Bridging Hospital and Community Care for Homeless Adults with Mental Health Needs: Outcomes of a Brief Interdisciplinary Intervention

Abstract

Objective:

This study examines health and service use outcomes and associated factors among homeless adults participating in a brief interdisciplinary intervention following discharge from hospital.

Method:

Using a pre-post cohort design, 223 homeless adults with mental health needs were enrolled in the Coordinated Access to Care for the Homeless (CATCH) program, a 4- to 6-month interdisciplinary intervention offering case management, peer support, access to primary psychiatric care, and supplementary community services. Study participants were interviewed at program entry and at 3- and 6-month follow-up visits and assessed for health status, acute care service use, housing outcomes, mental health, substance use, quality of life, and their working alliance with service providers. Linear mixed models and generalized estimating equations were performed to examine outcomes longitudinally. Additional post hoc analyses evaluated differences between CATCH participants and a comparison group of homeless adults experiencing mental illness who received usual services over the same period.

Results:

In the pre-post analyses, CATCH participants had statistically significant improvements in mental and physical health status and reductions in mental health symptoms, substance misuse, and the number of hospital admissions. The strength of the working alliance between participants and their case manager was associated with reduced health care use and mental health symptoms. Post hoc analyses suggest that CATCH may be associated with statistically significant improvements in mental health symptoms in the study population.

Conclusions:

A brief interdisciplinary intervention may be a promising approach to improving health outcomes among homeless adults with unmet health needs. Further rigorous research is needed into the effectiveness of brief interventions following discharge from hospital.

Among adults with complex health needs, homeless people are a particularly underserved population,¹ even within systems with universal health insurance, such as Canada.^2,3 Compared to the general population, people experiencing homelessness have higher rates of many acute and chronic health conditions, including mental illness and addictions, cognitive impairment,^4–6 and cardiovascular and infectious diseases.⁵ They face many barriers to accessing appropriate health services, including stigma and discrimination,^7,8 as well as service unavailability and fragmentation,^4,5,9 and are frequent users of hospital services.⁵ Expectedly, strategies to address their needs are a priority consideration in many jurisdictions.¹⁰

To date, a variety of interventions have been proposed to improve the health of adults experiencing homelessness,^11–13 although interventions aimed at improving coordination and continuity of care have focused primarily on homeless adults experiencing mental illness.^14,15 Care received after hospitalization has been cited as the most important factor in reducing psychiatric inpatient utilization,¹⁶ resulting in the emergence of “bridging” or “transitional” care coordination and case management strategies.¹⁷ Such strategies aim to facilitate continuity of care, including the provision of accessible, coordinated, and comprehensive services as patients move through different settings,^18,19 and can take place before or at the point of discharge, often including bridging to postdischarge services.²⁰

Critical time intervention (CTI) is a specific time-limited case management intervention for homeless people with mental illness, bridging care from hospital to community settings. Prior studies in the United States have shown that CTI is cost-effective²¹ and reduces homelessness,^22,23 psychiatric symptoms,²⁴ and psychiatric inpatient readmissions.²⁵ The model has been successfully adapted to support care transitions among people leaving prisons, shelters, and other institutions in the United States,^14,26,27 and testing in other delivery contexts is under way.

Better care for people with complex health conditions is a priority of the Government of Ontario’s health care strategy.²⁸ Consistent with this strategy, Toronto’s local health authority funded the Coordinated Access to Care for the Homeless (CATCH) program in 2010, aiming to improve health and health service use outcomes among homeless adults in Canada’s largest urban centre. CATCH offers one-stop access to comprehensive services, including psychiatric and medical care, peer support, and brief case management services over 4 to 6 months following discharge from local hospital emergency departments or inpatient units.

Given the paucity of Canadian studies, the primary goal of this study was to assess factors associated with health and service use outcomes of CATCH program participants 6 months after program enrolment. A secondary study goal was to evaluate, in a post hoc analysis, differences in health outcomes between CATCH participants and a comparison sample of homeless adults with mental illness receiving usual services over the same time period. We hypothesized that the brief intervention might lead to improved health status and other outcomes and that a stronger working alliance would be associated with improved outcomes; second, we anticipated that CATCH participants would experience superior outcomes compared to a group of homeless people with mental illness receiving usual care.

