Predicting Morbidity and Mortality for HIV-Related Opportunistic Infections

PROJECT SUMMARY This project uses machine learning to predict and improve outcomes for people with HIV (PWH). The research has been refined based on Year 1 findings, which highlighted significant regional variations in opportunistic infection (OI) patterns from the National Inpatient Sample (NIS) dataset. The project will now incorporate a national perspective to complement the original focus on the southern US. The research will predict the effects of simulated health insurance coverage expansion on OI-related hospitalizations and in-hospital mortality, using both national (NIS) and state-level (SID) data. These analyses will involve forecasting models and unsupervised machine learning to identify key trends, costs, and risk factors. Additionally, the project will develop and refine a predictive model for individual PWH clinical outcomes using the ADVANCE EMR network, with a new focus on health care access variables like insurance status and urban/rural location. The project's goal is to produce actionable risk scores and provide evidence to inform health policy, particularly regarding insurance coverage and geographic access, in alignment with national priorities for ending the HIV epidemic. The work will also focus on reusing these methods to develop open-source tools to make computational epidemiology more accessible to the broader HIV research community.