A Computable Definition of Sepsis Facilitates Screening and Performance Improvement Tracking

Sepsis kills almost 5,000 children annually, accounting for 16% of pediatric health care spending in the United States.

We sought to identify sepsis within the Electronic Health Record (EHR) of a quaternary children's hospital to characterize disease incidence, improve recognition and response, and track performance metrics.

Methods are organized in a plan-do-study-act cycle. During the "plan" phase, electronic definitions of sepsis (blood culture and antibiotic within 24 hours) and septic shock (sepsis plus vasoactive medication) were created to establish benchmark data and track progress with statistical process control. The performance of a screening tool was evaluated in the emergency department. During the "do" phase, a novel inpatient workflow is being piloted, which involves regular sepsis screening by nurses using the tool, and a regimented response to high risk patients.

Screening tool use in the emergency department reduced time to antibiotics (Fig. 1). Of the 6,159 admissions, EHR definitions identified 1,433 (23.3%) between July and December 2016 with sepsis, of which 159 (11.1%) had septic shock. Hospital mortality for all sepsis patients was 2.2% and 15.7% for septic shock (Table 1). These findings approximate epidemiologic studies of sepsis and severe sepsis, which report a prevalence range of 0.45-8.2% and mortality range of 8.2-25% (Table 2).^1-5.

Implementation of a sepsis screening tool is associated with improved performance. The prevalence of sepsis conditions identified with electronic definitions approximates the epidemiologic landscape characterized by other point-prevalence and administrative studies, providing face validity to this approach, and proving useful for tracking performance improvement.