Analyzing Behavioral Interventions and Integrated Care Models to Address Co-Occurring Depression and Diabetes: A Systematic Review and Meta-Analysis

Background: Depressive disorders frequently co-occur with diabetes and, when unaddressed, can exacerbate diabetes risks such as unregulated Hemoglobin A1C (HbA1c), diabetic retinopathy, and neuropathy. Integrated care (IC) models aim to address depression and diabetes simultaneously. IC models vary and utilize different behavioral health interventions. There are no meta-analyses that examine the overall effects of IC models using behavioral interventions for co-occurring depression and diabetes. Our study addresses this gap while examining the type of IC model and behavioral intervention as moderators.

Methods: The PRISMA guidelines for reporting were used for reporting and our Meta-Analysis was pre-registered with PROSPERO (Registration number: CRD42023491332). A systematic search was conducted utilizing PubMed, PsycINFO, and ProQuest. Two reviewers analyzed 517 abstracts. Inter-rater-reliability (IRR) statistics were captured at the abstract level (87.92%: near-perfect IRR). A total of 61 reports were acquired for full-text review. IRR statistics were captured at the full-text level (70.5%: moderately high IRR). A full team consensus meeting was held using the inclusion and exclusion criteria, leading to 100% consensus for the final 32 articles included in the analysis. Cohen’s d was utilized to calculate standardized mean differences.

Results: 32 studies with a total of 9,623 participants were included in our final analysis. The Collaborative care model was the most frequently used IC model with 14 studies including it. Cognitive behavioral interventions and behavioral activation were the most common behavioral interventions utilized. Studies used social workers, psychologists, and lay interventionists to deliver the interventions. There were 28 research reports with an effect size for HbA1c. Regarding the standardized mean differences, the pooled random effects model was statistically significant with a large effect size (Est. = -0.35, 95% CI: -0.51 - -0.19) with individuals in the treatment group experiencing an HbA1c reduction of 0.35 when controlling for the comparison group and standard errors. The moderating effects of the type of IC model (z = -1.31, p = 0.19), interventionists’ training (z=1.12, p=0.25), type of behavioral intervention (z = -0.69, p = 0.49), telehealth (z = 0.73, p= 0.46), and number of sessions (z= 0.68, p=.49) were not statistically significant. There was, however, a moderating effect for trials including multiple sites (Est.= .74, z=4.49, p<.01). There were a total of 23 reports that included sufficient information to calculate a standardized mean difference for depressive symptoms. The overall pooled random effects model was statistically significant with a large effect size (Est.= -0.72, 95% CI: -1.15 - -0.28) meaning that IC models averaged a 0.72-point reduction regarding depressive symptoms when controlling for the comparison group and standard errors.

Conclusions and Implications: Social workers frequent IC teams and can have an impact on diabetes and depression. IC models that use behavioral interventions produced large effects for both depression and diabetes. The type of IC model or behavioral intervention did not moderate the effects meaning that there are similar effects between treatment models. Social workers can lead in ensuring equal access to evidenced-based IC models and behavioral interventions.