Leveraging Sexually TransmittedInfection Data for Enhancing HIV Prevention and Management Guidelines: A Data Science Approach

Despite significant advancements in HIV prevention and treatment, there remains a critical gap in integrating sexually transmitted infection (STI) data into predictive models for HIV health care-related outcomes. STIs have been recognized as significant risk factors for HIV acquisition and poor HIV care outcomes through a range of biological (e.g., mucosal barrier disruption and inflammatory response) and behavioral (e.g., risky sexual networks) mechanisms. Incorporating rising real-world STI data into HIV surveillance and prediction via machine learning (ML)-based models can be an effective strategy to identify individuals at higher risk of HIV acquisition, disengagement from care, and failure to achieve viral suppression. Such models enable personalized risk assessment, facilitating tailored prevention and care plans based on individual-level data. The All of Us (AoU) research program is a national community-engaged initiative aimed at improving health by partnering with one million participants, more than 80% of whom belong to understudied populations in the US. The AoU program integrates a series of self-reported survey data (e.g., healthcare access and lifestyle) with relevant longitudinal EHR data (conditions, drug exposure, lab measurements, and procedures) to create a comprehensive data repository. By utilizing AoU data and ML, this proposed study will incorporate a comprehensive set of modifiable and non-modifiable factors from AoU to better understand the role of STIs in various HIV care outcomes. We will also target an understudied population in the AoU program to ensure that our findings are relevant and applicable to those most at risk. Then, we will develop clinical prevention and management guidelines based on the findings from ML-based prediction models regarding the roles of different STI patterns in HIV health care outcomes to advance the goals of Ending the HIV Epidemic. This guideline will inform our HIV management and prevention efforts by incorporating STI surveillance, diagnosis, and treatment into a comprehensive and tailored response to reduce HIV incidence and improve health outcomes for people with HIV. The innovations of the proposed research include the building of comprehensive computable phenotype algorithms to identify STI patterns, an innovative data science approach with longitudinal and multifactorial perspectives, and the transition from complex data insights to clinically practical workflows. With expanded data sources and advanced modeling approaches, the proposed study can influence public health strategies by developing clinical guidelines grounded in robust predictive models to improve early diagnosis, treatment retention, and viral suppression among PWH, contributing to the overarching goal of substantially reducing new HIV infections and improving health outcomes. The proposed research also has great potential to transition the PI from a Post-Doctoral Fellow to an independent researcher rapidly.