Mayo Clinic AI Study Shows Covid Shots Work Well in Real World
(Bloomberg) -- Two months after the first Covid-19 vaccines received U.S. clearance, the shots’ efficacy is holding up well in the real world, according to a study among more than 31,000 people immunized in four states.
The report, one of the first large-scale studies to assess how vaccines are performing in practice, used artificial intelligence software to scan the medical records of early vaccine recipients at Mayo Clinic facilities. The findings are important because vaccine and drug efficacy in the real world often differs from what is seen in the carefully controlled environment of experimental clinical trials.
Covid infection rates were almost 89% lower in people who received two doses of the Moderna Inc. or Pfizer Inc.-BioNTech SE vaccines, starting 36 days after their first shot. That compares with a matched control group of demographically similar people who lived in the same zip codes but weren’t yet immunized. Those diagnosed with Covid-19 were 60% less likely to be hospitalized if they had been vaccinated, the study found.
“It is very reassuring that we observed a high level of activity,” said co-author Andrew Badley, an infectious disease specialist at the Mayo Clinic in Rochester, Minnesota. “The efficacy is essentially unchanged from what we saw in the clinical trials.”
The study involving mostly health-care workers and long-term care residents hasn’t been published, but was posted without an independent review of the findings to OSF Preprints on Wednesday. However, it offers an indication of what to expect as vaccinations gain steam, with more than 56.1 million doses administered in the U.S. and 181 million given worldwide, according to the Bloomberg Vaccine Tracker.
Researchers were able to perform the study so quickly thanks to artificial intelligence software designed by the Cambridge, Massachusetts-based startup nference, which has been collaborating with the Mayo Clinic since early 2020. It’s raised $145 million in venture financing from the Mayo Clinic and others.
In addition to reading electronic physician notes from medical records, the software can instantly create matched control groups by finding people in the Mayo Clinic health system who are demographically and geographically identical to those who received a certain drug or vaccine.
“It is lightning fast,” said Venky Soundararajan, co-founder of nference and an author of the study. “You can click a button to create the cohort. You don’t need 40 graduate students. It takes 5 seconds.”
The company hopes to use the software to analyze fine grain questions, such as whether there are people for whom one dose is enough and what groups are most likely to experience breakthrough infections even after two doses, Soundararajan said.
©2021 Bloomberg L.P.