Diagnostic Accuracy of a Machine Learning-Derived Appendicitis Score in Children: A Multicenter Validation Study

Area of Science:

• Pediatric Medicine
• Medical Informatics
• Artificial Intelligence in Healthcare

Background:

• Accurate diagnosis of acute appendicitis in children is challenging due to varied symptoms and limitations of current scoring systems.
• Previous machine learning (ML) studies for appendicitis diagnosis were often limited by small sample sizes, single-center data, and lack of external validation.

Purpose of the Study:

• To develop and validate a machine learning-based diagnostic model for pediatric appendicitis using routinely available clinical and hematological parameters.
• To compare the performance of the ML model against traditional scoring systems like the Pediatric Appendicitis Score (PAS), Alvarado, and Appendicitis Inflammatory Response Score (AIRS).

Main Methods:

• A prospective, multicenter study involving 8586 pediatric patients for model development.
• External validation was performed on a separate, prospectively collected cohort of 3000 patients.
• The Random Forest algorithm was utilized, and diagnostic accuracy, sensitivity, specificity, and Area Under Curve (AUC) were evaluated.

Main Results:

• The ML model demonstrated superior performance compared to traditional clinical scores in both development and validation cohorts.
• In the external validation set, the Random Forest model achieved an AUC of 0.996, accuracy of 0.992, sensitivity of 0.998, and specificity of 0.993.
• Key predictors identified by feature-importance analysis included white blood cell count, red blood cell count, and mean platelet volume.

Conclusions:

• A machine learning scoring system utilizing accessible data significantly enhances the diagnosis of pediatric appendicitis.
• The developed model exhibits high accuracy and clinical interpretability, with the potential to minimize diagnostic delays and reduce the need for unnecessary imaging in children.
• This large-scale, prospectively validated study supports the clinical utility of ML in improving pediatric appendicitis diagnosis.