and Data Extraction and Synthesis

Two reviewers critically appraised the selected publications independently with criteria for grading internal validity derived from the US Preventive Services Task Force Methods Work Group,²⁹ by which evidence is classified as good, fair, or poor. We derived cutoff points for sample size, retention rate, and overall rating from Hwang et al.³⁰ and Altena et al.³¹ (Appendix B, Tables B and C, available as a supplement to this article at http://www.ajph.org).

Because we expected participants, settings, control group services, and outcome measures to differ markedly between studies, we could not conduct a meta-analysis. Instead, we focused on narrative descriptions of the evidence, with the goal to examine patterns across studies, provide information about applicability of results, and consider multiple explanations for differential findings across studies. We adapted the effect direction plot by Thomson and Thomas to prepare a visual summary of effect direction for all reported participant-level outcomes to accompany the narrative synthesis.³²

We first grouped the selected publications according to case management model and then according to study sample. One reviewer performed the data extraction, which a second reviewer checked. In addition to all participant-level outcomes, we extracted details of the intervention implementation, target population, recruitment setting, sample size, study design, and length of follow-up. We next tabulated outcome data and grouped them into 7 outcome domains derived from the extracted outcome measures, through a bottom-up approach. The 7 outcome domains, 4 of which were further divided into several outcome categories, were

1. service use (services provided by program staff and nonprogram inpatient, emergency, and outpatient services),

2. housing,

3. health (physical and mental),

4. substance use (alcohol and drugs),

5. societal participation (economic participation–security, criminal activity–legal problems, and social behavior–support),

6. quality of life, and

7. cost (service expenses and cost-effectiveness).

We determined the direction of effect impact (negative, positive, none, or unclear) and the level of statistical significance (P ≤ .05) for each extracted outcome measure (Appendix C, Tables D–J, available as a supplement to this article at http://www.ajph.org). We further synthesized the data to produce a single indication of overall impact in each outcome category for each publication, combining 2 or more measures where more than 1 outcome was reported for any outcome category. We used several synthesis techniques, such as tabulation, vote counting (as a descriptive tool), and concept mapping, in an iterative process as recommended by Popay et al. to conduct a narrative synthesis of the research-based evidence from the selected articles.³³

RESULTS

The results of the systematic search and selection process are summarized in Figure 1. We retrieved 3721 publications. Our review of titles and abstracts identified 133 publications that seemed to meet our criteria. Full-text versions of 5 publications could not be obtained, even by requesting them from libraries abroad or contacting the authors directly. Further examination of 128 full-text publications revealed that 33 satisfied our criteria for inclusion.^34–66 Interrater agreement for publication selection was moderate (Cohen κ = 0.49). Failing to include any of the 4 case management models as an experimental intervention was the most common reason for exclusion. Other publications had to be excluded because participants were not predominantly homeless or were younger than 18 years or because the studies described lacked a randomized or quasi-experimental design. Initially, we intended to include in our review strengths-based case management (SBCM), a model that emphasizes empowerment, self-direction, and the relationship between client and case manager,^67,68 but none of the publications that met our selection criteria studied SBCM. Therefore, we could only report results of studies on SCM, ICM, ACT, and CTI. Several publications reanalyzed previously published data, and others contained results from more than 1 research site; the 33 publications pertained to 21 unique study samples.

FIGURE 1—

Summary of database search and study selection in review of literature on models of case management for homeless adults, 1985–2011.

Study Quality

Agreement between reviewers, derived from the quality-rating items for a subsample of 6 articles, was substantial (weighted Cohen’s κ = 0.64). Of the 33 included publications, we rated 17 as having good internal validity, 15 as fair, and 1 as poor (Appendix B, Table C). The publication with a poor rating omitted important information about the study design. We were unable to determine whether comparable groups were assembled at baseline, whether groups suffered from differential attrition, whether valid and reliable measurement instruments were used and applied equally among groups, and whether an intention-to-treat analysis was performed.⁴²

Shortcomings encountered in study designs rated as fair were imprecisely defined interventions,³⁷ assembly of unequal groups,^37,38 and loss to follow-up of more than half of the sample or failure to report follow-up rates.^44,54,66 Other limitations were failing to maintain comparable groups during follow-up,^{37,43,54,60,62} failing to report details of measurement procedures and to adequately blind observers who assessed outcomes,^{35–39,41,48,66} and neglecting to perform an intention-to-treat analysis and to control for key confounders.^{35,39,43,45,52,62} We considered methodological limitations that could increase the risk of bias in our analysis.

