Capital Region of Denmark's AI-Assisted Emergency Call Triage — Corti's Speech-Recognition System for Cardiac Arrest and Stroke Detection

In Copenhagen, Denmark, Emergency Medical Services began using Corti, an AI-based speech-recognition and machine-learning assistant, to listen live to 112 emergency calls and flag likely cardiac arrests to dispatchers in real time, drawing on cues including keywords and abnormal breathing patterns.

An early evaluation covering 161,650 analyzed calls found Corti correctly identified cardiac arrest in 95% of cases, compared with a 73% recognition rate for human dispatchers in the same system, per reporting from the World Economic Forum and the Sudden Cardiac Arrest Foundation. Reporters flagged an important gap: the full 161,650-call study had not been peer-reviewed or published in detail at the time, leaving the false-positive rate undocumented.

Extension to stroke detection (peer-reviewed) — A 2022 retrospective study published in a peer-reviewed journal analyzed 9,049 stroke-related emergency calls from the Capital Region of Denmark (population 1.85 million, 2016-2018) and modeled that automatic speech-recognition software, calibrated on Corti's cardiac-arrest performance gains, could raise dispatcher stroke-detection rates and increase the share of patients receiving thrombolysis within the critical window from roughly 16% to 18% — a projection, not yet a measured deployment outcome.

Caveats — Researchers explicitly noted the stroke projection assumed the speech-recognition software would perform on stroke as well as it does on cardiac arrest, despite stroke symptoms being less specific (5-14.9% of cases are "stroke mimics", versus 0.6% for cardiac arrest), and that the AI "cannot explain how it makes decisions" — an acknowledged transparency limitation.

Read the full analysis: https://www.weforum.org/stories/2018/06/this-ai-detects-cardiac-arrests-during-emergency-calls/