There are long-standing disparities in treatment seeking for mental disorders among minority groups (1–11). Recent data suggest that the proportion of depressed adults who seek treatment is significantly lower among African Americans (53%) than among Caucasians (67%) (11). Likewise, in a study of individuals with depression, Alegráa and others (12) found that only 41% of African Americans and 36% of Hispanics sought treatment compared with 60% of Caucasians. Additionally, a majority of persons from minority groups who seek help for depression in the United States are seen in primary care settings (13,14), where their care is often inadequately managed (15,16).
Several barriers may contribute to existing disparities. Stigmatization continues to be one of the primary barriers facing minority populations who need depression care (17–20). Other barriers include mistrust of health care providers (21), negative beliefs and attitudes about treatments, and spiritual beliefs, all of which may explain why individuals from minority groups tend to delay treatment (22–26). In addition, many people who belong to minority groups have limited financial resources and poor access to affordable care (27). Travel barriers also represent substantial challenges to accessing depression care among individuals from minority groups (28–30). African Americans who suffer from depression are also more likely to be underdiagnosed (31) and experience poorer physician-patient communication than Caucasians (14,32–35).
In addition to having low rates of treatment seeking, African Americans are less likely than Caucasians to receive guideline-concordant depression care in primary care settings (14,32–35). Both Chermack and others (35) and Tiwari and others (36) found that relative to Caucasian veterans, African-American veterans were less likely to receive guideline-concordant depression care at Department of Veterans Affairs (VA) primary care clinics, despite facing relatively few financial barriers. Furthermore, African Americans are less likely to receive or adhere to pharmacotherapy (37,38), perhaps because the acceptability of antidepressants is lower among individuals from racial minority groups (39–41).
As a result, individuals from minority groups are also at risk for poorer clinical outcomes. The Sequenced Treatment Alternatives to Relieve Depression (STAR*D) clinical trial found that African Americans were twice as likely as Caucasians to experience worsening symptoms of depression during treatment with citalopram because of premature discontinuation of treatment, more taxing side effects, fewer clinic visits, and less acceptance of medications (42). Likewise, results from the Partners in Care study indicated that the proportion of individuals with probable depression after receiving usual-care treatment for depression for six months in a primary care setting was higher among African Americans (56%) and Hispanics (64%) than among Caucasians (41%) (43).
Efforts to improve the quality of depression care for minority populations have largely focused on collaborative care approaches. The Institute of Medicine's report on health disparities theorized that care management, by way of effective provider-patient communication and patient follow-up, should contribute to better health outcomes for minority groups (44). With respect to collaborative care for depression, promising results have been reported. Miranda and others (43) found that at the six-month follow-up, patients from minority groups in the Partners in Care study who received collaborative care had lower rates of probable depression compared with their counterparts who received usual care (24% and 56% for African-Americans and 47% and 64% for Hispanics, respectively). Miranda and others (42) also reported that patients from minority groups who received guideline-concordant, high-quality care demonstrated lower rates of probable depression at the six-month follow-up, compared with those who did not (21% and 71%, respectively).
Even with evidence that individuals from minority groups can make clinical gains when given high-quality depression care, few studies have examined potential factors contributing to these patients' improved responses to collaborative care interventions. This study evaluated racial-ethnic differences in clinical outcomes following a telemedicine-based collaborative care intervention among veterans from minority groups treated at VA community-based outpatient clinics in rural areas.
We tested several hypotheses. First, individuals from minority groups will have a worse treatment response to usual care than Caucasians. Second, individuals from minority groups will have a better treatment response to collaborative care than Caucasians. Third, minority group status will moderate the intervention's effect on treatment response, as will other characteristics of the veterans, such as having prior or current depression treatment, whether they perceived barriers to treatment, and attitudes toward antidepressant acceptability. Fourth, when the other characteristics are included as covariates, minority group status will no longer moderate the intervention effect. Our study extends the findings of Miranda and others (42,43) by evaluating a set of variable interaction terms to examine why members of racial-ethnic minority groups may respond at better rates than Caucasians to collaborative care to improve the quality of depression interventions.
Study setting and enrollment procedures
The Telemedicine Enhanced Antidepressant Management (TEAM) study was a randomized trial of telemedicine-based collaborative care tailored for small rural primary care practices, which demonstrated effectiveness (45,46). The study was conducted in seven community-based outpatient clinics that are satellite facilities of VA medical centers.
