Abstract:

Objective To establish a Nomogram model for assessing the risk of intestinal colonization by Carbapenem?Resistant Klebsiella pneumoniae （CRKP） to determine the specific probability of colonization and adopt individualized prevention strategies for the purpose of reducing the occurrence of colonization and secondary infection of neonatal CRKP. Methods A total of 187 neonates hospitalized between January 2021 and October 2022 and diagnosed with CRKP colonization by rectal swab/fecal culture as well drug sensitivity identification 48 h after admission were assigned to the CRKP group. Another 187 neonates without non?CRKP colonization during the same period were set as the non?CRKP group. All the data of the two groups were used for a retrospective analysis. The caret package in R 4.2.1 was used to randomly divide the 374 cases into the model group and validation group at a ratio of 3∶1. Then the glmnet package in R 4.2.1 was used to conduct a LASSO regression analysis over the data from the model group to determine the predictive factors for modeling and the rms software package was used to build a Nomogram model. The pROC and rms packages in R 4.2.1 were used to examine the data， analyzing the consistency indexes （Cindex）， receiver operating characteristic curves （ROC）， and area under the curves （AUC） and performing the internal and external validation of the efficacy of the Nomogram model via the calibration curves. Results LASSO regression analysis determined eight predictors from the 35 factors probably affecting neonatal CRKP colonization： gender， cesarean section， breastfeeding， nasogastric tube， enema， carbapenems， probiotics， and hospital stay. The Nomogram model constructed using these eight predictors as variables could predict CRKP colonization to a moderate extent， with the area under the ROC curve of 0.835 and 0.800 in the model and validation group， respectively. The Hos?mer?Lemeshow test showed that the predicted probability was highly consistent with the actual probability （the modeling group： P = 0.678 > 0.05； the validation group： P = 0.208 > 0.05）， presenting a higher degree of fitting. Conclusion The Nomogram model containing such variables as gender， cesarean section， breastfeeding， nasogastric tube， enema， carbapenems， probiotics， and hospital stay is more effective in predicting the risk of neonatal CRKP colonization. Therefore， preventive measures should be individualized based on the colonization probability predicted by the Nomogram model in order to keep neonates from CRKP colonization and reduce the incidence of secondary CRKP infections among them.

Key words: newborn, carbapenem?resistant klebsiella pneumonia, colonization, Nomogram model