Stanford (Calif.) Health clinicians are testing a machine-learning algorithm developed by Epic that may be able to identify whether a COVID-19 patient will need intensive care before their condition deteriorates, according to an April 1 Stat report.
Epic built the model, called the Deterioration Index, and trained it on data from hospitalized patients before the COVID-19 pandemic. Stanford chose to use Epic's model because it is already integrated with the hospital's EHR. Epic recently updated the model to help hospitals evaluate how well it works for COVID-19 patients, according to the report.
The machine learning model that Stanford is working with examines patient data and assigns patients a score based on how sick they are and how likely they are to require more intensive care. If the algorithm can be validated, Stanford plans to use it to initiative steps such as alerting a nurse to check in more frequently or order tests to help support physicians' decisions about COVID-19 care.
Prior to the coronavirus pandemic, Stanford was working to validate the model on its general population of hospitalized patients. The algorithm is designed for hospitalized patients who have not yet been admitted to the intensive care unit. The model analyzes patient data including vital signs, lab results, medications and medical history and then generates a composite score based on a 100-point scale. The higher number correlates with a higher concern that the patient's condition is becoming worse.
However, the researchers are more focused on dramatic changes in the score rather than a high number, said Ron Li, MD, a clinical informaticist at Stanford who is leading the project.
"If a patient's score is 70, which is pretty high, but it's been 70 for the last 24 hours — that’s actually a less concerning situation than if a patient scores 20 and then jumps up to 80 within 10 hours," Dr. Li said.