App uses biomarkers, AI to predict COVID-19 severity
A smartphone app created by researchers at the New York University College of Dentistry uses artificial intelligence and biomarkers in patients’ blood to assess risk factors and predict COVID-19 severity in patients.
“Identifying and monitoring those at risk for severe cases could help hospitals prioritize care and allocate resources like ICU beds and ventilators,” John T. McDevitt, PhD, chairperson of the New York University College of Dentistry, said in a press release. “Likewise, knowing who is at low risk for complications could help reduce hospital admissions while these patients are safely managed at home.”
To develop the point-of-care diagnostic tool for the app, McDevitt and colleagues evaluated data from 160 patients in Wuhan, China, according to a paper published in Lab on a Chip. They identified four biomarkers that were more prevalent in blood tests of patients with COVID-19 who died than in those who recovered — C-reactive protein, myoglobin, procalcitonin and cardiac troponin I.
These biomarkers indicate an increased risk for complications, including acute inflammation, various stages of cardiovascular disease and lower respiratory tract infections, according to the researchers.
Using these biomarkers, along with patient age and sex, the researchers created and trained a model with artificial intelligence (AI) to identify disease patterns and predict severity. Once a patient’s biomarker and other risk factors are entered into the model, it produced a severity score ranging from zero to 100, with the latter indicating a critical case of COVID-19.
McDevitt and colleagues validated the modeling using data from 12 patients hospitalized with COVID-19 in Shenzhen, China. They found that the severity scores produced by the model were significantly higher in patients who died compared with the scores of those who were discharged.
According to the press release, the researchers used data from more than 1,000 patients with COVID-19 in New York City to further validate the model.
The researchers then developed the smartphone app that quickly produces a patients’ severity score to make the point-of-care tool more accessible to clinicians. The app was evaluated and optimized in the Family Health Centers at NYU Langone, and researchers hope to make the app accessible nationally in the next few weeks, according to the press release.
The app is currently meant for research and informational purposes and can be used with existing laboratory tests with oversight from an authorized clinician, according to the release. However, McDevitt and colleagues plan to develop and evaluate the ability to quickly generate COVID-19 severity scores from a test where a drop of blood is placed in credit card-sized cartridge with bio-nanochips and tested for severity biomarkers with a portable analyzer.
“With COVID-19, point-of-care testing, coupled with a decision support system, could improve how clinicians triage patients — and potentially improve their outcomes, particularly for those who need more immediate and aggressive care,” McDevitt said in the press release.
Reference:
McDevitt JT, et al. Lab Chip. 2020;doi: 10.1039/d0lc00373e.
Perspective
Back to Top
It’s certainly an interesting idea, but I think we need more information about the ability to demonstrate that it will work on a going-forward basis and provide real information for nonadmitted patients. We know that you can use these biomarkers when patients are admitted to the hospital and get a gauge of how sick they are, and there are apps that already look at that. What seems interesting about this is that it can take the samples and do the testing as well as the analysis.
As described, this system provides biomarker information not currently readily available in the typical outpatient setting in a rapid fashion, plus it provides an analysis predicting which patients might have a poor outcome. Such patients could be watched more carefully, and as we develop guidelines, started earlier on anti-inflammatory and antiviral medications or plasma therapy. Right now, we usually do not admit such patients and would not typically start IV therapies or plasma therapy through an infusion. If we knew which patients didn’t require hospitalization but were at greater risk, instituting these IV treatments as an outpatient might be beneficial. It remains to be seen whether this product will be able to determine that.
If this methodology can be shown to really be useful and produces useful clinical results, and is something that could be operationalized in an outpatient setting at a reasonable cost — these are pretty big and expensive tests that we do when patients come to the hospital — it would be a valuable tool for the practicing clinician. If we could get information to determine whether a person is at a really low risk, moderate risk or greater risk, it would be significant.
Collapse