Artificial Intelligence Model Improves Sepsis Care in Pediatric Emergency Departments

Strategic Insights -

Researchers at Emory University School of Medicine in Atlanta, Georgia, have found that implementing an artificial intelligence (AI) model in pediatric emergency departments (EDs) can lead to significant improvements in sepsis care. According to a recent study, the model was able to reduce the time to first antibiotics and fluid bolus, and non-significant changes in mortality and ICU metrics were observed. However, the researchers caution that local workflows, documentation patterns, and patient populations make it challenging to generalize published or reported model performance metrics to real-world performance.

Key Takeaways:

  • The study involved a retrospective cross-sectional analysis of 599,163 ED visits in two pediatric health systems between January 1, 2021, and April 1, 2024.
  • The AI model was designed to predict sepsis cases in the ED and was implemented in two ED sites.
  • The model consisted of a nurse-facing interruptive alert that appeared when the model score exceeded a predetermined threshold, triggering clinicians to call a sepsis huddle.
  • The study found that implementing the model led to significant reductions in time to fluid bolus and borderline decreases in time to antibiotics.
  • Non-significant changes in mortality, ICU-free days, and ED to ICU admission rates were observed after the intervention.
  • The researchers conclude that implementing an externally developed model in real-world settings can be challenging due to local workflows, documentation patterns, and patient populations.
  • The study highlights the importance of tailoring AI models to specific clinical workflows and patient populations to achieve optimal performance.

Statistics:

  • 599,163 ED visits were analyzed in the study.
  • The AI model was implemented in two ED sites.
  • 268,102 ED visits with 741 (0.28%) sepsis cases were included in the pre-intervention cohort.
  • 331,061 ED visits with 1114 (0.34%) sepsis cases were included in the post-intervention cohort.
  • Mean time to first antibiotic decreased from 112 to 102 minutes (P = .05, 95% confidence interval of difference, -19.1 to 0.1).
  • Time to first bolus decreased by 16.7 minutes (P = .03, 95% confidence interval difference, -31.8 to -1.5) after the intervention.
  • Decreases in 30-day mortality (6% to 4%), ED to ICU admissions (87% to 84%), and ICU-free days (6 to 5) did not meet statistical significance.

Sources:

  • Emory University School of Medicine, Atlanta, GA, USA.
  • Journal of the American Medical Informatics Association.
  • Oxford Univ Press, Great Clarendon St, Oxford OX2 6DP, England.
  • Evan W. Orenstein, Pediatrics, Emory University School of Medicine, Atlanta, GA 30307, USA.
  • Swaminathan Kandaswamy, Naveen Muthu, Andrea McCarter, Nikolay Braykov, Jonathan M. Beus, Edwin Ray, Tal Senior, Sara P. Brown, Rebekah Carter, MaryBeth Gleeson, Hannah Thummel, John Cheng, Thuy Bui, Reena Blanco, Kiran Hebbar, and James F.