Effectiveness of Collaborative Care for Depression in Public-Sector Primary Care Clinics Serving Latinos

Depressive disorders are common in primary care settings (1) and result in disability and reduced quality of life (2,3). Rates of depression and associated disability burden may be higher among Latinos and low-income groups, partly because of their lack of access to timely, high-quality treatment (4–6).

Collaborative care improves depression care and clinical outcomes and is cost-effective, compared with usual care (7). Primary care–based interventions are well suited to reducing ethnic disparities in depression care and outcomes because treatment occurs in settings where Latinos are most likely to obtain care (8). Prior trials of collaborative care have found that compared with whites, Latinos with health insurance had similar improvements in rates of high-quality care but even greater improvements in clinical outcomes (9). The resulting reduction in disparities was maintained for nine years after the intervention (10). Furthermore, low-income, predominantly uninsured or publicly insured, Spanish-speaking Latinos seen in primary care respond to telephone-based cognitive-behavioral therapy (CBT) (11,12). Low-income women seen in county entitlement programs (13) or in diabetes care settings (14) similarly respond to both psychotherapy and antidepressant medications.

In previous collaborative care trials, interventions were delivered by trained mental health professionals rather than by providers available in public-sector clinics. Latinos with low incomes frequently lack knowledge about depression and its treatment, have significant concerns about stigma, are more likely to prefer psychotherapy to medication, face practical barriers to care (such as lack of transportation or child care), and have low rates of adherence to treatments (13–15). Providers and administrators in safety-net clinics often have competing demands, few expendable resources, and limited incentives for providing depression care.

We tailored, implemented, and evaluated collaborative care for depression in three public-sector, primary care clinics serving primarily low-income Latinos. Our patient-centered treatment model accommodated patient treatment preferences, used providers without prior mental health training, and included significant outreach. We hypothesized that patients assigned to the collaborative care intervention would have greater improvements in depression, receive more high-quality care, and have greater satisfaction with emotional health care at a 16-week follow-up.

Methods

Patients, Providers, and Clinics Together (PACT) to Improve Depression Care was a randomized controlled trial of collaborative care for depression among primary care patients in three public-sector clinics in Los Angeles. The study was approved by institutional review boards at the University of Southern California, the University of Washington, and the University of California, Los Angeles. All participants gave verbal informed consent for study screening and written informed consent for participation in the randomized controlled trial and study assessments.

Study Sites

Three primary care clinics serving low-income, uninsured, or publicly insured patients from racial-ethnic minority groups were included. The first is a community-based, comprehensive health center run by the Los Angeles County Department of Health Services and staffed by full-time internists. The second is an ambulatory care teaching clinic for internal medicine residents located at a county hospital. The third is a teaching clinic for family medicine residents located at a private, not-for-profit hospital. Before the study, the first clinic employed a social worker who provided limited case management and the third clinic had a staff psychologist to provide instruction to residents.

Study Sample

Study participants were recruited between November 2005 and June 2007 through waiting room screening or provider referral. [A figure in an online supplement to this article presents recruitment details.] In waiting rooms, systematic random sampling was based on combinations of study participants, the enrollment week, day of the week (Monday through Friday), and time of day (morning or afternoon). Patients were approached on the basis of random-number tables and seating charts. We approached 4,418 persons in waiting rooms, of whom 3,317 were eligible for study screening because they were registered clinic patients, were age 18 years or older, and spoke English or Spanish. A total of 1,354 (41%) completed study screening after providing verbal informed consent. A total of 413 (31%) screened positive for current major depressive disorder or dysthymia, which was determined by using the PRIME-MD Patient Health Questionnaire–9 (PHQ-9) (16) and two questions from the PRIME-MD (17). Of the 413 participants, 83 were excluded because they screened positive for bipolar disorder (assessed with the Mood Disorder Questionnaire [18]), had current or lifetime psychotic symptoms or disorder (assessed by using three stem items from the Composite International Diagnostic Interview Version 2.1 [19] psychotic disorders module and a question from the IMPACT study [20] regarding a history of schizophrenia or schizoaffective disorder), had a cognitive impairment (assessed by using a six-item screener derived from the Mini-Mental State Examination [21]), or had acute suicidal ideation. After written informed consent was obtained, 224 (68%) of the remaining 330 eligible participants completed the baseline assessment and were randomly assigned to the study intervention or to a waitlist control.

