Research on racial and ethnic disparities in the United States has consistently found that Blacks have impeded access to behavioral health services, particularly compared with non-Hispanic Whites. Although the 2003 Institute of Medicine report “Unequal Treatment” (
1) provided a blueprint for potential change, the “2014 National Healthcare Quality and Disparities Report” noted that disparities in access to behavioral health care remain a decade later and have shown little change from 2008 to 2012 (
2). Racial disparities also extend to access to behavioral health received through primary care providers. For example, Blacks with serious mental health concerns are significantly less likely than Whites to receive a primary care mental health visit and a prescription for a mental health condition (
3). Racial disparities may also play a role in determining the pathways through which clients are referred to mental health services, particularly those involving criminal legal systems or emergency or inpatient treatment settings (
4–
6).
Relatively little is known about whether similar patterns are observed among clients accessing services for early psychosis. Coordinated specialty care (CSC) is an intervention framework specifically targeting the early stages of nonaffective psychosis (
7). CSC programs typically have eligibility criteria related to age, presenting symptoms, and duration of untreated psychosis; moreover, these programs otherwise are intended to serve individuals across a range of demographic and racial-ethnic backgrounds. However, client populations in CSC programs can become skewed. For example, if a program is situated in a community mental health center that accepts only certain reimbursement methods, such as Medicaid, individuals who are above the income threshold could be excluded (
8). Given that Blacks are more likely than non-Hispanic Whites to receive Medicaid (
9), one potential result is that Blacks might be seen in higher numbers in CSC programs.
However, funding through the Community Mental Health Block Grant (MHBG) program is one way that CSC programs can serve all young adults experiencing first-episode psychosis (FEP), regardless of insurance. Following the Consolidated Appropriations Act of 2014, the MHBG includes a 10% supplement and set-aside for FEP services. MHBG funding has stimulated the growth of CSC programs nationally (
10). Notably, as a grant, MHBG funding also provides a mechanism of support that can be used for reimbursement of services for clients who are not covered under Medicaid, including those with private insurance. MHBG funding therefore reduces the chance that CSC programs have a skewed racial distribution in any specific direction, and MHBG-funded programs should reflect the demographic characteristics of the broader communities they serve. If they do not, what are the possible reasons for this disparity, and is there cause for concern?
In this study, we examined the distribution of clients by race among a set of CSC programs that received MHBG funding. Whereas earlier studies have focused on the disproportionately high rates of schizophrenia diagnoses among Black clients compared with White clients (
11–
14), in this study, we asked whether Black clients are disproportionately represented within CSC programs relative to the racial composition of the population in the surrounding service area.
Methods
Design, Sampling, and Procedure
Data presented here came from the MHBG 10% Set Aside Study, a 3-year, mixed-methods evaluation funded by the Substance Abuse and Mental Health Services Administration and the National Institute of Mental Health. The MHBG 10% Set Aside Study included special focus on 36 study sites. This analysis included 35 of these sites; we excluded a program in Puerto Rico because of complexities in Census race categories among Puerto Ricans living on the island (
15).
Table 1 presents the basic characteristics of the CSC sites.
Measures
Percentage of Black clients.
Each study site reported race data for consecutively enrolled clients between January 2018 and July 2019 (N=772). The metric “percentage Black” was based on the total number of clients enrolled at each site as the denominator. The number of clients per site ranged from 6 to 66 (mean±SD=22.1±15.0), and the percentage of Black clients across the programs ranged from 0% to 100%.
Background characteristics of clients.
We included several client-level variables, which we further elaborate in the Discussion, to explore possible contributions to the issue of disproportionality: whether the client received Medicaid, whether the client spent time in an inpatient hospital during the 6 months before entry into the CSC program, whether the client spent time in the emergency department (ED) or a crisis unit, and whether the client was on probation or parole.
Percentage of service area that is Black.
Each team lead was asked to identify the distance from the clinic (in miles) where most of their clients lived, designated as the “primary service area” for the purpose of this study. Sites reported distances between 4 and 50 miles (mean=18.4). Note that this “service area” reflects the population of CSC clients, and it may or may not be the same as the service area for the agency as a whole; data on agency service areas were not available. Percentage Black for the primary service area across programs ranged from 0.3% to 48.4%. Using geographic information system software and processing tools, we identified a boundary with the clinic location at the center and a radius of the identified distance served. We then aligned 2017 Census American Community Survey (ACS) data on race with this area, matching all the tracts within each of the primary service areas with the age brackets of young adults attending each specific program. For example, if the client population was between ages 18 and 30 years, we used the data on race corresponding only to the Census tracts representing ages “18 and 19,” “20–24,” and “25–29.” We note that ACS estimates were based on a sample, rather than the whole population, so the estimates had a degree of uncertainty associated with them. We addressed this uncertainty by using sensitivity analysis, described in the following section.
Service area characteristics.
