External Validation of Adult Prediction Models for Complicated Appendicitis After Contrast-Enhanced Computed...

Area of Science:

• Surgical diagnostics
• Medical imaging analysis
• Predictive modeling in medicine

Background:

• Accurate preoperative distinction between uncomplicated and complicated appendicitis is crucial for guiding treatment in adults.
• Existing prediction models for appendicitis require further validation.
• Divergent treatment pathways, including non-operative management, necessitate reliable diagnostic tools.

Purpose of the Study:

• To externally validate existing prediction models for complicated appendicitis in an adult cohort.
• To compare the diagnostic performance of the Atema 2015, Appendicitis Severity Index, and Mori 2024 models.
• To explore a modified SAS 2.0 model due to data limitations.

Main Methods:

• A single-center validation study included 496 adults undergoing appendectomy after contrast-enhanced computed tomography.
• The primary outcome was complicated appendicitis.
• Discrimination was assessed using the area under the receiver operating characteristic curve (AUC) and threshold metrics.
• Atemo 2015, Appendicitis Severity Index, and Mori 2024 were formally validated; a modified SAS 2.0 was explored.

Main Results:

• Mori 2024 exhibited the highest discrimination (AUC 0.783), followed by Atema 2015 (AUC 0.760) and the Appendicitis Severity Index (AUC 0.731).
• The Appendicitis Severity Index offered high specificity (98.6%) but low sensitivity (22.0%).
• Modified SAS 2.0 achieved an apparent AUC of 0.850 and an optimism-corrected AUC of 0.833.

Conclusions:

• Among the formally validated models, Mori 2024 performed best in this cohort.
• The Appendicitis Severity Index was highly specific but lacked sensitivity for complicated appendicitis.
• No single model evaluated should be used in isolation to exclude complicated appendicitis in this clinical setting.