Study Characteristics

Characteristics of included studies are shown in Table 2. Fourteen publications were issued in 1999 (when the review by Morse was published¹⁹) or later. Of the 21 study samples, 20 were recruited in the United States and 1 in the United Kingdom. The sample sizes ranged from 80 to 722 participants; the total sample size was 5618 participants. Varying definitions of homeless persons were employed across studies, and various homeless subgroups could be discerned: literally homeless persons, persons at risk for homelessness, homeless veterans, homeless ex-prisoners, homeless substance users, homeless persons with severe mental illness, and homeless persons with co-occurring mental and substance use disorders (dual diagnoses). These subgroups represent the large variety of recruitment settings where potential participants were approached. Because control group services often consisted of the usual care provided in a particular setting, services received by participants in control groups were also diverse (Table 2).

TABLE 2—

Sample Characteristics in Review of Research on the Effectiveness of Case Management Models for Homeless Adults

Case Management Model/Target population	Recruitment Setting and Location	Study Design	Intervention	Comparison Condition	Duration, Intervention/Comparison	Client-to-Staff Ratio, Intervention/Comparison	Baseline Sample Size, Intervention/Comparison	Study
Standard case management
Homeless addicted male veterans	Substance abuse and psychiatric inpatient units from VA hospital, Hines, IL	RCT	Case-managed residential care (residency phase and follow-up community phase)	Customary care (hospital program with referral to community services)	≤ 1 y/21 d	10–25 to 1/NA	178/180	Conrad et al.,34 Hultman et al.35
Homeless or marginally housed, poorly functioning people with severe, persistent mental illness	Local night shelters, hostels for the homeless, general practice clinic, and other organizations for the homeless, Oxford, UK	RCT	Case management by social services	Continued assistance as provided before the study	NR/NA	NR/NA	40/40	Marshall et al.36
Homeless substance abusers	Substance abuse treatment agencies and homeless shelters, Boston, MA	RCT	Case management services	Customary aftercare	≤ 9 mo/NA	30 to 1/NA	256/235	Orwin et al. (study 1)37
Homeless men with alcohol and other drug problems	Sobering-up station, Louisville, KY	CBA	Proactive case management	Self-initiated case management (same services as intervention, on request)	Ongoing/ongoing	NR/NR	142/37	Orwin et al. (study 2)37
Graduates of substance abuse programs who were recently or imminently homeless	Short-term inpatient substance abuse programs, Chicago, IL	Q-RCT	1: Case management only, 2: case management with supported housing	Access to aftercare in the community	≤ 8 mo/NA	NR/NA	1: 96, 2: 136/187	Sosin et al.38
Intensive case management
Homeless people with substance abuse problems	Substance abuse treatment center, Denver, CO	RCT	Intensive case management	Treatment and rehabilitation services by the treatment center	≤ 4 mo/NA	15 to 2a/NA	163/160	Braucht et al.39
High-frequency users of detox services who were homeless	Detox center, Seattle, WA	RCT	Intensive case management	Standard treatment by the detox center	Ongoing/NA	15 to 1/NA	150/148	Cox et al.40
Homeless people with a long-term, severe mental illness	Two state-operated mental hospitals, Chicago, IL	RCT	Assertive case management by psychiatric rehabilitation center	Community services (office-based outpatient care and case management)	NR/NA	10 to 1/NA	48/47	Korr et al.41
Homeless people with alcohol and other drug problems	County detox center and other agencies, Minneapolis, MN	RCT	Community-based intensive case management	Intermediate case management	NR/NR	15–20 to 1/40–50 to 1	82/117	Orwin et al. (study 3)37
Homeless substance users	Mobile medical outreach clinic, New York, NY	CBA	Intensive case management by social worker	Services as usual and self-referral to social worker	NR/NR	NR/NR	128 in total	Rosenblum et al.42
