A Modern Data Science Approach to Enhance Caregiver Support for Behavior Management in Children with ASD - SUMMARY Autism Spectrum Disorder (ASD) involves challenges in social communication and intensive, restricted, or repetitive behaviors and/or interests. Sensory atypicalities are prevalent in approximately 74% of individuals with ASD, can range from pleasurable experiences to distressing outbursts caused by hypersensitivity affecting self-control and decision-making. Severe outbursts may lead to self-harm. However, proactive strategies to prevent these outbursts are under-researched and underutilized compared to reactive approaches. This study aims to develop algorithms predicting outbursts using wearable biosensor data and advanced machine learning (ML) models. It is aimed at enhancing caregiver support for individuals with ASD using modern ML frameworks. We accomplish this goal across three aims. Aim one’s objective is the collection of a novel dataset of the peripheral physiological and movement precursors of outbursts. We accomplish this aim by observing such events naturally in children with ASD who are prone to such behaviors using wearable wristband-based biosensors along with recording devices to capture potential triggers. Aim two characterizes the peripheral physiological indicators as individuals progress into and through an outburst, exploring internal stages of dysregulation that comprise these behaviors. In this aim, we use statistical analysis to identify differences in the stress-reactivity profile, which inform later ML models. Aim three seeks to develop predictive ML algorithms capable of predicting outbursts that occur in individuals with ASD. Altogether, our study will lay technology capable of detecting the precursors of outbursts in real-life situations using state-of-the-art wearable devices and artificial intelligence (AI) algorithms. This technology could deliver immediate alerts for parents or caregivers, providing individualized recommendations to prevent or mitigate the severity of an imminent outburst, developed using evidence-based behavioral interventions. Moreover, as proactive strategies for preventing outbursts are under-researched compared to reactive strategies, and the fact that it is little known about the underlying mechanisms associated with an outburst, our results the foundations of our long-term aim: the development of will also contribute to a better understanding of the nature of outbursts and a more detailed and operationalized description of their physiological correlates.