We extracted ACS data on other potentially relevant community-level characteristics for each site service area, including percentage of individuals receiving Medicaid, the percentage of Blacks in the service area who were insured, and urbanicity.
Contextual program factors.
We conducted semistructured interviews with CSC program teams, covering a range of topics. A systematic content analysis of these interviews was not part of this study, although in Discussion, we draw on relevant reflections from team members.
Analyses
We adapted well-established methods for assessment of disproportionality that are used in U.S. federal special education policy (
16) and that center on two metrics. The first metric was the “relative difference in composition,” indicating the relative difference between the percentage of a specific racial-ethnic group’s composition within a CSC program and the composition of that group within the service area. When the relative difference in composition score is 0 or close to 0, it indicates no or low disproportionality. If the relative difference equals 1, the composition of Black clients is twice as much as would be expected from the composition of the Black population in the service area; if it equals 2, then it is three times as much, and so on. The second metric, the risk ratio, was used to assess the risk for site enrollment for Black clients. The risk was expressed as the percentage of the Black population in the service area who participated at the site, and a risk ratio was the ratio between the risk for the Black population compared with the risk for the non-Black population. A risk ratio that is equal to or close to 1 indicates that the risk for site enrollment for Black and non-Black populations is close or equal, that is, no or low disproportionality. A risk ratio of 0 indicates that there are no Black clients in a site, and it indicates high disproportionality of the non-Black population (details on both metrics are available in an
online supplement to this article).
In addition, we conducted a sensitivity analysis to address two issues. First, there was sampling error—a statistical error that arises from taking a sample from a population—associated with the ACS estimates of the percentage of the Black population in the service area. To address this error, 30 replications of each set of relative-difference and risk-ratio statistics were calculated with methods presented in Krenzke and Li (
17). The replicated-estimates approach provides a way to account for the sampling error from the ACS in the analysis so that a measure of uncertainty in the analysis results can be estimated. The variation in the site estimates (i.e., relative difference and risk ratio) across replicated tables aligned with the estimated sampling variance associated with the site estimates. This provided evidence that our measure of uncertainty in the analysis results was accurate. Second, some of the 35 sites had a small number of participating clients. To address this limitation, we adjusted the results from the replicated tables to reflect the next person or persons to come into the site according to the following three scenarios: the next client is Black, the next five clients are Black, and the next client is non-Black. This strategy allowed us to evaluate how much the results would be affected by making a small change (i.e., adding one or five additional clients). We include details on sensitivity analysis in
Table 2 (relative difference in composition) and
Table 3 (risk ratio) as well as in the
online supplement.
To compare Black and non-Black participants receiving Medicaid and having past experiences with agencies, we used chi-square tests. All aspects of this study were reviewed and approved by the Westat Research Ethics Board.
Results
The percentage of clients covered through Medicaid ranged from 11% to 92% across the 35 CSC sites. There were no statistically significant associations (in chi-square tests of independence) of being Black or non-Black with Medicaid receipt (p=0.356), being on probation or parole (p=0.234), or having received services from an inpatient setting (p=0.111) or ED or crisis center (p=0.303) in the 6 months before admission to the program.
Overall, 25 (71%) of the 35 sites had a percentage of Black clients that was higher than the percentage of Blacks living within the service area.
Figure 1 displays the sites in descending order by site size (as number of clients served), along with the percentage of Black clients at each site and the percentage of the Black population in the service area. Sites of all sizes exhibited this pattern of disproportionality, and site size was not significantly associated with whether a site served a higher percentage of Black clients relative to the percentage of Black clients residing within the service area (site size did not statistically significantly correlate with disproportionality). Among the 10 sites with a low percentage of Black clients, seven (sites 6, 24, 26, 27, 31, 33, and 35) did not have any Black clients. All these sites had a service area with a Black population of <5% among the age-eligible population. Therefore, the absence of Black clients in these programs was consistent with the percentage of Black residents within the service area.
We assessed the degree of disproportionality in Black clients served at each site by calculating both a relative difference in racial composition between service site and service area and a risk ratio for Black clients at each site (
Figure 2). Most sites had some degree of disproportionality according to each of the two metrics used. For site 28, for example, the relative difference in composition was 9.1, and the risk ratio was 26.1. This disproportionality could be interpreted as the composition of Black clients at this site being 10 times higher than the composition of Black residents within the site’s service area, and the risk for the Black population to enroll in the service being 26 times higher than for the non-Black population.
Four sites (14, 18, 22, and 23) had potentially moderate to very large disproportionality, with a relative difference in composition >3 and a risk ratio >5. Sites 4 and 21 each had a risk ratio >5 and a relative difference in composition between 1 and 3. Six additional sites (5, 7, 9, 20, 25, and 30) had a disproportionality in which the relative difference in composition was between −1 and 1, and the risk ratio was between 0 and 2. Considering the error bounds from the sensitivity analysis, the disproportionality status of site 13 was inconclusive.
