OP-4 DEVELOPMENT AND EXTERNAL VALIDATION OF A MODEL TO PREDICT MULTI-DRUG RESISTANT BACTERIAL INFECTIONS IN PATIENTS WITH CIRRHOSIS

Introduction and Objectives

Empirical antibiotic treatment for suspected infections in cirrhosis is crucial. This study aimed to develop and validate a model to predict the probability of infections by multi-drug resistant organisms (MDRO) in patients with cirrhosis.

Materials and Methods

Cross-sectional study (NCT05641025) of in-patients with bacterial infections from two prospective studies. A global transcontinental study was used for model development and internal validation (n = 1,302), and a study from Argentina and Uruguay (n=472) was used for external validation. Infection by MDROs was defined as an infection caused by at least one bacteria with acquired resistance to at least one antibiotic of three different families. A stepwise selection process was used for model development and bootstrapping for internal validation.

Results

The prevalence of infection by MDROs was 19% in the development and 22% in the external validation dataset. Most frequent infections were spontaneous bacterial peritonitis (SBP) and urinary tract infection (UTI). Half of the infections were community-acquired, and half were equally distributed among healthcare-associated and nosocomial origin. The model predictors are shown in the figure. Very good calibration was achieved in internal and external validation (Figure). Discrimination was adequate: area under the receiver operating characteristic curve (AUROC) of 0.73 (95% CI: 0.69 - 0.76) in internal validation and 0.67 (95% CI: 0.62 - 0.74) in external validation. When applying a probability cut-off point of 5% to the external dataset, a negative predictive value (NPV) of 93% (95% CI: 84% - 98%) was observed.

Conclusions

This easy-to-implement model achieved adequate performance for predicting infections by MDROs in patients with cirrhosis, offering costless bedside individualized risk estimates that might improve the selection of empiric antibiotics. Its high NPV suggests that it could be used as a rule-out tool, particularly in patients at higher risk of infection by MDROs, reducing the use of broad- spectrum antibiotics.