Study Suggests that AI May Help Identify Stroke More Effectively than Humans During Emergency Calls

According to results from a Danish retrospective study presented at the European Stroke Organisation Conference 2023, a new artificial intelligence (AI) framework identified stroke in emergency calls more effectively than human call handlers. The researchers used a dataset based on the Danish Stroke Registry which included over 1.5 million calls made to Copenhagen Emergency Medical Services (EMS) from 2015 to 2020. Over 7000 of these calls were related to stroke. Researchers developed a machine learning AI framework which first transcribed the audio calls into text and then analyzed the transcripts to determine risk of stroke.

Using emergency calls made in 2021, researchers next evaluated the AI’s precision (positive predictive value) and recall (sensitivity) in recognizing stroke. The AI system achieved a precision of 24.9% and a sensitivity of 63.0% compared with results from human call handlers who achieved a precision of 17.1% and a recall of 52.7%. Emergency calls, such as those made to Copenhagen EMS, are a critical first step in optimal stroke treatment. It is therefore essential to quickly and accurately assess whether callers to emergency service centers are experiencing a stroke.

“Many stroke cases go undetected at this stage, leading to delays in treatment that can have potentially life-threatening consequences for patients,” said study author Dr Jonathan Wenstrup. "With the implementation of this new, cost-effective supporting tool, we can enhance stroke identification and ensure more patients receive appropriate and timely care, ultimately improving patient outcomes."

Almost 800,000 Americans experience stroke each year. Researchers speculate that in the future healthcare telecommunications services may use AI tools to analyze emergency calls and help identify cases of stroke, thereby improving access to emergency medical care and potentially saving lives.