12. Development of Provider-Specific Antibiotic Prescribing Feedback for Inpatient Antibiotic Stewardship Programs in Veterans Affairs (VA) Facilities (ASP)

Abstract Background Provision of provider-specific outpatient antibiotic prescribing data has resulted in significant decreases in antibiotic use. We describe the development of reports of inpatient antibiotic prescribing by hospitalists attending on acute medical wards in VA medical facilities. Methods We created algorithms for determining the attending physician responsible for patient days present (DP), by considering changes of service (e.g., prior to admission from the emergency department) and transfers between services or physicians. Each antibiotic dose was assigned to a single attending, ward location, and service according to denominator assignment. Antibiotic use was grouped into Centers for Disease Control and Prevention drug categories and expressed as antibiotic days of therapy (DOT) per 1000 DP. Data were obtained from the VA Corporate Data Warehouse. Algorithms were iteratively refined based on reviews of medical records from three VA medical centers and applied to acute care patients at a single site for 2018-2020. Results In 2018-2020, 294 attendings oversaw acute inpatient care for >= 14 DP. 129 attendings with >= 300 DP oversaw 88.0% of all patient care and prescribed 87.6% of all antibiotics (480 DOT/1000 DP, IQR 375-559), 90.1% of broad-spectrum therapy for hospital-onset infections (55 DOT/1000 DP, IQR 31-72) and 88.3% of resistant Gram-positive therapy (70 DOT/1000 DP, IQR 39-89) in inpatient wards. The distribution of antibiotic use for acute care ward patients amongst these 129 staff is shown in the following figure. Conclusion We developed algorithms to attribute antibiotic therapy to inpatient attendings that can be broadly applied in facilities with electronic medical records. As with outpatient prescribing, we found large variation across inpatient attendings in overall antibiotic use and broad-spectrum antibiotic use. In future work, we will obtain provider feedback of report usability and interpretability and assess whether distribution of these reports allows antibiotic stewards to favorably influence provider prescribing practices. Disclosures Matthew B. Goetz, MD, Nothing to disclose Arjun Srinivasan, MD, Nothing to disclose

Background. Surveillance of Non-Ventilator Hospital-Acquired Pneumonia (NV-HAP) is limited by the ambiguity in diagnosing pneumonia. We implemented electronic surveillance criteria for NV-HAP across the VA healthcare system and tested for reliability, validity and meaning of the electronic criteria vs manual chart review.
Methods. We defined NV-HAP surveillance criteria as oxygen deterioration concurrent with fever or abnormal WBC count, ≥3 days of antibiotics, and orders for chest imaging. We applied these criteria to EHR data from all patients hospitalized ≥3 days at all VA acute care facilities from 1/1/2015-12/31/2020 and calculated NV-HAP incidence and inpatient mortality. Clinician reviewers used a consensus review guide to independently review and adjudicate 47 cases meeting NV-HAP surveillance criteria for 1) clinical deterioration, 2) CDC-NHSN pneumonia criteria, 3) treating clinicians' assessment, and 4) reviewer's diagnosis. All reviewers subsequently adjudicated all cases and conducted an error analysis to identify sources of discordance.
Conclusion. NV-HAP electronic surveillance criteria demonstrated high precision for identifying clinical deterioration and moderate concordance with CDC-NHSN pneumonia criteria or reviewer diagnosis. Agreement between electronic surveillance criteria vs manual chart review was low but similar to agreement amongst manual reviewers applying NHSN criteria. Electronic surveillance may provide greater consistency than human review while facilitating wide-scale automated surveillance.
Disclosures Background. Provision of provider-specific outpatient antibiotic prescribing data has resulted in significant decreases in antibiotic use. We describe the development of reports of inpatient antibiotic prescribing by hospitalists attending on acute medical wards in VA medical facilities.
Methods. We created algorithms for determining the attending physician responsible for patient days present (DP), by considering changes of service (e.g., prior to admission from the emergency department) and transfers between services or physicians. Each antibiotic dose was assigned to a single attending, ward location, and service according to denominator assignment. Antibiotic use was grouped into Centers for Disease Control and Prevention drug categories and expressed as antibiotic days of therapy (DOT) per 1000 DP. Data were obtained from the VA Corporate Data Warehouse. Algorithms were iteratively refined based on reviews of medical records from three VA medical centers and applied to acute care patients at a single site for 2018-2020.

Conclusion.
We developed algorithms to attribute antibiotic therapy to inpatient attendings that can be broadly applied in facilities with electronic medical records. As with outpatient prescribing, we found large variation across inpatient attendings in overall antibiotic use and broad-spectrum antibiotic use. In future work, we will obtain provider feedback of report usability and interpretability and assess whether distribution of these reports allows antibiotic stewards to favorably influence provider prescribing practices.
Disclosures. Matthew B. Goetz, MD, Nothing to disclose Arjun Srinivasan, MD, Nothing to disclose