Machine-learning-assisted Preoperative Prediction of Pediatric Appendicitis Severity

Area of Science:

• Pediatric Surgery
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Accurate preoperative diagnosis of acute appendicitis in children is crucial for effective management.
• Predicting the severity of appendicitis aids in tailoring treatment strategies and optimizing resource allocation.

Purpose of the Study:

• To evaluate the effectiveness of machine learning (ML) algorithms in improving the preoperative diagnosis of acute appendicitis in children.
• To accurately predict the severity of appendicitis using ML models.

Main Methods:

• Developed an ML pipeline using a retrospective dataset of 1980 children who underwent appendectomy (2014-2021).
• Employed imputation strategies for missing values and upsampling for infrequent classes.
• Tested various combinations of imputation, class balancing, and classification models to predict 5 appendicitis grades.

Main Results:

• The best ML pipeline achieved 70.1% accuracy and 0.77 AUROC for differentiating appendicitis severity grades.
• The model demonstrated superior performance compared to another pediatric appendicitis severity prediction tool (71.4% accuracy, 0.54 AUROC).
• Key predictive variables align with clinical experience and literature.

Conclusions:

• ML models can effectively predict appendicitis severity in children, outperforming existing tools.
• The developed ML model offers a novel approach for preoperative assessment of appendicitis severity.
• Potential for personalized, severity-based treatment and optimized resource allocation in pediatric appendicitis management.