Development and validation of a nomogram to predict the risk of vancomycin-related acute kidney injury in critical care patients.

Abstract

Vancomycin-associated acute kidney injury (AKI) leads to underestimated morbidity in the intensive care unit (ICU). It is significantly important to predict its occurrence in advance. However, risk factors and nomograms to predict this AKI are limited. This was a retrospective analysis of two databases. A total of 1,959 patients diagnosed with AKI and treated with vancomycin were enrolled from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. According to the 7:3 ratio, the training set (n = 1,372) and the internal validation set (n = 587) were randomly allocated. The external validation set included 211 patients from the eICU Collaborative Research Database (eICU). Next, to screen potential variables, the least absolute shrinkage and selection operator (LASSO) regression was utilized. Subsequently, the nomogram was developed by the variables of the selected results in the multivariable logistic regression. Finally, discrimination, calibration, and clinical utility were evaluated to validate the nomogram. The constructed nomogram showed fine discrimination in the training set (area under the receiver operator characteristic curve [AUC] = 0.791; 95% confidence interval [CI]: 0.758-0.823), internal validation set (AUC = 0.793; 95% CI: 0.742-0.844), and external validation set (AUC = 0.755; 95% CI: 0.663-0.847). Moreover, it also well demonstrated calibration and clinical utility. The significant improvement (P < 0.001) in net reclassification improvement (NRI) and integrated differentiation improvement (IDI) confirmed that the predictive model outperformed others. This established nomogram indicated promising performance in determining individual AKI risk of vancomycin-treated critical care patients, which will be beneficial in making clinical decisions.