Evaluating the efficiency of a nomogram based on the data of neurosurgical intensive care unit patients to predict pulmonary infection of multidrug-resistant Acinetobacter baumannii

Abstract

BackgroundPulmonary infection caused by multidrug-resistant Acinetobacter baumannii (MDR-AB) is a common and serious complication after brain injury. There are no definitive methods for its prediction and it is usually accompanied by a poor prognosis. This study aimed to construct and evaluate a nomogram based on patient data from the neurosurgical intensive care unit (NSICU) to predict the probability of MDR-AB pulmonary infection.MethodsIn this study, we retrospectively collected patient clinical profiles, early laboratory test results, and doctors’ prescriptions (66 variables). Univariate and backward stepwise regression analyses were used to screen the variables to identify predictors, and a nomogram was built in the primary cohort based on the results of a logistic regression model. Discriminatory validity, calibration validity, and clinical utility were evaluated using validation cohort 1 based on receiver operating characteristic curves, calibration curves, and decision curve analysis (DCA). For external validation based on predictors, we prospectively collected information from patients as validation cohort 2.ResultsAmong 2115 patients admitted to the NSICU between December 1, 2019, and December 31, 2021, 217 were eligible for the study, including 102 patients with MDR-AB infections (102 cases) and 115 patients with other bacterial infections (115 cases). We randomly categorized the patients into the primary cohort (70%, N=152) and validation cohort 1 (30%, N=65). Validation cohort 2 consisted of 24 patients admitted to the NSICU between January 1, 2022, and March 31, 2022, whose clinical information was prospectively collected according to predictors. The nomogram, consisting of only six predictors (age, NSICU stay, Glasgow Coma Scale, meropenem, neutrophil to lymphocyte ratio, platelet to lymphocyte ratio), had significantly high sensitivity and specificity (primary cohort AUC=0.913, validation cohort 1 AUC=0.830, validation cohort 2 AUC=0.889) for early identification of infection and had great calibration (validation cohort 1,2 P=0.3801, 0.6274). DCA confirmed that the nomogram is clinically useful.ConclusionOur nomogram could help clinicians make early predictions regarding the onset of pulmonary infection caused by MDR-AB and implement targeted interventions.