Hospitals are always looking for ways to swiftly recognize patients who may be at risk of sudden health declines. The time taken to identify such risks can mean the difference between saving a life or losing one. Fortunately, artificial intelligence (AI) is becoming a valuable asset in this area. A recent study from Canada has introduced a solution: an AI tool designed to alert healthcare professionals early. This tool, named CHARTwatch, was launched at St. Michael’s Hospital in Toronto to assist doctors and nurses in detecting warning signs of patient deterioration and reacting more quickly.
Impact on Patient Mortality
The study indicated that the General Internal Medicine (GIM) unit at the hospital experienced a reduction in specific types of patient fatalities when this system was utilized. The emphasis was on “non-palliative” deaths, which refer to deaths that occur without the patient being in palliative care. Palliative care focuses on improving the quality of life for patients with severe, often life-threatening conditions, rather than curing the illness. The aim of the AI tool is to prevent unexpected or unplanned deaths, excluding those that happen while patients are in palliative care.
The research, conducted from 2016 to 2022, included over 13,000 patient admissions in the GIM unit. Results showed a 26% relative decrease in non-palliative deaths during the implementation period, reducing the rate from 2.1% to 1.6%. Although this percentage may appear small, it represents a significant effect on patient outcomes. In a hospital environment, even slight decreases in mortality rates can lead to dozens of lives saved over time.
Real-Time Alerts for Better Care
CHARTwatch provides real-time notifications to healthcare providers, enabling them to take immediate actions when a patient displays signs of rapid decline. For high-risk patients identified by this system, non-palliative deaths fell from 10.3% to 7.1%. Additionally, following the introduction of CHARTwatch, there was an increase in the proactive care given to patients. For example, dosages of antibiotics and corticosteroids were raised accordingly, and vital signs were monitored more frequently than before.
Cautionary Notes from Researchers
While this development is certainly encouraging, researchers have emphasized the need for caution in their findings. The study was not randomized, which means other factors could have impacted the results. Moreover, the research was centered on one hospital unit, and outcomes may vary significantly in larger settings. Nonetheless, this study provides important evidence that machine learning tools will be essential in healthcare, improving patient care and potentially saving numerous lives in the future.
Canadian Medical Association Journal
