Predicting Situational Onset of Aggression in Minimally Verbal Youth with Autism Using Biosensor Data and Machine Learning Algorithms
Abstract
Unpredictable aggressive behavior by youth with autism isolates them from educational, social, and family activities. Approximately 2/3 of youth with autism display aggression, a common reason for treatment referral. However, evidence-based pharmacological and behavioral interventions for aggression in ASD are frequently ineffective. Aggression is particularly impairing in the 30-40% of youth with autism who are minimally verbal and cannot verbalize their distress. Aggression may represent a maladaptive attempt to express or modulate distress related physiological arousal. We hypothesized that physiological arousal precedes aggressive behavior. We aimed to predict aggression in minimally verbal autism participants before it occurs using data collected from a wrist-worn physiological sensor and time-synchronized behavior observation. Using sophisticated machine learning algorithms linking observable aggression to preceding physiological signals (heart rate, skin conductance), we demonstrate that aggression can be predicted three minutes before it occurs with 80-90% accuracy. These findings enable new opportunities for pre-emptive intervention.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2022
- Accession Number
- AD1195580
Entities
People
- Matthew S Goodwin
Organizations
- Northeastern University