Some hospitals using Epic's sepsis alert system experienced an increased number of alerts in the weeks after the COVID-19 pandemic began compared to the months leading up, according to a Nov. 19 report published in JAMA Network Open.
Researchers from Boston-based Mass General, Ann Arbor-based University of Michigan Medical School, New York University Grossman School of Medicine and St. Louis-based Washington University School of Medicine examined the number of alerts generated by Epic's sepsis model at 24 hospitals in the months leading up to the COVID-19 pandemic to evaluate possible links to overalerting and the variation in alerts across the hospitals.
The hospitals, analyzed from Nov. 3, 2019, to April 25, 2020, were part of four health systems: the University of Michigan in Ann Arbor, NYU Langone Health, Mass General Brigham in Boston and BJC HealthCare in St. Louis.
Five study insights:
- The sepsis model has been linked to causing potential alert fatigue. Three to four weeks after the hospital had its first COVID-19 hospitalization, the University of Michigan paused its Epic sepsis model alerts after nurses reported being overalerted.
- In the three weeks before and after the first COVID-19 cases in each health system, the proportion of patients generating sepsis alerts daily doubled from nine percent to 21 percent.
- The number of patients admitted to the hospitals declined on average by 35 percent, but the number of alerts increased on average by 43 percent.
- Larger hospitals generally experienced an increase in the proportion of patients generating sepsis alerts, whereas smaller hospitals experienced varying degrees of change in alert volume.
- The researchers said the increase in the number of alerts shows that the alerts can potentially be a burden. The study didn't examine the model's accuracy. However, researchers said that even if the alerts were accurate, many sepsis workflows are built around bacterial sepsis and may not apply in the context of COVID-19, according to the report.