995. Acute Kidney Injury Rates Before and After Implementation of Integrated Bayesian Software for Vancomycin Dosing

Abstract

Abstract Background Therapeutic drug monitoring of vancomycin with Bayesian models is recommended to maximize efficacy and minimize nephrotoxicity by targeting a 24-hour area under the curve (AUC24) of 400-600mg*h/L. Minimal real-world evidence exists on the impact of this strategy. In 2020, we integrated Bayesian software from DoseMeRx into Epic OneChart and converted from trough-based vancomycin to AUC-guided dosing. We assessed the incidence of acute kidney injury (AKI) in patients receiving vancomycin before and after this change and sought to establish an efficient method of assessment. Methods This was a retrospective, single-center quasi-experimental study. Patients were divided into pre-Bayesian software (trough guided dosing, Jan-Jun 2018) and post- (Jan-Jun 2021). Preliminary data showed 10% & 6% AKI rates, respectively. Therefore, a sample of 1442 patients was needed to detect a difference. Adults were included with a baseline Scr ≤2 mg/L, who received ≥3 vancomycin doses and had ≥1 vancomycin level measured. The primary outcome was AKI incidence assessed using AKIN criteria. Patients were reviewed in each period until 500 met inclusion criteria. This sample was used to determine the feasibility and accuracy of using EHR-generated reports. Results Of the 1,000 patients reviewed, baseline characteristics were similar; however, indications varied as pneumonia was more common in the pre-group and skin and soft tissue infection was more common in the post-group (Table 1). Initial daily dose and duration of vancomycin treatment were not different between groups. Incidence of AKI was decreased in the post-integration group, although this was not significant by bivariate analysis (15.4% vs. 12.6%, p=0.20). Fewer trough concentrations were drawn in the post-group (2 vs. 1.8, p=0.005), with overall lower vancomycin levels (13.2 vs. 12.1 mg/L, p=0.04). Conclusion The rate of AKI was numerically lower in patients monitored and dosed with Bayesian AUC-guided dosing software for vancomycin. Trough levels were lower, and vancomycin levels were needed less often. AKI incidence was higher than in our pilot but similar to what is reported in other studies. Further investigation is required to establish statistical significance and whether the assessment process can be automated. Disclosures Bryan T. Alexander, PharmD, BCIDP, AAHIVP, Astellas Pharma: Advisor/Consultant Trevor C. Van Schooneveld, MD, bioMerieux: Advisor/Consultant|bioMerieux: Grant/Research Support|Insmed: Grant/Research Support|Merck: Grant/Research Support|Thermo-Fischer: Advisor/Consultant Scott J. Bergman, PharmD, BCIDP, FCCP, FIDSA, Merck & Co., Inc: Advisor/Consultant|Merck & Co., Inc: Grant/Research Support|Merck & Co., Inc: Honoraria|Pfizer: Advisor/Consultant|Pfizer: Honoraria.