Background: Sepsis is the leading cause of death in US hospitals; prompt recognition and treatment are critical. Early sepsis recognition and treatment varies depending on location (ED/floor/ICU) and evaluation time (presentation vs during stay). “Sepsis response teams” (SRT) improve sepsis care. After our hospital system implemented a “best-practice alert” utilizing Epic’s “sepsis score” (SS) to affect ED and ICU sepsis care, our group focused on improving floor patient care. After a multidisciplinary, multifaceted root-cause-analysis, the team decided to pilot a “sepsis squashing squadron” (S3) – a combination of SRT and automated EHR-based risk scores and summary displays to improve workload efficiency and efficacy.
Methods: Through chart review, a S3 member screened medicine patients on a single unit. For patients with a likely and/or definitive infection and an adapted SOFA score (mSOFA) ≥ 2 and/or SIRS ≥2, adherence to Severe Sepsis/Septic Shock Early Management Bundle (SEP-1) components was assessed. If additional interventions were clinically appropriate, the reviewer contacted the primary team, primary nurse, charge nurse, and an on-site S3 member (for task-completion support) with specific recommendations. Automated scores were recorded for performance characterization: qSOFA, an institutional modification of the Modified Early Warning score (jMEWS), SS and Epic’s “deterioration index” (DI). Using an optimized, combined automated risk score threshold, a second pilot iteration was then performed.
Results: From 5/16/22-6/2/22, 190 patients were screened; 97 had a known or probable infection. Three patients’ care did not meet SEP-1. Average review time per patient was 2 minutes (range 1-19 minutes). Sepsis prevalence and SEP-1 adherence varied based on sepsis definition with prevalence/adherence–mSOFA ≥ 2: 24%/93%; SIRS ≥ 2: 16%/86%; and qSOFA ≥ 2: 5%/100%. Automated Epic scores demonstrated wanting performance with best sensitivity/specificity thresholds of 75%/59% (jMEWS), 32%/90% (SS), and 68%/53% (DI) compared to a SIRS-definition. Data analysis suggests a combined threshold jMEWS ≥ 2 and DI ≥ 30 and/or mSOFA ≥ 2 offers a sensitivity/specificity of >93%/>50%. Using the combined-score thresholds and with a more restrictive bacterial/fungal infection definition and an afternoon review time, a second pilot from 10/30/22 – 11/14/22 screened 109 patients, of whom 11 met SIRS-based criteria (10% prevalence). Bundle adherence was 100% and average per patient review time was 1 minute (range 0-13 minutes).
Conclusions: Commonly used individual, automated EHR scores failed to have satisfactory sensitivity and specificity for floor sepsis. Using optimized, combined EHR score thresholds and custom displays halved average patient review time relative to the first pilot, but opportunities to improve care were limited by sepsis prevalence and high SEP-1 adherence. This may be due to sample size, specific team and/or unit performance, and/or once-daily screening. Future iterations will refine EHR score thresholds, assess alternate units and provider teams, and/or review patients at alternate times to improve performance.