Strategy to Identify Paramedic Transported Sepsis Cases in an Emergency Department Administrative Database

Background: To evaluate a new strategy for identifying sepsis in Emergency Department (ED) patients that combines administrative diagnosis codes with clinical information from the point of first contact. Methods: This study linked clinical data from adult patients transported by a provincial Emergency Medical Services (EMS) system to ED and inpatient administrative databases. Sepsis cases were identified by searching ED databases for diagnosis codes consistent with infection and organ dysfunction. Organ dysfunction was further assessed using a partial Sequential Organ Failure Assessment (SOFA) score derived from EMS clinical information. Reliability was evaluated by comparing patients' ED diagnosis codes (ICD-10CA) to their inpatient diagnosis codes; criterion validity by comparing cases classified by the new strategy to an existing inpatient administrative algorithm; and construct validity by assessing for clinical characteristics typically associated with sepsis (e.g., mortality). Results: A total of 43,297 patients were included. ED infection codes were more reliable for classifying patients with infection than using ED sepsis codes alone (proportion of agreement with inpatient codes 79% vs. 74%; p-value < 0.001). The novel strategy requiring the presence of an infection code and either an organ dysfunction code or 2 or more SOFA points from EMS clinical information identified 1,379 more ED patients as having sepsis than the inpatient algorithm. These patients had high mortality supporting construct validity. Conclusions: Incorporation of a broader range of diagnostic codes and linking to an electronic database to obtain initial clinical information for the assessment of organ dysfunction improves reliability, criterion, and construct validity for identifying sepsis in ED patients.