M48. Digital Phenotyping in Schizophrenia Using Smartphones

Abstract

Background: The purpose of our study is to determine if self-reported symptom surveys and passive data (GPS, accelerometer, voice samples, call logs, text logs, and phone use data) collected from patients with schizophrenia in real time on their personal smartphone may be useful in predicting reoccurrence of psychotic symptoms. Methods: Research subjects are adults (age 18–45, both sexes) who have been diagnosed with schizophrenia or schizoaffective disorder and are currently in treatment. Inclusion criteria include owning a smartphone capable of running the study application. Subjects use the smartphone app for a 3-month period. During this period the app will always be constantly collecting data, up to 1 million data points per day per subject as outlined in the background section. In addition, subjects will have psychiatric batteries including the PANSS, CGI, BPRS, and GAF completed as part of monthly check-up meetings with study staff. Outcome measures include (1) Adherence to and Acceptability of the Smartphone Application. (2) Comparison of Smartphone Application Data to Clinician Collected Metrics 3) Sensitivity of the Smartphone Application in Predicting Symptom Change Through Individual Data Streams (Surveys, GPS, Accelerometer, Voice, Call Logs, Text Logs, Screen Data) and in combination. Results: To date 14 of 20 subjects have been recruited into the study. Three have completed the study, 2 have dropped out, and 9 are currently active. The subjects who dropped out of the study did so related to factors other than smartphone and app use. No subjects have complained about the app use or its tracking. We plan to recruit 6 more subjects in the next month. From data pooled from all subjects to date, the total duration of minutes not using the phone shows a negative correlation with the warning signs scale (P = .0001) as well as the length of outgoing phone calls (P = .0024). Taking medications is correlated with the number of outgoing text messages sent (P = .0004). Conclusion: Preliminary sensor data indicates that there are small but statistically significant correlations with psychiatric symptoms for patients with schizophrenia. Results to date also suggest that tracking patients with schizophrenia via their personal smartphone is acceptable in a research context and there is no evidence of harm to date. Using patients’ own phones to assess symptoms offers a potentially cost effective and scalable means to gather new streams of objective data regards patient’s state and real time functioning.