The NEJM Challenge: Reassessing Race in Clinical Algorithms

A clinician opens a calculator, enters a few values, and gets a risk estimate. That number can decide whether a patient is imaged, referred, or treated. What many clinicians don't realize, and most patients never learn, is that race is one of the inputs in a number of these tools.
Dr. Helen Burstin described these algorithms on MD Aware as "hidden in plain sight." The NEJM Challenge, part of a national initiative called ENGAGE, is meant to bring them into view and invite researchers to propose how to fix them. Host Dr. Shazia Siddique sat down with two of the people behind it.
Meet the guests
Helen Burstin, MD, MPH, MACP is CEO of the Council of Medical Specialty Societies, a coalition of more than 50 specialty societies that coordinates ENGAGE. She was previously Chief Scientific Officer of the National Quality Forum and held leadership roles at the Agency for Healthcare Research and Quality and Brigham and Women's Hospital. She is a member of the National Academy of Medicine.
Mary Beth Hamel, MD, MPH is Executive Editor of the New England Journal of Medicine and Associate Professor of Medicine at Harvard Medical School. She practices primary care at Beth Israel Deaconess Medical Center, where she also teaches research methods.
Race as a shortcut
Race often entered clinical algorithms because it was statistically significant in the data a tool was built on, not because there was a clear biological reason for it. Burstin explained that it frequently serves as a shortcut for factors that would explain more, such as social conditions and ancestry, and that some of those factors are things clinicians and health systems can actually change.
Hamel put the problem plainly: "When you use it in an algorithm, you don't know what it's measuring. And when you do that, you introduce bias." The revision of the eGFR kidney function equation showed both the harm that bias can cause and the progress that is possible when clinicians, researchers, and patients work on it together.
From the editor's chair, Hamel has seen real change. Authors are now more likely to explain why they included race in a study, give a rationale when they adjust for it, and point to the limitations when they interpret results. That wasn't common 15 or 20 years ago. Her verdict: it has changed, but not enough. Siddique added that the fix isn't as simple as taking race out or swapping in social determinants of health. The clinical context of each condition matters, which is why the work needs people who know both the methods and the medicine.
Announcing a field-wide effort, not a single study
ENGAGE (Eliminating Bias in Next-Generation Algorithms: Guidance and Evaluation) is coordinated by CMSS and funded by the Doris Duke Foundation, with MDCalc, NEJM, and the Association for Clinical and Translational Science as partners. Burstin said that efforts like this have often arrived "a bit late," after a flawed algorithm is already published and built into guidelines. But ENGAGE aims to reach researchers, statisticians, and educators earlier, so that the work reaching journals and specialty societies already handles bias and fairness well.
NEJM's role builds on work it has already done, including published research and commentary on how race in algorithms affects patients, and an eight-episode podcast series called The Race Equation. For MDCalc, the initiative builds on its Quality Rating System, which includes assessments of bias and fairness for the tools clinicians use.
Burstin pointed to what makes the Challenge different from past efforts: "A lot of this work has been done specialty by specialty, disease area by disease area." The Challenge looks across all of medicine to find the algorithms with the most potential to harm patients and communities.
How the Challenge works, and what makes a strong entry
The Challenge has two rounds. Round 1 is deliberately a light lift: a three-page proposal that identifies an algorithm still in use, explains why it matters, and makes the case for revising it. Twenty Round 1 winners each receive $1,250 and are invited to a convening hosted by CMSS. In Round 2, those finalists propose the methods they would use to revise the algorithm and show that the work is feasible, with mentorship and technical assistance from the ENGAGE team. Up to five final winners receive $20,000 each. Round 1 entries are due March 29, 2027.
Both guests offered advice for entrants:
Pick an algorithm that is in use and has real impact. Burstin suggested checking MDCalc to confirm a tool is still in use. Hamel asked clinician-researchers to think about the algorithms that shape their own decisions most often.
Bring the right expertise. "This is really about improving the care of patients," Hamel said. Methods experts should partner with clinicians, and the reverse.
Think about data early. Round 1 doesn't require analysis, but entrants should know what data they could use, since Round 2 is about feasibility.
Be concise and use the resources. Three pages rewards a crisp argument. ENGAGE's bias and fairness framework will be available when the Challenge opens.
Burstin also noted a larger goal. With what she called "a chill in the health equity research funding space," the Challenge is a way to draw early-career investigators into the field, keep the ones already there, and show other funders how much work remains.
Looking ahead
The Challenge runs through 2027, and its reach goes beyond the researchers who enter. Any clinician who uses these tools benefits from knowing which ones include race and why. In an upcoming episode of MD Aware, Dr. Felicity Enders and Dr. Jennifer Lin go deeper into how researchers actually assess an algorithm for bias and fairness.
To learn more about the Challenge or to enter, visit nejm.org/nejm-challenge.
MD Aware: Upgrading Clinical Judgment is a podcast from MDCalc. Listen on Apple Podcasts and Spotify.

