National Institute of Mental Health 5R21MH139049-02

Integrating Loneliness and Social Isolation Insights into Late-Life Suicide Risk Prediction Through the Digital Phenotyping of Electronic Health Records

The project aims to improve late-life suicide risk prediction by analyzing loneliness and social isolation markers in electronic health records using AI and NLP, resulting in better risk assessments and tools for healthcare.

Description

This study addresses the problem of predicting late-life suicide risk among older adults, driven by loneliness and social isolation, by utilizing the latest artificial intelligence and natural language processing methods. Existing risk assessment approaches often neglect these social factors, encapsulated in free text format, and underutilize longitudinal electronic health record (EHR) data. The overarching goal is to enhance suicide risk predictions and improve EHR data use, enabling more accurate assessments, timely interventions, and personalized care for lonely and socially isolated older adults. The project aims to develop and validate a tool that identifies linguistic markers of loneliness and social isolation in clinical notes, integrate these insights into a suicide risk prediction model for older populations, and evaluate the bias and interpretability of AI models to mitigate biases. Outcomes include an interactive dashboard and open-source toolkit for broad application.

View original record