Comparative Diagnostic Performance of RIPASA, ALVARADO, and AIR Scoring Systems in Acute Appendicitis: A Retrospective Evaluation Using Traditional Metrics and XGBoost Algorithm

Aim: This study aimed to evaluate the diagnostic performance of three clinical scoring systems—RIPASA, ALVARADO, and AIR—in identifying acute appendicitis. By comparing these methods against histopathological findings, the gold standard for diagnosis, the study further explored whether machine learning integration via the XGBoost algorithm could enhance classification accuracy. Material and Methods: A publicly available retrospective dataset containing records of 107 patients, including their RIPASA, ALVARADO, AIR scores, and histopathological diagnoses, was analyzed. Diagnostic performance metrics (accuracy, sensitivity, specificity, PPV, NPV, balanced accuracy) were computed for each scoring system. Additionally, an XGBoost classification model was constructed using the three scores as input variables to predict histopathological outcomes. Five-fold cross-validation and grid search optimization were applied to prevent overfitting and enhance generalizability. Results: RIPASA and AIR scores outperformed the ALVARADO score in both accuracy and sensitivity, with RIPASA demonstrating the most balanced diagnostic profile (accuracy: 86.0%, sensitivity: 87.8%, specificity: 76.5%). AIR achieved the highest sensitivity (97.8%) but had low specificity (29.4%). When combined with XGBoost, the AIR-based model achieved the best overall performance (accuracy: 96.0%, specificity: 100.0%, sensitivity: 90.9%, NPV: 93.3%). Conclusion: The RIPASA and AIR scoring systems demonstrate superior diagnostic capabilities compared to ALVARADO, particularly when combined with machine learning approaches like XGBoost. The integration of scoring systems with data-driven models holds promise for enhancing diagnostic accuracy and developing robust clinical decision support tools in emergency settings.