Prospective validation and implementation of a model to identify patients with carbapenem-resistant Enterobacterales (CRE) carriage on admission to acute care hospitals

OBJECTIVE: Use the electronic health record (EHR) to prospectively validate a risk prediction tool identifying patients at high-risk for carbapenem-resistant Enterobacterales (CRE). DESIGN: Prospective, cross-sectional analysis. PARTICIPANTS: Adults admitted to two hospitals in Atlanta, Georgia. METHODS: An EHR report calculated a CRE risk score on all admissions from 6/2024-3/2025. The risk score was determined from a prior model incorporating data from current and prior hospitalizations. Stool or perianal samples were cultured from a convenience sample of patients with the highest risk scores. A receiver operating curve analysis calculated sensitivity, specificity, positive predictive value (PPV), and negative predictive values (NPV) of various risk scores that could be used as a "threshold" for CRE admission testing. Using the threshold with the greatest Youden's index, we estimated the cost of implementing a CRE screening program. RESULTS: Of the 853 patients approached, 342 (40%) consented. Eleven (3.2%) tested positive for CRE. Patients with CRE had a higher median CRE risk score (0.19% vs 0.04%) than those who tested negative. The AUC of the model was 0.66. Using a testing threshold of 0.16% yielded a 55% sensitivity, 84% specificity, 10% PPV, and 98% NPV. The number needed to screen to diagnose 1 patient with CRE was 12 patients, and the screening program approximately costs $8,015/month. CONCLUSIONS: An EHR-based risk prediction tool can detect patients likely to be colonized with CRE. In facilities with a low CRE prevalence, identifying a high-risk subset of patients to test could be a cost-effective infection prevention initiative.