Pooling Patient Data to Answer Big Questions
Throughout the COVID-19 pandemic, there has been an urgent need to better understand who is at greatest risk for severe disease, how the disease and treatments have evolved, and how to better predict the need for resources. But answers require lots of data, such as what patients have experienced and what factors are associated with different treatment outcomes.
To provide this information, a research consortium quickly formed in the summer of 2020. The team built a system in which clinicians, researchers, patients — anyone, really — could submit questions that could be answered by COVID-19 patient record data from more than 200 participating hospitals. Questions are submitted and select answers are provided via a web portal at COVID19questions.org.
The consortium, called Reliable Response Data Discovery (R2D2), was led by Lucila Ohno-Machado, MD, PhD, chair of the Department of Biomedical Informatics at UC San Diego Health, and made possible by seed funding from the Gordon and Betty Moore Foundation. R2D2 comprised 14 health systems.
Here’s how the COVID19questions.org site works: Users submit questions about adults hospitalized with COVID-19. Consortium team members evaluate the submissions for clinical utility and the likelihood that available data can provide answers. Questions are then translated into a computer code that queries a variety of electronic medical records. Each health system runs the code on their own patient records and provides the results to the consortium. When sufficient results accrue to be statistically meaningful, the answers are posted back to COVID19questions.org — not as definitive conclusions, but as data in the form of charts or other graphics, which researchers can further pursue.
Among the questions explored:
• What is the mortality rate for hospitalized COVID-19 patients with each blood type?
• What is the mortality rate for COVID-19 patients with a history of hypertension who received anti-hypertensive medications?
• How does the mortality rate compare between hospitalized adult COVID-19 patients who received glucocorticoids and those who didn’t?
Between January 1, 2020 and March 13, 2021, the COVID19questions.org site drew from data on more than 55 million patients, including more than 3.4 million patients who were tested for COVID-19, 320,000 diagnosed with the disease, 80,000 hospitalized and 10,000 who died.
“The scientific community has talked about using electronic medical records for guiding research and for answering relevant questions for a long time,” Ohno-Machado said. “But until this pandemic, we hadn’t been doing it in a way that the public can see — this is much different than when only scientists can ask questions and publish their findings in academic journals.”
While the consortium’s seed funding has come to an end, the site continues to answer questions with the support of several consortium institutions.
Ohno-Machado said she’s looking forward to beginning to characterize “long COVID” — the long-term health effects experienced by many survivors. She’s curious about the types of symptoms people experience the most and whether they differ by gender, race and ethnicity.
Already, the consortium is finding that among COVID-19 survivors, mental health issues are most prevalent. And hair loss is more prevalent in female survivors than male.