253 Electronic Best Practice Advisories' Effectiveness in Detecting Sepsis in the Emergency Department

Abstract

The incidence of severe sepsis is 750,000 cases per year in the U.S. and Canada, causing significant morbidity and mortality. Current guidelines from the Surviving Sepsis Campaign (SSC) recommend administering appropriate antibiotics within the first 3 hours of recognition of severe sepsis and septic shock; delayed antibiotic treatment is associated with increased morbidity. Best practice advisories (BPAs) are reminder tools within the Epic electronic health record (EHR) system that provide clinical decision support. This study aims to evaluate if BPAs are effective at detecting potentially septic patients in an emergency department (ED) setting. A structured retrospective review of medical records was conducted at an academic urban emergency department (annual visits >85,000) of patients who initially met systemic inflammatory response syndrome (SIRS) criteria after an electronic BPA was implemented (March 1, 2014 to March 30, 2014 and the first 7-day monthly period of April 1, 2014 to September 30, 2014). The BPA Sepsis Risk Scoring System was created based on the SSC guidelines and guided by the Detecting and Treating Sepsis manual from Epic Systems. A score >2.5 meant that a patient met SIRS criteria, is potentially septic, and triggered the BPA. The definitive diagnosis of sepsis for ED SIRS visits was based upon final diagnostic related group (DRG) coding and further established by imaging modalities, urinalysis, and physical exam findings on chart review. BPAs’ sensitivity, specificity, PPV, NPV was calculated and 95% confidence intervals (CI) established. Over the course of the study period, 13,906 records were screened, 565 BPAs fired and 313 cases of sepsis were confirmed (2.3% prevalence). The BPAs’ sensitivity and specificity was 74.5% (95% CI 69.2-79.1%) and 97.6% (95% CI 97.3-97.8%) respectively, with positive and negative predictive values of 41.2% (95% CI 37.2-45.4%) and 99.4% (95% CI 99.3-99.5%) respectively. The BPA’s positive and negative likelihood ratios (LR) (weighted for prevalence) were 0.70 (95% CI 0.62-0.79) and 0.006 (95% CI 0.005-0.008). No significant changes were noted in the BPA sensitivity or specificity when confined to severe sepsis or septic shock. BPAs were an effective EHR-based tool that detected potentially septic patients with moderate sensitivity and high specificity in our ED. The test’s high negative LR and negative predictive value make it valuable in excluding sepsis as differential diagnosis in a general ED population. Future directions for follow-up studies include cost analysis, morbidity/mortality studies, and multicenter comparisons of other quality metrics that can be improved by having a system such as BPA that reduces time to appropriate medical intervention.