Machine Learning Tool Provides ED Clinical Decision Support 

University of Minnesota Medical School researchers evaluated the real-time performance of a machine learning tool that supported clinical decision-making for emergency department discharges at M Health Fairview Hospital.  

The multidisciplinary team of intensivists, hospitalists, emergency doctors, and informaticians studied the real-time performance of a machine learning-enabled COVID-19 prognostic tool. The tool can provide clinical decision support for emergency department providers to conduct shared decision-making with patients regarding emergency department discharge.  

“COVID-19 has burdened healthcare systems from multiple different facets, and finding ways to alleviate stress is crucial,” said an assistant professor at the U of M Medical School and Medical Director of M Health Fairview University of Minnesota Medical Center, Monica Lupei, MD, in a press release. 

The training data set included 1,469 patients that tested positive for COVID-19 within 14 days of acute care, hospital-based visit including emergency department, observation, and inpatient encounters between March 4 to August 21, 2020. The final training model included 2,041 patients.  

Led by Lupei, the research team successfully developed and implemented the COVID-19 prediction model in M Health Fairview’s health care system at 12 sites. The tool performed well across all genders, races, and ethnicities, according to researchers.  

Additionally, the logistic regression algorithm created to predict severe COVID-19 performed well in the patients under investigation, although developed on a COVID-19 positive population.  

As COVID-19 hospitalizations continue to rise in parts of the country, hospitals are facing resource shortages. The machine learning COVID-19 prognosis tool can assist physicians in decision-making and allocating resources to those that need them most.  

“Clinical decision systems through ML-enabled predictive modeling may add to patient care, reduce undue decision-making variations and optimize resource utilization — especially during a pandemic,” Lupei said.  

According to researchers, the machine learning-enabled logical regression model can be developed, validated, and implemented for clinical decision support across multiple hospitals while maintaining high performance and remaining equitable.  

“COVID-19 has burdened healthcare systems from multiple different facets, and finding ways to alleviate stress is crucial. Clinical decision support through machine learning-enabled predictive modeling may add to patient care, reduce undue decision-making variations, and optimize resource utilization, especially during a pandemic,” the study concluded. 

Lupei suggests that the effect on patient outcome and resource use needs to be examined and further researched with the machine learning model.