CHARACTERIZING THE CLINICAL COURSE IN SCHIZOPHRENIA WITH DIGITAL PHENOTYPING

Abstract Background Digital phenotyping methods offer the potential to better understand the lived experiences of patients with serious mental illnesses like schizophrenia. Yet to date it is unclear if the digital biomarkers offered from this method are unique to certain conditions like schizophrenia, or rather are shared by diverse populations, and to what degree digital phenotyping data are correlated with patient and clinician assessments. Methods 50 patients with schizophrenia and 50 matched healthy controls collected smartphone digital phenotyping data for a three month duration including measures of geolocation, physical activity, screen use, cognition, and self reported surveys. In-clinic assessments at study start and at three months assessed cognition (Brief Assessment of Cognition in Schizophrenia), psychosis symptoms (Positive and Negative Symptom Scale; PANSS) and other measures. Clustering and correlational methods were utilized to compare active and passive data streams both within and across groups. Results Adherence to active data (surveys and cognitive assessments) on the phone was roughly 50%, both for those with schizophrenia as well as for the healthy controls. Four unique clusters that included both active and passive data emerged for each group and the clusters were distinct with unique symptoms, cognition, and passive data metrics. Each group also possessed distinct correlations between active and passive data, with the schizophrenia group having more statistically significant findings especially around sleep. Discussion Digital phenotyping methods offer the potential to identify unique clusters of patients based on both their self reported as well as passive data. Future research will explore the utility of these clusters in predicting functional outcomes and offering personalized treatment.