Potential and limitations of large language models in acute chest pain triage. Response to Evaluating the predictive accuracy of ChatGPT in risk stratification for chest pain in the emergency department

Dear Editor-in-Chief, We were especially interested to find out about this recently published article by Malalan et al., which discussed the evaluation of ChatGPT 4.0 as a clinical decision-support tool for predicting Major Adverse Cardiac Events (MACE) in patients presenting with chest pain to the Emergency Department (ED).1 This novel study examines an innovative application of Large Language Models (LLMs) in a critical and time-sensitive clinical setting. With 178 patients, the authors conducted and have succeeded in a prospective observational study, where it evaluated ChatGPT’s score at three successive stages of patient evaluation — initial clinical history and ECG, first troponin test, and second troponin test. The results showed an encouraging pattern with ChatGPT’s predictive accuracy (Area Under the ROC curve) improving steadily from 0.699 to 0.776 (p=0.039) with the addition of clinical data. The model also showed some high negative predictive value, encouraging the possibility of targeting low-risk patients who could avoid avoidable hospitalisation. [...]