– Machine learning models have the ability to predict the possibility of critical illness or mortality in COVID-19 patients, according to a Journal of Medicine Internet Research (JMIR) study.
Researchers from Mount Sinai analyzed health record data from over 4,000 adult COVID-19 patients admitted to hospitals in the Mount Sinai Health System in New York to predict critical events or mortality three, five, seven, and ten days from admission.
They looked to study patient characteristics at admission and access the performance of machine learning models at multiple hospitals and time points.
Patient information included past medical history, comorbidities, vital signs, and laboratory test results at admission. This data would allow researchers to predict critical events, including intubation and mortality without clinically relevant time windows.
The machine learning model, XGBoost classifier, outperformed baseline models for mortality, with results of 0.89 at 3 days, 0.79 at 5 days, 0.80 at 7 days, and 0.84 at 10 days.
At week one, acute kidney injury, fast breathing, high blood sugar, and elevated lactate dehydrogenase were the strongest drivers in predicting critical illness.
On the other hand, old age, blood level imbalance, and C-reactive protein levels were the leading drivers in predicting mortality.
Overall, the machine learning model identified at-risk patients and uncovered underlying relationships that predicted outcomes, researchers said.
“We have created high-performing predictive models using machine learning to improve the care of our patients at Mount Sinai,” Girish Nadkarni, MD, assistant professor of medicine (Nephrology) at the Icahn School of Medicine, clinical director of the Hasso Plattner Institute for digital health at Mount Sinai, said in a press release earlier this week.
“More importantly, we have created a method that identifies important health markers that drive likelihood estimates for acute care prognosis and can be used by health institutions across the world to improve care decisions, at both the physician and hospital level, and more effectively manage patients with COVID-19.”
Patients with COVID-19 experience varying symptoms, which can oftentimes make effective patient triaging difficult, researchers noted. Although some patients are asymptomatic, others experience severe acute respiratory distress syndrome, multiorgan failure, or death.
Identifying key patient characteristics is important because it gives physicians and hospitals the potential ability to predict disease trajectory and leverage their resources efficiently to improve patient outcomes.
When the COVID-19 pandemic began rapidly spreading around the world, hospitals had to figure out how to accurately forecast the number of patients who would need hospitalization, who would need treatment in the intensive care unit and require a ventilator, how long the patients would stay, and how much personal protective equipment would be needed.
To combat these issues, top healthcare organizations have worked to develop different machine learning platforms to forecast staffing needs, hospitalization volumes, and the rate of confirmed COVID-19 cases
In mid-August, Cedars-Sinai developed a machine learning tool to help anticipate and prepare for increasing COVID-19 patient volumes with 85 percent to 95 percent accuracy.
The model intended to optimize the way care is provided at the hospital, including data points regarding patients’ vital signs and their length of stay, to predict the most effective treatments for patients.
The platform could also pinpoint the likelihood that a patient will be readmitted and the patients who will be most satisfied with their hospital experience
The new positive results from Mount Sinai’s machine learning model show the progress that the healthcare industry as a whole has made in developing these technologies to combat the pandemic and enhance patient-centered care.
“We have built machine learning models using patient data to predict outcomes,” Benjamin Glicksberg, PhD, assistant professor of genetics and genomic sciences at the Icahn School of Medicine at Mount Sinai and one of the study’s principal investigators, said in the press release.
“Now in the early stages of a second wave, we are much better prepared than before. We are currently assessing how these models can aid clinical practitioners in managing care of their patients in practice.”
Credit: Google News