Researchers Evaluate Safety-Constrained Multi-Agent AI for Emergency Department Dispositions
A study published in Nature Health Sciences assessed SMARTriage AI, a multi-agent language model system designed to help emergency departments make admission or discharge decisions.
Key points
- SMARTriage AI combined language models with deterministic safety rules for emergency department choices.
- Initial development tests showed high apparent accuracy, but performance dropped when physician summaries were withheld.
- A redesigned version under a locked protocol failed to outperform a standard single-agent language model.
Researchers at Tel Aviv University and Sheba Medical Center have evaluated a multi-agent artificial intelligence system called SMARTriage AI, intended to assist emergency departments with patient disposition choices such as admission or discharge. The retrospective single-center study analyzed 468 valid adult internal medicine and geriatric encounters from a single 24-hour period.
The system combined role-specialized language model agents with deterministic rules acting as safety gates. Initially, the development cohort showed an apparent accuracy of 88.2% and an admission recall of 96.1%. However, performance varied significantly depending on whether physician summaries were included in the prompts, dropping to 78.2% and further down to 65.0% when physician text was withheld. Furthermore, a redesigned system evaluated under a locked protocol performed worse than a single language model agent alone, achieving an area under the curve of 0.794 compared to 0.821–0.823.
The study authors noted important limitations, including that apparent performance was optimistically biased because the system was developed and evaluated on the same cohort without a held-out set. They emphasize that evaluation on unseen data, alongside external and prospective validation, is necessary before any conclusions can be drawn about clinical utility.
What to know: The AI system did not outperform simpler single-agent models in rigorous testing, and further prospective validation is required. Always consult healthcare professionals for medical decisions.
Source Nature Health Sciences — nature.com/articles/s41598-026-80974-4 ↗
Read next
🏃FitnessUnderstanding and Addressing Love Handles Through Lifestyle Changes
Experts clarify that spot reduction of fat is not physiologically possible, recommending a combination of calorie management and high-intensity interval training instead.
🏃FitnessCiele Releases New Woolmark-Certified Merino Winter Running Apparel
Montreal-based brand Ciele has launched its second collection of merino wool running gear, featuring performance-focused fabrics designed for cold-weather training.
🏥Public HealthPreventive Services Task Force Resumes Meetings Under New Leadership
The U.S. Preventive Services Task Force is set to reconvene this week for the first time in over a year following a significant membership overhaul by Health Secretary Ro…
❤️Conditions & CareBlood-Based Genetic Signature May Predict CAR T-Cell Therapy Success in Lymphoma
Researchers have identified a specific RNA signature in blood plasma that may help predict how well lymphoma patients respond to CAR T-cell therapy.
🧠Mind & Mental HealthAddressing Body Image and the Reality of Gluteal Anatomy
A personal reflection on the social and psychological impact of body shape, specifically regarding the gluteal region, and the limitations of consumer solutions for perce…
❤️Conditions & CareMigraine and Smoking Linked to Increased Early Stroke Risk in Women
A large-scale study suggests that people who have given birth and experience migraines while also smoking face a significantly higher risk of stroke before age 60.