1. An Investigation of Racial/Ethnic Disparities in Service Utilization Among Severely Mentally Ill Homeless Adults US
Marcela Horvitz-Lennon MD
Richard Frank PhD
Wesley Thompson PhD
Margarita Alegria PhD
Seo Hyon Baik
Sharon-Lise Normand PhD
Author affiliations: University of Pittsburgh Medical School (Dept. of Psychiatry), and Dept. of Statistics; Cambridge Health Alliance; and Harvard Medical School (Dept. of Health Care Policy) and Harvard School of Public Health (Dept. of Biostatistics). Supported by Grant P50 MH from the National Institute of Mental Health.

2. Severely mentally ill and homeless –can it get worse?

3. Talk Outline
Background
Study Aim
Methods
Main Results
Discussion

4. Background
Failure of research and policy efforts to eradicate homelessness among severely mentally ill people
Rates of severe mental illness among the homeless have increased in the past decades (North et al, 2004)
Racial & ethnic minorities over-represented among severely mentally ill homeless people (Burt et al, 2001)
Potential for sharp increase in the number of severely mentally ill and homeless people of Latino ethnicity

5. Background, II
Lower access and utilization of services for severely mentally ill minorities relative to non-Latino White persons (DHHS, 2001)
Evidence is larger and more consistent for Blacks
Less is known about service disparities among severely mentally ill people who are homeless
Most studies have controlled for race & ethnicity
Little evidence about utilization by non-Black minorities
Only some of those studies about naturalistic treatment

6. Study Aim
Investigate effect of race & ethnicity on utilization of services critical to treatment outcomes for severely mentally ill homeless adults
Cross-sectional design based on analyses of data collected at baseline by the Access to Community Care & Effective Services & Support (ACCESS) study (PI: R. Rosenheck)

7. Methods: Data Source
ACCESS: Quasi-experimental demonstration ( )
Study Aim: Evaluate outcomes of a systems integration intervention (Randolph et al, 1997)
18 sites (2 per state) intensively outreached severely mentally ill homeless adults not involved in treatment
Self-reported (SR) demographic, clinical, and treatment information collected at baseline, and 3 & 12 months
Our Data Source: Baseline data for Blacks, Latino, and non-Latino White subjects (N = 6,829)

8. Methods: Measuring Disparities
Heterogeneity in the definition & methods complicates interpretation of disparities findings
Conceptual framework proposed by the Institute of Medicine (IOM, 2002) provides a method to standardize assessment of disparities
Only adjustors are measures of need and preference
Socio-economic (SES), access, and geography are included in the disparity estimate

9. IOM’s Definition of Disparities
(Unequal Treatment: Confronting Racial & Ethnic Disparities in Health Care, 2002)
Need & Clinical
Appropriateness
Patient Preferences
Non-Minority
Operation of Healthcare
Systems and the Legal &
Regulatory Climate
Difference
Quality of Health Care
Minority
Disparity
Discrimination:
Prejudice, Stereotyping, &
Uncertainty
Populations with Equal Access to Health Care

10. Methods: Our Approach to Measuring Disparities
Outcome variables: SR utilization of psychiatric (outpatient), case management, and housing services (# visits in past 60 days)
Main explanatory variable: SR race/ethnicity
Adjustors: demographic and need (health, social)
We also adjusted for state because…
Treating geography as part of the disparity presumes that equal racial & ethnic geographic distribution is attainable
Adjusting by geography allows for targeted interventions
State was key aspect of ACCESS study design

11. Methods: Statistical Analyses
Sample divided into 2 cohorts: White & Black persons; White and Latino persons
Within each cohort: subjects stratified into comparable groups using propensity scores
Racial/Ethnic difference in mean self-reported service utilization in the previous 60 days computed as a function of adjustors (White minus minority)
Not adjusting for state [Method 1]
Adjusting for state [Method 2]: Main Analyses

12. Main Results

13. Table 1. Demographic & Need Variables
All
(n=6829)
Blacks
(n=3394)
Latinos
(n=381)
Whites
(n=3054)
P value
Age, mean (SD)
37.8 (9.50)
37.5 (8.7)
37.0 (9.5)
38.4 (10.2)
<.0001
Male (%)
62.3
60.5
71.4
63.2
Married (%)
5.4
4.7
7.9
5.8
0.014
Mental Health Need, mean (SD)
0.77 (0.66)
0.85 (0.69)
0.90 (0.70)
0.66 (0.61)
Subst. Use Dis. Diagnosis (%)
53.1
58.0
53.0
47.0
Medical Burden, mean (SD)
2.69 (2.69)
2.58 (2.57)
2.88 (2.98)
2.79 (2.77)
<.002
Chronic Homelessness (%)
53.7
57.1
53.5
50.0
Chronic Unemployment (%)
49.2
51.2
54.6
46.3