Administrative data were used to identify 24,882 patients due for annual depression screening, and 74% (N=18,306) were screened by phone by using the nine-item depression scale of the Patient Health Questionnaire (PHQ-9) (47). A total of 1,260 (7%) individuals screened positive for depression as defined by a PHQ-9 score ≥12. Exclusion criteria included having a diagnosis of schizophrenia; having current suicidal ideation; being recently bereaved; being pregnant; having a court-appointed guardian; having substance dependence, bipolar disorder, or cognitive impairment; or having a history of receiving or currently receiving specialty mental health treatment.
A total of 471 met the criteria and were eligible for baseline assessment, and 430 agreed to participate and were administered the baseline interview for baseline screening. Of the 430 patients who scheduled appointments, 395 (92%) attended the appointment and provided written consent. We enrolled 395 patients between April 2003 and September 2004. Patients were randomized to usual care (N=218) and to collaborative care (N=177). Data were collected via telephone interview. Follow-up telephone interviews were completed for 360 (91%) of the study participants at six months. The 35 patients who did not complete the six-month follow-up were not significantly different from those who completed the follow-up with respect to any baseline demographic or clinical characteristics.
Depression screening results were entered into the electronic health record for both intervention (N=177) and usual care (N=218) patients. Patients assigned randomly to the intervention received a stepped-care model of depression treatment for up to 12 months. Treatment intensity was increased for patients who did not to respond to lower levels of care by involving a greater number of intervention personnel with increasing mental health expertise.
The intervention involved on-site primary care providers located at the community-based outpatient clinics and an off-site depression care team, including nurses who managed the care of depressed patients (nurse care managers), clinical pharmacists, and psychiatrists, located at the parent VA medical center. The off-site intervention team focused on optimizing pharmacotherapy. A registered nurse serving as care manager used a scripted uniform protocol during telephone calls to patients to address specific treatment barriers, reasons for adherence problems (such as concern about addiction), and strategies for managing side effects. A pharmacist called patients who had not responded to treatment to discuss the patient's medication history and provide management of side effects. Psychiatrists supervised the off-site team and provided consultations via interactive video (for example, Skype).
At baseline, information about demographic characteristics (age, sex, race, income, education, and employment) was collected, and depression history (such as family history of depression, age of depression onset, and past and current depression treatment) were measured with the Depression Outcomes Module (48,49). Psychiatric comorbidity was measured with the Mini-International Neuropsychiatric Interview (MINI) (50,51). Social support was measured by using the Duke Social Support and Stress Scale (52,53). Acceptability of antidepressant treatment was measured by using an item developed for the Partners in Care study (54). The Depression Health Beliefs Inventory was used to measure perceptions about depression treatment, including barriers to, perceived need for, and effectiveness of treatment (55).
Health status was measured by using the physical health component score (PCS) and mental health component score (MCS) of the 12-Item Short Form Health Survey (SF-12v2) (56). Depression severity was measured at baseline and six months with the Hopkins Symptom Checklist (SCL-20) (57,58). Treatment response was measured dichotomously as a 50% improvement in depression severity between baseline and six-month follow-up.
Explanatory variables with missing values were imputed using the MI and MIANALYZE procedures in SAS, 9.2, which uses multiple imputation methods. Sampling and attrition weights were calculated from administrative and baseline data, respectively, to adjust for the potential bias associated with nonparticipation, loss to follow-up, or both.
To test the hypotheses that individuals in the racial minority group, relative to Caucasians, would have a poorer response to usual care and a better response to collaborative care, we conducted bivariate analysis after stratifying the sample according to intervention group and to determine differences in baseline sample characteristics between racial groups.
If both hypotheses were supported, it would suggest that the intervention was more effective for the racial minority group than for Caucasians. Accordingly, we hypothesized that minority group status would moderate the intervention's effect on response. We further hypothesized that prior depression treatment, current depression treatment, a perception of barriers to treatment (“perceived barriers”), and a belief that antidepressants are acceptable would moderate the effect of the intervention on treatment response and that minority group status would no longer be a significant predictor of treatment response after controlling for these variables. Because prior depression treatment (59), current depression treatment (60), perceived barriers (61,62), and a belief that antidepressants are acceptable (63) have been empirically associated with positive treatment response and were all correlated with minority group status at baseline, we considered them as potential moderators.
