Mobile Behavioral Sensing for Relapse Prevention in Schizophrenia
Dror Ben-Zeev, PhD
Our interdisciplinary research team at the University of Washington, Dartmouth College, Zucker Hillside Hospital and Cornell are collaborating on the development of a mobile system that uses smartphone-embedded sensors (i.e. microphone, accelerometer, GPS, light sensor) coupled with Ecological Momentary Assessment (EMA) computer generated self-reports, to track a range of behaviors (i.e. paralinguistic aspect of speech, physical activity, location, sleep, mood, symptoms) that are relevant for relapse in schizophrenia. Using machine learning techniques, the system leverages behavioral data and patient self-reported clinical updates to generate personalized early warning models.
The models are adaptive and evolve with patients’ use of the system over time, focusing on variability from one’s typical behavioral patterns to calibrate a unique patient relapse signature. Treatment teams are informed about patients’ clinical status on an ongoing basis. When the mobile system “flags” trends that are consistent with one’s relapse signature, it will trigger patient functions (i.e. illness self–management strategies, prompts for help-seeking behavior) and provider functions (i.e. real-time notification, prompts to initiate contact, timesensitive treatments) to help prevent progression to full psychotic relapse.
Funded by: National Institute of Mental Health