Development and validation of a nomogram-based risk prediction model for carbapenem-resistant Klebsiella pneumoniae in hospitalized patients

Abstract

Carbapenem-resistant Klebsiella pneumoniae (CRKP) poses one of the major challenges in clinical anti-infective therapy worldwide. This retrospective cohort study at a tertiary general hospital in Wuhan aimed to identify risk factors for hospital-acquired CRKP infections among 1,113 patients. All participants were aged 18 years and above, and had confirmed positive cultures for KP isolated within 48 hours post-hospitalization. Independent risk factors were identified using LASSO logistic regression and incorporated into a predictive nomogram. The factors included in the nomogram were prior carbapenem exposure, prior β-lactams-β-lactamase inhibitor combination (BLBLI) exposure, prior intensive care unit (ICU) stay, and prior mechanical ventilation. The areas under the receiver operating characteristic curve (AUC) for the nomogram were 0.793 in the training group (70% of patients) and 0.788 in the validation group (30% of patients), demonstrating its discriminatory power and predictive accuracy. The P values for the Hosmer-Lemeshow test were 0.333 and 0.684, indicating good calibration. The clinical utility of the nomogram was further supported by decision curve analysis (DCA) and clinical impact curve (CIC), demonstrating its potential to guide clinical decision-making. Our retrospective analysis identified key risk factors for CRKP infection and developed a nomogram that could effectively predict CRKP infections in hospitalized patients. Although the single-center nature of this study limits generalizability, the nomogram provides a foundation for future prospective, multicenter validations.IMPORTANCEWe established a nomogram scoring system that incorporates four key risk factors: prior carbapenem exposure, prior β-lactams-β-lactamase inhibitor combination (BLBLI) exposure, prior intensive care unit (ICU) stay, and prior mechanical ventilation. This nomogram demonstrated strong discriminatory power, excellent calibration, and significant clinical utility. This study highlights the critical risk factors associated with hospital-acquired carbapenem-resistant Klebsiella pneumoniae (CRKP) infections, providing valuable insights for clinicians to identify high-risk patients.

Keywords: Klebsiella pneumoniae; carbapenem resistance; nomogram; risk factor.

Fig 1

(A) LASSO coefficient profiles of the 35 risk factors. Each curve in the figure presents the change of each variable in coefficient. (B) Tenfold cross-cross validation fitting and then selecting the mode.

Fig 2

The nomogram for predicting the risk of CRKP infection including ICU stay, mechanical ventilation, exposure to carbapenems, and exposure to BLBLIs. The probability of CRKP infection was calculated by adding the scores for each variable. CRKP, carbapenem-resistant Klebsiella pneumoniae; ICU, intensive care unit.

Fig 3

Receiver operating characteristic (ROC) curve, calibration plots, decision curve analysis (DCA), and clinical impact curves (CIC). (A) ROC of the predictive model in the training and validation cohorts. (B) Calibration plots of the predictive model in the training and validation cohorts. (C) DCA of the predictive model in the training and validation cohorts. (D) CIC of the predictive model in the training and validation cohorts. The black dashed lines indicate the net benefit when all hospitalized patients are assumed not to have developed CRKP infections and are not treated. The blue dashed lines indicate the net benefit when all hospitalized patients are assumed to have developed CRKP infections and have received treatment. The purple curves (number of high risk) indicate the number of people classified as positive (high risk) by the predictive model at each threshold probability. The yellow curves (number of high risk with event) show the number of true positives for each threshold probability. CRKP, carbapenem-resistant Klebsiella pneumoniae.