Factors Influencing the Participation in an Emergency Department Based Diabetes Screening Program

This dissertation includes five separate chapters. In the first chapter, we provide a background in type 2 diabetes, health disparities, and the benefit of conducting diabetes screening within the emergency department (ED). We describe the parent project, the Innovating Diabetes Screening in Emergency Departments and Linkage Services (IDEAL) program, which aimed identify patients with undiagnosed prediabetes or diabetes within the ED and link them to care. Finally, we discuss the gaps in the literature, provide a conceptual framework, and propose the study aims that will be evaluated in this thesis. The next chapter, we used electronic medical record (EMR) data from patients who visited the ED in May 2021 to validate a best practice advisory (BPA) diabetes screening algorithm in identifying eligible patients for hemoglobin a1c (HbA1c) testing. We found that the algorithm performance adequately given the ease of implementing within the ED workflow, relatively minimum cost to run the HbA1c test, and volume of patients ultimately identified with abnormal HbA1c results that would need follow up to care. The third chapter we developed and internally validated a predictive model of abnormal HbA1c among ED patients who were screened for diabetes. Our work was guided by the TRIPOD guidelines. We split our sample into a derivation (80%) and test (20%) cohort and used a 10-fold cross validation to derive our final model which was then internally validated on the test cohort. Our final model appeared to have adequate performance in discriminating between normal versus abnormal hbA1c. We found that substance misuse is an important predictor of abnormal HbA1c among patients screened for diabetes in the ED, which may be an important factor to consider for linkage to care efforts and treatment following ED screening. In the fourth chapter, we conducted a qualitative study on primary care provider (PCPs) perspectives on linkage to care and management of patients identified with potentially undiagnosed prediabetes during ED diabetes screening. We found that factors such as patient experienced SDOH, lack of time, and inadequate health system resources make it challenging for providers to deliver diabetes prevention services in their practice. Additionally, providers noted potentially challenges in being informed of HbA1c lab results within and across health systems, despite EMR integration. The last chapter summarizes and discusses the implications of our findings from this thesis. Our findings will immediately inform the continued improvement of the IDEAL program and development of programs beyond in identifying important factors that impact the screening and linkage to care process of ED based diabetes screening programs. Future initiatives should focus on updating the EMR-based algorithm, validating the prediction model in larger ED population, and identifying patient related perspectives on diabetes prevention services.