Researchers at the Feinstein Institutes for Medical Research, part of Northwell Health, have published a study in Nature Communications examining the use of artificial intelligence (AI) and wearable devices to predict patient deterioration in hospitals. The study found that an AI-powered system could identify patients at risk of serious clinical events up to 17 hours before those events occurred.
The research was led by Theodoros P. Zanos, PhD, head of Northwell’s Division of Health AI and associate professor at the Feinstein Institutes’ Institute of Bioelectronic Medicine and the Institute of Health System Science. Dr. Zanos explained: “Currently, hospital staff rely on intermittent vital sign checks of a patient, like their heart rate or temperature, to identify worsening conditions. Combining clinical wearables with the predictive AI algorithm we have developed, we can help clinicians identify these deteriorations a lot earlier and more accurately than standard-of-care early warning scores, improving patient outcomes.”
In the study, 888 adult non-ICU patients at Northwell hospitals wore VitalPatch biosensors from Vital Connect, Inc., which recorded heart rate, respiratory rate, temperature, and movement data. The AI model continuously processed this information along with demographic details to alert clinicians about patients who were likely to experience rapid health declines.
According to the findings, the AI predicted half of all rapid response team activations—emergency interventions when a patient’s condition suddenly worsens—and over 83% of unplanned ICU transfers involving previously stable patients whose health declined unexpectedly. It also identified every case that would have required intubation or involved cardiac arrest or death within 24 hours.
The researchers noted that their AI system works independently of which clinical wearable is used as long as accurate measurements are available.