Method

Study Design and Setting

The study design and setting have been described in detail elsewhere.²⁹ Briefly, the study was designed as a mixed-methods case study evaluating the program’s uniqueness and complexity.³⁰ The current article focuses on participant self-reported quantitative survey data. Qualitative findings are described elsewhere.¹⁸

The study took place in Toronto, home to the largest cohort of homeless people in Canada, with over 5,000 people experiencing homelessness on any given night.³¹ Despite a variety of health and social services, including homelessness services, past surveys indicate that a large number of homeless adults in Toronto lack a usual source of health care³² and describe their health needs as a barrier to achieving housing stability.³¹

CATCH Intervention

The CATCH program is a 4- to 6-month interdisciplinary intervention, offering homeless adults discharged from hospital case management and additional health supports. Case management services include community visits, crisis intervention, supportive therapy, assistance in obtaining income supports and housing, and transfers of care to long-term community services. Furthermore, CATCH offers participants peer support as well as weekly low-barrier clinics staffed by a nurse, a primary care physician, and 2 psychiatrists. Clinic staff work with case managers to conduct interdisciplinary assessments and develop comprehensive care plans. The program has partnerships with community health centers, social services, and other community agencies to improve access to income, health, and social supports available locally.

CATCH Study Participants

CATCH is a partnership of 3 local hospitals, a large community mental health agency, a homeless shelter, a consumer-driven agency, and a physician practice plan, and it offers “one-stop” interdisciplinary services. Program participants are referred from participating hospitals or from the local homeless shelter following discharge from hospital. All participants in the current study met CATCH program eligibility requirements, including current homelessness (living on the street or crisis/emergency shelters or couch surfing), unmet mental health needs (identified by health providers), and unmet support needs identified by participants (no current connection to case management or primary or psychiatric care). Furthermore, they met study-specific criteria, including being ≥18 years of age, having no prior receipt of CATCH program services, and living in the province of Ontario. Program exclusion criteria included need for residential care or a history of severe aggression. All study participants provided written informed consent. The study received Research Ethics Board approval at St. Michael’s Hospital.

Data Collection

Research staff met with participants for an hourlong face-to-face interview at study baseline and at 3- and 6-month follow-up visits. Baseline interviews were completed within 2 weeks of program enrolment while follow-up visits were completed within a 4-week window of their due date. Participants received a $25 honorarium for each completed interview.

Study Outcome Measures

Study outcomes have been previously described in detail.²⁹ The primary study outcome was the change in participant health status from baseline to 6 months, evaluated by the physical and mental health component scores (PCS and MCS, respectively) of the Short-Form 36 (SF-36) health survey,³³ which has previously been used among homeless adults.^34,35

Secondary study outcomes included mental health symptoms, disease-specific quality of life, substance use, acute service use, and housing, assessed by the following instruments, respectively:

1. modified Colorado Symptom Index (CSI),³⁶ a 14-item instrument specifically designed for individuals with mental health problems;

2. the total score of the Lehman Quality of Life Interview (QoLI-20), a 20-item instrument that assesses 7 subscales (living situation, everyday activity, family, social relationships, finances, safety, and overall satisfaction) among people living with severe mental illness³⁷;

3. composite scores of the drug (DCOMP) and alcohol (ACOMP) modules of the Addiction Severity Index (ASI) measuring recent (last 30 days) severity of use^38,39;

4. self-reported number of hospitalizations and the number of emergency department (ED) visits in the 3 months prior to each follow-up visit; and

5. a modified version of the Residential Time Line Follow-Back Calendar⁴⁰ (RTLFB), which captured participant housing history in the 3 months prior to each follow-up visit, including the number of days spent in the following residence types: street, stable, temporary/unstable, emergency/crisis shelter, hospital, other institution, or other places. The following housing categories were defined as “homeless”: streets, unstable residence, or emergency/crisis shelters.^40–42

Finally, a 12-item self-report measure, the Working Alliance Inventory–Participant (WAI-PAR) measured the working-alliance construct and assessed client-service provider agreement on therapy goals, tasks, and the development of a strong relational bond between client and provider.⁴³ The WAI-PAR was administered at the 3- and 6-month interviews, provided the participant had a service provider at the time of the visit. Three domain subscores (goals, task, and bond) and a total summary score were tabulated, with greater scores indicating stronger alliance or agreement between the participants and their service provider.

