Development and validation of nomograms for personalized risk assessment of ICU-acquired carbapenem-resistant...

Carbapenem-resistant Enterobacterales (CRE) pose a grave and escalating threat within the intensive care unit (ICU), associated with alarmingly high mortality rates, prolonged hospitalization, and substantial healthcare costs.

The aim of this study was to develop and validate nomograms predicting ICU-acquired CRE infection and/or colonization.

In this prospective, multi-centre cohort study with active surveillance, 2149 ICU patients were enrolled and subjected to comprehensive monitoring, including systematic screening for clinical infections and rectal swab collection. Predictor selection for ICU-acquired CRE infection and/or colonization was performed via least absolute shrinkage and selection operator regression and the Akaike information criterion to derive an optimal predictor pool. The resulting nomogram was constructed in a training cohort (6 ICUs) and externally validated in a distinct cohort (4 ICUs), with evaluation encompassing discriminative ability, calibration, and clinical benefit.

The incidence of ICU-acquired CRE infection/colonization was 14.5%. The primary model for predicting CRE infection/colonization outcome incorporated 11 independent predictors and demonstrated robust and generalizable performance. It achieved areas under the curve (AUC) of 0.757 in the training set and 0.764 upon external validation. The model was well-calibrated (calibration slope: 0.874) and provided a net clinical benefit within a defined risk threshold range of 12%-55%. Specific models for predicting CRE infection and colonization alone also showed good discriminative ability, with external validation AUCs of 0.794 and 0.730, respectively.

This study presents nomograms for identifying high-risk patients for ICU-acquired CRE, enabling personalized risk assessment to facilitate targeted screening and preventive measures.