Racial-Ethnic Differences in Health Service Use in a Large Sample of Homeless Adults With Mental Illness From Five Canadian Cities

Compared with the general population, homeless individuals experience a higher prevalence of acute and chronic health conditions (1), cognitive impairment (2), and premature mortality (3,4). In addition, this population experiences considerable barriers to accessing health services and has high rates of unmet needs for health care (5–9). Among homeless adults, those from ethnoracial minority or immigrant groups face more access barriers than their white counterparts, including language barriers, lack of knowledge about services, stigma, and perceived discrimination (10–13).

An extensive body of literature documents disparities in access to and use of general medical and mental health services among ethnoracial minority groups in the United States and the United Kingdom (14–16), where these groups have been found to receive poorer quality care, even when analyses have adjusted for access-related and socioeconomic factors (14,17). Although a number of U.S. and U.K. studies have investigated the role of race and ethnicity in health care access among homeless adults with mental illness (18–23) and have examined specific homeless subpopulations, including veterans (24,25) and women (26), their findings may not be generalizable to other health care contexts or to homeless individuals who face unique health and social challenges.

Our knowledge about ethnic and cultural group disparities in health care access and use among homeless adults with mental illness in Canada is scant. Given the diversity of the population, the growing numbers of homeless people, and the disproportionately high rates of Aboriginal Canadians experiencing homelessness, there is an urgent need for greater understanding of this important area (18–21,27). This study adopted the framework of the behavioral model for vulnerable populations (28) to investigate the role of racial-ethnic or cultural group membership in health care access and use in a large Canadian sample of homeless adults with mental illness in the context of other predisposing, need, and enabling factors.

Methods

Design

We used data from the At Home/Chez Soi Project (AH/CS), a two-year randomized trial evaluating the effectiveness of Housing First in five cities across Canada. Housing First is a consumer-driven intervention rooted in harm reduction principles that provides immediate access to housing and support services that is not contingent on acceptance of psychiatric treatment or abstinence from drugs or alcohol (29). A detailed description of the trial can be found elsewhere (30).

Study Population

Data for this analysis were obtained from the trial’s baseline interview, conducted between October 2009 and June 2011. Study participants were homeless adults with mental illness from Vancouver, Winnipeg, Toronto, Montreal, and Moncton who were recruited through referrals from community agencies, including shelters, drop-in centers, outreach teams, inpatient programs, and criminal justice programs. Eligibility criteria included the following: legal adult status (age ≥18 or ≥19 years, depending on the province), absolute homelessness (lacking a regular, fixed, physical shelter for more than seven days and having little likelihood of obtaining accommodations in the upcoming month) or precarious housing (having as a primary residence a single-room-occupancy hotel, rooming house, or hotel or motel and having had two or more episodes of being absolutely homeless in the past year), and presence of a mental disorder (bipolar disorder, depression, panic disorder, posttraumatic stress disorder, mood disorder with psychotic features, and psychosis) with or without a co-occurring substance use disorder as determined at study entry by DSM-IV criteria with the MINI International Neuropsychiatric Interview (MINI) (31). Participants were excluded from the study if they were currently served by an assertive community treatment or intensive case management program or did not have Canadian citizenship or legal immigrant or refugee status.

At study entry, participants were stratified into a high- or moderate-needs group (30). Briefly, high-needs participants had a score of ≤62 on the Multnomah Community Ability Scale (32) and a MINI diagnosis of current psychotic or bipolar disorder, in addition to one of the following: two or more hospitalizations for mental illness in any one year of the last five, comorbid substance use (based on the MINI), or a recent arrest or incarceration. All other participants were considered to have moderate needs.

All participants provided written informed consent and were compensated for their baseline interview. Ethical approval was obtained from the research ethics boards at each study site.

