# A risk prediction model for medical conflict in emergency departments

## TL;DR

This paper develops a high-accuracy model to predict and prevent medical conflicts in emergency departments based on patient-related factors.

## Contribution

A novel risk prediction model using neural networks and SVM for early warning of medical conflicts in emergency departments.

## Key findings

- The model achieved 96.4% accuracy with high sensitivity and specificity in predicting medical conflicts.
- Key risk factors include waiting time, patient non-compliance, and skepticism toward care.
- SVM validation confirmed the model's robustness with consistent performance on independent testing samples.

## Abstract

This study aims to identify and analyze patient-related risk factors that lead to medical conflict in emergency departments and to build a patient-related risk prediction model, which can serve as a tool for medical professionals to prevent medical conflict in emergency departments.

At present, the research on the countermeasures of medical conflict in emergency departments mainly focuses on post-event coping strategies and rarely addresses the early warning of conflict occurrence.

Through a retrospective analysis of medical conflict events in the emergency pre-examination and rescue areas, 105 conflict cases and 525 non-conflict cases were collected. Univariate analysis, neural network analysis and support vector machine (SVM) were performed to analyze patient-related risk factors in medical conflict, thereby constructing a prediction model.

Neural network analysis yielded an accuracy of 96.4% (sensitivity: 91.2%, specificity: 97.7%) and an area under the ROC curve of 0.966. The order of importance of risk factors included in the prediction model is in descending order of importance: waiting time; patient non-compliance to the treatment process; patient’s skepticism toward care; number of caregivers; medical professionals’ failure to respond to patients’ needs; and history of psychosomatic diseases. The SVM utilizing 10-fold cross-validation demonstrated an accuracy of 96.4% in independent testing samples, with a sensitivity of 91.2% and a specificity of 97.2%.

This model may accurately predict the occurrence of medical conflicts in emergency departments with high accuracy, providing a theoretical basis for preventing medical conflicts.

We have developed a risk prediction model to help emergency medical professionals identify factors that may contribute to preventing medical conflict, reducing conflict events, and enhancing the doctor-patient relationship.