Since the start of the Covid-19 pandemic, clinical researchers have been rushing to discover new drugs and therapies, novel uses for old drugs, and other strategies for combating Covid-19. At the same, a silent partner has been quietly churning away in the background: artificial intelligence (AI). No longer relegated to the sidelines, AI technology is driving new applications to improve Covid-19 planning, treatment, and outcomes, especially on the pandemic frontlines.

Early Detection Helps Preserve Frontline Resources

Early detection and diagnosis of Covid-19 infections rank as one of the most important (and timely) uses for AI technology, according to a recent review in Diabetes & Metabolic Syndrome Journal, with currently available AI platform options ranging from enhanced to AI-guided cardiac ultrasound.

Until now, Covid-19 diagnosis has primarily relied upon commercially available diagnostic platforms that use molecular detection as a framework. Still, gaps remain concerning identifying patients with high mortality risk or with low risk for complications, which can impact resource utilization. To address these gaps, New York University researchers created an app that uses AI to help assess patient risk factors (e.g., age, sex) and key serum biomarkers (e.g., C-reactive protein, myoglobin, procalcitonin [PCT], and cardiac troponin), all critical drivers of Covid-19 complications and mortality. These data are then fed into a statistical learning algorithm — the Covid-19 Severity Score — to predict mortality; scores range from 0 (mild to moderate) to 100 (critical).

In a study published in Lab on a Chip journal, the researchers explain that they first developed the Covid-19 Disease Severity model based on data from 160 hospitalized Wuhan Chinese patients. Findings showed that males accounted for 70% of deaths, had significantly higher biomarker levels, and were substantially older than discharged patients. These data were then validated in 12 hospitalized Shenzhen, Chinese patients, with significantly higher median Covid-19 Severity Scores in patients who died vs. those who were discharged.

What’s Next?

Preliminary data from a general New York City patient population demonstrates that other biomarkers might help discern patients at most significant risk for severe Covid-19 illness (e.g., D-dimer for thrombotic events or PCT for bacterial co-infections). However, additional research is needed to evaluate others driving serious manifestations, such as cytokine storms. Meanwhile, the study’s lead investigator John McDevitt, Ph.D., says that plans to develop and scale the app for point-of-care serum testing are underway. Future work may also involve a test for population-based Covid-19 surveillance in clinical or public settings to support healthcare workers on the frontlines tasked with rapidly diagnosing and triaging patients needing the most care.