Clinical, laboratory, and computed tomography predictors of complicated acute appendicitis in adults selected for surgery: An internally validated cohort study

Distinguishing complicated from uncomplicated acute appendicitis before surgery can support perioperative risk stratification among adults selected for appendectomy. We conducted an observational cohort study using de-identified data from an ongoing doctoral research program at a tertiary hospital in Ho Chi Minh City, Vietnam. Adults with contrast-enhanced computed tomography (CT)-confirmed acute appendicitis who were selected for surgery and subsequently underwent appendectomy were included. In the parent study, 2291 emergency appendectomies were screened, 849 patients consented to participate, and 496 patients formed the analytic cohort after prespecified exclusions. The primary analysis was a prespecified complete-case bias-reduced logistic regression model; discrimination was summarized by the area under the receiver operating characteristic curve (AUC) and internally validated by bootstrap resampling. Sensitivity analyses included multiple imputation and alternative coding of fat stranding. Of 496 patients, 200 (40.3%) had complicated appendicitis. The primary complete-case model included 473 patients with 192 events. Independent predictors were older age (adjusted odds ratio [aOR]: 1.027 per year, 95% confidence interval [CI]: 1.011-1.044), guarding (aOR: 3.755, 95% CI: 1.972-7.149), rebound tenderness (aOR: 2.248, 95% CI: 1.031-4.901), higher C-reactive protein (aOR: 1.378 per doubling of [CRP + 1], 95% CI: 1.197-1.585), larger appendiceal diameter (aOR: 1.123 per mm, 95% CI: 1.026-1.230), periappendiceal free fluid (aOR: 2.452, 95% CI: 1.122-5.359), and appendicolith (aOR: 2.084, 95% CI: 1.220-3.559). The apparent AUC was 0.864 (95% CI: 0.831-0.897), with an optimism-corrected AUC of 0.840. Calibration analysis showed a Brier score of 0.144, calibration intercept of -0.005, and calibration slope of 1.053. Multiple imputation and alternative fat-stranding coding were directionally consistent. These findings may support CT-integrated preoperative risk stratification among adults selected for surgery. The model should not be used as a stand-alone bedside decision rule, and external validation is required before routine clinical use.