The MDDP randomized trial, approved by the University of Southern California Institutional Review Board, was conducted in Los Angeles County Department of Health Services community clinics. Patients with diabetes, identified from medical charts, provided verbal consent to depression screening by bilingual study recruiters between August 2005 and August 2007; 387 study participants provided written informed study consent and completed a structured baseline interview prior to computer-driven randomization to MDDP collaborative care (INT) or modestly enhanced usual care (EUC). Eligible patients were ≥18 years, English or Spanish speaking, and endorsed one of the 2 cardinal depression symptoms for more than half the days to nearly every day over the last two weeks and scored ≥10 on the PHQ-9 indicating a high likelihood of clinically significant depression. Patients with acute suicidal ideation, excessive current alcohol use, or recent lithium/antipsychotic medication use were ineligible for the trial.
MDDP is based on evidence-based practice guidelines for primary care and is designed to reduce known patient and provider barriers to depression care in the study community public safety net clinics.19–21
The 12-month, team delivered, intervention (INT) included: 1) a choice of starting treatment with either antidepressant medications (AMs) prescribed by the treating primary care physicians (PCPs) (who provided care to both study groups and attended a psychiatrist didactic on the algorithm and depression treatment) or Problem-Solving Therapy (PST) (if initially preferred by patients over AM) provided by bilingual graduate social work diabetes depression clinical specialists (DDCS); 2) application of a stepped care treatment algorithm; 3) DDCS monthly telephone follow-up on symptom monitoring, treatment maintenance, relapse prevention, behavioral activation, and additional booster in-person or telephone PST sessions if indicated; 4) care and service system navigation by the DDCS and/or an assistant patient navigator. A study psychiatrist and the study principal investigator provided weekly telephone DDCS supervision and, if requested, the psychiatrist provided PCP antidepressant medication telephone consultation. During the intervention year, 84% of INT patients received depression treatment (49 PST, 9 AM, 104 both) and 35 (23%) PST-graduated patients also attended ≥1 optional open-ended PST monthly support group sessions. Over 12 months, a dosage increase or change in type of medication was provided to 63 (56%) patients prescribed AM. During the intervention year, 63% of INT patients with baseline moderately severe to severe depression symptoms (i.e., PHQ-9 score ≥15) and 44% of INT with moderate depression symptoms (i.e., PHQ-9 score 10–14) received both PST and AM (chi-sq=6.94, df=1, p=0.01).
EUC patients were informed of their depression and were given patient and family focused depression educational pamphlets (Spanish or English) and a community mental health, financial, social services, transportation, and child care resource list. EUC PCPs were informed of patient PHQ-9 scores and study participation and could prescribe antidepressant medications, provide counseling or refer patients to community mental health care. During the initial year, 32.5% of EUC patients received depression treatment (37 AM, 11 self-reported counseling, 15 both), while 10 (19%) patients prescribed AM had a dosage increment or change in antidepressant.
Of 387 patients enrolled at baseline, 24 month attrition totaled 123 (31.8%). Of these, 48 (12.4%) declined continuing study participation, 29 (7.5%) were no longer receiving care at the clinic, 9 (2.3%) no longer living in the US, 3 (<1%) were deceased, and 34 (8.8%) could not be located. Although average HbA1C values did not differ between attrition and retention groups (9.38% vs. 8.73%, P=0.20), relatively more attrition patients had an HbA1C greater than 7% compared to retention patients (56.9% vs. 43.1%, P=0.02). All other baseline depression, clinical and demographic characteristics were similar between attrition and retention groups. In this report, major depressive disorder is defined as: 1) at least one of the 2 cardinal depression symptoms more than half the days to nearly every day over the past 2 weeks, and 2) having PHQ-9 scores ≥10. Symptom deterioration over a one year post-intervention period was defined as meeting major depression criteria at 18 or 24 months among those who did not meet PHQ-9 criteria for major depression at 12-month follow-up. Standard definitions have defined relapse as occurring prior to 6 months after reaching symptom remission and recurrence as occurring ≥6 months after symptom remission. However in most depression treatment trials only a minority of patients meet remission criteria at study completion, with the majority having residual symptoms. Thus we use the term “symptom deterioration” in this article.
Of 193 patients in the study, 158 (82%) had both 18- and 24-month follow-up data, 24 (12%) had only 18-month, and 11 (6%) had only 24-month follow-up. If participants completed only one follow-up interview post 12-month, the available follow-up data were applied.
We examined at three predictors of interest, i.e., baseline demographics and measures of clinical severity, 12-month characteristics, and worsening clinical status between baseline and 12 month follow-up. Demographic data on age, gender, marital status, ethnic group, birthplace, time living in the US, language, education, and employment were collected at baseline interview. Follow-up telephone outcome interviews at 12, 18, 24 months were conducted by bilingual interviewers blinded to treatment assignment. The 20-item Symptom Checklist (SCL) depression scale24
was used as a reliable and valid measure of depression in medical populations that has been shown to be sensitive to change in primary care studies (Cronbach α =0.91). The Patient Health Questionnaire-9 (PHQ-9) was used because it provides both a dichotomous diagnosis of major depression,25
a continuous severity score and has been found to have 73% sensitivity and 98% specificity to a diagnosis of probable major depressive disorder (MDD) based on structured psychiatric interview.26
The Brief Symptom Inventory (BSI)27
assessed anxiety; the Sheehan Disability Scale (SDS) assessed functional impairment.28
Health-related quality of life (QOL) was assessed using the MOS Short-Form Health Survey (SF-12).29
Chronic pain was defined as pain present most of the time for 6 months or more during the past year. The Summary of Diabetes Self-Care Activities Questionnaire assessed self-reported adherence;30
the Whitty 9-item questionnaire assessed diabetes symptoms.31
Glycated hemoglobin (HbA1c
) was obtained from medical records (the last test done within 3 months of each follow up interview). Patient self-report of socioeconomic stressors (financial, work, family, legal problems, and community violence worry) were also obtained during each outcome assessment. The dichotomized clinical status changes, worsening versus not worsening, were identified by comparing the clinical outcomes at 12 month follow-up to their baseline values, such as decreasing functional status or increasing diabetes symptoms indicating a worsening status. International Classification of Diseases (ICD-9) codes between 6 months prior to baseline to 12 months post baseline were obtained from electronic medical records in order to identify new medical diagnoses over initial 12 months.
Because the intervention ended at 12 months, this Year 2 sample is no longer an intent-to-treat population. We conducted bivariate analyses to compare baseline demographic characteristics, 12-month clinical characteristics, and changes in health status in the initial 12 months such as increasing diabetes symptoms or new ICD-9 diagnoses (from baseline to 12 month follow-up) between patients with and without depressive symptom deterioration with t-test for continuous data and chi-square test for dichotomous data. Variables that were significantly associated with symptom deterioration were jointly entered into a logistic regression model with stepwise selection to identify predictors. To assess whether predictors were affected by randomization status, we also examined randomization group and the interactions between randomization group and each predictor. Randomization group was not a significant covariate, and there were no significant interactions between randomization group and these predictors. Therefore we combined data of subjects from both intervention and control groups for the analysis. All analyses were conducted at 0.05 significance level (2-tailed) using SAS version 9.1 (SAS Institute Inc, Cary, NC).