Background: Carbapenem-resistant Enterobacterales (CRE) colonisation at liver transplantation (LT) increases the risk of CRE infection after LT, which impacts on recipients' survival. Colonization status usually becomes evident only near LT. Thus, predictive models can be useful to guide antibiotic prophylaxis in endemic centres.
Aims: This study aimed to identify risk factors for CRE colonisation at LT in order to build a predictive model.
Methods: Retrospective multicentre study including consecutive adult patients who underwent LT, from 2010 to 2019, at two large teaching hospitals. We excluded patients who had CRE infections within 90 days before LT. CRE screening was performed in all patients on the day of LT. Exposure variables were considered within 90 days before LT and included cirrhosis complications, underlying disease, time on the waiting list, MELD and CLIF-SOFA scores, antibiotic use, intensive care unit and hospital stay, and infections. A machine learning model was trained to detect the probability of a patient being colonized with CRE at LT.
Results: A total of 1544 patients were analyzed, 116 (7.5%) patients were colonized by CRE at LT. The median time from CRE isolation to LT was 5 days. Use of antibiotics, hepato-renal syndrome, worst CLIF sofa score, and use of beta-lactam/beta-lactamase inhibitor increased the probability of a patient having pre-LT CRE. The proposed algorithm had a sensitivity of 66% and a specificity of 83% with a negative predictive value of 97%.
Conclusions: We created a model able to predict CRE colonization at LT based on easy-to-obtain features that could guide antibiotic prophylaxis.
Keywords: CLIF-SOFA score; carbapenem-resistance; machine learning; peritonitis prophylaxis; prediction model.
© 2022 The Authors. Transplant Infectious Disease published by Wiley Periodicals LLC.