Background Acute cholangitis, a severe biliary infection that is frequently caused by gram-negative pathogens, presents health challenges owing to the increasing prevalence of antimicrobial resistance, particularly among extended-spectrum beta–lactamase (ESBL)-producing bacteria. Accurate risk stratification of patients with extended-spectrum beta-lactamase–producing bacteria is crucial for optimizing antimicrobial therapies. Aims This study aimed to develop and internally validate a risk stratification scoring system for identifying ESBL-producing bacteria in patients with acute cholangitis, with the goal of supporting clinical risk stratification. Methods This retrospective cohort analysis included adult patients with acute cholangitis admitted between 2019 and 2023. Patients were excluded if they had incomplete medical records, missing microbiological data, or non-adherence to the Tokyo Guidelines 2018 diagnostic criteria. Predictors of positivity for ESBL-producing bacteria were identified using multivariate logistic regression and integrated into a scoring system, and the performance metrics were evaluated. Results A total of 303 patients with positive blood or bile cultures were included in the analysis, of whom 111 (36.6%) had ESBL-positive cholangitis. Independent predictors of positivity for ESBL-producing bacteria included prior antibiotic treatment (3.5 points), chills with rigors (2.5 points), alanine aminotransferase levels 300 U/L (2.0 point). At cutoff scores of 7.5 and 11, the scoring system demonstrated sensitivity of 56% and 16%, specificity of 75% and 98%, positive predictive values of 56% and 82%, and overall accuracy of 68% and 70%, respectively. Conclusions This preliminary scoring system demonstrated high specificity at higher cutoffs, supporting its use as a rule-in tool for identifying patients at high risk of ESBL-producing bacteria in acute cholangitis. However, its moderate sensitivity limits its role as a rule-out strategy. This preliminary internally validated model should not be used to guide routine clinical decision-making without external multicenter validation, and any application must consider local resistance patterns and patient context.
Sornsenee et al. (Tue,) studied this question.