The relationship between the Alvarado risk group and final histopathological findings in acute appendicitis, with internally validated machine-learning models for complicated appendicitis and intraoperative abscess

The Alvarado score is widely used in acute appendicitis, but its ability to identify complicated disease is limited. In operated adults, we examined the relationship between the Alvarado risk group and histopathological severity and whether machine learning (ML) models combining it with routine preoperative data improve the prediction of complicated appendicitis and intraoperative abscess. We retrospectively reviewed 420 consecutive adults who underwent appendectomy for histopathologically confirmed appendicitis at a single center. Predictors were age, sex, leukocyte count, neutrophil count, C-reactive protein (CRP) and the Alvarado risk group (scores 5–6, 7–8, 9–10). Outcomes were complicated appendicitis (histopathological grade 3–4; primary) and intraoperative abscess (secondary). A random forest (main model), nine other algorithms and an ensemble were evaluated with 5-fold nested cross-validation repeated five times, averaging each patient’s out-of-fold predictions. AUROCs were compared with the Alvarado risk group using the DeLong test with Holm correction; calibration and decision curves were assessed. Complicated appendicitis occurred in 104 patients (24.8%) and intraoperative abscess in 90 (21.4%). The Alvarado risk group was associated with histopathological grade (χ²=79.5; p < 0.001; odds ratio for complicated appendicitis 14.3, 95% CI 6.6–31.0, group 3 vs. group 1). For complicated appendicitis, the main model (AUROC 0.759, 95% CI 0.707–0.810) did not differ significantly from the Alvarado risk group (0.725, 0.673–0.775; Holm-adjusted p = 0.089). For abscess, the main model showed higher discrimination (0.808 vs. 0.739; ΔAUROC 0.069, 95% CI 0.013–0.129; Holm-adjusted p = 0.043), although the difference reached p < 0.05 in only two of five repetitions. The other algorithms performed similarly, and none was significantly better than the Alvarado risk group after correction. After within-fold isotonic recalibration, calibration slopes were 0.85 and 0.86; decision curves showed no consistent net-benefit advantage over the Alvarado risk group. In operated adults, the Alvarado risk group was significantly associated with histopathological severity but discriminated complicated disease only moderately. Adding routine laboratory variables did not improve the prediction of complicated appendicitis and yielded a modest, marginally significant gain for intraoperative abscess. These internally validated findings are hypothesis-generating and do not support clinical use without external validation.