Theoretical Framework

Although racial-ethnic or cultural identity was our key variable of interest, we adopted the behavioral model for vulnerable populations as a framework for this study (28). According to the model, health behavior, including use of health services and barriers to care, can be explained in terms of predisposing factors (for example, demographic characteristics and health beliefs), enabling factors (for example, personal, family, and community resources), and need factors (for example, perceived and evaluated health status) (28,33,34). These factors have been shown to be associated with health care use among homeless persons (20,35–39).

Outcomes

Emergency department, medical, other clinical and social provider visits.

We extracted four self-reported service use outcomes from a questionnaire about health, social, and justice service use developed for AH/CS (30,40): the number of emergency department (ED) visits in the past six months and the number of visits to medical, other clinical (nonmedical), and social service professionals in the past month. Medical visits included those to family or specialist physicians (including psychiatrists). Other clinical visits included visits to an addictions counselor, case manager, chiropractor, Individual Placement and Support counselor, mental health worker, midwife, natural healer, nurse, nutritionist or dietician, occupational therapist, optometrist, pharmacist, physiotherapist, prosthetist, psychoeducator, psychologist, radiology technologist, clinical social worker, therapist, or counselor. Social service providers included child and family workers, community workers, conditional release officers, court workers, delegated curators or trustees, follow-up workers, housing workers, street and outreach workers, peer support workers, probation officers, public trustees, welfare workers, and work or vocational counselors.

Hospitalizations.

We derived the number of nights spent in a hospital in the past six months from a questionnaire about demographic characteristics and service and housing history developed for AH/CS (30), which asked participants to self-report the number of nights spent in institutions, including hospitals, in the six previous months.

Explanatory Variables

Racial-ethnic or cultural identity.

Participants reported their racial-ethnic or cultural group membership as part of the demographic questionnaire, modified from instruments previously used in the Canadian population (41,42). Participants who self-identified as Aboriginal were included in the Aboriginal group. White participants were those who self-identified as white or of a mixed ethnicity that did not include an ethnoracial minority group. Ethnoracial minority participants were grouped on the basis of the Statistics Canada definition of “visible minorities” (non-Caucasian in race or nonwhite in color and also not First Nations) (43) and included individuals from one of the following racial-ethnic or cultural groups: black African, black Canadian/American, black Caribbean, East Asian, Indian-Caribbean, Latin American, Middle Eastern, South Asian, and South East Asian. Participants who reported a mixed background that included at least one of the racial-ethnic groups listed above were also considered as belonging to ethnoracial minority groups. Because of the large number of participants who identified as black, a separate group (“black”) comprised individuals of black African, black Canadian/American, and black Caribbean identity, and all other participants from ethnoracial minority groups were placed into the group “other ethnoracial minority.”

Other predisposing factors.

Age, gender, length of lifetime homelessness, education, marital status, and arrests by police were collected from the demographic questionnaire (Table 1).

Need factors.

Need (high versus moderate) for mental health services was assessed at study entry by using the algorithm described above. Presence of an alcohol or drug use disorder (present versus absent) was established at study entry via the MINI (31). Chronic conditions were assessed by using a questionnaire on comorbid conditions developed for AH/CS, which elicited the presence of a comprehensive list of chronic general medical conditions (up to 31, depending on gender); chronic conditions were scored as less than 3 versus 3 or more. Finally, both mental and physical health status at study entry were assessed by using the two summary measures of the Short-Form 12 (SF-12) (44)—the physical component summary and mental component summary score—with norms based on the U.S. general population.

Enabling factors.

A short questionnaire on health service access, adapted for AH/CS from existing instruments (45), assessed access to health care in the past six months, including access to a family physician or general practitioner, access to a regular health care source, and unmet need for care.

Statistical Analysis

We examined differences in demographic, need, and enabling factors among the racial-ethnic and cultural groups by using chi square tests for categorical variables and one-way analysis of variance or the Kruskal-Wallis test for continuous variables, as appropriate.

