AI Tool Reduces Hospital Deaths by 26%
The use of an artificial intelligence (AI) system called Chartwatch led to a 26% drop in the number of unexpected deaths among patients in hospital, a study has found.
The research team looked at 13,000 admissions to the general internal medicine ward in St Michael’s Hospital in Toronto. The ward is an 84-bed unit that cares for some of the hospital’s most complex patients, and researchers wanted to see the impact Chartwatch had on deaths. Over a period of 18 months, they compared the effect of the AI tool among that patient population with thousands of patients admitted to other subspecialty units. The findings have been published in the Canadian Medical Association Journal.
Unity Health, which runs St Michael’s, has an AI team, which started developing Chartwatch in 2017, based on suggestions from staff that predicting deaths or serious illness could be key areas where machine learning could make a positive difference. The technology underwent rigorous development and testing before it was deployed in October 2020.
Dr Amol Verma, the lead author of the study and professor of AI research and education in medicine at University of Toronto, explained how the system works: “Chartwatch measures about 100 inputs from [a patient’s] medical record that are currently routinely gathered in the process of delivering care. So a patient’s vital signs, their heart rate, their blood pressure … all of the lab test results that are done every day.”
Working in the background alongside clinical teams, the tool monitors any changes in someone’s medical record “and makes a dynamic prediction every hour about whether that patient is likely to deteriorate in the future,” Verma added. That could mean someone getting sicker, or requiring intensive care or even being on the brink of death, giving doctors and nurses a chance to intervene.
In some cases, those interventions involve escalating someone’s level of treatment to save their life, or providing early palliative care in situations where patients can’t be saved. In either case, the researchers said, Chartwatch appears to complement clinicians’ own judgment and leads to better outcomes for fragile patients, helping to avoid more sudden and potentially preventable deaths.
While there was a drop in the number of deaths in the unit using Chartwatch, there was no similar drop in the number of deaths in the units not using it.
Verma said the success rate of Chartwatch made it particularly notable: “Very few AI technologies have actually been implemented into clinical settings yet. This is, to our knowledge, one of the first in Canada that has actually been implemented to help us care for patients every day in our hospital.”
In one example, a patient suffering from a cat bite and a fever otherwise appeared fine until an alert from Chartwatch showed that his white blood cell count was unusually high. The cause turned out to be cellulitis, a bacterial skin infection, which can be highly dangerous and sometimes fatal. Because it was flagged early, however, the nursing team were able to give the patient antibiotics promptly before he deteriorated.
Verma also acknowledged the limitations of the study, however. It took place during the Covid-19 pandemic, at a time when health care faced an unusual set of challenges. The hospital’s patient population is also distinct, with a high level of complex patients, including individuals facing homelessness, addiction and overlapping health issues.
“Our study was not a randomised controlled trial across multiple hospitals. It was within one organisation, within one unit,” Verma said. “So before we say that this tool can be used widely everywhere, I think we do need to do research on its use in multiple contexts.”
