2000. Utilization of a ‘Never Event’ Framework to Classify Antimicrobial Appropriateness
Abstract Background Contemporary strategies can be leveraged to predict antimicrobial overuse, yet little information is gained on the appropriateness of antibiotics prescribed. Classifying appropriateness is complicated by the lack of a standard definition for appropriateness. Thus, we created and implemented a novel ‘antibiotic never event’ (NE) framework to systematically classify the most inappropriate usages of vancomycin and correlated these NE to abnormal consumption trends (i.e., antibiotic outbreaks). Methods Vancomycin use was categorized by an algorithm using data query from the electronic medical records. Extracted data included vancomycin use, relevant patient demographics, and microbiological data. Electronic classifications placed each vancomycin therapy into type 1 (use for non-susceptible organism after susceptibility finalization) or type 2 (use exceeding 48h after susceptibility report when a safe de-escalation is possible) NE. Patients were categorized as cases or controls (no NE) at Northwestern Memorial Hospital (NM) and Henry Ford Hospital (HF) between January 2014 and October 2017. A manual chart review was performed. Sensitivity (SEN), specificity (SPEC), PPV, and NPV were calculated for NE prediction. Vancomycin use was quantified during the same period. Linear models with prediction intervals (PI) were generated to identify potential outbreaks, which were linked to monthly NE counts defined as a binary factor. Results A total of 220 NE cases were electronically identified for vancomycin at NM (n = 197) and HF (n = 23). Random cases were matched 1:1 (NM = 200) and 1:5 (HF = 115) to controls for manual review. At NM and HF, 35 and 24 true positives were identified, respectively. Thus, overall SEN and SPEC were 93.7% and 75.1% and PPV and NPV were 45.7% and 98.1%, respectively. Linear models revealed 11 potential outbreak periods at HF and 5 at NM. A PI of 80% showed a combined SEN below 10% and SPEC above 90%, respectively. Conclusion The methodology was generalizable across two centers. In the pilot review, our method was highly sensitive and an effective screening tool for NE identification. Antibiotic consumption trends did not correlate with NE. In summary, the NE classification was sensitive in assessment of antibiotic appropriateness, whereas consumption alone does not predict NE. Disclosures All authors: No reported disclosures.
