Poster

Design of an AI Support for Diagnosis of Dyspneic Adults at Time of Triage in the Emergency Department

Tolestam Heyman, E., Ashfaq, A., Ekelund, U., Ohlsson, M., Björk, J., Khoshnood, A. M. & Lingman, M.

European Emergency Medicine Congress

Organizer:: European Society for Emergency Medicine (EUSEM)
Location:: Barcelona, Spain
Date:: 17 – 20 September 2023

Abstract

We developed an AI-based diagnostic support tool for adults presenting with dyspnea at emergency department triage. Using complete regional healthcare data, the model identified key predictors, including prior heart failure or COPD, daily smoking, atrial fibrillation or flutter, life difficulties, and maternal healthcare contacts. Sensitivity was 75% for acute heart failure, 93% for COPD exacerbation, and 54% for pneumonia, with specificity above 75%. Each visit generated an individual graph showing the basis of the AI decision.

PDF