Evaluation of the Efficacy of a Nomogram to Predict Multidrug-Resistant Pulmonary Infections Based on Data from Neurosurgery Ward Patients

This study aimed to construct and evaluate a nomogram based on data from neurosurgery ward patients to predict the probability of multidrug-resistant (MDR) pneumonia occurrence. We retrospectively collected clinical data, early laboratory test results, and physician prescriptions for 35 variables from patients. Univariate and stepwise regression analyses were used to screen variables to determine predictive factors, and a nomogram was constructed in the training group based on the results of the logistic regression model. Using the validation group, discrimination, calibration, and clinical applicability were assessed based on the receiver operating characteristic curve, calibration curve, and decision curve analysis (DCA). Among 3397 patients admitted to the neurosurgery ward from January 1, 2021, to September 30, 2024, 438 patients had pulmonary infections, including 208 patients with MDR pneumonia and 230 patients with non-MDR pneumonia. We randomly divided these patients into a training group (70%, N = 307) and a validation group (30%, N = 131). The nomogram consisted of only six predictive factors (creatinine clearance rate (CCR)≥130 mL/min/1.73 m2, the Day 1 neutrophil-to-lymphocyte ratio (NLR), albumin≤30 g/L, hemoglobin, combination of antibacterial drugs, and tracheostomy), which demonstrated significantly higher sensitivity and specificity in the early identification of MDR pneumonia (AUC of the training group = 0.816 (95% CI: 0.760-0.862), AUC of the validation group = 0.797 (95% CI: 0.720-0.874)) and good calibration. DCA confirmed the clinical applicability of this nomogram. We propose for the first time that augmented renal clearance (ARC) is an independent risk factor for the occurrence of MDR pneumonia in neurosurgical patients. Moreover, we successfully established a convenient prediction model that consists of six prediction factors, which can assist neurosurgeons in making early predictions of the incidence of MDR pneumonia.