How does an AI diagnose dyspnoea in ED triage without human guidance?

The study aimed to improve ED triage for dyspnoeic adult patients using AI, analyzing data from two Swedish EDs. Key predictive variables prioritized by the AI included previous heart failure diagnosis, atrial fibrillation on ECG, COPD complaints, and specific medications. Surprisingly, veterinary medication intake showed predictive value, while traditional indicators like vital signs and sex were disregarded. This suggests potential insights for AI-driven diagnosis in ED settings, aligning with previous knowledge but with some unexpected findings.