1227. Development of a Clinical Prediction Model for Mortality in Methicillin-Resistant Staphylococcus aureus Bacteremia

Abstract

BackgroundMethicillin-resistant Staphylococcus aureus bloodstream infection (MRSA BSI) is associated with high mortality despite advances in medical care. Mortality prediction may have a profound impact on clinical decision making and risk stratification. Widely used scoring systems such as the Acute Physiology and Chronic Health Evaluation (APACHE) II Score and the Pitt Bacteremia Score were derived in the general critical care and Gram-negative BSI populations, respectively and may be less precise in MRSA BSI. We sought to develop a predictive model (PM) for 30-day mortality in patients with MRSA BSI based on characteristics readily assessable at initial evaluation.MethodsRetrospective, singe-center, cohort study in adults with MRSA BSI 2008 to 2018. Patients who did not receive active therapy within 72 hours of index culture were excluded. Independent baseline demographic, clinical and infection predictors of 30-day mortality were identified through multivariable logistic regression analysis with bootstrap resampling and coefficient shrinkage. The PM was derived using a regression coefficient-based scoring method. PM discriminatory ability was assessed using the c-statistic. The optimal threshold score was determined using the Youden Index (J).ResultsA total of 455 patients were included and 30-day mortality was 16.3%. The PM consisted of five variables and a potential total score of 33. Points were assigned as follows: age (9 points ≥90 years, 6 points 80–89 years, 5 points 70–79 years, 0 points <70 years); Glasgow Coma Scale (8 points ≤9, 5 points 10–13, 0 points ≥14); 7 points infective endocarditis or pneumonia; 5 points serum creatinine ≥ 3.5 dl/L; and four points respiratory rate <10 or >24. The PM c-statistic was 0.860 (95% CI 0.818, 0.902). The PM score with the maximum J value was 13. Thirty-day mortality was 5.2% vs. 44.5% for PM score <13 vs. ≥13 points, respectively (P < 0.001). The sensitivity, specificity, positive predictive value (PV), negative PV, and accuracy using a threshold of 13 points were 77.0%, 81.4%, 44.5%, 94.8%, and 80.7%, respectively.ConclusionOur findings demonstrate a weighted combination of five independent variables readily assessable at initial evaluation can be used to predict, with high discrimination, 30-d mortality in MRSA BSI. External validation is required before wide-spread clinical use.Disclosures M. J. Rybak, Allergan: Consultant, Grant Investigator and Speaker’s Bureau, Research grant and Research support. Achaogen: Consultant, Grant Investigator and Speaker’s Bureau, Consulting fee, Research grant and Research support. Bayer: Consultant, Grant Investigator and Speaker’s Bureau, Consulting fee, Research grant and Research support. Melinta: Consultant, Grant Investigator and Speaker’s Bureau, Consulting fee, Research grant and Research support. Merck: Consultant, Grant Investigator and Speaker’s Bureau, Consulting fee, Research grant and Research supportTheravance: Consultant, Grant Investigator and Speaker’s Bureau, Consulting fee, Research grant and Research support. Sunovian: Consultant, Grant Investigator and Speaker’s Bureau, Consulting fee, Research grant and Research support. Zavante: Consultant, Grant Investigator and Speaker’s Bureau, Consulting fee, Research grant and Research support. NIAID: Consultant, Grant Investigator and Speaker’s Bureau, Consulting fee, Research grant and Research support.