Upon completion of this activity, participants will be able to:

EDITOR

Camille Martin, Technical Writer/Editor, Preventing Chronic Disease. Disclosure: Camille Martin has disclosed no relevant financial relationships.

CME AUTHOR

Charles P. Vega, MD, Clinical Professor of Family Medicine, University of California, Irvine. Disclosure: Charles P. Vega, MD, has disclosed the following relevant financial relationships: Served as an advisor or consultant for: McNeil Pharmaceuticals.

AUTHORS AND CREDENTIALS

Disclosures: Brian Wu, BS; Haomiao Jin, MS; Irene Vidyanti, MEng; Pey-Jiuan Lee, MS; Kathleen Ell, DSW; Shinyi Wu, PhD have disclosed no relevant financial relationships.

Affiliations: Brian Wu, Haomiao Jin, Irene Vidyanti, Pey-Jiuan Lee, Kathleen Ell, University of Southern California, Los Angeles, California; Shinyi Wu, RAND Corporation, Santa Monica, California.

The DCAT team conducted a quasi-experimental trial that examined the effects of implementing depression monitoring in a diabetes disease management program for low-income urban populations in the Los Angeles County Department of Health Services (DHS) Ambulatory Care Network, the second-largest safety-net care system in the United States. Before DCAT, DHS had a diabetes disease management program with nurse-driven and physician-supervised care management for high-risk or high-service-use patients. DHS applies evidence-based diabetes care management components and uses structured tools (case management, patient education and self-management support, care coordination, depression screening and physician notification, an electronic disease registry, and integrated clinical decision support systems) to deliver more than 80% of the care by nurses under protocol and was the model for the supportive care group. These tools support clinical assessment and decisions in a limited care-management period of 6 months. The integration of team staff, including physicians, nurse practitioners, nurses, and social workers, provided an intensive care model with strict guidelines for follow-up and monitoring of diabetes symptoms and comorbid risks, such as depression. Participants received weekly telephone calls from care team members and were seen by nurses and social workers who provided comprehensive team-based care to improve disease management and quality of care.

During implementation of DCAT, from October 2011 to May 2013, diabetes disease management was supplemented with periodic screening and monitoring of depression symptoms with the Patient Health Questionnaire 9-item scale (PHQ-9), a standard tool in each clinic’s disease registry, and the DHS depression care protocol and treatment guideline. The program also designated a social worker to provide problem-solving therapy, an evidence-based treatment of depression. All care providers were offered training in problem-solving therapy via a 1-day workshop; they were also trained in the collaborative depression care model and adaptive treatment approach via 1 of 3 webinars.

This study involved 5 DHS primary care clinics, selected by DHS leaders on the basis of criteria that reflected geographic and diabetes care model diversity. The usual care group included 2 community clinics that represented standard clinical practice, in which primary medical providers and their staff translated and adopted evidence-based depression care. The supportive care study group included 2 care teams from the DHS diabetes disease management program. These teams practiced in 2 community clinics and 1 hospital-based outpatient clinic. We hypothesized that at 6 months after enrollment in the study, patients who received the diabetes supportive care would have improved depression and diabetes outcomes compared with patients receiving usual care.

Patients were recruited from 5 DHS primary care clinics. The patients were predominantly low-income, low-literacy, middle-aged, Spanish-speaking Hispanic or Latino women who had been diagnosed with diabetes for more than 5 years. Approximately one-third of participants were depressed, and approximately one-third of the patients were men.

Patients were eligible for the study if they were aged 18 years or older with type 2 diabetes, had a working telephone number, spoke English or Spanish, and read and understood the consent form. Patients were ineligible for the trial if they presented with baseline acute suicidal ideation (as measured by PHQ-9, item 9), cognitive impairment (Short Portable Mental Status Questionnaire scores less than 5) (18), alcohol abuse (2 or more CAGE items from a quantity–frequency index, and patient perceptions of substance use) (19), or if they had recently used lithium or antipsychotic medication. Patients were not required to have depression to be eligible for the study because DCAT addressed the elevated risk of depression among patients with diabetes by testing a care approach that incorporates depression screening, symptom monitoring, and treatment follow-up for diabetes patients.

Approval was obtained from the University of Southern California and the Los Angeles Biomedical Research Institute human subjects review boards. The enrollment period was from April 2011 to May 2012 in the 5 study clinics. Patients with type 2 diabetes were identified for recruitment from database and clinic records. Patients provided verbal consent during study eligibility screening to bilingual research assistants. Of the 1,704 patients screened, 1,066 (63%) were women and 638 (37%) were men. Men had a significantly lower enrollment rate than women (83% vs 88%, respectively; P = .003), which was associated with poor alcohol use scores (5% vs 1%, respectively). A total of 964 diabetes patients (86% of patients screened) provided written informed consent and completed a structured baseline interview that included both PHQ-9 (scores range from 0–27, where higher scores indicate worse depression) and Hopkins Symptom Checklist-20 depression symptom assessments (individual items scored 0–4, where higher scores indicate worse depression) (Figure). After excluding 87 participants with missing data, the baseline sample had 877 patients: 416 patients in the usual care group and 461 patients in the supportive care group. Patient and family educational materials concerning depression, including a comic book fotonovela (20) designed for patients and family members with low health literacy, were provided to all study patients in Spanish or English by bilingual study recruiters. The study clinic physicians were notified of the baseline depression screening results for patients whose PHQ-9 scores were 10 or higher or who exhibited suicidal ideation (score greater than 1 for item 9 of PHQ-9). Participants received no monetary incentive for the study enrollment and baseline assessment. However, they did receive a $10 gift card for each follow-up assessment they completed. The study clinics participated in the study pro bono.

