Artificial Intelligence Predicts Severe Disease in COVID-19 Patients

Using artificial intelligence, NYU researchers accurately predicted which patients newly diagnosed with COVID-19 would go on to develop severe respiratory disease, according to a study published in Computers, Materials & Continua.

For more coronavirus updates, visit our resource page, updated twice daily by Xtelligent Healthcare Media.

The rapid spread of COVID-19 around the world has given way to an urgent need to identify which cases will escalate to critical illness, the research team said. While over 80 percent of cases seem to be mild, those who develop more severe symptoms often need oxygen and prolonged ventilation.

Acute respiratory distress syndrome (ARDS), fluid buildup in the lungs that can be fatal in the elderly, is a key feature in COVID-19 patients who experience declining outcomes. The team wanted to determine whether AI techniques could help accurately predict which patients with the virus would go on to develop ARDS.

Researchers collected laboratory, demographic, and radiological findings from 53 patients who tested positive for COVID-19 at two Chinese hospitals, with an average age of 43 years. The team then used the data to train AI models designed to get smarter with the more data they consider.

“Our goal was to design and deploy a decision-support tool using AI capabilities—mostly predictive analytics—to flag future clinical coronavirus severity,” said co-author Anasse Bari, PhD, a clinical assistant professor in computer science at NYU Courant Institute of Mathematical Science.

“We hope that the tool, when fully developed, will be useful to physicians as they assess which moderately ill patients really need beds, and who can safely go home, when hospital resources are stretched thin.”

The results showed that characteristics thought to be hallmarks of COVID-19, like certain patterns in lung images, fever, and strong immune responses, were not useful in predicting which patients with initial mild symptoms would go on to develop severe lung disease.

Age and gender were also not helpful in predicting critical illness, although past studies have found men over 60 to be at higher risk.

The AI tool found that changes in three features, including levels of the liver enzyme alanine aminotransferase (ALT), reported myalgia, and hemoglobin levels, were most accurately predictive of subsequent, severe disease. With these factors, the team was able to predict risk of ARDS with up to 80 percent accuracy.

The researchers noted that ALT levels rise dramatically as diseases like hepatitis damage the liver. ALT levels were only a bit higher in patients with COVID-19, but still played a major role in predicting disease severity. Myalgia, or deep muscle aches, were also more common in COVID-19 patients, and past research has linked myalgia to higher inflammation in the body.

Higher levels of hemoglobin, the iron-containing protein that enables blood cells to carry oxygen to bodily tissues, were also strongly linked to subsequent respiratory distress. The team pointed out that this could be explained by other factors, such as unreported smoking of tobacco, which has been associated with increased hemoglobin levels.

Some limitations of the study included the relatively small dataset and the limited clinical severity of disease in the population studied. The latter could be due to an unexplained lack of older patients admitted into the hospitals during the study period.

Further refining the model using more data from different settings will help improve its predictive power, the team said.

“While work remains to further validate our model, it holds promise as another tool to predict the patients most vulnerable to the virus, but only in support of physicians’ hard-won clinical experience in treating viral infections,” said corresponding study author Megan Coffee, MD, PhD, clinical assistant professor in the Department of Medicine and member of the Division of Infectious Diseases and Immunology at NYU Langone.

The study demonstrates the ability for AI and predictive analytics to augment and support clinicians, especially in the midst of a global health crisis.

“This study shows that predictive analytics can play a role in augmenting clinical skills in distinguishing between ‘sick’ from not ‘sick,’” researchers said.

“Just as predictive text is intended to augment, but not replace writers, the goal is not to create a black box to supersede clinical reasoning, but to create models that can provide insight. Clinical acumen is based on both personal learning and collective professional learning; machine learning can add further insight.”

The advent of AI in healthcare has opened up new possibilities for tracking and monitoring the pandemic, which could help improve outcomes and clinical care responses.

“I will be paying more attention in my clinical practice to our data points, watching patients closer if they for instance complain of severe myalgia,” said Coffee. “It’s exciting to be able to share data with the field in real time when it can be useful. In all past epidemics, journal papers only published well after the infections had waned.”