Post Hoc Analysis

The comparison sample was selected among homeless adults with mental illness participating in the usual care arm of the At Home/Chez Soi (AH/CS) randomized controlled trial in Toronto, which took place during the same time period.^41,44 Briefly, AH/CS participants met the following study inclusion criteria: were ≥18 years old, were absolutely homeless or precariously housed, and had a serious mental disorder with or without coexisting substance use disorder, assessed by the Mini International Neuropsychiatric Interview 6.0 (MINI)⁴⁵ prior to study entry. Participants randomized to usual care could access housing, health, and social services available locally and were provided with information on existing services. Participants completed longer interviews at 6-month intervals up until 24 months postrandomization and shorter interviews in the intervening 3-month intervals.

Participants selected for inclusion in the comparison group reported at least 1 visit to a hospital emergency room in the 6 months prior to entering the study or at least 1 psychiatric or medical hospital admission in the 3 months prior to entering the study. CATCH program and comparison cohorts were mutually exclusive.

The following outcomes available for both studies at the baseline and 6-month follow-up time points were examined: mental health symptom severity (CSI); disease-specific quality of life (QoLI-20); health status (SF-12, a shorter 12-item instrument approximating the longer SF-36 instrument⁴⁶); substance use, using the following ASI questions: “How many days in the past month did you have alcohol problems?” and “How many days in the past month did you have drug problems?” (problems could include craving, withdrawal symptoms, disturbing effects of use, or wanting to stop and being unable to)⁴⁷; self-reported number of ED visits in the past 6 months; number of days spent in hospital in the past 6 months; and the number of days spent in stable housing in the past 6 months (using the RTLFB inventory).⁴⁰

Multiple Imputation

Prior to statistical analyses, the level of missing data was evaluated and data were imputed using a sequential regression multivariate imputation approach, often referred to as multivariate imputation by chained equations (MICE).⁴⁸ Sources of missing data included loss to follow-up, nonresponsive on specific items (particularly on multi-item instruments), skipped interviews, or study withdrawal. The proportion of missing data for outcomes ranged from 0% to 29% (median, 3.6%) at baseline and from 22% to 42% (median, 25%) at 6 months.

Forty imputed data sets were generated by MICE in STATA 13 (StataCorp, College Station, TX) using the “mi impute chained” function. The resulting data sets were combined and analyzed using PROC MIANALYZE procedure in SAS 9.4 (SAS Institute, Cary, NC).

Multiple imputation via MICE was also implemented for the AH/CS study and is detailed elsewhere.⁴⁹

Statistical Analysis

Pre-post analysis

For estimating changes in the primary outcomes of physical and mental health (PCS and MCS of the SF-36, respectively) and secondary outcomes of mental health symptomatology (CSI), quality of life (QoLI-20), and the alcohol (ACOMP) and drug (DCOMP) component scores of the ASI, we used linear mixed models with random intercepts (SAS PROC MIXED),⁵⁰ which treated time continuously (in months), adjusting for baseline predictors. Adjusted changes in primary outcome scores from baseline to 6 months were estimated from the final models.

Data on housing and acute care service use outcomes were measured at each of the 3 study visits, and all 3 outcomes comprised rates, which calculated the number of events or days during the observation period. For the baseline visit, the observation period included 6 months prior to study entry. For both the 3- and 6-month visits, the observation period included the 3-month period prior to each visit. We modeled the baseline, 3-month, and 6-month outcomes using general estimating equation (GEE) models using either a Poisson (hospital admissions) or negative binominal (ED visits, days homeless) distribution with an exchange covariance matrix, treating time categorically (6 months vs. baseline and 3 months vs. baseline) (PROC GENMOD), adjusting for baseline covariates, and, where required, including an offset in units of 3 months to account for participants’ amount of person time at baseline and each follow-up. Estimated rate ratios (RR) were generated from the models, which compared the number of visits or number of days in the past 3 months at each follow-up visit to baseline.

All pre-post models were adjusted for a variety of baseline variables, based on their potential influence on the outcome of interest, and included age, sex, ethnicity, chronic health conditions and baseline measures of alcohol and drug use, severity of mental illness, and percentage of time spent homeless (Suppl. Table S1).

Finally, Spearmen correlation coefficients were calculated to explore the relationship between the WAI-PAR total score and each subscore and study outcomes at 6 months.

Post hoc analysis

In the absence of randomization, models comparing 6-month outcomes between CATCH and comparison group participants resembled those used in the pre-post analyses and included baseline age, sex, ethnicity, chronic health conditions, alcohol and drug use, severity of mental illness, and the baseline values for each corresponding outcome.