No statistically significant relationships were identified between service area urbanicity and either measure of racial disproportionality. No significant association was detected between the percentage of Medicaid recipients within a program and the degree of disproportionality. However, we noted significant associations between the percentage of individuals receiving Medicaid in the service area and disproportionality (for the relative difference in composition metric, ρ=−0.339, p=0.046; for the risk ratio metric, ρ=−0.330, p=0.053). A negative correlation means that a lower percentage of Medicaid recipients is associated with a higher degree of disproportionality.
Discussion
Using well-established metrics for assessing disproportionality in racial composition at service sites, in conjunction with adjustments to account for potential sources of error, we found that approximately 71% of CSC programs in this study served a disproportionately Black population relative to the surrounding service area. We discuss potential explanatory factors for this finding in the following.
Site Location and Population Served
In some cases, disproportionality could be an artifact of site location and population. Namely, a site may be in an area of the city that is largely non-Black, but it may still serve a predominantly Black population by drawing heavily from Medicaid clients. We did not, however, observe an association between percentage of Medicaid recipients within a program and degree of disproportionality.
Referral Sources
Young adults enter CSC programs through a variety of pathways, and some of these channels may have an aspect of racial bias. In Canada, for example, Black youths disproportionately enter the mental health system through emergency care systems and judicial avenues, compared with non-Black youths (
4). We did not find differences between Black and non-Black clients in service receipt in an inpatient setting, in treatment in an ED or crisis center, or in experience of a legal issue within the 6 months before the study. These results alone, however, do not rule out the possibility of differences in referral patterns because it is possible that a higher proportion of Black clients than non-Black clients experienced one of these situations and were referred to the CSC programs. For example, a difference in referral patterns by inpatient physicians might occur if attending physicians are more likely to encourage private care (including referral to the physician him- or herself) for patients who are not Black and to recommend referral to CSC for Black patients.
Family and Client Choice of Clinic
Depending on demand for services, more than one CSC program may serve individuals with FEP in a community. This scenario is most likely in urban areas where the population can support multiple programs and where there may be a university-based program as well as a program in a community mental health center. In these situations, family choice could contribute to disproportionality. For example, site 34 in this study had no White clients, because according to team members at this site, White families with private insurance chose the reportedly more prestigious university-based CSC program over their community-based clinic. As a counterexample, site 10 team members believed that the stigma of going to a community mental health center was a reason that White families did not choose their clinic; however, this site also did not display evidence of disproportionality. Among the seven sites with the highest disproportionality, three were in areas with another CSC program, and four were not. Given these findings, it is difficult to determine whether client and family preferences played a role in driving disproportionality in this study.
Outreach and Recruitment
A core element of the CSC model is to conduct outreach and activities in the community to raise awareness of FEP and to ensure that clients reach the clinic as early as possible (
7). Disproportionality could theoretically result from more assertive outreach in predominantly Black schools and neighborhoods within a service area. We did not have data on specific outreach activities among these 35 sites, although we did not hear about such activities during the site visit interviews. Given the high number of sites that displayed some degree of disproportionality, we do not believe that targeted outreach was a likely explanatory factor for disproportionality.
Clinical Implications and Future Directions
Across many fields, disproportionality in racial makeup at a CSC site is considered a negative phenomenon. For example, decades of data show that Black students are more likely to be identified for special education under the “subjective” classifications, such as emotional disturbance and intellectual disability (
18). After being placed in special education, Black students spend considerably more time in separate classrooms than do White students (
19), a placement that is considered less beneficial than placement in general education. In contrast, disproportionality in this study could be viewed positively. Namely, Black young adults seem to be finding their way to evidence-based, specialized programs that many consider the best option available for someone experiencing early symptoms of psychosis, provided that the diagnosis is accurate. The primary clinical implication, therefore, is that programs must ensure that they are particularly responsive and racially sensitive in their work with these clients, especially if they are not accustomed to working with Black families and clients. Moreover, clinicians and other team members should reflect the race of the population they serve.
Limitations
A primary limitation of this study was that we were unable to precisely compare the racial composition of clients in the CSC programs with that of the larger agency or clinic; moreover, at least in some sites, what appeared to be disproportionality in the program may have been comparable to the racial composition in other units or programs within the agency. We also could not distinguish between different subgroups of Blacks, such as African American, Black African, and Black Caribbean groups, even though such distinctions may be relevant for interpreting the findings (
20).
Conclusions
In this study, using data from across diverse CSC programs, all of which were supported through MHBG funds, we found that a disproportionately higher percentage of Blacks received services through many CSC programs. By using Census data that were matched by age to the actual population served, and by conducting sensitivity analyses to address potential limitations due to sampling error and low client counts, we strengthened the validity of these findings. We saw no evidence of a relationship between this disproportionality and factors such as receipt of Medicaid as well as recent experience with hospitalizations, crisis services, and probation. In the absence of clear explanatory factors for our findings, we suggest that this area may be ripe for discussion and further investigation.