Project Summary/Abstract The CDC recommends PrEP (a biomedical prevention strategy in which people without HIV take HIV medication daily to decrease the probability of becoming infected by more than 90% for MSM with a history of inconsistent or no condom use. In Los Angeles, Latino MSM are less like than White MSM to have used PrEP in the past 12 months or to have used PrEP consistently for two or more continuous months. Reasons for discontinuing PrEP include fear of side effects and struggling with daily adherence. This has contributed to a widening disparity in access to, and use of, PrEP to prevent HIV infection. We focus on PrEP because even a modest increase in PrEP use can result in significant declines in HIV incidence. Siblings can impact wellbeing. Sibling relationships contribute to a person’s development early in life and can provide a lifelong source of social and emotional support. However, we know little about whether their siblings can be engaged in interventions to support PrEP adherence—or any health behavior. Machine learning in HIV prevention strategies is in a nascent stage. Machine learning algorithms build a model based on sample data in order to make predictions. It has been used to identify candidates for PrEP use. However, to our knowledge, these innovative approaches have not been used to analyze the medical records of LMSM already using PrEP in order to predict who might discontinue using PrEP. Understanding how LMSM might stop using PreP or who is likely to be non-adherent can inform how we approach PrEP- adherence interventions. Specific Aims: 1. To develop PrEP adherence messages to increase PrEP adherence in high-risk Latino MSM that can be delivered by their siblings. We will use the Information-Motivation-Behavior (IMB) model and conduct in-depth interviews with 40 LMSM-Sibling pairs to identify factors relevant for the development and delivery of messages to support adherence to PrEP. Messages will be tested with 3 focus groups. 2. To use machine learning and ensemble classifiers to develop models to predict and identify who will not adhere to, or stop using, PrEP. We will work with our community partner, St. John’s Community Health, to examine different machine learning approaches using medical records from their clinics to develop models to predict and identify who is likely to stop adhering or stop using PrEP. These two aims provide a unique opportunity to bring together theory-based, evidence-driven behavioral intervention research and machine learning analyses. An important aim of this project is to develop an effective and brief intervention that can be sustained by our community partner, St. John’s Community Clinic, and ultimately scaled-up by health departments.