We tested a two-stage logistic regression model evaluating the possible moderation of minority group status and other potential moderators on treatment response. The moderation hypotheses were tested by using the analyses methods specified by Kraemer and colleagues (64), in which a moderator is defined as a baseline or prerandomization characteristic that can be shown to have an interactive effect with treatment on the outcome. Kraemer and Blasey (65) also suggested that all measured variables be centered. However, for this study, we chose not to center minority group status and intervention status because we did not want to estimate the overall effect of the interaction across both the minority and the Caucasian groups. Rather, we wanted to estimate the specific treatment effects separately for the minority group and Caucasians.
The first stage of the model included intervention status and minority group status as predictors of treatment response as main effects. An intervention by minority group status interaction term was also included in the model to evaluate whether minority group status was a significant moderator of the intervention effect on response. We controlled for case mix variables that were found to be significant at the p≤.2 level in bivariate analyses. They included age, male gender, married status, income <$20,000, prior or current depression treatment, perceived barriers, and attitudes about whether antidepressants are acceptable forms of treatment.
The second stage of the model included analysis of the interactions of prior depression treatment, current depression treatment, perceived barriers, and views about acceptability of antidepressants and intervention group in the regression model predicting response. We expected that if these variables were significant moderators of the intervention effect on treatment response, minority group status would no longer be considered a significant moderator.
The study was approved by the Research and Development Committees of the Central Arkansas Veterans Healthcare System in Little Rock, the Overton Brooks VA Medical Center in Shreveport, Louisiana, and the G. V. (Sonny) Montgomery VA Medical Center in Jackson, Mississippi, and their affiliated institutional review boards at the University of Arkansas Medical Sciences and the University of Louisiana Health Sciences Center at Shreveport.
Of the 360 patients, 272 (75%) were Caucasian and 88 (25%) were from a minority group, including 64 (18%) African Americans, 11 (3%) Native Americans, and 13 (3.6%) individuals from other groups. As shown in Table 1, those in the Caucasian group were more likely than those in the minority group to be male (94% and 87%, respectively; χ2=4.9, df=1, p=.03) and to be married (66% and 49%, respectively; χ2=8.1, df=1, p=.01). They were also more likely to report that antidepressants are an acceptable treatment (mean±SD ratings of 1.8±.8 and 2.2±.9, respectively; p=.001) and more likely to have had prior depression treatment (70% and 57%, respectively; χ2=4.9, df=1, p=.03) or current depression treatment (45% and 30%, respectively; χ2=6.3, df=1, p=.01). No other variables were found to be statistically different across the two groups.
Figure 1 shows response rates among minority and Caucasian patients to usual care and the intervention. Chi square analyses revealed no significant differences in the response rates among individuals of the minority racial group (8%) and among Caucasians (18%) in the usual care group, a finding that did not support our hypothesis that individuals of minority race would have worse outcomes associated with usual care. However, individuals in the minority group had a significantly higher rate of response than Caucasians to the intervention (42% and 19%, respectively; χ2=8.2, df=1, p=.004), a finding that supported our second hypothesis.
The first stage of the regression analysis is reported in Table 2. Model 1 revealed that neither the intervention nor minority group status predicted treatment response as main effects, an indication that the intervention was not effective for Caucasians. However, the interaction of minority group status and intervention significantly predicted treatment response (odds ratio [OR]=6.18, p=.009), an indication that the intervention was effective for the minority group. Therefore, we found support for our hypothesis that minority group status would moderate the intervention's effect on treatment response.
The second stage of the regression analysis, as shown in Table 2, model 2, revealed no correlation between response to the intervention and any of the other potential moderators, including prior depression treatment, current depression treatment, perceived barriers, and a belief in the acceptability of antidepressants as treatment. Therefore, we did not find support for the hypothesis that factors correlated with minority group status would also moderate the intervention effect and that minority group status would no longer moderate the intervention effect when those factors were included as covariates. Instead, we found that minority group status remained a significant moderator of the intervention's effect on treatment response (OR=6.02, p≤.01).
Although our findings are preliminary, the data supported our hypothesis that the minority group would respond better than Caucasians to the collaborative care intervention, as was demonstrated by Miranda and others (42) in the Partners in Care study. Our findings and the earlier study found that minority group status moderated the treatment effect of collaborative care, such that individuals in the minority group were significantly more likely than Caucasians to respond to the intervention. However, it is important to note that most of the gains made by minority groups in the Partners in Care intervention were largely attributed to the component related to quality improvement in psychotherapy, not medication (66). Collectively, these findings suggest that individuals from racial minority groups can make remarkable clinical gains in primary care settings offering collaborative care.
