The National Institutes of Health's All of Us research program announced this past week that it had made a significant increase in the COVID-19 information available in its precision medicine database.  

The expanded COVID-19 survey dataset now includes responses from nearly 100,000 participants, which could allow researchers to glean new insights from the wide-ranging information.   

This means the program's researcher workbench now includes data from 329,000 people, 80% of whom represent communities underrepresented in medical research.  

"It's a really diverse community," All of Us Chief Executive Officer Josh Denny told Healthcare IT News. "They're very engaged, and we're excited about where things are."

The program, which was founded in 2015 and launched nationwide enrollment in 2018, aims to eventually have one million participants representing the United States and reflecting population heterogeneity.   

Its goal with that "universal cohort" – many of them outfitted with connected-health technology – is to build a database that can inform studies on a variety of health conditions.   

Before the pandemic, Denny explained, the program had recruited engagement partners as trusted members in the community to deliver its message to potential recruits. 

"We also had 350 clinics where people did enrollment," he said. "It was a lot about high-touch – exactly what COVID-19 isn't."  

Many of those partners have found ways to be virtual, and 200 clinics have reopened with COVID-19 protocols in place.  

"We send out a lot of saliva tests," he said.  

Participants are asked to make their electronic health records available, as well as to share DNA results, answer health surveys and potentially visit a partner center to have physical measurements taken along with biosamples.  

The program recently gave public access to researchers, who can use the data to conduct studies. Anyone with an internet connection can also access the data browser.  

Already, said Denny, "We have about 1,100 researchers who have signed up."

He noted that the strength that comes from working with a diverse dataset, having done machine learning work himself.   

"Diversity of sites, EHR vendors, ages, genders, races – every kind of diversity – you need it all for the algorithm to learn effectively," said Denny.  

He cited an example from Dr. Sally Baxter, who used the All of Us database to build a model predicting the risk of progressing toward glaucoma surgical intervention. 

Baxter initially trained the model on data from a single institution – but when she used All of Us data to validate it, she found that the area under the receiver operating characteristic curve, or AUC, was only 0.49.  

By contrast, when she trained the model on All of Us data, AUCs ranged from 0.80 to 0.99.  

"It just shows the power of pulling that information," said Denny. With regard to the COVID-19 data, Denny notes the advantage of having longitudinal perspectives captured from across the country.  

"It means we have this population of individuals [for whom] we have data from before COVID," he said. "It means you can assess some of the characteristics of COVID-19 in a potentially less biased way."  

The individuals already in the cohort who have been diagnosed with COVID-19 may also have contributed Fitbit data or their genomics, he said.  

"Some will have 'long COVID,' and we'll be able to learn potentially more about that. And we'll look at things about the COVID survey in the [context of] bigger population," he explained.

"You'll be able to combine that information with other studies and answer important questions," he said.   

Denny emphasized that those who want to get involved with All of Us should sign up.

"The programs from a participant and researcher standpoint are open," he said. "Anyone from the United States can enroll, and we encourage them to do so."

Kat Jercich is senior editor of Healthcare IT News.
Twitter: @kjercich
Email: kjercich@himss.org
Healthcare IT News is a HIMSS Media publication.