Validation of a carbapenem-resistant Enterobacteriaceae colonization risk prediction model: A retrospective cohort...

To assess the external validity of a carbapenem-resistant Enterobacteriaceae colonization risk prediction model (CREP-model).

This retrospective cohort study included 414 patients admitted to the intensive care units of a tertiary hospital from November 1, 2017 to May 31, 2018. Data were collected via medical record review, and we analyzed the performance of the CREP-model by assessment of calibration, discrimination, clinical usefulness, and uniformity-of-fit.

The validation subjects showed differences in age, Acute Physiology and Chronic Health Evaluation II score, mechanical ventilation days, and carbapenem-resistant Enterobacteriaceae colonization rate from those of the CREP-model development subjects. The calibration-in-the-large was 0.069 (95% confidence interval [CI], 0.065-0.074), and calibration slope was 1.114 (95% CI, 1.091-1.136). The area under the receiver operating characteristic curve was 0.883 (95% CI, 0.838-0.928). At the predicted risk of 0.25, the sensitivity, specificity, and correct classification rates were 81.3%, 79.8%, and 80.0%, respectively, and the net benefit according to the model was 0.035 with 64 fewer false-positive results per 100 patients. The calibration, discrimination, and clinical usefulness showed similar results among subjects stratified according to sex, age group, medical department, and admission source.

The CREP-model showed good performance in the validation sample; therefore, we recommend introducing the CREP-model into intensive care units of tertiary hospitals to improve decision-making.