We also offered to screen 306 clinic patients who were referred by primary care providers, 282 (92%) of whom agreed to be screened. Of these, 212 (75%) screened positive for depressive disorders, 21 of whom met exclusion criteria. Of the remaining 191 eligible participants, 176 (92%) completed the baseline assessment and were randomly assigned to the intervention or control group.

In total, 400 participants were randomly assigned, including 193 at the first clinic, 74 at the second clinic, and 133 at the third clinic. Randomization was stratified by referral source, clinic site, and gender. Enrolled study participants completed baseline assessments (95% [N=379] in person and 5% [N=21] by telephone) in English or Spanish (26% [N=105] in English and 74% [N=295] in Spanish). The response rate for 16-week follow-up assessments was 83% (N=331)—84% (N=171) for intervention participants and 82% (N=160) for enhanced usual care participants. Twenty-five percent (N=82) of follow-up assessments were completed in English and 75% (N=249) in Spanish.

Intervention Conditions

In the collaborative care arm, three bilingual, master’s-level social workers without previous mental health care experience functioned as depression care specialists. They received general reading materials and attended one week of in-person training in collaborative care and a 12-week, manualized CBT intervention (22). This culturally adapted version of CBT was available in Spanish and used simplified language and graphic depictions to help with low literacy. The depression care specialists subsequently had joint weekly psychotherapy supervision with two of the authors (ITL and JM). Each depression care specialist videotaped all therapy sessions with one or two patients. These videotapes were reviewed and then discussed during supervision, as were ongoing therapy issues and questions.

The depression care specialists also completed clinical assessments of patients randomly assigned to the intervention, educated patients about depression and its treatment—dispelling culturally based misconceptions as needed—and elicited treatment preferences for psychotherapy or antidepressant medication. Intervention patients were then allowed to choose to receive psychotherapy, antidepressant medication, or both. Depending on initial treatment response, treatments could be switched or augmented during the course of the study. For patients who preferred therapy, depression care specialists provided the 12-week CBT intervention (22). For patients who preferred medication treatment, depression care specialists communicated with primary care providers to facilitate initial prescriptions or changes in prescriptions. They also encouraged patients to adhere to medication, assessed side effects and treatment response, and communicated their findings to primary care providers.

All intervention patients were routinely assessed for depression symptom severity with the PHQ-9. During joint weekly caseload supervision with one author (ITL), new patients were discussed, PHQ-9 scores for all patients were reviewed, and if necessary, recommendations were made to adjust treatment by using a stepped care algorithm adapted from the IMPACT study (20). Depression care specialists also provided active outreach and limited case management. They placed reminder calls for study appointments, assisted with scheduling medical appointments, and made referrals to social service agencies if indicated.

In enhanced usual care, patients received an educational pamphlet about depression and its treatment, a letter that they could take to their primary care provider stating that they screened positive for depression, and a list of local mental health resources. They were free to engage in any mental health treatment. Following their 16-week assessments, patients initially assigned to the enhanced usual care intervention were offered the study intervention.

Study Measures

At baseline and 16 weeks, depression severity was assessed with the PHQ-9 (16), and mental- and physical-health–related quality of life was assessed with the Short Form-12 (23). In addition to using the PHQ-9 score as a continuous outcome measure, we calculated the proportion of participants with PHQ-9 scores <10 (indicating that they no longer met threshold criteria for major depression) (16) and the proportion of participants with a ≥50% reduction in scores (considered indicative of treatment response) (24).

Use of depression care services was measured with self-report items previously validated in the Partners in Care study (PIC) (25). Lifetime and four-month use of psychotherapy was determined with items assessing one or more visits to a psychiatrist, psychologist, social worker, or counselor. Lifetime and four-month use of medication for depression was determined with items assessing the use of any prescribed medication for personal, mental, or emotional problems, such as depression, anxiety, or nerves. Patients were asked to provide information regarding their medication (name, dose per pill, and number of pills per day) and medication adherence (number taken in past 30 days and number of months in which the medication was taken in the past four months). Minimally adequate psychotherapy was defined as four or more visits to a mental health specialist (25). Minimally adequate antidepressant treatment was defined as use of an antidepressant, given at the lowest therapeutic dose or higher, for at least 25 of the past 30 days or for at least two of the past four months (25).

