Researchers are developing algorithms to accurately identify people with Long COVID by examining their electronic health records.

Researchers are developing algorithms to accurately identify people with Long COVID by examining their electronic health records.

What you need to know

Researchers supported by the National Center for Advancing Translational Sciences (NCATS) and the National Heart, Lung, and Blood Institute are developing models that can potentially find people who have Long COVID based on their medical records.

What did the researchers do?

Using the NCATS National COVID Cohort Collaborative (N3C), researchers collected deidentified data voluntarily shared by a group of 1.5 million adults who had been diagnosed with COVID-19 or had a positive COVID-19 test more than 90 days before.

Researchers started by examining the records of patients at three of the 59 sites that N3C gathered data from — about 100,000 people who had COVID-19. Nearly 600 of those patients had visited a Long COVID clinic. By comparing these patients with patients who had COVID-19 but did not go to a Long COVID clinic, the researchers built machine learning models that could identify the differences between the groups of patients, such as differences in the medications they were taking, how often they saw other doctors, and other conditions the patients had been diagnosed with. The researchers then tested the models on health records from a fourth N3C site.

In total, the researchers created three models — one for identifying potential Long COVID patients across the whole dataset and two that focused more specifically on people who had or had not been hospitalized for COVID-19. After testing, the researchers found that each model was highly effective at identifying people who likely had Long COVID.

Why is this research important?

With more refinement, models like these could help researchers determine whether a person with a positive COVID-19 test may be likely to develop Long COVID. Once they can identify those people, researchers can determine what they have in common and what differentiates them from those who do not have Long COVID, paving the way for better and faster treatment of patients with Long COVID.

Where can I go learn more?

Predicting Long COVID with Artificial Intelligence

  • To understand which patients develop Long COVID, the NIH Rapid Acceleration of Diagnostics (RADx®) initiative is supporting research that uses artificial intelligence and machine learning tools.

Researching COVID to Enhance Recovery (RECOVER) Initiative

  • NIH created the RECOVER Initiative to learn about the long-term effects of COVID-19. Whether or not you have had COVID-19, you may be able to participate in RECOVER research.

Studying Long COVID Might Help Others With Post-Viral Fatigue Ailments

  • Avindra Nath, M.D., the clinical director of the National Institute of Neurological Disorders and Stroke, discusses Long COVID research and how it can benefit people with other diseases.

Sources

Pfaff, E. R., Girvin, A. T., Bennett, T. D., Bhatia, A., Brooks, I. M., Deer, R. R., Dekermanjian, J. P., Jolley, S. E., Kahn, M. G., Kostka, K., McMurry, J. A., Moffitt, R., Walden, A., Chute, C. G., Haendel, M. A., and the N3C Consortium. (2022). Identifying who has long COVID in the USA: A machine learning approach using N3C data. Lancet Digital Health, 4(7), e532–e541. https://doi.org/10.1016/S2589-7500(22)00048-6

Pointer

News and Stories

Read stories about the efforts underway to prevent, detect, and treat COVID-19 and its effects on our health.

NIH COVID-19 Resources by Topic

COVID-19 research information and resources by topic from NIH institutes and centers

Page last updated: November 7, 2022