Visualizing and predicting new and late HIV diagnosis in South Carolina: A Big Data approach

点击翻译按钮获取中文摘要

Abstract Although many notable strides have been made in HIV prevention, the current pace of progress in the US, especially in the Southern states, is too slow to achieve the goals laid out in “Ending the HIV Epidemic: A Plan for America” (EtHE). The success of effective HIV interventions demands more precise and timely surveillance of the epidemic, especially for the HIV clusters and outbreaks as indicated by a large or an increase in number of new and/or late diagnoses among certain key populations or geolocations (“hotspots”). However, some critical gaps remain in our current HIV surveillance and targeted prevention efforts. These gaps include a lack of timely prediction of HIV risk that is needed for a rapid public health response, reliance on data from limited sources in identifying the hotspots and predictors of new infection clusters, and a lack of efforts in utilizing various data sources to inform the selection and delivery of targeted prevention and control at state and local levels. To develop more effective, timely, and targeted approaches for HIV prevention, it is critical to develop a data-driven surveillance and prediction system of new and late diagnoses so that the local health departments and healthcare systems can rapidly identify priority populations and geolocations where HIV is spreading, or our prevention efforts are lagging behind. Building on an integrative multi-level data structure and a strong academic-government partnership, we propose the current study to strengthen the understanding and visualization of HIV infection clustering, enable the prediction of outbreaks, and contribute to the optimization of strategies to deliver evidence-based prevention. In close collaboration with the South Carolina Department of Health and Environmental Control (SC DHEC) and other stakeholders, we will integrate multi-level data sources, including statewide electronic health records data, county-level contextual data, geospatial data, and social media data, to predict new and late HIV diagnoses in SC and develop an interactive web portal to visualize the spatiotemporal patterns and trends of new and late HIV diagnoses in SC across geolocations and over time, particularly in the context of unanticipated HIV service disruptions by COVID-19 and other future public health crises. Using a Big Data approach and the integration of multi-level data sources, the proposed research will provide a better understanding and visualization of the dynamic spatiotemporal patterns and new case predictions. The prediction models and the interactive web portal will assist SC DHEC, AIDS service organizations, and other healthcare systems to rapidly identify, characterize, and predict new HIV clusters and to deploy targeted HIV prevention and control efforts in a timely fashion.