Items adapted from the PIC study (25) were used to assess knowledge and attitudes about depression and its treatment, perceived stigma, and satisfaction with care. Knowledge and attitudes were assessed with ten items—four items related to general depression knowledge, four related to medication, and two related to psychotherapy. Scores for overall knowledge and attitudes and for each subtopic were the percentage of items answered correctly. Perceived stigma was measured with items assessing how difficult or embarrassing it would be if others knew about a depression diagnosis or visits to a psychiatrist. The proportion of participants who reported either a lot of or some difficulty or embarrassment was calculated. Overall satisfaction with health care and satisfaction with emotional care were similarly assessed with one item each. The proportion who responded that they were satisfied or very satisfied was calculated.

The baseline interview assessed age, gender, race-ethnicity, education, employment status, insurance status, language preference, and country of origin. The presence of comorbid medical illnesses was assessed with items from the PIC study (25), and the possible presence of a comorbid anxiety disorder (panic disorder, generalized anxiety disorder, or posttraumatic stress disorder) was assessed by using items from the PRIME-MD (17).

Data Analyses

Demographic and baseline clinical characteristics were computed. Baseline characteristics of the groups were compared by using t tests (numerical variables) and chi-square tests (categorical variables).

The impact of the intervention on clinical and process-of-care outcomes was assessed with intent-to-treat analyses for 16-week outcome measures. Analysis-of-covariance models were fitted for continuous outcomes, logistic regression models for dichotomous outcomes, and Poisson regression models for count variables. The independent variable was intervention status; covariates included baseline measures for the same outcomes plus study site. Sensitivity analyses were conducted by including additional covariates (age, gender, anxiety, and depressive disorder status at baseline), with no substantive changes in results. A sensitivity analysis for count variables based on generalized negative binomial regression models produced similar results. To show effect sizes, we present unadjusted means and proportions for each group. The results of linear regression analyses are presented as between-group differences, the results of binary outcomes are presented as odds ratios, and the results of Poisson regression analyses of count data are presented as rate ratios.

Nonresponse weighting (26,27) was used to account for missing data for the 17% of patients who did not complete 16-week follow-up assessments (16% [N=33] of the intervention group and 18% [N=36] of the enhanced usual care group]. The objective of nonresponse weighting is to extrapolate from the observed 16-week sample to the original intent-to-treat sample. Nonresponse weights were constructed by fitting logistic regression models to predict follow-up status from baseline sociodemographic and clinical characteristics. Separate models were fitted for the intervention and enhanced usual care groups. The reciprocal of the predicted follow-up probability was used as the nonresponse weight for each participant. Intent-to-treat analyses for intervention effects, weighted by nonresponse weights, were conducted with SUDAAN Version 11.1 (28). Weighted and unweighted analyses yielded very similar results.

Results

Sample Characteristics

As shown in Table 1, no significant differences were seen at baseline between the intervention and enhanced usual care groups. Most participants were female, and the average age was about 50. Most participants were from racial-ethnic minority groups (85% Latino and 4% non-Latino white), and two-thirds were predominantly Spanish speaking. More than three-quarters were born outside the United States. Educational attainment was quite low, with 48% reporting less than six years of formal education. Over a third were employed full- or part-time. More than half were uninsured. The sample reported a high illness burden from comorbid medical illnesses, depression, and anxiety disorders. Recent use of mental health services was low.

Clinical Outcomes

The intervention had an impact on clinical outcomes. As shown in Table 2, PHQ-9 scores were significantly lower at 16 weeks in the intervention group, compared with enhanced usual care (8.6 versus 13.3, p<.001). The proportion of participants with a PHQ-9 score <10 at follow-up was significantly higher in the intervention group (57% versus 26%, p<.001), as was the proportion experiencing a ≥50% improvement in depressive severity (56% versus 19%, p<.001). At follow-up, a significantly smaller proportion of patients in the intervention group had thoughts of death or suicide, as measured by the ninth item on the PHQ-9 (12% versus 21%, p=.025). The collaborative care intervention also led to significant improvements in overall health-related quality of life (p<.001), mental health–related quality of life (p<.001), and physical health–related quality of life (p=.016). Although the intervention did not significantly change knowledge, attitudes, and perceived stigma related to depression, it did produce significant improvements in satisfaction with overall health care (p=.001) and with emotional health care (p<.001).