Homeless men with stable mental health and a problem with alcohol or drug use	Men’s homeless shelter, Philadelphia, PA	RCT	Shelter-based intensive case management	1: Integrated comprehensive services by residential treatment facility, 2: usual care shelter services (case management)	≤ 9 mo/1: ≤ 6 mo, 2: nonspecific	15 to 1/1: NA, 2: 50–75 to 1	200/1: 220, 2: 302	Stahler et al.43
Homeless people	Local human service agencies (shelters, soup kitchens, crisis and housing assistance services), Buffalo, NY	RCT	Intensive case management	Free to seek services in the community	≤ 8 mo/NA	NR/NA	101/101	Toro et al.44
Assertive community treatment
Homeless people with serious mental illness	Emergency rooms and inpatient units of the public acute care psychiatric hospital, St Louis, MO	RCT	1: Assertive community treatment only, 2: assertive community treatment with community workers	Broker case management	Ongoing/NR	10 to 1/85 to 1	165 in total	Burger et al.,45 Calsyn et al. (study 2),46 McBride et al. (study 2),47 Morse et al.,48 Wolff et al.49
Homeless people with a severe mental illness	Local emergency shelters, St Louis, MO	RCT	Assertive community treatment	1: Daytime drop-in center (assistance by social workers), 2: outpatient treatment (office-based outpatient therapy, medication, and assistance with social services)	Ongoing/1: NR, 2: NA	10 to 1/1: 40 to 1, 2: NA	52/1: 62, 2: 64	Calsyn et al. (study 1),46 McBride et al. (study 1),47 Morse et al.50
Homeless people with severe mental illness and substance use disorder	Variety of settings (emergency shelters, psychiatric hospitals, street locations frequented by homeless persons), St Louis, MO	RCT	1: Integrated assertive community treatment by community mental health agencies, 2: assertive community treatment only by community mental health agencies	Standard care (linkage assistance to access community services)	NR/NA	NR/NA	1: 61, 2: 65/65	Calsyn et al.,51 Fletcher et al.,52 Morse et al.,53 Morse et al.54
Homeless veterans with a substance abuse disorder, dual disorder, or psychiatric disorder	VA medical centers, San Francisco and San Diego, CA; New Orleans, LA; and Cleveland, OH	RCT	1: Modified assertive community treatment by VA case managers, 2: modified assertive community treatment with special access to housing subsidies	Standard VA homeless services (broker case management)	NR/short term	25 to 1/NR	1: 90, 2: 182/188	Cheng et al.,55 Rosenheck et al.56
Homeless or unstably housed people with severe mental illness, substance use disorder, high service use, and poor independent living skills	Two state-operated outpatient community mental health centers, Connecticut	RCT	Integrated assertive community treatment	Integrated standard clinical case management	NR/NR	10–15 to 1/25 to 1	99/99	Essock et al.57
Homeless people with severe, persistent mental illness, most with dual diagnoses	Inner-city psychiatric hospitals and community agencies for homeless people, Baltimore, MD	RCT	Assertive community treatment	Usual community services (generic case management services)	Ongoing/NR	10–12 to 1/NR	77/75	Lehman et al.,58 Lehman et al.59
Homeless and seriously mentally ill people being released from jail	Jail system of a large US urban center	RCT	Assertive community treatment by team of forensic case managers	1: Intensive case management by individual forensic case managers, 2: referral to aftercare by community mental health centers	1 y/1: NR, 2: NA	NR/NR	60/1: 60, 2: 80	Solomon et al.60
Critical time intervention
Severely mental ill people leaving a men’s shelter	On-site psychiatry program in a men’s shelter, New York, NY	RCT	Critical time intervention	Usual services only (referral to community agencies)	9 mo/NA	NR/NA	48/48	Herman et al.,61 Jones et al.,62 Jones et al., 63 Lennon et al.,64 Susser et al.65
Recently or imminently homeless veterans with serious mental illness	Inpatient units of VA medical centers, Chicago and Hines, IL; Houston, TX; Lyons, NJ; Montrose, NY; Richmond and Salem, VA; and San Diego, CA	HCT	Critical time intervention	Usual discharge planning services by inpatient unit staff and standard referral to outpatient services	6 mo/NA	15 to 1/NA	206/278	Kasprow et al.66

