Anomaly detection to predict relapse risk in schizophrenia.

Area of Science:

• Digital health
• Psychiatry
• Behavioral science

Background:

• Technology integration in clinical care is rapidly expanding, accelerated by the COVID-19 pandemic.
• Smartphone-based digital phenotyping offers a novel approach to monitor subtle, moment-to-moment behavioral and symptomatological changes.

Purpose of the Study:

• To investigate the efficacy of smartphone-based anomaly detection for predicting relapse in serious mental illness.
• To assess the potential clinical utility of analyzing real-time behavioral data collected via smartphones.

Main Methods:

• Utilizing integrated smartphone sensors to collect longitudinal data on user behavior and symptomatology.
• Defining and detecting anomalies as significant deviations from an individual's established baseline.
• Evaluating the performance of anomaly detection in predicting relapse events.

Main Results:

• The study achieved 89% sensitivity in detecting relapse.
• A specificity of 75% was recorded for predicting schizophrenia relapse.
• Demonstrated the potential for individualized analysis of real-time behavioral data.

Conclusions:

• Smartphone-based anomaly detection shows promise for early identification of serious mental illness relapse risk.
• Longitudinal collection and analysis of real-time behavioral data via smartphones have significant clinical utility.
• Further research is needed to enhance specificity and explore clinical integration, including just-in-time adaptive interventions.