Organized Self-Management Support Services for Chronic Depressive Symptoms: A Randomized Controlled Trial

Between January 2010 and October 2011, participants were recruited from four primary care clinics of Group Health (GH) Cooperative and from the Swedish Family Medicine Residency Clinic at Cherry Hill. Both GH and Swedish are nonprofit health care organizations in the Seattle, Washington, area. GH clinics provide coverage and care to approximately 600,000 Washington residents; the Cherry Hill clinic provides primary care to an economically disadvantaged population numbering about 20,000 per year.

Potential participants were identified through population-based screening and provider referral. Electronic medical records data from the GH and Cherry Hill clinics were used to identify potentially eligible patients. Identified patients were at least 18 years old, had at least two visits for depression associated with antidepressant treatment in the past year, had no diagnosis of bipolar or psychotic disorder in the past two years, and were not currently participating in other depression-related studies. Providers at participating GH clinics and Cherry Hill were also notified about the study, and patients at both organizations were allowed to self-refer to the study. Brochures and flyers were designed to help providers at Cherry Hill inform other potentially eligible patients (missed by electronic identification) about the study and how to self-refer.

Study staff at each site mailed invitation letters to potentially eligible patients that included a $2 bill as a “preincentive” and an option to decline further contact. Staff followed up with a phone call to assess interest in participation. GH study staff were responsible for conducting eligibility assessment by phone with all patients expressing interest in participation. Staff at Cherry Hill asked interested patients for verbal permission to release their contact information to the GH study staff.

Inclusion criteria were assessed by telephone screening. Eligibility requirements included significant residual symptoms (score of ≥10 on the Patient Health Questionnaire–9 [PHQ-9]) (38,39), history of recurrent major depression (three or more episodes in the past five years) or dysthymia, and willingness to attend groups. Exclusion criteria included history of mania or hypomania, cognitive impairment, metastatic cancer, chronic renal failure, or having a plan to move out of state in the next 18 months. Co-occurring alcohol or drug use, personality, or anxiety disorders were not exclusion criteria.

Following telephone screening, participants were asked to complete an in-person baseline assessment. Research study staff obtained written informed consent, and participants received $20 for time and travel.

At baseline, trained interviewers assessed current depression with the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I) (40) and the 20-item Hopkins Symptom Checklist (SCL-20) for depression (41). The Recovery Assessment Scale (RAS) was included to assess participants’ perceptions of recovery and included five subscales (personal confidence and hope, willingness to ask for help, goal and success orientation, reliance on others, and no domination by symptoms) (42). Co-occurring anxiety disorders were assessed with the Psychiatric Diagnostic Screening Questionnaire (PDSQ) (43), and borderline personality disorder was assessed by using the self-report screener from the Structured Clinical Interview for DSM-IV Axis II Personality Disorders (SCID-II) (44). Cutoff scores for the SCID-II and PDSQ were selected to maximize specificity. Bipolar spectrum disorder was assessed with the Mood Disorder Questionnaire (45). We included additional questions regarding use of depression services (including inpatient and outpatient mental health specialty services, medical visits addressing depression, and antidepressant medication).

Following the baseline interview, participants were randomly assigned to the intervention or usual care by using electronically generated randomization tables. Randomization was stratified by site and blocked in randomly allocated block sizes ranging from 4 to 12 through the use of a scheme that was designed by the study statistician and incorporated into the tracking database; interviewers were blinded to treatment group. Participants assigned to the intervention group were notified by a study care manager. Participants assigned to usual care were notified by mail.

Participants assigned to either group were free to use any primary care or mental health specialty services normally available at GH or Cherry Hill or available elsewhere.

The 18-month intervention included regular telephone or in-person contacts with a care manager and a structured group program. The meetings were co-led by a care manager and a trained peer specialist. After random assignment, participants who were assigned to the intervention were invited to attend an engagement session modeled on the work of Grote and colleagues (46) and Zuckoff and colleagues (47). During the engagement sessions, the care manager and peer specialist followed an outline and a suggested script that employed evocative and motivational strategies to increase initial engagement in all aspects of the intervention program (46,47). As part of the group program, participants met weekly for ten weeks, twice a month for two months, and then once a month for maintenance of progress.

Over the course of the study, four clinicians with varied professional training functioned as care managers, including a social worker, a marriage and family therapist, a health educator, and a psychologist. All had extensive experience working with patients with depression. Three individuals with personal experience of depression served as paid peer specialists. Each specialist completed a five-day training and certification program from the Depression and Bipolar Support Alliance. Peer specialists held drop-in office hours and made group reminder calls, in addition to cofacilitating groups. During these contacts, peer specialists engaged in conversations focused on goals and strategies to facilitate recovery. Peer specialists referred all clinical concerns to the care manager, treating providers, or supervisory team.

The tracking database supported systematic tracking of patient contacts, assessment of depressive symptoms (PHQ-9), documentation of self-care and recovery goals, and participation in the group program. The care manager contacted patients at specified intervals no longer than a month apart during the first 12 months, after which the intervals between contacts varied according to symptoms, treatment adherence, and patient engagement with treatment providers. During each contact, care managers helped participants identify, plan, and troubleshoot actions to meet personal self-management goals.

Care managers also sent quarterly reports to treating providers via e-mail (for GH providers) or by fax (for Swedish providers). Reports included information about patients’ current depressive symptoms (PHQ-9), participation in the group program, and self-care and recovery goals. The care manager also provided care coordination and facilitated follow-up care. All intervention staff received weekly supervision by the study psychologist and psychiatrist. These weekly sessions were used to discuss patients with moderate or more severe depressive symptoms or who were overdue for contact or to respond to concerns by care managers about a particular patient.