Process-of-Care Outcomes

The intervention also had an effect on process-of-care outcomes. As shown in Table 3, the proportion of patients making at least one psychotherapy visit was significantly larger in the intervention group, compared with enhanced usual care (83% versus 8%, p<.001), as were the total number of psychotherapy visits (7.9 versus .2, p<.001) and the proportion of patients receiving minimally adequate psychotherapy (73% versus 4%, p<.001). The proportion of patients receiving any antidepressant was also higher in the intervention group (45% versus 26%, p<.001), as was the proportion receiving minimally adequate antidepressant treatment (37% versus 18%, p<.001). The intervention resulted in significant gains in quality of depression care; 77% of intervention participants received either minimally adequate psychotherapy or antidepressant treatment in the past four months, compared with 21% of enhanced usual care participants (p<.001).

As shown in Table 4, data from depression care specialists’ records for 196 patients randomly assigned to the intervention indicated that 83% attended the initial assessment session after significant outreach (an average of 6.1 telephone calls). The average subsequent treatment visits with the depression care specialist was 6.7, most for psychotherapy. The average number of primary care provider appointments for depression care was .5.

Discussion

The collaborative care intervention for depression that was tailored for clinics serving low-income, mostly Spanish-speaking patients from racial-ethnic minority groups and that used resources typically available to these clinics had a large impact on clinical outcomes and quality of depression care. Participants in the intervention had reduced depressive symptomatology, increased satisfaction with overall and emotional health care, and a much higher likelihood of receiving a minimum level of adequate depression care, compared with patients in enhanced usual care. Clinically meaningful findings were that intervention participants were four to five times as likely as those receiving enhanced usual care to experience a ≥50% improvement in symptoms or to fall to below-threshold levels for major depressive disorder.

The largest intervention effect on quality of care was an increase in psychotherapy visits. Low-income Latinos are more likely to prefer psychotherapy or combined care over medication alone (29–32). The intervention was available in Spanish. It provided patient education and active outreach, similar to other trials of collaborative care with primary care patients from racial-ethnic minority groups (13,33,34). In combination, these factors led to improved outcomes, and the intervention has the potential to decrease mental health care disparities.

Psychotherapy can be difficult to provide in public-sector, primary care settings, given the relative lack of availability of trained mental health professionals. Our study demonstrated that social workers without prior mental health experience can be trained to provide effective CBT. Our social workers participated in one week of psychotherapy training followed by joint weekly supervision with use of videotapes of patient sessions. They also received weekly caseload supervision. Social workers typically available in public-sector clinics may help extend mental health care to vulnerable and underserved populations. The social workers were extremely effective: 69% of patients who came to a first session completed treatment by attending at least four sessions.

Although this study provides strong evidence that care for depression can be substantially improved in public-sector clinics, several limitations should be noted. First, the patient sample consisted primarily of low-income, Spanish-speaking Latinos and may not be representative of patient populations in other public-sector settings. Furthermore, many potential participants refused either screening or involvement in the intervention, further limiting generalizability. Second, the study social workers were trained and supervised by well-established clinicians, who may be less readily available in public-sector settings. Third, the follow-up period was short, and most measures, although preestablished, were self-report. Finally, although the intervention was designed to be feasible for public-sector clinics, the use of study personnel rather than clinic providers limited intervention sustainability.

Conclusions

Despite study limitations, our findings confirm that collaborative care interventions that are culturally relevant, accommodate patient treatment preferences, use typically available providers, and include outreach can substantially reduce the burden of depression in public-sector settings. Our findings also underscore the importance of offering psychotherapy as a treatment and of training more readily available health care professionals to deliver patient-centered care. Collaborative care, now feasibly funded through the Affordable Care Act, could greatly reduce depression-related disparities.