Predicting Multidrug-Resistant Gram-Negative Bacterial Colonization and Associated Infection on Hospital Admission

Abstract

OBJECTIVEIsolation of multidrug-resistant gram-negative bacteria (MDR-GNB) from patients in the community has been increasingly observed. A prediction model for MDR-GNB colonization and infection risk stratification on hospital admission is needed to improve patient care.METHODSA 2-stage, prospective study was performed with 995 and 998 emergency department patients enrolled, respectively. MDR-GNB colonization was defined as isolates resistant to 3 or more classes of antibiotics, identified in either the surveillance or early (≤48 hours) clinical cultures.RESULTSA score-assigned MDR-GNB colonization prediction model was developed and validated using clinical and microbiological data from 995 patients enrolled in the first stage of the study; 122 of these patients (12.3%) were MDR-GNB colonized. We identified 5 independent predictors: age>70 years (odds ratio [OR], 1.84 [95% confidence interval (CI), 1.06–3.17]; 1 point), assigned point value in the model), residence in a long-term-care facility (OR, 3.64 [95% CI, 1.57–8.43); 3 points), history of cerebrovascular accidents (OR, 2.23 [95% CI, 1.24–4.01]; 2 points), hospitalization within 1 month (OR, 2.63 [95% CI, 1.39–4.96]; 2 points), and recent antibiotic exposure (OR, 2.18 [95% CI, 1.16–4.11]; 2 points). The model displayed good discrimination in the derivation and validation sets (area under ROC curve, 0.75 and 0.80, respectively) with the best cutoffs of<4 and ≥4 points for low- and high-risk MDR-GNB colonization, respectively. When applied to 998 patients in the second stage of the study, the model successfully stratified the risk of MDR-GNB infection during hospitalization between low- and high-risk groups (probability, 0.02 vs 0.12, respectively; log-rank test, P<.001).CONCLUSIONA model was developed to optimize both the decision to initiate antimicrobial therapy and the infection control interventions to mitigate threats from MDR-GNB.Infect Control Hosp Epidemiol 2017;38:1216–1225