Bile Over Blood: A Predictive Model for Guiding Antimicrobial Therapy When Blood Cultures Miss the Pathogen in Biliary Infections

Background Blood culture (BC) is considered the gold standard for diagnosing bacteremia; however, its sensitivity is notably limited in cases of acute biliary tract infections, often resulting in under-detection. Despite this, the patterns and contributing factors leading to such under-detection have not been systematically investigated. This study aims to quantify the discrepancies in pathogen profiles between bile and paired blood cultures, and to identify pathogen-specific under-detection rates and independent predictors of blood culture leakage. Methods This retrospective study included 398 patients with radiologically confirmed biliary obstruction who underwent concurrent bile and blood culture collection between January 2017 and December 2024. Patients were categorized into four groups based on culture results: leakage (bile+/blood − OR bile+/blood + without identical pathogen) non-leakage (bile+/blood + with ≥ 1 identical pathogen), double-negative, and blood-only-positive. Multivariable logistic regression was used to identify independent factors associated with blood culture leakage. Results The positivity rate of bile cultures (83.67%, 333/398) was significantly higher than that of blood cultures (36.18%, 144/398) ( P  < 0.001). The overall leakage rate was 70.27% (234/333). Significant differences in leakage rates were observed among different pathogens ( P  < 0.001), with Enterococcus faecalis exhibiting the highest leakage rate (94.23%), while Escherichia coli (31.93%) and Klebsiella pneumoniae (25.88%) showed higher concordance. Multivariable analysis identified fever (aOR = 0.45, 95% CI: 0.26–0.79, P  = 0.005), presence of E. coli in bile (aOR = 0.50, 95% CI: 0.30–0.82, P  = 0.007), and polymicrobial infection (aOR = 0.50, 95% CI: 0.30–0.82, P  = 0.006) as protective factors, whereas antibiotic use (aOR = 2.08, 95% CI: 1.23–3.52, P  = 0.006) was an independent risk factor. The predictive model exhibited moderate discriminative capacity with an AUC of 0.693. Conclusion This study confirms that the under-detection of blood cultures in biliary tract infections is highly pathogen-specific, with E. faecalis being most frequently missed. A clinical prediction model based on fever, antibiotic use, E. coli colonization, and infection complexity was developed. These findings emphasize that in afebrile patients, those receiving antibiotic therapy, or those with Enterococcus -dominated bile cultures, clinical decision-making should rely more heavily on bile culture