Sepsis Alert Routine Implementation in the Emergency Department

Sepsis is a life-threatening condition associated with high morbidity and mortality, affecting millions of people globally each year. Timely recognition and early initiation of treatment are critical for improving outcomes. In response to this need, a structured sepsis alert routine was introduced in October 2017 at the Emergency Department (ED) of Karolinska University Hospital Huddinge, a large tertiary care hospital in Stockholm, Sweden.

The sepsis alert routine was designed to identify patients at risk of sepsis early in their ED visit, using predefined triage criteria. The alert was triggered by ED nurses based on suspected infection and specific triage parameters using the RETTS system (Rapid Emergency Triage and Treatment System). Patients with triage priority 1, or triage priority 2-3 combined with a blood lactate level of >3.2 mmol/L, were eligible. Once activated, a multidisciplinary response team-including physicians from the ED, Infectious Diseases, and Intensive Care Unit-was rapidly assembled to coordinate patient assessment and initiate early, protocolized sepsis care, including prompt antibiotic treatment and fluid resuscitation. The response followed a standardized checklist aligned with the Surviving Sepsis Campaign guidelines.

This study evaluates the clinical impact of the sepsis alert routine by comparing outcomes before and after its implementation. It uses a retrospective, observational cohort design with two data collection periods: a 2-year control period before the intervention (October 2015 - September 2017) and a 2.25-year period during which the sepsis alert was active (October 2017 - December 2019). All patients included were ≥18 years of age, admitted to hospital from the ED with suspected bacterial infection (defined by antibiotic therapy initiated within 48 hours and continued for at least 96 hours or until death, discharge, or ICU transfer), and met sepsis alert criteria.

To assess the effectiveness of the intervention, the primary outcome is 28-day all-cause mortality. Secondary outcomes include 90-day mortality, time to antibiotic administration, and length of hospital stay. Multivariable logistic regression will be used to analyze the full study population, with the intervention period (pre vs. post-implementation) as the main exposure variable, adjusting for covariates including sex, age, comorbidities (Charlson Comorbidity Index), SOFA score, triage priority, infection focus, body temperature, and do-not-resuscitate orders.

To further reduce confounding by indication, a propensity score matching analysis will be performed comparing patients who triggered the sepsis alert with matched historical controls. Propensity scores will be estimated using a logistic regression model including the aforementioned covariates as well as bacteremia status. Balance diagnostics and sensitivity analyses will be conducted to evaluate and account for any residual imbalances. Time-to-event outcomes will be analyzed using Kaplan-Meier survival curves.

This study will contribute important real-world data on the implementation of a sepsis alert system in a high-volume academic ED setting, and its potential to improve sepsis care and reduce mortality.