Contrary to our hypothesis, there was no significant difference between the groups in response to usual care, even though small differences in the groups' response rates consistent with those reported in the literature were found (42,66). Because of the relatively small number of minority participants in the usual care group and the low response rates for both groups, there was insufficient statistical power to detect small between-group differences.
We also evaluated other potential treatment moderators that correlated with or otherwise related to minority group status, including prior depression treatment, current depression treatment, perceived barriers, and a belief in antidepressants' acceptability. Contrary to our hypothesis, none of these factors significantly moderated the effectiveness of the intervention, and minority group status remained a significant moderator of the intervention's effect on response. We were surprised to find that none of these other moderators were significant, even though three of them—all but perceived barriers—were significantly correlated with minority group status. Failure to detect significant findings may be due to the fact that the measures we used to assess these constructs were imprecise. For example, our assessment of antidepressant acceptability included one question about treatment preferences, and we did not evaluate other factors that have been found to contribute to an unaccepting attitude toward medication among minority groups, such as tolerance of side effects (39,40,42).
Further, our measures of prior and current treatment were dichotomous and may not have captured the extent to which patients of minority race or ethnic group were engaged in or were adherent to treatment in the past. Last, it is possible that unmeasured constructs associated with race and ethnicity may also explain why minority group status was a significant treatment moderator.
Our main finding that the collaborative care intervention had a greater effect among the minority group than among Caucasians may be explained by prior experiences with antidepressant treatment. A majority of patients in our sample had received prior depression treatment or were receiving depression treatment at baseline. Yet they remained symptomatic enough to meet study eligibility criteria (PHQ-9≥12). Receipt of prior or current depression treatment varied significantly by racial group, but neither prior nor current depression treatment moderated the intervention effect. Therefore, racial differences in the receipt of any prior depression treatment alone cannot explain our findings.
However, we suggest that racial difference in the quality of prior depression treatment may explain our findings. Previous research clearly indicates that veterans in minority groups are less likely to receive adequate depression care (35,36). We speculate that the prior care afforded patients in minority groups in our study may also have been inadequate; once they were enrolled in an intervention that provided high-quality depression care, they responded. On the other hand, the response rates among Caucasian patients to usual care may have reflected true treatment resistance, given that they were nearly identical to the group's response rate to the intervention.
If true, the improvement in quality of pharmacotherapy treatment resulting from the collaborative care intervention would be expected to improve outcomes for minority groups but not for Caucasians. Patients with treatment-resistant depression will likely need supplemental specialty treatments, such as evidence-based psychotherapies, to achieve remission of symptoms. We must also note that many of the 475 patients who were excluded at stage 1 were already in treatment and disproportionately included Caucasians, which might have affected the differential positive outcome for minority patients.
We also speculate that specific aspects of the intervention may have contributed to the success of the intervention in the minority group. As shown in the STAR*D studies, proper prescribing addresses only some of the barriers to receiving adequate care, whereas collaborative care models may address more multifaceted barriers faced by minority groups and lead to a reduction of depression symptoms. More study should be given to specific elements of collaborative care, such as education, activation, and side-effect monitoring, that may improve depression outcomes for minority groups.
Limitations of our study include the possibility that our results cannot be generalized to other populations, such as nonveterans or female veterans. However, in contrast to other depression studies, which include mostly women (14,32–35), our study of mostly men confirms that interventions to improve quality of depression treatment may be as effective in closing racial gaps among men as they are among women.
Also, combining African Americans and members of other minorities into one group because of inadequate sample sizes may have oversimplified our findings.
Overall, our findings indicate that racial-ethnic disparities in depression care may be addressed successfully by implementing collaborative care programs in primary care settings and lend support for the use of these programs in an array of populations and health care systems. Most notably, we replicated findings of Partners in Care with a different population and in a different setting. We hope that our findings will promote research into other potential factors that may explain the enhanced treatment responses we found among veterans from minority populations.
This research was supported by grant VA IIR 00-078-3 to Dr. Fortney and grant VA NPI-01-006-1 to Jeffrey Pyne, M.D., of the VA Health Services Research and Development Center for Mental Health and Outcomes Research and the VA South Central Mental Illness Research Education and Clinical Center.
The authors report no competing interests