Statistical methods for the assessment of social engagement in psychosis using digital technologies

摘要

ABSTRACT This is an application for a K01 award for Dr. Linda Valeri, a biostatistician at McLean Hospital and the Harvard Medical School. Dr. Valeri is establishing herself as a young investigator in psychiatric biostatistics focusing on psychotic disorders. This K01 award will provide Dr. Valeri with the support necessary to accomplish the following goals: (1) to become expert in psychiatric biostatistics focusing on mobile health (mhealth) research for psychotic disorders (2) to conduct investigations using mhealth technologies in patients with psychosis; (3) to develop automated software for advanced machine learning methods in mhealth studies; and (4) to develop an independent research career. To achieve these goals, Dr. Valeri has assembled a team comprised of three mentors, Dr. Dost Öngür, Chief of the McLean Hospital Psychotic Disorders Division, who leads a neuroimaging laboratory studying the biology of psychotic illness, and co-mentors Dr. Russell Schutt, Professor of Sociology at University of Massachusetts in Boston, who has extensive experience in the study of social interactions in patients with severe mental illness, and Dr. Jukka-Pekka Onnela, Associate Professor of Biostatistics at Harvard T.H. Chan School of Public Health, who has developed a platform for collection of raw sensor data from mobile devices, called “Beiwe”, and conducts research in the fields of digital phenotyping and network science. Dr. Valeri’s research will focus on the development of statistical methods for the analysis of mhealth data to shed light on the role of social engagement in psychosis. The proposal builds upon the hypothesis that social interactions captured by passive mobile data streams (call and text logs) and mobile surveys are potential targets of intervention and could lead to a sustained recovery by promoting perceived social support and improving psychiatric symptoms. In Aim 1(a) we propose to extend a machine learning approach, Bayesian Kernel Machine Regression, for the analysis of mobile data streams accounting for time- varying confounding. The approach will allow in Aim 1(b) to establish (i) reliable links between a high dimensional time series of passive measures of social and mobility behaviors with self-reported measures of social interaction and (ii) the effect of social interaction dynamics on perceived social support and psychiatric symptoms measured in clinical settings. Further, we will extend the approach to correct for selection bias introduced by missing data in mobile surveys (Aim 2). For both aims, Dr. Valeri will develop software (Aim 3) and apply the approaches to investigate these scientific questions using data from an ongoing study based at McLean Hospital Psychotic Disorders Division that employs the smartphone platform for collection of sensor data developed by Dr. Onnela. Dr. Valeri’s investigation will provide preliminary evidence on features and timing of social interaction behaviors that can improve psychiatric symptoms along with understanding of potential mechanisms of action. This research will form the basis for an intervention study that encourages social interactions of patients with psychosis, to be proposed in a future R01 application. !