Using Structural Equation Modeling to Understand the Process of Change in a Community Health Worker Intervention for African Americans and Latinos with Type 2 Diabetes

Background. Community health worker (CHW) are community residents who are trained as social justice and health advocates who are vital to linking underserved and disenfranchised populations to health and social service systems.  Although CHW interventions have demonstrated promise in improving health behaviors and outcomes, particularly for racial and ethnic minority communities, very little is known as to how these interventions produce positive outcomes.  Thus, the purpose of this paper is to test a conceptual model for a CHW intervention for improving type 2 diabetes outcomes for African Americans and Latinos in Detroit, Michigan. 

Research Questions. In our community-based participatory research (CBPR) intervention, CHWs use an empowerment-based, culturally tailored approach to provide diabetes education (individual and group formats) and support to patients in health clinics, conduct home visits, and accompany patients to clinic visits.  This study asks whether components of our intervention had a differential impact on improving self-efficacy, diabetes-related distress, and self-management behaviors.  Furthermore, we ask whether these changes lead to a reduction in our primary outcome, hemoglobin A1c (HbA1c), which is a measure of blood glucose.   

Methods.  African American and Latino participants with type 2 diabetes (N=326) were recruited from medical records at three healthcare sites in Detroit. We use Structural Equation Modeling (SEM) to test a conceptual model that CHW education and support serves a means for improving self-efficacy, distress, self-management behaviors and ultimately HbA1c from baseline to 6 months. SEM was chosen because it enables us to test the order in which measurable variables affect each other.  The model was estimated by Full Information Maximum Likelihood (FIML).  Goodness of fit was evaluated with Joreskorg-Sorbom GFI (Goodness of Fit Index) for absolute fit, Bentler’s CFI (Comparative Fit Index) for comparative fit, RMSEA (Root Mean Square Error of Approximation ) for parsimony, and SRMR for prediction. Power was calculated by the method of MacCullum, Browne, and Sugawara (1996).

Results. All post-intervention measures were highly correlated with pre-intervention values.  Post-intervention HbA1c dropped by 0.55 per unit increase in self-management behavior. Improvement in self-management behavior was significantly associated with lower diabetes distress. Program attendance in group, versus individual format, was associated with a 6 point decrease in diabetes distress and significant increase in knowledge of diabetes management.  Greater self-efficacy was associated with higher attendance.  Based on GFI=.9928 and CFI=.935, the model explained 99.25% of the generalized covariance and was a 93.5% improvement over the null model.  RMSEA was 0.067, indicating a reasonable fit, and the SRMR was 0.0464, which indicates that the model has good predictive ability.

Conclusion.  Structural Equation Modeling is an effective method to understand the process of change in key outcome variables and to measure the effects of a CHW intervention.  This study has implications for social work by informing researchers and practitioners of the differential effect of various components of a CHW intervention that has the potential for reducing complications related to diabetes.  The study also informs social work about the value of CHWs and their role in eliminating health disparities.