Dozens of hospitals including University of Michigan and Stanford (Calif.) University are using a repurposed artificial intelligence tool developed by Epic to help identify risk levels of hospitalized COVID-19 patients, STAT reports.
Epic's index uses machine learning to analyze patient data such as vital signs, lab results and nursing assessments to determine how likely the patient is to require intensive care. The tool assigns patients a score, ranked 0 to 100, with a higher number indicating an elevated risk of deterioration. This allows clinicians to determine ahead of time whether the patient may need a ventilator, so clinicians can ensure an intensive care unit bed and breathing assistance is available.
Epic built and trained the model on data from hospitalized patients before the COVID-19 pandemic. Hospitals typically would take more time to test the tool on hundreds of patients to refine the algorithm, but because COVID-19 is a new disease, physicians are shortening the validation process to begin using the tool to support care. These hospitals include Fort Wayne, Ind.-based Parkview Health, Wenatchee, Wash.-based Confluence Health and Toledo, Ohio-based ProMedica. Stanford University and Ann Arbor-based University of Michigan are also testing the technology.
"Nobody has amassed the numbers to do a statistically valid" test of the AI, said Mark Pierce, MD, chief medical informatics officer at Parkview Health, according to the report. "But in times like this that are unprecedented in U.S. healthcare, you really do the best you can with the numbers you have, and err on the side of patient care."
At University of Michigan, clinicians found that about 9 percent of patients whose scores were low during the first 48 hours of hospitalization were unlikely to experience a life-threatening event, and clinicians were able to consider moving them to a field hospital for lower-risk patients. Conversely, the health system found that 10 percent to 12 percent of patients with higher scores were much more likely to need intensive care.