Linear regression models were fit for SF-12 PCS, SF-12 MCS, CSI, and QoLI-20 Total. Negative binomial regression models were fit for number of ED visits, days with alcohol problems, days with drug problems, days in hospital, and days in stable housing. An offset in units of 6 months was included in models for the last 2 outcomes (days in hospital and days in stable housing) to account for individual person time over the 6-month period. Model results were combined over the imputed data sets (PROC MIANALYZE) to estimate mean differences or RR and 95% confidence intervals.

All analyses were conducted using SAS 9.4. Supplemental Table S2 outlines the covariates included in all models run as part of the post hoc comparison analyses.

All statistical tests were 2-sided, and statistical significance was set at a P value of less than .05.

Results

Sample Demographics

The CATCH sample characteristics are summarized in Table 1 and described in detail elsewhere.²⁹ Briefly, from January 1, 2013, to April 9, 2014, the program received 281 referrals, 240 of whom were eligible for the CATCH study and 225 of whom agreed to participate (94%). Data for 2 participants were removed due to interviewer comments regarding poor reliability of responses. As a result, the final analytic sample comprised 223 participants. Follow-up rates at 3 and 6 months were 85% (n = 190) and 78% (n = 174), respectively. Compared to those who completed the study at 6 months, participants lost to follow-up were less likely to report English as their main language. No other differences were noted between those completing the study and those who did not.

Table 1.

Sample Demographics and Clinical Variables for CATCH Participants (n = 223) and a Comparison Group of Homeless Adults with Mental Illness (n = 168).

Characteristic	CATCH Participants (n = 223)	Comparison Group (n = 168)	P
Age, mean ± SD, y	40.0 ± 12.0	41.2 ± 12.0	0.30
Male	173 (77.6)	116 (69.1)	0.06
Canadian born	165 (74.0)	102 (60.7)	0.01
English main language	187 (83.9)	111 (66.1)	<0.001
Ethnicity or cultural identitya
White	153 (68.6)	75 (44.6)	<0.001
Single, never married	135 (60.5)	106 (63.1)	0.61
Has at least 1 child under 18 years of age	70 (31.4)	48 (29.1)	0.63
Education
Less than high school	109 (49.3)	68 (41.0)	0.10
Completed high school	40 (18.1)	44 (26.5)
Completed college/university/grad school	72 (32.6)	54 (32.5)
Unemployed	206 (92.4)	162 (96.4)	0.09
Total income past month ($CAD)
<$500/month	76 (35.4)	50 (29.9)	0.28
≥$500 to $1000/month	88 (40.9)	82 (49.1)
≥$1001 or more/month	51 (23.7)	35 (21.0)
Single longest period of homelessnessb
<1 month to <1 year	132 (60.6)	66 (39.3)	<0.001
≥1 year to <3 years	40 (18.4)	52 (31.0)
≥3 years	46 (21.1)	50 (29.8)
Baseline mental health symptom severity (CSI), mean (95% confidence interval)c	42.0 (40.4 to 43.5)	41.1 (39.3 to 42.9)	0.45
Has 3 or more chronic health conditions	117 (52.5)	104 (61.9)	0.06

Values are presented as number (%) unless otherwise indicated. The following variables had missing data: children younger than 18 years (n = 3), education (n = 4), unemployment (n = 1), total monthly income (n = 9), and longest period of homelessness (n = 5). CATCH, Coordinated Access to Care for the Homeless; CSI, Colorado Symptom Index.

^aFour ethnic/cultural categories were used (white, ethnoracial, aboriginal, and other); however, due to small cell size in some categories, they are not shown here.

^bThe mean ± SD for single longest duration of homelessness was 1.8 ± 2.9 years and 3.0 ± 2.9 years for CATCH and comparison group participants, respectively. The median (interquartile range) was 0.58.0 (0.17 to 2.00) and 1.00 (0.42 to 3.00) for CATCH and comparison group participants, respectively.

^cCalculated after multiple imputation because of the high proportion of missing data.