Because all outcomes were overdispersed count data—that is, had variance greater than the mean—we employed a negative binomial regression model, which accounts for this characteristic. For each outcome, an initial unadjusted negative binomial model was fit to the data, with racial-ethnic or cultural identity as the sole predictor. The adjusted model included additional predisposing, need, and enabling factors (28). For each factor, we estimated the adjusted rate ratio and 95% confidence intervals to compare the rate of outcome occurrence during the specific time frame (for example, number of visits to a medical provider in the past month) in one factor category with the rate of outcome occurrence in the reference category (for example, participants with unmet needs versus those without unmet needs).

Statistical significance was set at <.05. All analyses were performed using SAS, version 9.4.

Results

Characteristics of the total study sample and of each racial-ethnic or cultural group are shown in Table 1. Nearly half (49%) of the study sample was white. In total, 397 participants (18%) were immigrants to Canada, but the proportion of immigrants was substantially higher among black participants (66%, N=162) and other ethnoracial participants (43%, N=169), compared with white (6%, N=64) or Aboriginal (≤5%, N=≤5) participants. All variables except for recent arrest, level of need for mental health services, and having a usual source of care showed statistically significant differences by racial-ethnic or cultural group membership.

Racial-Ethnic or Cultural Identity and Access to Care

The overall proportion of participants who had a usual source of care was surprisingly high (84%) and did not differ by racial-ethnic or cultural group. The proportion of participants with access to a family physician was highest among Aboriginal participants and lowest among white participants. Unmet need for health care was highest among Aboriginal participants and lowest among black participants (Table 1).

Racial-Ethnic or Cultural Identity and Health Service Use

Figure 1 shows the distribution of participants who had at least one instance of use of each health service, by racial-ethnic or cultural group. Only ED visits and visits to social service providers showed statistically significant differences by group membership in analyses using chi square tests (unadjusted model).

In adjusted negative binomial regression models (Table 2), racial-ethnic or cultural group membership was associated with the frequency of ED, medical, and social service visits. Compared with white participants, black participants had fewer ED visits, and Aboriginal participants had fewer medical visits. The rate of social service visits was higher among Aboriginal participants and those from other ethnoracial minority groups compared with white participants.

In addition to racial-ethnic or cultural identity, several other factors were associated with health care use. A higher number of ED visits was associated with three or more years of homelessness, a higher level of education, a recent arrest, high needs for mental health services, diagnosis of a substance use disorder, access to a family physician, and unmet need for care. Having a higher physical health component score, being male, and being older were associated with fewer ED visits. A higher number of medical visits was associated with a higher level of education, high needs for mental health services, a greater number of chronic conditions, and access to a family physician, whereas being single, being older, and having higher physical and mental health component scores were associated with fewer visits. A higher number of visits to social service providers was associated with having a recent arrest and having an unmet need for care.

Having high needs for mental health services was associated with more days in the hospital, whereas a higher number of nonmedical clinical visits was associated with having a substance use disorder, a greater number of chronic health conditions, a higher physical health component score, access to a family physician, and an unmet need for care.

Discussion

A key limitation of previous studies of racial-ethnic disparities in health care use and access is their limited generalizability to diverse homeless populations and health care contexts, with scant attention given to homeless people in Canada. Even though Canada has a universal health care system, which would be expected to minimize access barriers, we found persistent ethnoracial differences in access to and use of health care in a large Canadian sample of homeless adults with mental illness. Rates of reported unmet need for health care were highest among Aboriginal participants and lowest among black participants. Rates of health care use varied by service type. Compared with white participants, black participants had less use of ED visits, Aboriginal participants had a lower rate of medical visits and a higher rate of social service visits, and participants in the “other” group had a higher rate of social service visits.

Research has consistently demonstrated racial-ethnic disparities in health care use in the general population, particularly in the United States. However, the impact of race-ethnicity or cultural identity on health care use among homeless adults, who face additional health care needs and barriers, is less clear (34,46). The ACCESS study (Access to Community Care and Effective Services and Support) examined a large number of homeless adults with severe mental illness in 15 U.S. cities and reported that compared with white participants at study entry, black participants had fewer psychiatric outpatient visits and Latino participants had more case management visits (47). However, after the analysis adjusted for need and socioeconomic status, neither black nor Latino participants differed from white participants in the probability of accessing care. A large study of homeless women from Los Angeles County found the largest disparities in health care use (including hospitalization) among Latina women, compared with both white and black women (26). In unadjusted analyses, a recent smaller study of homeless adults from Toronto noted that the odds of at least one ED visit were lower among both black adults and those from other ethnic minority groups and higher among Aboriginal adults, compared with white adults (48).

