Abstract Introduction Epidemiological surveillance of surgical site infection (SSI) is labor-intensive and costly. Transitioning to automated surveillance can improve its efficiency. Methods Retrospective cohort study in colorectal surgery. SSIs were classified according to CDC-NHSN criteria: superficial incisional (SSI-S), deep incisional (SSI-D), and organ/space (SSI-O/S). A predictive alert algorithm model was established using K-fold cross-validation, based on a logistic regression model with elastic-net regularization. The model was used to select the most relevant variables and manage correlated predictors. The postoperative alerts selected were: microbiology, abdominal CT, antibiotic administration, 7 days of hospitalization, readmission, reoperation, and mortality. The diagnostic accuracy of the algorithm was compared with manual surveillance. Univariate and bivariate analyses were performed for each detected alert, and sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), AUC, and ROC curves were calculated. Results A total of 1,213 patients were analyzed between 2010 and 2023 (1,085 colon surgeries and 128 rectal surgeries). Manual surveillance identified an SSI rate of 11.2% (SSI-S 3.1%, SSI-D 1.2%, SSI-O/S 6.8%). For overall SSI detection, the algorithm yielded a sensitivity of 72%, specificity 88%, PPV 42%, NPV 96%, and AUC 85.9 (82.2–89.6). For SSI-S, SSI-D, and SSI-O/S, the algorithm achieved sensitivities of 82%, 87%, and 83%; specificities of 46%, 70%, and 91%; PPVs of 46%, 36%, and 41%; NPVs of 99%, 100%, and 99%, respectively; and AUCs of 57.1, 80.5, and 91.9. Conclusions The algorithm showed good accuracy for detecting overall SSI, SSI-D, and SSI-O/S, supporting its use for semi-automated surveillance of colorectal SSI.
Badia et al. (2026) studied this question.