AI, EHR Data Can Help Predict COVID-19 Outcomes, Mortality

February 09, 2021 - Artificial intelligence could help identify factors in EHR data that are predictive of poor COVID-19 outcomes and mortality, according to a study published in npj Digital Medicine.

The vast amount of data contained in EHRs can provide a valuable asset to researchers during crises like the current pandemic, the team noted.

For the study, the research group used the COVID-19 datamart, a frequently refreshed snapshot of longitudinal data on patients with a COVID-19 infection flag from many data sources across the Mass General Brigham hospital system.

Using EHRs from more than 16,000 students, the team applied an artificial intelligence algorithm to identify 46 clinical conditions representing potential risk factors for death after a COVID-19 infection.

"Despite relying on only previously documented demographics and comorbidities, our models demonstrated performance comparable to more complex prognostic models requiring an assortment of symptoms, laboratory values and images gathered at the time of diagnosis or during the course of the illness," said Zachary Strasser, MD, co-lead author and postdoctoral fellow at MGH.

The study found that age was the most important predictor of mortality in COVID-19 patients. Researchers also identified a history of pneumonia as a significant risk factor, as well as histories of diabetes with complications and cancer – particularly breast and prostate – among COVID-19 patients between the ages of 45 and 65.

In patients between the ages of 65 and 85, diseases affecting the pulmonary system, including interstitial lung disorders, chronic obstructive pulmonary disease (COPD), lung cancer, and a history of smoking, were strong predictors of poor outcomes.

Comorbidities registering the highest odds ratios for death irrespective of age were chronic kidney disease, heart failure, abdominal aortic aneurysm, hypertension, and aortic valve disease.

The study also found that females were at lower risk of death from COVID-19. Even after adjusting for age and chronic diseases, researchers found that women benefit from an unknown form of underlying protection against the worst outcome of COVID-19. However, in the oldest cohort of patients the team did find that being African American was associated with a higher chance of mortality.

Researchers noted that EHR data and predictive analytics tools become especially critical in the wake of a healthcare crisis like the COVID-19 pandemic, when access to large-scale clinical trial grade data isn’t available.

"The ability to quickly utilize data that has already been collected across the country to compute individual-level risk scores is crucial for effectively allocating and distributing resources, including prioritizing vaccination among the general population," said Shawn Murphy, MD, PhD, senior author and chief research informatics officer at Mass General Brigham.

The approach used in the study could help hospitals and health systems manage limited therapeutic and preventive resources to treat the virus, researchers noted.

"By combining computational methods and clinical expertise, we developed a set of models to forecast the most severe COVID-19 outcomes based on past medical records, and to help understand the differences in risk factors across age groups," said co-lead author Hossein Estiri, PhD, an investigator in the Laboratory of Computer Science at MGH and an assistant professor of Medicine at Harvard Medical School (HMS).

"Many prior studies have isolated small subsets of EHR data from after the infection, but ours is the first and largest to use entire historical medical records to try to untangle the role of age as the most important risk factor for COVID adverse outcomes."

Throughout the pandemic, researchers have leveraged EHR data to uncover insights about patient outcomes from COVID-19. In April 2020, a study published in JAMA showed that EHR data could reveal the most common chronic diseases in patients hospitalized with COVID-19.

The researchers found that hypertension, diabetes, and obesity are key chronic diseases in the acuity of the virus.

“To our knowledge, this study represents the first large case series of sequentially hospitalized patients with confirmed COVID-19 in the US,” researchers stated.

“Older persons, men, and those with pre-existing hypertension and/or diabetes were highly prevalent in this case series and the pattern was similar to data reported from China. However, mortality rates in this case series were significantly lower, possibly due to differences in thresholds for hospitalization.”