Study on medical dispute prediction model and its clinical-application effectiveness based on machine learning

Area of Science:

• Medical Informatics
• Public Health
• Machine Learning

Background:

• Medical disputes pose a global public health challenge, necessitating innovative solutions.
• Increasing attention is being paid to understanding and mitigating medical disputes.

Purpose of the Study:

• To develop a machine learning (ML) model for predicting medical disputes.
• To evaluate the clinical utility of the ML model in reducing medical dispute occurrence.

Main Methods:

• A retrospective study analyzed medical disputes from 2019-2021, with a 1:1 control group.
• Univariate feature selection identified 11 key factors influencing disputes.
• Six ML models were trained and evaluated using metrics like AUC, sensitivity, and accuracy, with DCA for clinical utility.

Main Results:

• The random forest model demonstrated superior performance with an AUC of 0.945, sensitivity of 0.887, and accuracy of 0.887.
• Key predictors identified include inpatient department, hospitalization expenses, and discharge type.
• Decision curve analysis confirmed the model's significant clinical benefits.

Conclusions:

• Inpatient department, hospitalization expenses, and discharge type are primary drivers of medical disputes.
• The random forest model shows strong potential for dispute prediction and clinical application.
• This ML approach is recommended for promoting offline dispute prediction and prevention.