World Trade Center (WTC) related disorders inflict a major toll on the health and well-being of WTC responders. These debilitating conditions can be ameliorated or even reversed with lifestyle medicine programs targeting nutrition and diet, physical activity, sleep, and stress management. Despite their efficacy, lifestyle medicine programs are constrained by high attrition rates. Adjunctive interventions to increase retention can be deployed if accurate forecasting algorithms are developed to identify who is most likely to drop out. Previous studies found a host of demographic and behavioral risk factors for attrition. However, none are sufficiently strong individually to accurately predict future dropout, and they have not yet been combined into a validated predictive algorithm. Furthermore, recent technological advances made it possible to identify new powerful predictors of dropout using objective, passively sensed data (natural language, sleep, and movement). However, the utility of these passively sensed data for predicting attrition has not been rigorously proven. To address these gaps, the goal of the current project is to identify psychological and behavioral factors assessed from self-report, medical exams, and passive sensing that predict attrition in a lifestyle medicine program. We will use existing infrastructure within the WTC Health Program at Stony Brook University to assess 800 WTC responders as they enroll in an established 3-month lifestyle medicine program. At program intake, we will (a) assess established risk factors, (b) record the first treatment visit to obtain natural language predictors, and (c) give participants a FitBit to wear for 1 week to measure physical activity and sleep patterns via passive sensing. Our aims are to (1) identify individual predictors of attrition in a lifestyle medicine program and (2) combine predictors to develop an algorithm to forecast attrition (using machine learning methods). This study will improve our ability to identify patients at risk of attrition and also characterize profile of dropout risk across many potential predictors, revealing pathways to attrition that can be targeted by supportive interventions in future studies (e.g., motivational interviewing, just-in-time adaptive interventions based on passive sensing data). The proposed project takes the first step to address the problem of high attrition in lifestyle medicine programs among WTC responders and other patient populations. It will pave the way for randomized clinical trials of supportive interventions with patients identified by the algorithm as likely to drop out.