Obstructive sleep apnea (OSA), the most common sleep disorder, is a major problem in the Veteran community that poses risks for decreased physical and mental health. The predominant treatment for OSA is positive airway pressure (PAP) therapy. While PAP therapy is highly effective, adherence is suboptimal with approximately 50% stopping therapy by year 2. Our group has shown that breathing patterning is predictive of long-term adherence, but the method used requires a 15-minute procedure while awake. The goal of this project is to refine breathing patterning metrics and enhance clinical integration by repurposing clinical data obtained as part of routine OSA evaluation (sleep studies) and treatment (PAP machine data). This retrospective analysis will first apply an already-developed breathing pattering metric to be more applicable for both in-lab polysomnography and home sleep apnea testing. Comparison of the breathing patterning metric in sleep and wake will determine applicability for home sleep apnea testing, which does not gather sleep staging data. Because PAP machines only record flow, we will also evaluate whether the breathing patterning metric has similar performance characteristics in different respiratory signals (plethysmography and flow) for predicting adherence by comparing concordance index. We will then evaluate for the optimal breathing patterning metric and machine learning algorithm based on accuracy in predicting PAP adherence. Breathing patterning can be measured in multiple ways; to optimize adherence, a panel of breathing patterning metrics will be evaluated. Multiple machine learning algorithms will be compared to determine which has the best discrimination for predicting PAP adherence. Breathing patterning changes with age will be evaluated. In addition, we will evaluate the added utility of breathing patterning to a model that predicts adherence using patient demographics, past medical history, and sleep study summary data. This project could provide a practical, cost-effective method to identify patients that are likely to become nonadherent. My short-term career goal is to develop a foundation of focused research in optimizing obstructive sleep apnea (OSA) treatment via patient-centered solutions that will inform my clinical practice. To fulfill this goal, the developed research and career development goals of this CDA-2 project complement each other. Additional instruction in modern machine learning techniques will allow for adherence model development and validation. Coursework in clinical informatics will assist me in optimizing the design and data collection from the VA electronic medical record. Instruction on managing research records will help with managing the large databases necessary for this project. Proposed training in research team leadership and responsible conduct of research is particularly relevant to a transdisciplinary project such as this and will have long-term career benefits. Over the course of...