Artificial Intelligence-Driven Cost Savings from Emergency Department Chest Pain Patient Evaluation: A Monte Carlo...

Chest pain patients presenting to the emergency department (ED) often require risk stratification through stress testing, which may necessitate hospitalization. This practice contributes to ED crowding by occupying beds with low-risk patients who may derive limited benefit from testing, thereby increasing bed occupancy, prolonging ED boarding times, amplifying crowding, and resource strain. Use of an artificial intelligence (AI)-driven clinical decision support tool improves the detection of obstructive coronary artery disease and focuses diagnostic testing on those most likely to benefit.

We created a Monte Carlo simulation informed by available published inputs and ran 1,000 trials to estimate the national impact of using an AI-driven decision-support tool to reduce avoidable downstream cardiac testing among eligible U.S. ED patients with chest pain. Model inputs included patient demographics, chest pain history, electrocardiogram (ECG) findings, medication use, and lab values (including some nonroutine tests). Our primary outcome was U.S. annual cost savings from reclassifying eligible ED patients using an AI-driven decision-support tool. Secondary outcomes included reductions in short-stay hospitalizations, cancer cases, and cancer deaths due to averted radiation exposure.

Universal adoption of an AI-driven decision-support tool was estimated to save a mean (standard deviation) of $675 million ($340 million) by avoiding 688,000 (147,000) downstream cardiac diagnostic tests. This resulted in annual decreases of 537,000 (115,000) hospitalizations, 8.30 million (1.96 million) bed hours, 490 (150) new cancer diagnoses, and 250 (80) cancer deaths.

Assuming similar recategorization of patients seen in a European outpatient cohort, widespread adoption of an AI-driven clinical decision-support tool to evaluate U.S. ED patients with suspected obstructive coronary disease could yield substantial benefits by reducing avoidable downstream cardiac testing. By accurately identifying patients who do not require urgent diagnostics, this approach could help mitigate crowding, relieve resource strain, and improve patient flow-supporting a more efficient emergency care system.