Note. CBA = controlled before-and-after study; HCT = historically controlled trial; NA = not applicable; NR = not reported; Q-RCT = quasi-randomized controlled trial; RCT = randomized controlled trial; VA = Veterans Affairs.

^aThis program employed a dyad structure, with pairs of case managers sharing caseloads.

Outcome measures also varied widely between studies. Frequently, different instruments were used to measure the same outcome (Appendix C, Tables D–J). Table 3 presents the overall impact in each outcome category for each publication, combining 2 or more measures where more than 1 outcome was reported (a visual interpretation of the synthesis is available as a supplement to this article at http://www.ajph.org).

TABLE 3—

Impact of Case Management Models on Outcomes for Homeless Adults

		Service use						Health		Substance use			Societal participation					Cost
Intervention/Study	Quality rating	Program	General Nonprogram	Inpatient Nonprogram	Emergency Nonprogram	Outpatient Nonprogram	Housing	Physical	Mental	General	Alcohol	Drugs	General	Economic	Criminal	Social	Quality of life	Service	Cost-Effect
Standard case management
Conrad et al.34	Good						Mixeda	Positive	No difference		Positivea	Mixed		Positive	No difference
Hultman et al.35	Fair	No difference20	Mixed8
Marshall et al.36	Fair		No differencea	No difference			No difference2		No differencea				No differencea	No difference2		Mixeda2	No differencea
Orwin et al. (study 1)37	Fair						Positive2	No difference2	Mixed2		Mixed2	Mixed2		Mixed3
Orwin et al. (study 2)37	Fair						No difference2	No difference2	No difference2		No difference2	No difference2		Mixed3
Sosin et al.38	Fair						Positive			Positive
Intensive case management
Braucht et al.39	Fair		No difference2	No difference5		No difference	No difference2	No difference2	No difference		No difference2	Mixed2	No difference	No difference2	No difference2	No difference2	Mixed
Cox et al.40	Good			Positive			Positive2				Mixed7			Mixed3
Korr et al.41	Fair			No differencea			Positive
Orwin et al. (study 3)37	Fair						No difference	No difference2	Mixed2		Mixed2	Negative		Mixed3
Rosenblum et al.42	Poor	Positive	Positive2		Positive		No difference	No difference				No difference		Positive
Stahler et al.43	Fair	Positive					Mixed2		No difference		Mixed3	No difference3		No difference2	No difference2	No difference
Toro et al.44	Fair						Mixeda2	No differencea	Mixeda4		No differencea			No differencea2		No differencea3
Assertive community treatment
Burger et al.45	Fair								Mixed2
Calsyn et al. (study 2)46	Good		Positive2													No difference10
McBride et al. (study 2)47	Good						Positive
Morse et al.48	Fair	Mixed10					Positive		Mixed6	No difference	No difference2	No difference2		No difference
Wolff et al.49	Good	Positive2		No difference3		No difference7	No difference2		Positive					Mixed3				Mixedb7
Calsyn et al. (study 1)46	Good		Positive													No difference9
McBride et al. (study 1)47	Good						Positive
Morse et al.50	Good	Positive2	Mixed10				Positivea		No difference2		No difference		No difference	No difference		No difference
Calsyn et al.51	Good														No difference6
Fletcher et al.52	Fair	Positive					Positive		No difference	No difference
Morse et al.53	Good	Positive3					Positive		No difference	No difference2				No difference				Mixed4
Morse et al.54	Fair	Mixed5					Positive		No difference		No difference2	No difference2
Cheng et al.55	Good						No difference4	No difference	No difference		No difference3	No difference3		No difference		No difference3	Negative
Rosenheck et al.56	Good	Mixedb15				No difference	No difference6	No difference	No difference2		No difference2	No difference2		No difference3	No difference3	No difference3	No difference6	No difference18
Essock et al.57	Good			Mixeda			Mixeda		No differencea	Positivea	Mixeda2	No differencea2	No differencea		Mixeda		No differencea
Lehman et al.58	Good			Mixed3	Mixed3	Mixed3	Mixed4	Mixed	Positive				No difference		No difference		Mixed7
Lehman et al.59	Good			No difference5	Mixed3	Mixed4												Mixed13	No difference
Solomon et al.60	Fair	Mixed3													No difference
Critical time intervention
Herman et al.61	Good								Mixed3
Jones et al.62	Fair			Mixedb3	No differenceb3	Mixedb6	Mixedb3								Mixedb2
Jones et al.63	Good						Positive3							No difference4				No difference21	Mixed6
Lennon et al.64	Good						Positivea
Susser et al.65	Good						Mixed6
Kasprow et al.66	Fair			Positivea		No differenceb	Mixed2		Positivea		Positive2	Positive2		No difference2

Note. Mixed = mixed or conflicting study findings; negative = intervention had negative impact on outcome; no difference = intervention and comparisons outcomes were the same; positive = intevention had a postive impact on outcome. All differences between control and intervention group at follow-up were significant at P ≤ .05 (unless otherwise indicated). Synthesis of multiple outcomes within same outcome category: where multiple outcomes all reported effects in the same direction, we reported this effect direction. Where direction of effect varied across multiple outcomes, we reported the majority effect direction if 70% of outcomes reported a similar direction. If less than 70% of outcomes reported a consistent direction of effect, we reported the effect direction as mixed. Where availability of statistics or data varied, we considered effects as statistically significant if statistical significance was available for more than 60% of the outcomes. Number of outcomes within each category synthesis was 1 unless indicated in subscript beside effect direction.

^aDifference in change between control and intervention group.

^bNo statistics–data reported.

Service Use

The 2 studies of SCM that examined service utilization detected few differences between SCM and the control conditions.^35,36 In a sample of substance-dependent homeless veterans, SCM participants received more support from program staff and other participants and were better prepared for program completion than participants in the control program.³⁵ However, SCM did not increase the use of other Veterans Affairs services, as had been expected at the outset of this study. With the exception of reporting more substance abuse treatment at the 3-month follow-up, SCM participants reported similar service use as controls.³⁵ In line with these findings, a second study showed that SCM for mentally ill people who were homeless or marginally housed did not significantly affect participants’ needs for psychiatric and social care or reduce the length of hospital stays.³⁶

Two studies, both of which examined samples of homeless people with substance use problems, compared the services offered by ICM programs to usual case management services. Participants recruited from a homeless shelter who received ICM were more satisfied than control participants, although a minority (29%) of the ICM participants completed the program.⁴³ In the other study, participants recruited at a medical van who were assigned to the ICM group had significantly more contacts with the medical van’s case manager than did control participants, who had the opportunity for self-referral to the same case manager.⁴² Although these 2 studies found that program services significantly differed between conditions, findings from 3 other studies on the impact of ICM on nonprogram service utilization were mixed.

