Precision applied behavior analysis intervention for autism spectrum disorder using natural language processing and graph centrality

The increased prevalence of Autism Spectrum Disorder (ASD) and the urgent need for personalized treatment have highlighted the role of data science in enhancing clinicians’ capacity and treatment quality. Application of Natural Language Processing (NLP) has created new paradigms by analyzing and finding similarities between the treatment prescriptions extracted from Electronic Health Records (EHRs). Social Network Analysis (SNA) and centrality computation methods have opened new avenues to identify behavior patterns and mental health symptoms, forecasting therapy progression and personalization trajectories. In this paper, we develop a novel SNA graph model by preprocessing longitudinal Applied Behavior Analysis (ABA) treatment data of 29 patients using NLP methods and computing various centrality scores. We perform community detection at various temporal points during the six-month intervention duration and find patient similarity based on prescription and socio-demographic similarity-building edge weights. We develop a treatment recommendation model and match its outcome on recommendation and effectiveness measures with the ground truth. Our contribution explores novel approaches in determining the node influence of centrality measures on patient-level skill acquisition and treatment recommendation.