Pre-exposure prophylaxis (PrEP) is effective at reducing the acquisition of HIV; however, achieving the full impact of this intervention is contingent on maintaining engagement in care throughout periods of high and low risk. Currently, in the United States, only 270,000 out of 1.2 million individuals with indications for PrEP receive it. Additionally, even for those who start PrEP, adherence, and persistence in comprehensive HIV prevention care is poor, characterized by multiple discontinuations and restarts. Furthermore, HIV risk changes over time and individual-level reasons for engagement, very early disengagement (<2 visits after initiation), early disengagement (2-4 visits after initiation), and late disengagement (>4 visits after initiation) are not well described; social and behavioral determinants of health (SBDH) like insurance, housing status, substance use, and mental health, are increasingly recognized as key factors. As PrEP programs expand to meet the projected need, the cost of providing support services will be substantial and thus identifying patients at highest risk of loss-to-follow-up and selecting optimal services to support persistence in care and PrEP adherence is essential. Furthermore, many individuals receiving HIV prevention services are known to receive highly fragmented care typified by multiple providers and institutions, thereby creating a challenge to accurately characterize persistence in care. Guided by the information-motivation-behavioral skills (IMB) model, we propose to address this critical area of research by focusing on factors that are associated with very early, early, and late disengagement from HIV prevention care. We will identify baseline factors through a comprehensive questionnaire conducted at the time of enrollment into comprehensive HIV prevention care. We will layer onto that longitudinal factors, such as social and behavioral determinants of health (SBDH), which can vary over time, to get a more comprehensive and precise picture of factors affecting persistence in prevention care. To address the issue of fragmented care and better characterize persistence in care, we will utilize information contained in a large Health Information Exchange (HIE), Healthix, to capture all health care visits in New York. We propose to use machine-learning methods to design predictive models of disengagement from HIV prevention care. This comprehensive assessment of persistence in HIV prevention care will inform the development of high quality, scalable models of HIV prevention care, making it possible to target limited resources towards individuals at the highest risk of disengagement.
Pre-exposure prophylaxis (PrEP) is effective at reducing the acquisition of HIV; however, achieving the full impact of this intervention is contingent on keeping patients engaged in care throughout periods of high and low risk. HIV risk changes over time and a person?s reasons for engagement, disengagement, and re- engagement in care are not well understood, but likely change over time. In this project, we propose to utilize self-reported and clinical data to elucidate reasons for very early disengagement (<2 visits after initiation), early disengagement (2-4 visits after initiation), and late disengagement (>4 visits after initiation) and use machine learning methods to design predictive models for disengagement from HIV prevention care.