The number of days spent in psychiatric hospitals by homeless mentally ill people did not differ between ICM and control participants.⁴¹ Among homeless substance users, ICM did not have a significant differential effect on the number of days spent in residential treatment facilities or the number of in- and outpatient services received.³⁹ We found some evidence, however, that ICM was more effective than standard detoxification treatment in reducing subsequent detox admissions.⁴⁰

Six studies, as reported in 8 articles, compared the services received by ACT and control participants.^{48–50,52–54,56,60} These studies indicated that, for several different homeless subpopulations, ACT increased contacts between participants and case managers or other program staff,^{48–50,53,54,56,60} enhanced the level of assistance directly provided by program staff,^48,53,56 and improved participant satisfaction.^{48–50,52,53}

One article, which reported findings from 2 studies with severely mentally ill homeless participants, showed that participants in the ACT programs had relatively larger professional networks.⁴⁶ In 1 of these same samples, Morse et al. found that ACT participants also had more contacts with service agencies than did control participants.⁵⁰ In a reexamination of the other sample, however, Wolff et al. found no significant differential effects with regard to in- or outpatient services for these mentally ill homeless participants.⁴⁹ This reanalysis may have lacked sufficient power or may have been biased by differential attrition, because service utilization data were available for approximately half of the original sample.

For participants with dual diagnoses, we found evidence that ACT is effective in shortening the length of psychiatric hospital stays^57–59 and reducing the number of emergency room visits for mental health problems.^58,59 ACT was not found to significantly reduce other inpatient service use by these participants, such as inpatient medical care,^58,59 residential substance abuse treatment,^58,59 or mental health rehabilitation.⁵⁹ Two articles on the same study reported that ACT participants with dual diagnoses visited outpatient mental health services and substance abuse treatment more often than participants receiving generic case management services.^58,59 Among substance-using homeless veterans, however, ACT did not have a differential effect on outpatient service use.⁵⁶

None of the articles examined the differences between program services provided by CTI and control conditions. Preliminary results indicated that CTI increased use of nonprogram outpatient services.⁶² In another study, CTI reduced the length of hospital and other institutional stays for mentally ill homeless veterans.⁶⁶

Housing

Of 5 studies examining the impact of SCM on housing stability, 3, all reporting on homeless substance users, showed statistically significant effects.^34,37,38 The 2 studies that did not find a positive impact of SCM on housing outcomes differed in several important ways. In 1 study, high-risk participants in the sample of homeless substance users were all assigned to the SCM condition, which as implemented did not differ in type or intensity of services from the control condition.³⁷ In the other, the sample consisted of mentally ill people instead of substance users, and sample size of 80 may have provided insufficient power to reveal a significant difference between groups in the housing variables.³⁶

Seven studies of ICM produced mixed results on housing outcomes. Of the 5 studies investigating the effects of ICM on homelessness or residential stability in substance-abusing populations,^{37,39,40,42,43} 1 reported a significantly better result for ICM than for the control condition.⁴⁰ These mostly nonsignificant findings could have been attenuated by treatment nonadherence and lack of differential service utilization between groups. For example, 71% of participants assigned to shelter-based ICM services for substance-using homeless men did not complete the program.⁴³

The 2 studies of ICM that did not examine homeless substance users showed a positive impact on housing. In a study with severely mentally ill participants, ICM significantly improved housing stability.⁴¹ In a more heterogeneous sample of homeless persons, participants receiving ICM experienced better living conditions during follow-up than did control participants.⁴⁴

Most articles examining the effect of ACT on housing outcomes found positive effects.^{47,48,50,52–54,57,58} Homeless persons with severe mental illness who received ACT spent fewer days homeless or more days in stable housing than did participants who received drop-in center services, office-based outpatient treatment, or less proactive case management.^47,48,50 For homeless participants with dual diagnoses, ACT programs also improved housing stability more than SCM and other forms of linkage assistance.^{52–54,57,58} As reported in 2 articles, 1 study with homeless veterans with substance abuse disorders did not find a significant positive effect of ACT on housing. For this sample, ACT did not have a significant impact on any of the housing-related outcome measures, unless participants were also supplied with special access to subsidized housing (Section 8 vouchers).^55,56 However, the model integrity of this ACT program was debatable because the case managers had relatively high caseloads of 25 clients.