We observed lower ED use among homeless black adults with mental illness than among their white counterparts. Our study used number of chronic health conditions as a proxy of overall health, and the proportion of participants reporting three or more chronic conditions was lowest among black participants (48%), indicating a substantially reduced burden of health problems. Black participants also had the lowest proportion of participants indicating an unmet need for care (32%) or a substance use disorder (53%). This group’s better health was further evidenced by higher scores on both the mental and physical component scores of the SF-12. Finally, two-thirds of black participants were immigrants to Canada, suggesting the so called “healthy immigrant effect,” a phenomenon previously noted among homeless immigrants (49).

As in previous studies, Aboriginal participants were more likely to have demographic or clinical characteristics associated with poorer health outcomes, including a diagnosis of a substance use disorder, more chronic health conditions, and a longer duration of homelessness (50). Aboriginal participants reported high rates of unmet need for care (52%). Although the proportion of participants who reported access to a regular medical doctor was highest in the Aboriginal group (68%), Aboriginal participants had lower rates of medical visits than white participants in adjusted analyses. Compared with their white counterparts, Aboriginal participants may prefer social sources and models of care over medical sources and models, a possibility supported by greater use (ARR=1.54) of social service providers among Aboriginal compared with white participants. The history of collective trauma, stigma, and discrimination experienced by Aboriginal peoples in Canada (50,51) likely plays a role in service use, particularly in the context of institutional settings. These associations should be explored in greater detail in future studies.

It is notable that rates of ED and medical visits decreased with older age in our sample, contrary to what is observed in the general population (52). However, homeless adults age at an accelerated rate; by middle-age many have accrued a number of chronic and “geriatric” conditions (including memory loss and falls) that are more common among those 15 to 20 years older in the general population (53,54). Analysis of 2005–2009 data from a nationally representative survey of hospital ED visits in the United States found that homeless adults age 50 and older had fewer ED discharges resulting from psychiatric or substance use disorders, compared with their younger counterparts; however, rates of diagnoses related to alcohol abuse were higher among the older participants (54). In our sample, ED visits were associated with substance use disorders, mental health symptom severity, and psychosis. Compared with those under 50, older participants had a lower risk of some key factors associated with ED visits in this population, including alcohol and drug misuse and greater psychiatric symptomatology (data not shown). Future studies should further explore age-associated patterns of service use among homeless adults from diverse racial-ethnic groups.

This study had several strengths, including a large well-characterized sample of homeless adults with mental illness. Nonetheless, several limitations should be noted. Health care use data were based on participant self-report and thus could have been influenced by recall error. The cross-sectional design did not allow us to draw causal inferences between predictors and service use outcomes. Study participants had to meet eligibility criteria for AH/CS project, and they may not be representative of the general homeless population in Canada. We did not explore how the interactions between the predictors examined may have moderated health care use, which may be a focus of future work. In particular, recent research has shown that gender moderates use of mental health services (54,55) and that a large number of ED visits are potentially preventable (55). Future analyses using administrative data sources will allow for a better understanding of the patterns of health care utilization in this sample, including the role of access to primary care in preventing ED use in homeless populations. Despite these limitations, our sample offers a unique opportunity to begin exploring disparities in health care use among disadvantaged individuals in the Canadian context of universal health insurance and to identify areas in need of attention.

Conclusions

Ethnoracial differences were found in both access to and use of medical and social services among homeless adults with mental illness in five Canadian cities. Such disparities in access to and use of services in a system of universal health insurance point to the need for further studies to identify the source of these disparities and opportunities to address them through appropriate services and supports.