A Predictive Model for Estimating Risk of Harm and Aggression in Inpatient Mental Health Clinics

Area of Science:

• Psychiatry
• Clinical Psychology
• Health Informatics

Background:

• Serious mental illness is a significant predictor of aggression and violence.
• Accurate prediction of inpatient aggression is crucial for patient and staff safety.
• Existing methods for predicting harm risk in psychiatric settings require enhancement.

Purpose of the Study:

• To develop and validate a predictive algorithm for inpatient aggression with potential for harm.
• To identify key risk factors associated with harmful aggressive incidents in psychiatric inpatients.
• To evaluate the performance of the developed predictive model.

Main Methods:

• Retrospective analysis of inpatient data from St. Joseph's Healthcare Hamilton (2016-2017).
• Development of a predictive model using identified risk factors for harm.
• Evaluation of the model's accuracy, specificity, and sensitivity.

Main Results:

• The predictive model demonstrated an overall accuracy of 75%.
• Specificity for identifying non-harmful incidents was high at 91.85%.
• Sensitivity for detecting harmful incidents was 28.57%.

Conclusions:

• The developed algorithm represents a foundational step towards predicting inpatient harm risk.
• Further refinement is needed to improve sensitivity in identifying harmful aggression.
• The model offers potential for enhanced safety and care in inpatient psychiatric settings.