Implementation Fidelity of a Complex Behavioral Intervention to Prevent Diabetes Mellitus in Two Safety Net Patient‐Centered Medical Homes in New York City

Abstract

Research ObjectiveAssessing implementation fidelity is critical for complex interventions to understand the reasons for their success or failure. However, few interventions systematically document and report implementation processes. Therefore, we sought to conduct concurrent process evaluation of CHORD (Community Health Outreach to Reduce Diabetes), a pragmatic, cluster‐randomized, controlled trial aimed at evaluating the impact of a Community Health Workers (CHW) led, health coaching intervention (in‐person and remotely) on preventing the onset of type 2 Diabetes Mellitus (DM).Study DesignThe Conceptual Framework for Implementation Fidelity (CFIF) was applied to measure implementation fidelity and factors moderating fidelity of the three core intervention components: patient goal setting, education topic coaching with CHWs, and referrals made to address social determinants of health. Descriptive statistics and regression models were computed to determine factors related to fidelity.Population StudiedThe study population included primary care patients with prediabetes range glycemia at two New York City (NYC) safety‐net hospitals (Veterans Administration NY Harbor ‐ VA and Bellevue Hospital ‐ BH). Primary care teams were randomized to receive the CHW‐driven, one‐year intervention. Of the 559 patients enrolled in the intervention arm, 79.4% completed an intake survey, constituting the analytic sample for fidelity assessment.Principal FindingsThe 2 study sites contributed 60% (BH) and 40% (VA) of the sample. Protocol adherence was as intended in the protocol for the 3 core components with more than 80% of patients setting ≥1 goal and receiving coaching on ≥1 education topic, and 45.0% receiving ≥1 referral. After adjusting for patient gender, language, race, ethnicity, and age, the study site moderated adherence to goal setting (77.4% BH vs. 87.7% VA) and having ≥1 education session (78.9% BH vs. 88.3% VA). Study site also moderated rate of encounters (median number of encounters 6 BH vs 4 VA). Scores on the Patient Activation Measure had no impact on fidelity of core intervention components.ConclusionsThe fidelity of CHORD implementation varied across its three core components and was moderated by implementation site.Implications for Policy or PracticeThe CFIF was a useful approach to collect and analyze data concerning implementation fidelity of a complex behavioral intervention. Despite being implemented in a research setting, interventions may not completely adhere to their core components, which can influence outcomes. Our study emphasizes the importance of examining implementation fidelity of complex interventions and of assessing moderating factors. Our study also empirically tested the CFIF using quantitative concurrent process evaluation of core intervention components.Primary Funding SourceNational Institutes of Health.