Ju Yeon Song1, Ihn Sook Jeong2. 1. Pusan National University Yangsan Hospital, Yangsan, Republic of Korea. 2. College of Nursing, Pusan National University, Yangsan, Republic of Korea. Electronic address: jeongis@pusan.ac.kr.
Abstract
BACKGROUND: Few hospitals have conducted surveillance cultures to identity carbapenem-resistant Enterobacteriaceae (CRE) colonization in Korea because of high test costs, long wait time to get results, and lack of personnel. This study aimed to identify risk factors of CRE colonization and to develop a model for predicting the risk of CRE colonization. METHODS: This retrospective cohort study was done with medical record review at a tertiary hospital between October 1, 2016, and October 31, 2017, in 444 adult patients admitted to the intensive care unit. The identification of risk factors and risk model development were done by multiple logistic regression. The area under the receiver operator characteristic curve and correct classification were measured. RESULTS: The risk prediction logistic model for CRE colonization was composed of 4 factors among 22 significant variables on univariate analysis and 1 interaction term; E (logit of CRE colonization) = -2.821 + 1.606 (isolation of multidrug-resistant organisms) + 1.347 (≥15 days of cephalosporin administration) + 0.980 (≥15 days of carbapenem administration) + 0.544 (score ≥21 on Acute Physiology and Chronic Health Evaluation II). The risk prediction model showed 0.80 (95% confidence interval, 0.74-0.85) for the area under the curve, and 68.9% correct classification. CONCLUSIONS: It is recommended to utilize the risk prediction model to screen out for the colonization of CRE, and to do validation studies for verifying the generalizability.
BACKGROUND: Few hospitals have conducted surveillance cultures to identity carbapenem-resistant Enterobacteriaceae (CRE) colonization in Korea because of high test costs, long wait time to get results, and lack of personnel. This study aimed to identify risk factors of CRE colonization and to develop a model for predicting the risk of CRE colonization. METHODS: This retrospective cohort study was done with medical record review at a tertiary hospital between October 1, 2016, and October 31, 2017, in 444 adult patients admitted to the intensive care unit. The identification of risk factors and risk model development were done by multiple logistic regression. The area under the receiver operator characteristic curve and correct classification were measured. RESULTS: The risk prediction logistic model for CRE colonization was composed of 4 factors among 22 significant variables on univariate analysis and 1 interaction term; E (logit of CRE colonization) = -2.821 + 1.606 (isolation of multidrug-resistant organisms) + 1.347 (≥15 days of cephalosporin administration) + 0.980 (≥15 days of carbapenem administration) + 0.544 (score ≥21 on Acute Physiology and Chronic Health Evaluation II). The risk prediction model showed 0.80 (95% confidence interval, 0.74-0.85) for the area under the curve, and 68.9% correct classification. CONCLUSIONS: It is recommended to utilize the risk prediction model to screen out for the colonization of CRE, and to do validation studies for verifying the generalizability.