PROJECT SUMMARY Sickle cell disease (SCD) affects over 20 million people living worldwide and approximately 100,000 individuals living in the United States. Individuals with SCD are at increased risk of mortality, poor health-related quality of life, and high health care utilization. Pain is the primary factor linked to poor health outcomes and increased medical costs for individuals with SCD. The majority of SCD pain episodes are unanticipated; leading to a lack of prophylactic pain management, increased use of opioids and other health care, and poor quality of life. Accurate mathematical models to predict SCD pain in pediatric patients would facilitate the development, testing, and maximizing of the timing of implementation of interventions to improve their effectiveness, and reduce the use of opioids; thus, minimizing the risk of opioid dependence. Our central hypothesis is that dynamic mathematical models that incorporate time-varying mobile health (mHealth) variables will increase the accuracy of prediction of individual daily pediatric SCD pain features – pain severity, onset, and exacerbations. We also hypothesize that changes in mHealth data are important precursors to changes in pediatric SCD pain. Our short-term goal is to develop and test a dynamic mathematical modeling framework that includes combinations of mHealth variables to identify the best model formulations for predicting individual daily pediatric SCD pain features. The proposed study will leverage existing data from a previous project focused on the relationship between sleep and SCD pain – to date, the largest study that incorporates ecological momentary assessments (EMAs) and actigraphy measures for youth with SCD. The previous analyses of these data did not consider the dynamic nature of the relationships or examine the range of mHealth data available. To accomplish this goal, we aim to (1) construct a framework consisting of a dynamic mathematical model that focuses on predicting pediatric SCD pain severity that can utilize various combinations of mHealth variables, and (2) determine which modeling framework instances – mHealth data combinations coupled with the model – are effective for predicting individual SCD pain severity patterns. Then (3) we will use the framework instances selected to predict risk of pediatric SCD pain onset and pain exacerbations and determine which mHealth variable combinations are most successful at predicting these pain features. Specifically, we will apply machine learning algorithms and assess the ability of each instance of the modeling framework to predict pain onset or exacerbation the next day.