Implementing Machine Learning Models for Suicide Risk Prediction in Clinical Practice: Focus Group Study With...

Area of Science:

• Clinical Informatics
• Artificial Intelligence in Healthcare
• Mental Health Research

Background:

• Machine learning models for suicide risk prediction using electronic health record data are proliferating.
• Clinical implementation and utility of these models remain largely unexplored.
• Stakeholder partnership, especially with frontline providers, is crucial for successful deployment.

Purpose of the Study:

• To inform the deployment of suicide risk prediction models in clinical practice.
• To understand current suicide risk assessment and management practices.
• To identify provider perspectives, barriers, facilitators, and recommendations for automated suicide risk prediction tools.

Main Methods:

• Conducted 10 focus groups with 40 healthcare providers across various departments.
• Transcribed and coded audio recordings for recurrent themes.
• Utilized consensus meetings for discrepancy resolution.

Main Results:

• Providers expressed dissatisfaction with current suicide risk assessment tools.
• General attitudes towards automated suicide risk prediction models were positive.
• Key facilitators included presenting specific risk factors and developing systematic workflows; barriers included liability and alert fatigue.

Conclusions:

• Providers are open to machine learning-based suicide risk prediction systems despite dissatisfaction with current methods.
• Concerns regarding barriers and facilitators for implementation were raised.
• Future efforts should incorporate systematic qualitative feedback from diverse stakeholders.