Identifying Clusters of Vision and Eye Health Problems Using Secondary Surveillance Data - To address population vision and eye health problems and disparities, information is needed on the location and scope of these problems and possible explanatory factors. The NEI Strategic Plan and Notice of Special Interest ‘Research Addressing Eye and Vision Health Equity/Health Disparities’ both highlight the critical need for public health disparities research to understand and address these problems. The CDC-funded Vision and Eye Health Surveillance System (VEHSS) aggregates prevalence information from most eyecare in the United States, as well as from large national surveys. However, there is yet no systematic mechanism to analyze these data to identify and locate geographic clusters of vision and eye problems. In this project, we will (Aim 1) identify and locate geographic clusters with persistently high and/or spiking rates of vision problems and diagnosed eye disease, or low rates of eyecare service utilization, and (Aim 2) identify and rank community- level factors associated with clusters of different vision and eye health indicators. In Aim 1, we will develop an application to access VEHSS public use files (PUFs), conduct scan statistical analyses, and report identified clusters. For eye disease and eyecare services, we will use county-level VEHSS public use files aggregated from Medicare Fee for Service, Medicare Advantage, and Medicaid (including state Children's Health Insurance Programs CHIP); each of which contain prevalence information for 17 eye disorder categories, 79 clinical subcategories, and 28 eyecare services. For vision problems, we will use ZIP-code level self-reported vision difficulty from the American Community Survey. We will conduct scan statistical analyses using the NIH-funded SaTScan software to identify clusters in space (overall high prevalence in groups of counties or ZIP codes) and space-time (spikes in prevalence in these locations). We will report significant clusters on the VEHSS website to ensure wide dissemination. In Aim 2, we seek to understand causal or correlated factors associated with identified clusters. First, we will assemble an analytic file containing community-level characteristics including sociodemographic profiles, social determinants of health, chronic health risk factors including diabetes, smoking, and hypertension, correlated conditions including hearing loss and dementia, and environmental risk factors such as pollution. We will then use machine learning models to predict the geographic clusters and use SHapley Additive exPlanations (SHAP) analysis to determine which community-level factors had the most positive and negative predictive power for identifying the presence of clusters of different vision and eye health conditions. The findings of this project will enhance current vision and eye health surveillance practice by addressing a critical lack of information on the existence, location, and explanatory factors associated with geographic clusters of vision and eye problems.