14. Table 1, Cont’d. SES and State
Variable
All
Blacks
Latinos
Whites
P value
Income ($), mean (SD)
392 (517.4)
393 (555.0)
337 (432.9)
397 (480.3) NS
Education (yrs), mean (SD)
11.6 (2.5)
11.5 (2.3)
9.6 (3.4)
11.9 (2.6)
<.0001
State, N (column %)
Connecticut
778 (11.4)
333 (9.8)
138 (36.2)
307 (10.1)
Pennsylvania
788 (11.5)
614 (18.1)
24 (6.3)
150 (4.9)
Virginia
755 (11.1)
501 (14.8)
15 (3.9)
239 (7.8)
North Carolina
759 (11.1)
465 (13.7)
9 (2.4)
285 (9.3)
Texas
757 (11.1)
257 (7.6)
57 (15.0)
443 (14.5)
Illinois
742 (10.9)
462 (13.6)
32 (8.4)
248 (8.1)
Missouri
751 (11.0)
377 (11.1)
25 (6.6)
349 (11.4)
Kansas
803 (11.8)
211 (6.2)
567 (18.6)
Washington
696 (10.2)
174 (5.1)
56 (14.7)
466 (15.3)

15. Table 2. Unadjusted Utilization
Mean (SD)
Service
Whites
Blacks
P value
Latinos
Psychiatric Outpatient
4.2
(9.35)
3.9
(8.43) NS
3.7 (9.40)
Case Management
2.5
(5.25)
3.1
(6.20)
<0.0001
3.9 (8.52)
Housing
0.8
(2.70)
(2.99)
0.8 (3.72)

16. Table 3. Adjusted Differences (Method 1)
Mean Difference (SE)
Reference
(Mean Overall Unadjusted Utilization)
Service
Whites minus Blacks
Whites minus Latinos
Psychiatric Outpatient
0.36
(0.22)
0.54
(0.51)
4.0
Case Management
- 0.55*
(0.14)
- 1.46*
(0.31)
2.5 – 3.9
Housing
- 0.04
(0.06)
0.07
(0.15)
0.8
* P value < 0.01

17. Table 4. Adjusted Differences (Method 2)
Mean Difference (SE)
Reference
(Mean Overall Unadjusted Utilization)
Service
Whites minus Blacks
Whites minus Latinos
Psychiatric Outpatient
3.2
(3.68)
2.8*
(0.08)
4.0
Case Management
2.3
(6.51)
0.5*
(0.14)
2.5 – 3.9
Housing
0.5
(4.24)
0.4*
(0.10)
0.8
* P value < 0.01

18. Discussion Low levels of utilization of 3 critical services
However, not in treatment was a study inclusion criterion
Main analyses showed that Latinos utilized fewer services than Whites of comparable need
Consistent with available evidence (Leda & Rosenheck, 1995; Lehman et al, 1999), no evidence of service disparities for Blacks
Possible explanation for Method 1 findings
Given unequal distribution of racial & ethnic groups, geographic disparities possibly obscured ethnic disparities

19. Discussion, II. Why the White-Latino Disparity?
Site as proxy for characteristics of local service system
Differences in quality of local providers? (Hasnain-Wynia, 2007)
Differences in linguistic accomodations & organizational cultural competency? (Betancourt et al, 2006)
Site strongest predictor of perceived service barriers among outreached ACCESS subjects (Rosenheck & Lam, 1997)
Latinos less receptive to outreach and treatment efforts?
Lowest rates of enrollment (Lam & Rosenheck, 1999)
However, opposite evidence also available (Shern et al, 1996)
Access differences?
Socio-economic differences??
Clinician-level factors?? (Balsa & McGuire, 2003)

20. Discussion, III. Limitations
Use of SR service utilization data
However, evidence exists that severely mentally ill persons provide reliable information (Goldberg et al, 2002)
Inability to elucidate role of language
Generalizability of findings
Others

22. Adjustment Variables Demographic Health Need Social Need
Age
Sex
Marital status
Health Need
Composite measure of mental health need: (1) Addiction Severity Index [ASI] -Drug Use; (2) ASI -Alcohol Use; (3) ASI -Psychiatric + Diagnostic Interview Schedule -Depression + Psychiatric Epidemiology Research Interview -Psychosis; (4) n of psychiatric diagnoses
Self-reported Medical burden
Clinician-formulated Substance use disorder (SUD) diagnosis
Social Need
chronic homelessness
chronic unemployment
Geography (state)

23. (2) Disparity Estimate: Combine stratum-specific estimated differences
(3) State-Adjusted Disparity Estimate: Combine state-specific estimates across states
ALL STATES
ALL STATES
STATE “I”
(2) State-Specific Disparity Estimate: Combine stratum-specific estimated differences within the state
Stratum 1 …
Stratum 10
Stratum 1 …
Stratum 10
(1) Stratum-Specific Disparity Estimate: Calculate difference in outcomes within each stratum
(1) Within-State, Stratum-Specific Disparity Estimate: Calculate difference in outcomes within each stratum
Method 1
Method 2