AI tool shows promise in predicting deteriorating patient condition

An artificial intelligence tool accurately predicted which inpatients would have worsening conditions during the COVID-19 pandemic, an Agency for Healthcare Research and Quality study found. Published in Physiological Measurement, the randomized controlled trial included more than 10,400 patient visits and is one of the first to externally validate an AI predictive tool in a prospective cohort after being implemented. The tool, developed five years before the onset of the pandemic, particularly was accurate in predicting declines due to respiratory issues and remained accurate in spite of changes in clinical practice over time. The researchers concluded real-world validation is important for predictive models. (AHRQ News article, 7/23/24)