Development of a Risk Prediction Model for Carbapenem-resistant<i>Enterobacteriaceae</i>Infection After Liver Transplantation: A Multinational Cohort Study
Maddalena Giannella, Maristela Pinheiro Freire, Matteo Rinaldi, Edson Abdala, Arianna Rubin, Alessandra Mularoni, Salvatore Gruttadauria, Paolo Grossi, Nour Shbaklo, Francesco Tandoi, Alberto Ferrarese, Patrizia Burra, Ruan Fernandes, Luís Fernando Aranha Camargo, Ángel Asensio, Laura Alagna, Alessandra Bandera, Jacques Simkins, Lilian M. Abbo, Márcia Halpern, Evelyne Santana Girão, Maricela Valerio, Patricia Muñóz, Ainhoa Fernández Yunquera, Liran Statlender, Dafna Yahav, Erica Franceschini, Elena Graziano, Maria Cristina Morelli, Matteo Cescon, Pierluigi Viale, Russell E. Lewis, CRECOOLT study group, Michele Bartoletti, Renato Pascale, Caterina Campoli, Simona Coladonato, Francesco Cristini, Fabio Tumietto, Antonio Siniscalchi, Cristiana Laici, Simone Ambretti, Renato Romagnoli, Francesco Giuseppe De Rosa, Antonio Muscatello, Davide Mangioni, Andrea Gori, Barbara Antonelli, Daniele Dondossola, G. Rossi, Federica Invernizzi, Maddalena Peghin, Umberto Cillo, Cristina Mussini, Fabrizio Di Benedetto, Débora Raquel Benedita Terrabuio, Carolina Devite Bittante, Alexandra do Rosário Toniolo, Elizabeth Balbi, José Huygens Parente Garcia, Ignacio Morrás, António Ramos, Ana Fernández‐Cruz, Magdalena Salcedo
Abstract
BACKGROUND: Patients colonized with carbapenem-resistant Enterobacteriaceae (CRE) are at higher risk of developing CRE infection after liver transplantation (LT), with associated high morbidity and mortality. Prediction model for CRE infection after LT among carriers could be useful to target preventive strategies. METHODS: Multinational multicenter cohort study of consecutive adult patients underwent LT and colonized with CRE before or after LT, from January 2010 to December 2017. Risk factors for CRE infection were analyzed by univariate analysis and by Fine-Gray subdistribution hazard model, with death as competing event. A nomogram to predict 30- and 60-day CRE infection risk was created. RESULTS: A total of 840 LT recipients found to be colonized with CRE before (n = 203) or after (n = 637) LT were enrolled. CRE infection was diagnosed in 250 (29.7%) patients within 19 (interquartile range [IQR], 9-42) days after LT. Pre- and post-LT colonization, multisite post-LT colonization, prolonged mechanical ventilation, acute renal injury, and surgical reintervention were retained in the prediction model. Median 30- and 60-day predicted risk was 15% (IQR, 11-24) and 21% (IQR, 15-33), respectively. Discrimination and prediction accuracy for CRE infection was acceptable on derivation (area under the curve [AUC], 74.6; Brier index, 16.3) and bootstrapped validation dataset (AUC, 73.9; Brier index, 16.6). Decision-curve analysis suggested net benefit of model-directed intervention over default strategies (treat all, treat none) when CRE infection probability exceeded 10%. The risk prediction model is freely available as mobile application at https://idbologna.shinyapps.io/CREPostOLTPredictionModel/. CONCLUSIONS: Our clinical prediction tool could enable better targeting interventions for CRE infection after transplant.