Real-time AI Predicts Adverse Outcomes in Acute Pancreatitis in the Emergency Department: Comparison with Clinical Decision Rule
Authors: Ching-Hung Chang, Chia-Jung Chen, Yu-Shan Ma, Yu-Ting Shen, Mei-I Sung, Chien-Chin Hsu, Hung-Jung Lin, Zhih-Cherng Chen, Chien-Cheng Huang, Chung-Feng Liu
Published in: Academic Emergency Medicine, 2023 October 27
Conclusion:
- The study successfully integrated a real-time AI prediction model into HIS, which demonstrated superior predictive performance for adverse outcomes in AP patients when compared to the conventional BISAP score.
- Despite promising initial results, further validation is required to confirm the model’s reliability and generalizability across different clinical settings.
Methods:
- Retrospective analysis of 8,274 ED patients with AP from three hospitals over 2009-2018.
- Evaluated data included demographics, comorbidities, lab results, and adverse outcomes.
- Six AI algorithms were tested; the best-performing algorithm, based on Area Under the Curve (AUC), was integrated into the Hospital Information System (HIS) for real-time predictions.
- The AI model’s predictive accuracy was compared to the Bedside Index for Severity in Acute Pancreatitis (BISAP).
Results:
- Mean patient age: 56.1±16.7 years; 67.7% male.
- Light Gradient Boosting Machine (LightGBM) and eXtreme Gradient Boosting (XGBoost) showed the highest AUC for predicting sepsis, ICU admission, and mortality.
- AI model outperformed BISAP in predicting sepsis (AI: 0.961 vs. BISAP: 0.785), ICU admission (AI: 0.973 vs. BISAP: 0.778), and mortality (AI: 0.975 vs. BISAP: 0.817).
Chang, C.H., Chen, C.J., Ma, Y.S., Shen, Y.T., Sung, M.I., Hsu, C.C., Lin, H.J., Chen, Z.C., Huang, C.C. and Liu, C.F., Real‐time AI Predicts Adverse Outcomes in Acute Pancreatitis in the Emergency Department: Comparison with Clinical Decision Rule. Academic Emergency Medicine.