Multiple reports examined CTI housing outcomes in 2 unique samples.^62–66 For mentally ill men leaving a homeless shelter, adding CTI to community services for 9 months was effective in decreasing homeless nights.^62–65 Interestingly, the difference between groups became more pronounced after the time-limited intervention had ended.⁶⁵ In a study with homeless veterans who were leaving inpatient care, CTI significantly increased days housed, although the CTI and control groups did not differ in nights spent homeless.⁶⁶

Physical and Mental Health

Two articles examined the impact of 3 SCM programs on physical health problems.^34,37 In 1 study, the SCM program for homeless substance users, provided in a residential setting, helped to lessen the severity of participants’ medical problems.³⁴ Two other studies employed similar samples and used the same measurement instrument but did not detect any effects, although results might have been weakened by attrition bias and program nonadherence.³⁷ The 4 studies that addressed mental health problems did not find a significant positive effect of SCM.^34,36,37

Homeless substance users did not benefit from ICM in their physical or mental health in 4 studies.^37,39,42,43 In a heterogeneous group of homeless persons, ICM had a positive impact on interviewer ratings of psychiatric symptom severity.⁴⁴ Although participants’ ratings of symptom severity did not differ between groups, ICM participants reported fewer stressful life events, which have been well established as a predictor of several mental disorders, particularly depression.^69–71

Because the 3 articles that discussed the impact of ACT on participants’ physical health generally did not report differential effects, little evidence exists that ACT affects this outcome.^55,56,58 Two of 6 studies that assessed mental health impacts found statistically significant reductions in psychiatric symptoms.^48,58 In a mentally ill sample, Morse et al. found that interviewer ratings for some symptoms were lower for the ACT group, although their unmasked interviewers may have introduced bias.⁴⁸ Another study found that homeless participants with dual diagnoses reported fewer symptoms if they had received ACT.⁵⁸

A significant effect of CTI on the reduction of psychiatric symptoms was demonstrated in both studies with mental health problems as an outcome.^61,66 No article reported inclusion of a physical health measure.

Substance Use

Four studies of SCM, whose participants were homeless people with substance abuse problems, assessed substance use outcomes, as reported in 3 articles.^34,37,38 All but 1 found differential effects, suggesting that SCM was significantly more effective than referral to community services in reducing alcohol and drug use among homeless substance users.^34,37,38 In the 1 study that did not replicate these results, all participants at high risk for relapse were assigned to the SCM group. Moreover, the services received hardly differed between groups. Therefore, it is not surprising that SCM participants did not reduce their substance use more than control participants.³⁷

Four of 6 studies of the effect of ICM on alcohol or drug use did not show a positive impact.^37,39,42,44 One study provided some evidence that ICM decreased days drinking and the severity of alcohol problems for homeless substance users.⁴⁰ This was confirmed by a per-protocol analysis of a second study with a similar sample: program completers reported less alcohol use than did control participants.⁴³ However, 2 other studies failed to replicate these findings,^39,44 and another study's results significantly favored the control condition.³⁷ These nonsignificant and negative results could have been biased by methodological limitations, such as high rate of attrition⁴⁴ and lack of differentiation in services received.^37,39

Five studies, which produced 8 articles with alcohol and drug use outcomes, concluded that ACT did not significantly affect substance use or related problems.^{48,50,52–57} One study on CTI looked at substance use variables.⁶⁶ In a sample of mentally ill homeless veterans, those offered CTI improved more with regard to alcohol and drug use than participants who received usual services. Furthermore, participants in the CTI group spent less money on these substances.⁶⁶

Societal Participation and Quality of Life

Three articles with findings from 4 studies on SCM included measures related to societal participation.^34,36,37 Two studies found that mentally ill or substance-using homeless participants who received SCM did not spend more days in employment.^36,37 Homeless substance users receiving SCM also did not experience more economic security,³⁷ although they generally reported fewer problems that interfered with employment.^34,37 One study looked at the impact of SCM on the severity of legal problems experienced by homeless substance users and did not find a differential effect.³⁴ In a sample of mentally ill homeless persons, interviewers observed less deviant behavior among SCM than among control participants, although participants’ perceptions of their social behavior and interviewer ratings of general functioning did not differ between groups.³⁶

We found very little evidence that ICM improved economic participation or security,^{37,39,40,43,44} reduced criminal activity or legal problems,^39,43 or promoted social behavior or support.^39,43,44 Two studies indicated that ICM could help homeless substance users to access sources of public assistance.^40,42 A study with a heterogeneous sample of homeless people, however, did not yield the same result.⁴⁴

