Hospital App Alerts On Deadly Condition

The Montefiore Health System and its affiliated medical school are using artificial intelligence to better predict a common emergency among patients in their care, among other conditions.
 The first application focused on using AI to predict which patients will develop respiratory failure that requires mechanical ventilation, a project that has so far yielded better results than traditional alerting systems.
 Parsa Mirhaji heads the entity involved, the Center for Health Data Innovations at Montefiore Health System and Albert Einstein College of Medicine. Dr. Mirhaji said his team has three AI applications that are live and four more under development, and the results have been promising.
 Montefiore has 11 hospitals in the Bronx, Westchester and the Hudson Valley. Three hospitals use the AI application that predicts acute respiratory failure. Officials plan to expand it to more hospitals.
 The AI application analyzes more than 40 data points for each patient every few hours, including vital signs such as blood pressure, and assigns a score for patients at risk of intubation or death. Patients whose scores exceed a certain threshold trigger an alert.
 The application is part of a study that is expected to wrap up this summer. The initial results were published last fall.
 The application was able to correctly identify 63% of the patients in the 2013 data who died or had a lengthy intubation within 48 hours. That compares with an accuracy rate of 16% to 28% for traditional hospital alert systems that warn about patients at risk of death or other conditions, the study said.



Add new comment