The University of Virginia Health System has rolled out an AI-based predictive analytics tool that monitors COVID-19 patients continuously and predicts whether they are likely to take a turn for the worse.
The system, CoMET, draws moment-to-moment data from a patient’s EHR, including EKG, lab results, and vital signs, to create a graphic summarizing their risk of experiencing a serious event over the next 12 hours. The model updates Itself every 15 minutes. UVA Health uses CoMET to monitor its Medical Intensive Care Unit, Special Pathogens Unit, Cardiovascular Intensive Care Unit, Critical Care Unit, Surgical Intensive Care Unit, and Intermediary Care Unit.
CoMET generates a graphic display (known as a “comet”) allowing clinicians to see how individual patients are doing without reading reports or digging through the EHR. While the comet produced by stable patients is small, yellow and sits close to the X-Y axis on the display, the comets grow, turn bright orange or deep red and expand across the display. These changes can indicate cardiovascular instability, respiratory instability or both.
According to CoMET creator and cardiologist Dr. Randall Moorman, the system is designed to track rapid, unpredictable changes in status which are difficult to capture using traditional means. “Vital sign measurements and labs can come too late,” Moorman said in a prepared statement.
UVA Health just began a two-year study of the CoMET software, funded by a $600,000 grant, embracing its entire fourth floor, Over the next two years, cardiologist Jamie Bourque and School of Nursing professor Jessica Keim-Malpass are assigning CoMET displays to half the beds and comparing the outcomes of patients in the experimental and control groups of patients.
CoMET is one of a rapidly expanding list of tools designed to predict COVID risks and manage COVID patient flow.
For example, a number of healthcare organizations are working with an Epic tool known as the Deterioration Index, including dozens of US health systems and hundreds of hospitals.
More recently, the Cleveland Clinic worked with Epic to create a COVID-19 risk prediction model drawing on both patient entered data and clinical information. Researchers at the Cleveland Clinic developed and tested the model using clinical data from more than 11,000 of its patients. When in use, the tool integrates both medical records from Epic and patient information collected by the patient using its MyChart portal.
It will be interesting to see whether CoMET’s use of a graphical display makes it easier for physicians to check in on patients at a glance. Hopefully, UVA will know well before the two year study is completed whether this approach improves treatment for pandemic patients.