Pre-Post Analyses

Unadjusted means and standard errors for study outcomes at baseline, 3-month, and 6-month visits are found in Supplemental Table S3, while Table 2 shows the results of the pre-post adjusted analyses. From baseline to the 6-month visit, CATCH participants had a statistically significant improvement in both mental (mean difference, 7.1; 95% confidence interval [CI], 4.8 to 9.4) and physical health (mean difference, 1.5; 95% CI, 0.004 to 2.9). Compared to baseline, at 6 months, intervention participants also had statistically significant reductions in mental health symptoms (CSI) (mean difference, –7.3; 95% CI, –9.0 to –5.7), drug use (mean difference, –0.029; 95% CI, –0.041 to –0.020), and alcohol use problems (mean difference, –0.045; 95% CI, –0.070 to –0.020). Gains from baseline to 6 months in quality of life were also statistically significant (10.4; 95% CI, 6.8 to 13.9). Significant reductions from baseline to 6 months were observed in the rate of hospitalization (RR, 0.61; 95% CI, 0.43 to 0.87) but not of homelessness (RR, 0.96; 95% CI, 0.83 to 1.11) or emergency department visits (RR, 0.77; 95% CI, 0.55 to 1.08).

Table 2.

Pre-Post Analyses: Adjusted Changes from Baseline to 3 and 6 Months for CATCH Participant Outcomes.

	3 Months vs Baseline		6 Months vs Baseline
Outcome	Mean Difference (95% CI)	P Value	Mean Difference (95% CI)	P Value
Physical component score (SF36-PCS)	0.73 (0.002 to 1.5)	0.049	1.5 (0.004 to 2.9)	0.049
Mental component score (SF36-MCS)	3.5 (2.4 to 4.7)	<0.0001	7.1 (4.8 to 9.4)	<0.0001
Mental health symptoms (CSI)	—	—	–7.3 (–9.0 to –5.7)	<0.0001
Quality of life (QoLI-20 total score)	—	—	10.4 (6.8 to 13.9)	<0.0001
Alcohol component (ASI)	—	—	–0.045 (–0.070 to –0.020)	0.0005
Drug component (ASI)	—	—	–0.029 (–0.041 to –0.017)	<0.001
	Rate Ratio (95% CI)	P Value	Rate Ratio (95% CI)	P Value
Number of hospital admissions in past 3 months	0.72 (0.55 to 0.96)	0.025	0.61 (0.43 to 0.87)	0.006
Number of ED visits in past 3 months	0.97 (0.75 to 1.26)	0.816	0.77 (0.55 to 1.08)	0.131
Number of days homeless in past 3 months	1.04 (0.93 to 1.18)	0.469	0.96 (0.83 to 1.11)	0.594

Mean values correspond to the adjusted monthly change from baseline (slope), whereas the rate ratios correspond to the estimated rate of events in the past 3 months at each time point divided by the rate of events in the past 3 months at baseline. Model results are based on data generated by multiple imputation; model covariates for each outcome are detailed in Supplemental Table S1. Higher values are associated with better outcomes for SF36-PCS, SF36-MCS, and QoLI-20 total score and with poorer outcomes for the ASI and the CSI. CATCH, Coordinated Access to Care for the Homeless; CI, confidence interval; ED, emergency department; QoLI-20, Lehman Quality of Life Interview; SF36-MCS, Short-Form 36 mental health component score; SF36-PCS, Short-Form 36 physical component score.

Study outcomes were associated with a number of baseline covariates, detailed in Supplemental Table S1. For the primary study outcomes, physical health status was negatively associated with baseline age (–0.22; 95% CI, –0.32 to –0.12) and having at least 3 chronic health conditions (–8.68; 95% CI, –10.97 to –6.38), while mental health status was negatively associated with baseline alcohol use problems (–7.85; 95% CI, –13.35 to –2.34).

Associations with Working Alliance

Completed WAI-PAR instruments were available for a total of 140 CATCH participants. Table 3 shows the distribution of the total and component WAI-PAR scores, as well as Spearman correlations between WAI-PAR values and study outcomes at 6 months. The WAI-PAR was significantly and positively associated with the number of hospital admissions and ED visits and mental health symptomatology but none of the other 6-month outcomes.

Table 3.

Working Alliance Inventory (WAI) among CATCH Study Participants.

