External Validation of Adult Prediction Models for Complicated Appendicitis After Contrast-Enhanced Computed Tomography: A Single-Centre Study.
Source: PubMed, NCBI / U.S. National Library of Medicine
Preoperative distinction between uncomplicated and complicated appendicitis is important when treatment pathways diverge, including non-operative management for adults. Several adult prediction models exist, but validation remains limited. We performed a single-centre validation study using a dataset from Nhan Dan Gia Dinh Hospital. Adults undergoing appendectomy after contrast-enhanced computed tomography were included. Complicated appendicitis was the primary outcome. Atema 2015, the Appendicitis Severity Index and Mori 2024 were prespecified for formal external validation. Because the pain score was unavailable, SAS 2.0 was not formally validated; instead, a modified SAS 2.0 using a surrogate abdominal examination variable was explored. Discrimination was assessed with area under the receiver operating characteristic curve, and threshold metrics at cut-offs. The cohort comprised 496 adults; 200 (40.3%) had complicated appendicitis. Mori 2024 showed the highest discrimination (area under the receiver operating characteristic curve 0.783, 95% confidence interval: 0.741-0.826), followed by Atema 2015 (0.760, 0.715-0.804) and the Appendicitis Severity Index (0.731, 0.686-0.776). Mori outperformed the Appendicitis Severity Index (p = 0.0146). At published cut-offs, sensitivity/specificity were 51.6%/89.7% for Atema, 22.0%/98.6% for the Appendicitis Severity Index and 75.0%/65.4% for Mori. In 474 complete cases, modified SAS 2.0 achieved an apparent area under the
Abstract
Preoperative distinction between uncomplicated and complicated appendicitis is important when treatment pathways diverge, including non-operative management for adults. Several adult prediction models exist, but validation remains limited. We performed a single-centre validation study using a dataset from Nhan Dan Gia Dinh Hospital. Adults undergoing appendectomy after contrast-enhanced computed tomography were included. Complicated appendicitis was the primary outcome. Atema 2015, the Appendicitis Severity Index and Mori 2024 were prespecified for formal external validation. Because the pain score was unavailable, SAS 2.0 was not formally validated; instead, a modified SAS 2.0 using a surrogate abdominal examination variable was explored. Discrimination was assessed with area under the receiver operating characteristic curve, and threshold metrics at cut-offs. The cohort comprised 496 adults; 200 (40.3%) had complicated appendicitis. Mori 2024 showed the highest discrimination (area under the receiver operating characteristic curve 0.783, 95% confidence interval: 0.741-0.826), followed by Atema 2015 (0.760, 0.715-0.804) and the Appendicitis Severity Index (0.731, 0.686-0.776). Mori outperformed the Appendicitis Severity Index (p = 0.0146). At published cut-offs, sensitivity/specificity were 51.6%/89.7% for Atema, 22.0%/98.6% for the Appendicitis Severity Index and 75.0%/65.4% for Mori. In 474 complete cases, modified SAS 2.0 achieved an apparent area under the receiver operating characteristic curve of 0.850 and optimism-corrected area under the curve of 0.833. In this restricted-spectrum cohort, Mori performed best among the formal models, whereas the Appendicitis Severity Index was most specific but poorly sensitive. None of the formal models should be used alone to exclude complicated appendicitis in this setting.