The manualized self-management group program was adapted from the cognitive-behavioral therapy program used in our pilot study (48) and supplemented with recovery-oriented content (49,50). [The content of weekly classes is summarized in an online supplement to this article.] The peer specialist acted as both a cofacilitator and a participant role model for the groups.

Interviewers blinded to treatment assignment contacted participants by telephone for outcome assessments at three, six, 12, and 18 months after randomization. Participants were paid $20 for completion of each interim assessment and $30 for completion of the final assessment (regardless of treatment assignment or level of participation in treatment). Each assessment included the SCL-20 (41); the current depression module of the SCID-I (40); the RAS (42); the Patient-Rated Global Improvement (PGI) (51), a 7-point rating of treatment effectiveness from a patient’s perspective; and questions about mental health care utilization and antidepressant use since the previous assessment.

All participants provided documented oral consent for participation in telephone screening and written informed consent prior to study enrollment. Study procedures and materials were approved by both the GH and Swedish Medical Center institutional review boards. Study safety was monitored by a Data Safety Monitoring Board, which met every six to 12 months to review reports on recruitment, data collection, patient safety, and outcomes.

We compared the intervention and control groups at baseline by using t tests for continuous variables and chi square tests for binary or categorical variables.

Our primary outcome measure was the mean score on the SCL-20. Secondary outcome measures included percentage of participants with major depression on the basis of SCID-I diagnosis, percentage of participants with a PGI rating of “much improved” or better, and mean score on the RAS. The primary analysis assessed the long-term effect of the intervention by comparing the average of the six-, 12- and 18-month outcomes for each group. We estimated this effect by using repeated-measures methods, including random-effects mixed models for continuous or ordinal outcomes (such as the SCL-20), and generalized estimating equations (GEE) for binary outcomes. GEE models used an independent working correlation with empirical standard errors. We also compared the groups on rate of improvement over the first six months by using random-effects mixed models and GEEs and at each time point (cross-sectional analysis) by using t tests and chi square tests. All analyses were completed with and without adjustment for site, age, gender, and baseline value of the outcome variable (except for the model for initial rate of change).

Analyses classified participants according to initial treatment assignment without regard to their degree of participation in the intervention (intent-to-treat approach.) The study had 80% power to detect an 11.3% difference on the SCL-20. All analyses used SAS 9.1.

Between January 2010 and September 2011, a total of 2,166 potentially eligible patients were identified either by review of computerized records or by referral by treating providers. Of those invited for screening, 1,150 completed screening interviews; 302 patients were eligible to participate, completed baseline interviews, and were randomly assigned to the intervention (N=150) or to a usual care control group (N=152). [A flowchart documenting participation in the intervention is available in an online supplement to this article.] Enrollment continued until we reached our enrollment goal of 300, chosen on the basis of power analyses.

Of the 150 patients assigned to the intervention, 138 (92%), 137 (91%), 135 (90%), and 134 (89%) completed the three-, six-, 12-, and 18-month follow-up interviews, respectively. Of the 152 patients assigned to usual care, 145 (95%), 140 (92%), 138 (91%, and 140 (92%) completed the three-, six-, 12-, and 18-month follow-up interviews, respectively. There were no significant differences between the treatment groups in follow-up rates overall or at any time point.

As shown in Table 1, the modal participant was a middle-aged female with moderate symptoms of depression. Those assigned to the intervention group were more often female, but treatment groups were similar in all other baseline characteristics.

Of the 150 participants randomly assigned to the intervention, 148 (99%) had at least one care manager contact, 133 (89%) had four or more care manager contacts, and 127 (85%) remained engaged (received some care management contacts) during the last six months of their enrollment. A total of 124 (83%) participants attended at least one group session, 71 (47%) completed seven or more weekly group sessions, and 54 (35%) attended some group sessions in the last six months of their enrollment. Study participants attended 10±11 group sessions, including approximately 6±4 weekly sessions and 4±5 maintenance sessions. Thirteen groups were initiated during the course of the study. All sessions were audiotaped; a sampling of sessions (N=7) were independently rated for fidelity, and all were rated excellent.

To determine if the intervention was efficacious, we examined the long-term effects on four outcomes (selected a priori): the SCL-20, the RAS, percentage with major depression (according to the SCID-I), and percentage who reported being greatly improved (PGI). The intervention had significant long-term effects—defined as the average of the six-, 12-, and 18-month follow-ups—on all four outcomes in repeated-measures analyses, with or without adjustment for age, gender, site, and baseline value of the outcome variable (adjusted p<.002, except for the RAS and the SCL-20 [p=.03]) (Table 2).

We also examined whether the intervention affected the rate of improvement over the first six months of the trial (“initial improvement”). The intervention resulted in faster rates of initial improvement on the SCL-20 and the RAS (adjusted p=.003 for the SCL and .013 for the RAS). Initial improvement for the control and intervention groups did not differ significantly for either the percentage with major depression or the percentage who reported being greatly improved.

We tested whether the effect of the intervention on the SCL depended on receipt of specialty mental health care at baseline. We found no evidence of such an interaction (data not shown).

Rates of self-reported utilization of nonstudy services for depression and antidepressant medication are shown in Table 3. Of GH participants for whom complete computerized pharmacy records were available, five of 122 in the intervention group and six of 128 in usual care initiated antidepressant treatment between baseline and three months (pharmacy records were not available at the Cherry Hill site). Intervention participants were more likely than members of the control group to be taking antidepressant medication at six, 12, and 18 months. Participants in the intervention group had significantly more nonstudy-related visits for depression at six months, but otherwise the rates of mental health visits between the two groups were comparable.