SHM Converge Scientific Abstract ePoster Gallery-Real-World Performance of Emergency Department Admission Prediction Ai Model

Pdf Summary

This Mayo Clinic study evaluated the real-world performance of an AI model designed to predict which emergency department (ED) patients would be admitted to the hospital. The model was deployed across six ED sites within the same healthcare system, including one tertiary referral center, three community hospitals, and two critical access hospitals. Implementation occurred between January and May 2023, and the authors analyzed the most recent six months of data to compare performance across sites and over time.<br /><br />The study included substantial patient volumes at each site, ranging from 1,524 visits in Lake City to 34,038 visits in Rochester. Admission rates also varied by site. Performance was measured using the area under the receiver operating characteristic curve (AUC). Across the six EDs, AUC values ranged from 0.81 to 0.86, indicating good predictive accuracy at each location. When performance was examined by month across all sites, AUC remained stable at 0.82 to 0.83, suggesting little or no decline in model accuracy over time.<br /><br />The authors concluded that the AI model showed consistent live performance over six months across diverse ED settings, with no evidence of model drift despite ongoing workflow and practice improvement efforts. They suggest that well-designed admission prediction models can remain robust in real-world clinical use across multiple implementation sites.

Asset Subtitle

Riddhi S. Parikh

Meta Tag

Author List Alexander Ryu, Benjamin Hinton, Derick Jones, Heather Heaton, Jens Boyum, Ray Qian, Riddhi S. Parikh, Shant Ayanian

Category Research

Concept AI Model

Concept Emergency Department Admission

Concept Predictive Performance

Concept AUC

Concept Model Discrimination

Distinguished Non-Finalist

Presenter Organization Mayo Clinic

Presenting Author Riddhi S. Parikh

Track Technology in Hospital Medicine

Keywords

Mayo Clinic

artificial intelligence

emergency department

hospital admission prediction

real-world performance

model deployment

AUC

model drift

clinical implementation

predictive accuracy

AI Model

Emergency Department Admission

Predictive Performance

AUC

Model Discrimination