All subjects received comprehensive assessments at baseline and at 6, 12, and 18 months by independent English–Spanish bilingual interviewers. Primary outcomes included 5 depression outcomes and 7 diabetes care measures (including satisfaction with care and disability reduction) (Table 1).

The target sample size was based on power analysis for 2 primary outcomes: reduction of prevalence of major depression (PHQ-9 score ≥10) and depression remission (PHQ-9 score ≤8 with a reduction ≥50% for patients with major depressive disorder at baseline). Power analyses were conducted using nQuery (Statistical Solutions) to estimate effect sizes of the treatment with preintervention and postintervention comparisons and longitudinal statistical approaches for repeated measures comparing the trend of depression-related outcomes in the DCAT study. The calculations assumed an α level of .05 and a power of .80. With the assumption that attrition rates would be less than 20% for patients at each 6-month follow-up assessment — up to 18 months for preintervention and postintervention comparisons — a sample size of 51 patients with depression in each study group would allow the detection of a small effect size of less than .01. A previous trial in 2008 of the Multifaceted Depression and Diabetes Program established that 25% to 30% of diabetes patients also experience depression (21); for DCAT, we added a 25% cushion for differences in baseline characteristics because of the quasi-experimental trial design. Therefore, DCAT required a sample size of approximately 500 type 2 diabetes patients in each study group.

Initial statistical tests were performed to assess differences in differences (DID). However, because the study design defined individual treatment centers as separate arms, we aimed to improve statistical testing by calculating propensity scores to interpret the probability of treatment assignment conditional on observed baseline characteristics. Both tests were performed with assumptions of P ≤ .05. A multinomial logistic regression model was used to estimate the propensity score; the model used study group as the dependent variable and all 27 measured baseline characteristics as the independent variables. The baseline characteristics were 1) age, 2) sex, 3) preferred language, 4) body mass index, 5) education level, 6) employment status, 7) economic status, 8) total stressor number, 9) sum of the stress level, 10) predicted future health cost, 11) age at onset of diabetes, 12) insulin use, 13) Medical Outcomes Study Short Form-12 Health Survey (SF-12) physical score (scored 0–100, where higher scores indicate a higher level of physical health), 14) SF-12 mental score (scored 0–100, where higher scores indicate a higher level of mental health), 15) number of diabetes complications, 16) Whitty-9 diabetes symptoms scale score (scored 1–5, where higher scores indicate more severe diabetes), 17) diabetes emotional burden, 18) diabetes regimen distress, 19) mean Toolbert diabetes self-care score (scored 0–7, where higher scores indicate better diabetes self-care), 20) PHQ-9 score (scored 0–27, where higher scores indicate worse depression), 21) Brief Symptom Inventory total score (scored 0–24, where higher scores indicate worse anxiety) , 22) mean Sheehan Disability Scale score (scored 0–10, where higher scores indicate more significant functional impairment), 23) dysthymia, 24) previous diagnosis of major depressive disorder, 25) chronic pain, 26) overall patient satisfaction, and 27) glycated hemoglobin (A1c) value. We subsequently checked the distribution of the estimated propensity scores because between-group comparisons would be suspect if there had been substantial separation between treatment arms; the propensity score method is a more conservative and parsimonious statistical modeling method for nonrandomized study samples (22).

Comparative treatment effects were estimated using linear or logistic regression models featuring outcomes at 6 months as the dependent variable; the independent variables were study group, care team, outcome variable at baseline, estimated propensity scores, insulin use, A1c, age, sex, and preferred language. Regression that includes estimated propensity scores as covariates is an effective tool to adjust sample biases in observational or quasi-experimental studies (23). The coefficients of study groups predicted comparative treatment effects. Three care team variables were used to adjust for differences among providers. Analyses were performed by using SAS version 9.1 (SAS Institute, Inc).

The DCAT study enrolled 964 low-income, predominantly Hispanic or Latino patients with diabetes to test and compare the translational models of depression care management. Of these patients, 484 were in the usual care group and 480 in the supportive care group.

Because DCAT used a quasi-experimental design comparing study groups, we first examined whether major baseline characteristics that could influence the outcome measures were balanced between the study groups. There were no significant differences in baseline PHQ-9 depression or SF-12 mental scores, Sheehan Disability Scale ratings, or body mass index in pairwise comparisons between groups (Table 2).

The DID test with the main outcome variable of PHQ-9 change score at 6 months was significant (P = .01). When adjusted for the same covariates, DID and propensity score methods had consistent results (P = .01 and P = .05, respectively).

Compared with usual care, supportive care significantly decreased PHQ-9 scores (least squares mean [LSM] = 6.34, standard error [SE] = 0.49 vs LSM = 5.08, SE = 0.48, respectively; P = .047), reduced the number of patients with moderate or severe depression (a PHQ-9 score ≥10; P = .04), and improved depression remission (P = .05) (Table 3). Supportive care also significantly improved patient satisfaction with care for emotional problems (P = .01). Scores on the Sheehan Disability Scale (P = .03) were significantly lower in the supportive care group compared with the usual care group. There were no significant differences between the 2 groups in terms of satisfaction with care, emotional distress among patients with a baseline PHQ-9 score of 10 or more, cholesterol, diabetes self-care, exercise, satisfaction with diabetes care, or A1c levels.