Machine Learning Classification of Psychiatric Data Associated with Compensation Claims for Patient Injuries.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Psychiatric Patient Safety

Background:

• Adverse events are prevalent in healthcare, yet compensation claims in psychiatric treatment are less frequent than in other specialties.
• Common psychiatric patient injury claims involve diagnostic errors, suicide, or inappropriate coercive treatments.
• Understanding and categorizing these claims is crucial for improving patient safety and healthcare quality.

Purpose of the Study:

• To categorize patient injury types within psychiatric compensation claims using machine learning.
• To develop AI-based classification models for these injury categories.
• To achieve binary classification of compensation claim decisions (accepted/declined).

Main Methods:

• Finnish psychiatric evaluations from compensation claims were classified into six categories using AI machine learning.
• A secondary binary classification was performed to distinguish accepted from declined claims.
• Artificial data generation was employed as a preprocessing step to enhance classification accuracy.

Main Results:

• The AI model achieved good classification results for the six injury categories.
• Binary classification of claim decisions was more complex but feasible.
• Preprocessing with artificial data improved classification accuracy to 88% for six classes and 89% for binary classification using random forests.

Conclusions:

• The study successfully demonstrated the feasibility of using machine learning to categorize psychiatric patient injury claims.
• AI-based classification offers a viable method for analyzing compensation claims and potentially improving patient safety in psychiatry.
• The results indicate that AI can reasonably address the defined objectives in analyzing psychiatric compensation claims.