Predicting Diabetic Retinopathy from Risk Factor Data and Digital Retinal Images

African American and Latinx communities nationally and in California not only bear a disproportionate burdenof COVID-19 positive cases and deaths but are also not taking part in COVID-19 testing for a wide range ofunderstudied reasons. This can have profound implications in safety net health care settings where vulnerablepatients, who are in need of clinical procedures to prevent significant morbidity, are refusing such potentiallylifesaving procedures because of fear of COVID-19 testing and/or contracting COVID-19. The Los AngelesCounty Department of Health Services (LACDHS) is the second largest publicly operated county safety nethealth care system in the United States, serving more than 750,000 patients annually. Timely access to healthcare in this under-resourced, high-need setting has been an ongoing challenge for its majority Latinx andAfrican American patients. With the current pandemic, COVID-19 testing for patients has become an essentialfirst step in the provision of critical procedural care. However, the range of reasons why patients refuseCOVID-19 testing is little understood. To this end, we propose to explore the obstacles to COVID-19 pre-procedural testing and provide COVID-19 specific training to LACDHS Community Health Workers (CHWs)from these same communities to effectively address: a) the primary goal of increasing COVID-19 testing forindividual patients, and the secondary goals of b) facilitating needed procedural care in a timely manner for thesafety net health system, and c) developing a sustained public health presence in these communities to buildtrust and preparedness for critical COVID-19 related future needs. Trained CHWs can help to more effectivelyovercome obstacles to COVID-19 testing, including historical barriers of mistrust, provide COVID-19 healtheducation, help address social determinants of health and help facilitate technological literacy to improvepatient access to testing and care in a telehealth environment. The proposal uses a multidisciplinary, mixed-methods approach including unsupervised machine learning and qualitative interviews to systematicallyexplore barriers and facilitators to COVID-19 testing among vulnerable safety net patients. We will then trainclinically based, ethnically/linguistically matched CHWs to implement a hypothesis-driven interventionconsisting of six group classes and six personalized patient encounters with African American and Latinxsafety net patients. This study has the following specific aims: Aim 1- Utilize machine learning methods toassess whether there are characteristics that define African American and Latinx safety-net patients whoengage in or refuse COVID-19 testing; Aim 2 - Conduct in-depth interviews with African American and Latinxpatients who either declined or accepted COVID testing to explore contextual, behavioral, and attitudinalfactors shaping patient circumstances and concerns; Aim 3 - Develop, implement, and pre-test a CHWintervention with the information from Aims 1 and 2, utilizing a randomized control design among AfricanAmerican and Latinx safety net patients to assess the effect of the CHW hypothesis-driven intervention ontrust, self-efficacy, and intent to participate in COVID-19 testing.