Performance and external validation of prediction models for gangrenous cholecystitis in emergency surgical patients: a systematic review

Background: Gangrenous cholecystitis complicates emergency surgical assessment because routine clinical findings overlap with nongangrenous inflammation. Preoperative prediction offers a basis for risk stratificati n and operative planning. This review evaluated model performance and the strength of external validation. Methods: Two authors searched PubMed, Scopus, Web of Science, and the Cochrane Library through Jan 2026, screened studies, and extracted data. Original multivariable development and validation studies using operative or pathological gangrene outcomes were synthesized narratively. Discrimination, threshold accuracy, calibration, and validation design were examined separately. Results: eleven original articles described clinical equations, bedside scores, imaging approaches, and machine learning. Independent evaluations of established scores yielded areas under the receiver operating characteristic curve of 0.72-0.78. A six-factor clinicoradiological score reported an area of 0.83, although cohort-specific validation reporting required interpretation. A random forest achieved 0.818 in another-centre validation, while a fusion computed tomography model achieved 0.879 and 0.887 in two external cohorts. Same-centre temporal testing was distinguished from geographical validation. Conclusions: External validation supports several approaches, without establishing a universally transferable tool or improved clinical outcomes. Representative multicentre evaluation, transparent calibration, and prospective assessment of model-guided decisions remain priorities for emergency surgical implementation.