Clinical Note-Extracted Psychosocial Factors for Predicting Suicide Attempt Among ED Patients With Suicidal Ideation

Area of Science:

• Emergency Medicine
• Psychiatry
• Data Science

Background:

• The Joint Commission recommends universal suicide screening in emergency departments (EDs).
• Current suicide risk prediction models using clinical data have limited performance.
• The role of psychosocial information in enhancing predictive accuracy is understudied.

Purpose of the Study:

• To determine if adding psychosocial factors to clinical data improves suicide attempt (SA) prediction.
• To evaluate the performance of models using clinical data alone, psychosocial factors alone, and a combination of both.

Main Methods:

• Retrospective analysis of 4661 ED patients presenting with suicidal ideation (SI).
• Electronic health record data was used to extract clinical data (Vanderbilt Suicide Attempt and Ideation Likelihood [VSAIL] score) and psychosocial factors (homelessness, financial insecurity, chronic stress, social isolation, loneliness, adverse childhood experiences).
• Cox proportional hazards regression model was employed to assess the predictive performance (AUROC, AUPRC, PPV) of different model combinations.

Main Results:

• Models combining VSAIL and psychosocial factors showed significantly higher predictive performance (AUROC, AUPRC, PPV) compared to VSAIL alone.
• Psychosocial factors significantly improved prediction accuracy across both Vanderbilt University Hospital and Regional Health Systems datasets.
• Chronic stress was identified as the most significant predictor of SA (β = 0.643, P < .001).

Conclusions:

• Augmenting clinical data-based suicide risk prediction with psychosocial factors extracted from clinical notes significantly enhances predictive performance.
• These findings support the use of psychosocial factors for improved risk stratification in ED patients.
• Targeted interventions, particularly those addressing chronic stress, can be more effectively implemented based on enhanced risk assessment.