Similarly, ACT did not seem to have any impact on economic participation or security.^{48–50,53,55,56} Of 5 studies exploring the effect of ACT on measures of criminal activity or legal problems,^{51,56–58,60} 1 showed a significant differential effect: for participants with dual diagnoses, ACT significantly decreased the likelihood of being incarcerated.⁵⁷ In addition, ACT did not affect variables related to social support, except for an effect on perceived material support at 1 research site.⁴⁶ No other differences were found between ACT and control groups in the size or quality of participants’ social network for 2 samples of mentally ill and 2 samples of homeless people with dual diagnoses.^46,50,55,56

Studies of CTI found no differential effect regarding days in employment⁶⁶ or income.^63,66 One study found that CTI participants spent fewer days in jail than control participants. The article, however, did not report any statistics.⁶² None of the studies included measures of social behavior or support as outcomes.

Few studies considered quality of life as an outcome. One study, on the effectiveness of ACT for homeless persons with dual diagnoses, found a significant improvement.⁵⁸ ACT participants were generally more satisfied with life than those receiving SCM services at a 6-month follow-up. Over longer periods of follow-up, however, no evidence was found for an effect of ACT on this outcome in samples of dually diagnosed homeless persons^57,58 or substance-abusing homeless veterans.^55,56 Similarly, SCM did not seem to improve the quality of life of homeless and marginally housed people with mental disorders relative to usual community care. In a study with homeless substance users, general life satisfaction was higher among control participants, who received usual services, than among participants who received ICM.³⁹

Service Costs and Cost-Effectiveness

No study examined the costs associated with SCM or ICM. In line with the differential effects of ACT on service utilization, studies confirmed that costs for outpatient services, including case management services, were higher for ACT than for standard services offered by psychiatric hospitals and agencies serving homeless persons.^49,53,56,59 Costs for inpatient services, however, were lower, which led Lehman et al.⁵⁹ and Wolff et al.⁴⁹ to conclude that their ACT programs were not more expensive than usual services and achieved better results.

Similarly, the pattern of service use associated with CTI was reflected in its cost. Acute mental health costs, which include charges for inpatient and emergency services, were lower for CTI than for standard shelter services, although this difference was not significant.⁶³ The only significant differential effect was a substantial reduction in shelter costs among CTI participants.⁶³ Jones et al. demonstrated that the cost of resources used by CTI participants did not differ from the costs for usual care participants and that the costs for providing CTI were compensated for by long-term improvements in housing stability.⁶³

DISCUSSION

Our systematic review is the first to our knowledge to provide a comprehensive overview of the evidence provided by randomized controlled trials and quasi-experimental studies for the effectiveness of 4 models of case management in homeless populations. Because the case management models perform the same functions, they are not mutually exclusive.²⁰ Many studies we reviewed did not ascertain whether services were delivered in accordance with the criteria of the model. We categorized the studies by model according to definitions provided in the articles; these categories were correct only to the extent that the studies reported accurate information about the models.

Except for 1 study conducted in the United Kingdom with severely mentally ill people, all studies concerned with SCM recruited homeless substance users as participants. Although we found little evidence for a differential effect on service utilization, the findings provided some evidence that SCM is effective for this homeless subpopulation in improving housing stability, reducing substance use problems, and removing employment barriers. For the mentally ill sample, however, few of these results were replicated. Thus, SCM seemed to be more beneficial than usual care for substance-using homeless persons.

Five out of 7 studies that assessed the effect of ICM also focused on homeless substance users. For this group, findings were nonsignificant or mixed in all outcome categories except for access to public assistance. Study quality ratings and service utilization data suggested that these largely nonsignificant findings could be the result of treatment nonadherence and lack of between-group differentiation in the services received. The 2 other ICM studies provided some evidence for a positive effect of ICM on housing outcomes for severely mentally ill homeless persons and the general homeless population. However, more research is needed before any conclusions can be drawn about the consistency of these findings.

The samples in studies of ACT consisted of homeless persons with dual diagnoses, severely mentally ill persons, substance-using veterans, and mentally ill ex-prisoners. Results indicated that ACT improved the housing stability of severely mentally ill as well as dually diagnosed homeless participants more than less proactive case management models. Outcomes related to housing were not included in the study with mentally ill ex-prisoners and did not improve in the sample of substance-using veterans; however, this could be attributable to a lack of model fidelity in this study. For all subpopulations, findings in the other outcome categories were largely nonsignificant or inconsistent. Although ACT appeared to influence patterns of mental health service use, most studies did not show a differential effect of ACT on psychopathology or other mental health outcomes. Nevertheless, the improvements in housing stability and reductions in inpatient and emergency mental health service use seemed to be sufficient to compensate for the higher costs associated with ACT.

