3073 — Prediction Rules to Identify Veterans with Methicillin-Resistant Staphylococcus Aureus or Vancomycin-Resistant Enterococcus at Hospital Admission

Morgan DJ (VA Maryland Health Care System), Day HR (VA Maryland Health Care System), Strauss SM (VA Maryland Health Care System), Schweizer ML (VA Maryland Health Care System), Roghmann M (VA Maryland Health Care System), Perencevich EN (VA Maryland Health Care System)

Objectives:
Clinical prediction rules have been used to efficiently identify patients at high risk of colonization with Methicillin-Resistant Staphylococcus Aureus (MRSA) or Vancomycin-Resistant Enterococci (VRE). In 2007, the VA mandated active surveillance for all patients admitted to acute-care hospitals. This mandate, while targeting patient safety, is expensive. If high-risk patients can be identified using a simple prediction rule and targeted for MRSA (and also VRE) screening, it might be more cost-effective. We have developed, validated and implemented a clinical prediction rule at the University of Maryland that identifies 80% of patients with MRSA, while limiting expensive culture to only 50% of admitted patients, saving the hospital approximately $200,000 annually. We aimed to determine the clinical efficacy of similar prediction rules in a Veterans Hospital.

Methods:
We are conducting a prospective cohort study of all adult inpatients admitted to the medical and surgical wards of a tertiary-care VA hospital. In the first 48 hours of admission, patients were approached for consent, administered a 44-item questionnaire, and received nasal and peri-rectal cultures for MRSA and VRE, respectively. We report an interim analysis.

Results:
Out of the initial 341 patients enrolled, 331 underwent nasal cultures and 241 underwent peri-rectal cultures. Overall, 14% were MRSA positive and 7% were VRE positive. Patient self-report of having received antibiotics in the past year was the most sensitive single predictor for MRSA (71%); specificity was 50% (relative risk (RR) = 2.2 (95% CI 1.2 - 4.0), p = 0.014. This rule had a sensitivity of 81% for VRE (RR = 3.7 (95% CI 1.1 – 12.6), p = 0.04. Use of this predictor would require swabbing only 53% of admissions. A prediction rule using self-report of hospitalization or receiving antibiotics in the past year would have identified 80% and 94% of patients colonized with MRSA or VRE, respectively, and require swabbing 70% of admissions.

Implications:
Patient self-report of receiving antibiotics within the past year identifies a group of patients at high risk for colonization with MRSA or VRE that could be considered for targeted active surveillance culturing.

Impacts:
This approach has the potential for significant cost-savings compared to the current practice of universal active surveillance.