Machine Learning Helps Identify Pneumonia as Driver of COVID-19 Deaths

Northwestern University researchers found that secondary bacterial pneumonia was a major driver of COVID-19 deaths in critically ill patients following a machine learning (ML)-based analysis of medical record data. 

The findings, which were published last month in the Journal of Clinical Investigation, examined the association between secondary bacterial pneumonia that does not resolve and death in patients with COVID-19. According to the study, secondary bacterial pneumonia was common among these patients, affecting approximately half of critically ill COVID-19 patients. 

The researchers analyzed data from 585 patients with severe pneumonia and respiratory failure in the intensive care unit (ICU) at Northwestern Memorial Hospital enrolled in the Successful Clinical Response to Pneumonia Therapy (SCRIPT) study. Of these, 190 had COVID-19.  

Lung samples collected from participants as part of SCRIPT to help diagnose and assess the outcomes of secondary pneumonia events were also incorporated into the study. 

Using an ML tool known as CarpeDiem, the research team grouped similar ICU patient-days into clinical states using information from each patient’s electronic health record (EHR). The approach, based on the concept of ICU daily rounds, enabled the researchers to evaluate how complications like pneumonia ‘superinfection’ impacted the trajectory of a patient’s COVID-19 infection. 

Thus far, research in this area is somewhat limited, the researchers noted. 

“The importance of bacterial superinfection of the lung as a contributor to death in patients with COVID-19 has been underappreciated, because most centers have not looked for it or only look at outcomes in terms of presence or absence of bacterial superinfection, not whether treatment is successful or not,” explained study co-author Richard Wunderink, MD, who leads the Successful Clinical Response in Pneumonia Therapy Systems Biology Center at Northwestern, in the news release detailing the study. 

Overall, the research team found that the success or failure of a treatment was significantly correlated with outcomes for patients in the study cohort. 

“Those who were cured of their secondary pneumonia were likely to live, while those whose pneumonia did not resolve were more likely to die,” said senior author Benjamin Singer, MD, the Lawrence Hicks Professor of Pulmonary Medicine in the Department of Medicine and a Northwestern Medicine pulmonary and critical care physician. “Our data suggested that the mortality related to the virus itself is relatively low, but other things that happen during the ICU stay, like secondary bacterial pneumonia, offset that.” 

The researchers also found no evidence of cytokine storm, a phenomenon the study authors indicated has been proposed as a major driver of COVID-19 mortality. 

“The term ‘cytokine storm’ means an overwhelming inflammation that drives organ failure in your lungs, your kidneys, your brain and other organs,” Singer stated. “If that were true, if cytokine storm were underlying the long length of stay we see in patients with COVID-19, we would expect to see frequent transitions to states that are characterized by multi-organ failure. That’s not what we saw.” 

Moving forward, the researchers will continue work in this area focused on using molecular data from participant lung samples combined with ML to investigate why some pneumonia is resolved or cured in some patients, but not in others. 

The research team also aims to expand their technique to larger cohorts and datasets to further improve the model’s predictions and use these to improve care, the press release notes.