Sepsis
Johns Hopkins University: an early warning system tested on 590,000 patients in 5 hospitals cut deaths from sepsis by 20% and found severe cases almost 6 hours earlier; more than 4,000 clinicians used it (2022). A university hospital in San Diego: an algorithm using more than 150 parameters in two emergency departments cut deaths from sepsis by 17% (2024).
Patient decline
A Toronto hospital: unexpected deaths in the general medicine ward fell by 26% among 13,649 patients. High-risk cases numbered 482 in 19 months with the system, against 1,656 in 43 months before it (CMAJ, 2024). South Korea: according to the developer, a 24-hour cardiac arrest forecast runs on more than 50,000 beds in more than 20 top-level hospitals (2025). Russia: the Almazov Center assesses the risk of complications for all its hospital patients every day, with 89% accuracy.
Ambulance and primary care
The Yaroslavl Region: AI ECG in ambulances and rural first-aid posts, 848 patients with acute coronary syndrome sent to emergency care, and feedback from a cardiologist within minutes.
What a clinic can take from this
Fewer alerts, but more accurate ones: a system that pings the doctor for every reason loses the doctors’ trust. Rollout is half the result: at Johns Hopkins, the system worked once thousands of doctors used it. For a private clinic without a hospital ward, the same principle applies to notifications for the administrator.