# Digital Phenotyping for Computational Models of Relapse Prediction in Early Course Psychosis

> **NIH NIH K23** · BETH ISRAEL DEACONESS MEDICAL CENTER · 2021 · $191,000

## Abstract

Project Summary
 The candidate requests support for a four-year program of training and research to better understand
how smartphone based digital phenotyping and computational methods can predict relapse and create digital
phenotypes of symptoms and clinical outcomes in early course psychosis.
 In the proposed training plan, the candidate will build upon his previous experiences in engineering,
clinical informatics, and clinical psychiatry to perform a multidisciplinary project at Beth Israel Deaconess
Medical Center. His training plan includes training in: 1) statistical methods for multivariate longitudinal analysis
and predictive inference 2) the neuropsychiatric assessment of schizophrenia 3) longitudinal clinical research
methodology with a focus on mobile technologies, and 4) the responsible conduct of research.
 Even with appropriate care, relapse is common in early course psychosis and each episode is
associated higher costs of care, poorer lifetime outcomes, and chronicity of the disease. There is a need to
learn more about the personal factors associated with relapse for individual patients in order to improve risk
predictions, ensure appropriate early interventions, and support coordinated specialty care services for
schizophrenia. This study proposes that smartphones sensors eg (GPS, accelerometer), wearable devices like
smartwatches collecting physiology, and smartphone based surveys and cognitive tests, when combined with
appropriate statistical methods, can capture digital biomarkers, refereed to here as digital phenotypes, of early
course psychosis that can offer personalized relapse prediction and augment population level risk factors.
 This candidate's research plan seeks to: 1) propose digital phenotypes and relapse models of early
course psychosis captured in an affordable and scalable manner from subject's personal smartphones as well
as a wearable sensor in order to automatically collect self-report of symptoms, behaviors, cognition, and
physiology 2) and evaluate the accuracy of digital phenotypes and the relapse prediction models.
 This study proposes to address this hypothesis by utilizing smartphone based digital phenotyping
methods, primarily through running the Beiwe app on subjects' own smartphones, to capture longitudinal data
on symptoms, behaviors, cognition, and physiology across subjects' natural environments. These studies will
be performed across 3.5 years in subjects with early course psychosis and range between 6 to 12 months.
 The broader aim of this research is to understand the systems and processes, both personal and
environmental, which contribute to relapse in early course psychosis. An understanding of the computational
basis of relapse will inform better nosology, allow development of biomarkers of illness that may offer better
targets for biological research, inform development of personalized interventions for psychotic illnesses, and
help support early interventions for schizophrenia.

## Key facts

- **NIH application ID:** 10133145
- **Project number:** 5K23MH116130-04
- **Recipient organization:** BETH ISRAEL DEACONESS MEDICAL CENTER
- **Principal Investigator:** John Torous
- **Activity code:** K23 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2021
- **Award amount:** $191,000
- **Award type:** 5
- **Project period:** 2018-04-01 → 2022-03-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10133145

## Citation

> US National Institutes of Health, RePORTER application 10133145, Digital Phenotyping for Computational Models of Relapse Prediction in Early Course Psychosis (5K23MH116130-04). Retrieved via AI Analytics 2026-05-23 from https://api.ai-analytics.org/grant/nih/10133145. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*