	Distributiona		Correlationsb
Variable	Mean ± SD	Median (IQR)	SF36-PCS	SF36-MCS	Mental Health Symptoms (CSI)	QoLI-20 Total Score	Alcohol Component (ASI)	Drug Component (ASI)	No. of Hospital Admissions in Past 3 Months	No. of ED Visits in Past 3 Months	% of Days Homeless in Past 3 Months
WAI TASK (range 4.0-28.0)	23.1 ± 4.5	24 (21.0-26.5)	−0.030	0.107	−0.097	0.159	0.064	0.049	−0.211c	−0.141	−0.034
WAI BOND (range 9.5-28.0)	23.4 ± 4.2	24 (21.5-27.0)	–0.108	0.123	–0.094	0.177	0.151	0.098	–0.138	–0.105	0.013
WAI GOAL (range 11.0-28.0)	23.8 ± 3.9	24.5 (21.5-27.0)	0.029	0.075	–0.090	0.243	0.119	0.065	–0.180c	–0.172c	–0.078
WAI TOTAL (range 34.5-84.0)	70.5 ± 11.2	73 (65.0-78.0)	–0.025	0.102	–0.109c	0.206	0.112	0.073	–0.186c	–0.166	–0.044

Higher values are indicative of improved working alliance between client and worker for the total score and each component. ASI, Addiction Severity Index; CATCH, Coordinated Access to Care for the Homeless; CSI, Colorado Symptom Index; ED, emergency department; IQR, interquartile range; QoLI-20, Lehman Quality of Life Interview; SF36-MCS, Short-Form 36 mental health component score; SF36-PCS, Short-Form 36 physical component score.

^aMean (SD) and median (IQR) for the total and component scores.

^bSpearman correlations coefficients of the WAI-PAR subscales and total scores and study outcomes at 6 months. The WAI instrument was completed by 140 CATCH study participants who indicated that they had met at least once with a case manager, mental health care professional, or worker and completed a 6-month interview.

^cP < 0.05.

Post Hoc Analyses

Compared to CATCH participants, the comparison group was less likely to be Canadian born, English speakers, and white and more likely to be homeless for more than a year (Table 1).

CATCH participants, compared to the usual care group, had statistically significant reductions in mental health symptom severity (mean difference, –2.50; 95% CI, –4.84 to –0.16), the number of days experiencing alcohol (rate ratio, 0.43; 95% CI, 0.19 to 0.99), and drug problems (rate ratio, 0.21; 95% CI, 0.09 to 0.52) at 6 months. CATCH participants also had an increased rate of ED visits (rate ratio, 2.14; 95% CI, 1.41 to 3.25) and days spent in hospital (rate ratio, 5.79; 95% CI, 2.87 to 11.70) at 6 months compared to the cohort of adults receiving usual care (Table 4).

Table 4.

Post Hoc Comparison Analyses: Results of Analyses Comparing Outcomes at 6 Months among CATCH participants with a Comparison Group of Homeless Adults with Mental Illness Who Had Access to Usual Care.

Outcomes	Mean Difference at 6 Months (95% CI)	P Value
Physical component score (SF12-PCS)	0.59 (–1.93 to 3.10)	0.648
Mental component scores (SF12-MCS)	1.04 (–2.13 to 4.21)	0.519
Mental health symptoms (CSI)	–2.50 (–4.84 to –0.16)	0.036
Quality of life (QoLI-20 total score)	–0.19 (–4.81 to 4.44)	0.936
	Rate Ratio (95% CI)	P Value
Number of days in hospital in past 6 months	5.79 (2.87 to 11.70)	<0.0001
Number of days in stable housing in past 6 months	0.92 (0.43 to 1.97)	0.839
Number of ED visits in past 6 months	2.14 (1.41 to 3.25)	0.0004
Number of days with alcohol problems in past month	0.43 (0.19 to 0.99)	0.047
Number of days with drug problems in past month	0.21 (0.09 to 0.52)	0.0007

Model results are based on data generated by multiple imputation; model covariates for each outcome are detailed in Supplemental Table S2. CATCH, Coordinated Access to Care for the Homeless; CI, confidence interval; CSI, Colorado Symptom Index; ED, emergency department; QoLI-20, Lehman Quality of Life Interview; SF12-MCS, Short-Form 12 mental health component score; SF12-PCS, Short-Form 12 physical component score.

