1034. Automating Assessments of Vancomycin Appropriateness

BACKGROUND: Assessing appropriateness of hospital antibiotic use is typically a labor-intensive task for antimicrobial stewardship teams and relies heavily on clinician judgement rather than a systematic process. Vancomycin is a frequently used agent that is a common stewardship target. We developed an algorithm to automatically classify the appropriateness of vancomycin days of therapy (DOTs) based upon electronic health record data. METHODS: We constructed a retrospective cohort of Oregon Health and Science University (OHSU) Hospital and Doernbecher Children’s Hospital patients admitted August 1, 2017 to July 31, 2018 receiving vancomycin. Data were collected on demographic, encounter, pharmacy, microbiology, and surgery data. An electronic algorithm was applied to classify vancomycin DOTs as appropriate, inappropriate, or indeterminate. Inappropriate use was defined as any case in which there was an opportunity for de-escalation as identified using microbiology data, ICD-10 codes, and procedure codes. RESULTS: We included 4,231 encounters; 493 (12%) were pediatric patients. Our algorithm automatically classified 59%, 3%, and 38% of encounters as having either appropriate, inappropriate, or indeterminate DOTs, respectively. Forty-four percent of all encounters received no more than a 24-hour course of vancomycin and were considered appropriate empiric therapy; half of these were attributed to surgical prophylaxis. Nine percent of all encounters had vancomycin administered within 3 days of a blood, sputum or tissue culture in which either a methicillin-resistant Staphylococcus species or an ampicillin-resistant, vancomycin-susceptible Enterococcus species was isolated and were classified as appropriate. Six percent of all encounters had cultures in which only Gram-negatives, fungi, or yeast were isolated and were therefore considered appropriate in the empiric period (≤48 hours) but inappropriate thereafter. CONCLUSION: Automated assessments of antibiotic appropriateness could facilitate more informed antimicrobial stewardship initiatives and serve as a valuable stewardship metric. Characterization of indeterminate vancomycin use may inform increased automated classification. Further effort is needed to validate these assessments. DISCLOSURES: All authors: No reported disclosures.