What happened

Researchers analysed data from 27,697 patients across multiple hospitalisations, identifying 2,552 cases of sepsis. However, the Epic Sepsis Model (ESM) only detected 183 cases that clinicians had missed, while failing to identify 1,709 patients who had been diagnosed with sepsis by healthcare providers. 

The finding indicates that the model's sensitivity is very low, as it only identified 7 percent of the clinician-diagnosed cases. Additionally, the ESM generated alerts for 18 percent of all hospitalised patients, many of whom did not have sepsis, leading to alert fatigue among medical staff. 

This situation could potentially delay timely treatment for actual sepsis cases and contribute to poorer patient outcomes.

Why it happened

The algorithm was trained to flag when doctors would submit bills for sepsis treatment, which does not always line up with patients’ first signs of symptoms.