Discussion

Based on our findings, a brief multidisciplinary intervention may be a promising approach to improving health outcomes among homeless adults with unmet health needs following discharge from hospital. In pre-post analyses, program participants experienced improvements in mental and physical health and quality of life and reductions in mental health symptoms, substance misuse problems, and hospital admissions over 6 months. Both primary study outcomes were negatively associated with baseline factors: physical health status at 6 months was associated with age and comorbid conditions at baseline, while mental health status was associated with baseline alcohol use problems. The strength of the working alliance between CATCH participants and their case managers was associated with reduced hospital admissions and mental health symptoms, highlighting the importance of training and the interpersonal characteristics of service providers working with disadvantaged populations. Qualitative work with this and other disadvantaged populations has highlighted the importance of offering accessible and welcoming services, including service providers who are familiar with the realities of homelessness, attentive to individual service needs and preferences, responsive, and nonjudgmental.^18,51

Additional analyses with a comparison cohort of homeless adults with mental illness receiving usual care confirmed the program’s successful reduction in mental health symptoms and alcohol and drug use problems, as well as noted increased emergency department visits and days spent in hospital over 6 months. This finding may reflect differences between the groups, including the acuity of presentation at baseline, as well as the appropriate receipt of needed care.

Our findings are consistent with prior research on the effectiveness of brief interventions in reducing psychiatric symptoms²⁴ among homeless adults discharged from hospital. Furthermore, our qualitative findings¹⁸ suggest that CATCH can promote continuity of care by offering timely, responsive, accessible, and coordinated services, including connections to long-term services. Novel findings of our study include the negative association of substance use with mental health status outcomes, as well as the significant reductions in substance misuse among CATCH participants in both the pre-post and the comparison group analyses, highlighting groups that may require additional targeted interventions. Earlier studies examining brief interventions among homeless adults have not focused on substance misuse as a study outcome,^22,26 and further research, using a randomized design, will be required to further investigate this potential benefit.

Prior studies of interventions for homeless adults rarely have included physical health outcomes⁵²; a trial of housing assistance among homeless and unstably housed people living with HIV was an exception.⁵³ We observed that a brief multidisciplinary intervention may result in improved physical health outcomes among homeless adults with mental health and addiction needs and requires further confirmatory research. Health outcomes were negatively associated with baseline age and chronic health conditions. Future studies of interventions for homeless people should consider evaluating physical health outcomes and further explore these relationships.

Studies of CTI following discharge from hospital or shelters have been shown to improve housing outcomes over 18 months of follow up.^22,23 In contrast, CATCH was not associated with decreased homelessness or improved housing stability in either analysis. This is not surprising given the short duration of follow-up and the scarcity of affordable housing in the local community. Our qualitative work highlighted the importance of the broader service landscape and key additional services needed by homeless people, including affordable housing and accessible mental health services.¹⁸ Of note, CATCH participants did not have access to rent supplements or dedicated housing units. In contrast, pairing case management with access to housing, such as the case with Housing First programs, results in significantly improved housing outcomes among homeless adults with mental illness in Canada.⁴⁹

The current study was unique in that it examined the effect of a one-stop community-based intervention and included examination of the health status of homeless adults. While possible improvements in both the physical and mental components of health status, observed in the pre-post analysis, were not confirmed in the post hoc analyses with a comparison group receiving usual care, this lack of confirmation could have resulted from several causes. First, while we made every effort to find comparable outcomes between the cohorts, this study was not originally designed as a matched cohort study. The CATCH study used the SF-36 as the main outcome measure, while the comparison cohort completed the SF-12. Although the SF-12 is a shorter version of the SF-36,⁴⁶ the removal of two-thirds of the scale items may have resulted in the loss of the domains of improvement observed among CATCH participants. Alternatively, the lack of observable differences in the health status outcomes could result from sample differences; most notably, all CATCH participants had a history of psychiatric hospitalization while AH/CS participants had on average a longer history of homelessness.

This study has some additional limitations. The lack of a randomized trial design limits causal inference. Although we attempted a post hoc matched cohort analysis, several differences between the cohorts may limit generalizability. First, the comparison group participants were not exclusively recruited at the point of discharge from hospital and had less frequent health service use histories at study entry. Furthermore, they differed from CATCH participants in terms of demographic and clinical characteristics. However, we attempted to address this by selecting comparison group participants with a recent history of self-reported acute care utilization and adjusted for baseline differences. Finally, we were unable to collect WAI-PAR data from all study participants, and it is possible that those with poorer alliances severed the relationships with their workers prior to data collection.

Conclusions

In summary, this study used a pre-post cohort design to examine factors associated with health outcomes among participants of a brief interdisciplinary intervention for homeless adults with unmet health needs following discharge from hospital. Brief interdisciplinary interventions offer a promising approach to improving health outcomes for this population, and further rigorous research is needed to examine the effectiveness of such interventions in the Canadian context.