Development of a machine learning model for predicting compulsory psychiatric care using clinical notes.

Area of Science:

• Psychiatry
• Machine Learning
• Healthcare Informatics

Background:

• Early identification of patients at high risk for compulsory psychiatric care is crucial for timely intervention.
• Proactive measures can potentially avert the necessity of compulsory psychiatric care.

Purpose of the Study:

• To develop a machine-learning algorithm for the early prediction of compulsory psychiatric care.
• To enable timely interventions and preventive strategies for at-risk patients.

Main Methods:

• Utilized known risk factors and clinical notes from electronic health records.
• Developed a machine-learning model using data from a Dutch mental health institution.

Main Results:

• The predictive model achieved a precision of 0.91.
• Successfully identified over 50% of compulsory care trajectories 60 days prior.
• Previous compulsory care episodes were the most significant predictor; incorporating clinical notes improved performance.

Conclusions:

• Predictive modeling for adverse psychiatric events shows promise for clinical practice.
• Understanding model function and input variables is key to enhancing clinical relevance.