Our findings are consistent with the 4 previous reviews of the literature on this topic. All these reviews found ACT superior to other services, including other models of case management, in helping severely mentally ill homeless persons to achieve housing stability.^19,25–27 Contrary to our findings, however, Coldwell and Bender also concluded that ACT participants demonstrated greater improvement in psychiatric symptom severity.²⁶

CTI was examined in 2 samples of severely mentally ill homeless persons, 1 group leaving a homeless shelter and the other leaving inpatient care for veterans. For both groups, CTI was significantly better than usual services in supporting housing stability and reducing psychiatric symptoms and substance use. The improved level of housing stability experienced by these severely mentally ill participants appeared to be linked to the positive impact of CTI on the length of hospital, shelter, and other institutional stays. CTI achieved better long-term results than usual care with similar associated costs. CTI was the least researched model in our review, so consistent results from further studies are needed. Nevertheless, results from these 2 studies were very promising.

Across the 4 different models, case management generally seemed to have a positive impact on housing stability and patterns of service use. Findings about substance use outcomes were mixed, and effects on variables measuring health, societal participation, and quality of life were largely nonsignificant.

Limitations

Because of great variability between studies, comparisons of research findings could only be undertaken with great caution. In addition, caution was warranted by heavy reliance on participants’ self-reports. Self-report data can give rise to over- or underreporting of treatment effects and a distortion of the differences between experimental and control groups.^72–74

We excluded reports that were not published in English between 1985 and 2011, possibly giving rise to selection bias. We also excluded studies that did not include a control group, such as descriptive reports and studies with a pretest–posttest design. The qualitative information that we may have omitted as a result might have elucidated the quantitative data provided by randomized and quasi-experimental research.

Generalizing Research Findings

Our review showed that case management has produced favorable effects in homeless populations, but also revealed gaps in the evidence. Because the evidence for the effectiveness of case management has been collected in the context of specific times, locations, and service settings, this research cannot easily be extrapolated. Most of the trials in our review that assessed the effectiveness of case management were conducted in a particular time frame as part of multisite demonstration programs in the United States that addressed the specific problems of homelessness of that time and focused on individuals with chronic or severe mental illness, substance abuse problems, or dual diagnoses.^75,76 Although the prevalence of mental illness and substance use among homeless people is still high, many are not mentally ill.¹¹ More evidence is needed to establish which model is most suitable and cost-effective for homeless people who are not mentally ill and are not substance users but who often have fewer or other support needs.

Similarly, we found examples of limited generalizability across countries. All but 1 of the studies we reviewed were conducted in the United States. The only European study, from the United Kingdom, could not replicate many of the findings from earlier studies. Because other countries have marked differences in social welfare systems, housing and labor markets, and health care systems, in addition to differences in the nature of their homeless populations,^10,77 it is highly likely that evidence-based practices from the United States will not produce the same results in European countries. In a review of the research literatures in the United States and other developed nations, Toro also notes that systematic research evaluating interventions for homeless people is virtually nonexistent in Europe.¹⁰

Implications for Future Research and Practice

To properly inform policymakers in the European Union, experimental trials should be conducted among different homeless groups in a variety of service settings and countries. These studies should be carefully designed. They should aim for more uniformity in outcomes examined and for more standardization of measurement instruments. Moreover, several important outcomes have received insufficient study. Few publications in our review included outcomes related to quality of life, societal participation, physical health, or community integration.

Future studies should summarize or refer to key components of the intervention being studied and present results of model fidelity assessment.^19,27 Inclusion of model fidelity in the study design is vital to explore relationships between case management models, homeless subgroups, service settings, and outcomes. Rather than comparing competing models, it may be more fruitful to attempt to predict which well-defined components of a given case management model will facilitate favorable outcomes for certain homeless subpopulations and what approach will be most cost-effective.⁷⁸

Our results suggest that practitioners could employ case management to assist homeless persons with improving their housing stability and changing their service use patterns. We found little evidence for the effectiveness of ICM, but this could very well be attributable to factors not related to this model. SCM seems to improve housing stability, reduce substance use problems, and remove employment barriers for homeless substance users more than referral to community services. Compared with SCM and other case management services, ACT has consistently produced positive effects on housing stability and has been found to be cost-effective. However, this model seems to be suitable mainly for mentally ill or dually diagnosed homeless persons with multiple and complex needs.²³ CTI has also produced promising results and seems to be more applicable for a variety of settings and populations because of its practical and time-limited nature.²⁴ Only when the evidence gaps have been addressed can we establish which case management models or which specific components of these models are most suitable to accompany housing, as part of a rapid-rehousing approach to homelessness, for specific homeless subgroups.