HMH Med School Professor Publishes AI Research   

Hackensack Meridian Health Med School Professor Publishes AI Research

Charles E. Binkley, MD

With sweeping innovations in healthcare information and technology, it’s easy for patients, providers, and healthcare leaders to get swept up en masse in the new world of predictive AI modeling. One physician urges matching expediency with deliberation in adopting such advancements.

Charles E. Binkley, MD, director of AI ethics and quality and associate professor of surgery at Hackensack Meridian School of Medicine (HMSOM), published a recent review study in the popular science journal, Nature Medicine, weighing both potential benefits and distinct harms the use of AI in public healthcare might pose for disabled patient populations.

In the journal article, Dr. Binkley implored caution in applying predictive AI—typically obtained through training on huge sample sizes and datasets—to the much smaller and more variable conditions of underrepresented patient populations such as disabled people.

“Predictive AI could fail these patient populations by making predictions based on results skewed toward data from the larger able-bodied population,” said Dr. Binkley. “This could perpetuate historical biases and discriminatory practices of denying potentially life-saving treatment for part of the population, in favor of misrepresented data gleaned from the whole.”

According to Dr. Binkley, this problem could also swing the pendulum too far the other way.

“Such large-scale, predictive AI could inaccurately overestimate the perceived physiologic frailty of a disabled patient,” said Dr. Binkley, “because of inconsequential factors in data adding up to a larger discrepancy in the patient’s condition than what actually exists.”

To address or prevent these issues, Dr. Binkley suggested adopting AI models specifically trained on the disabled patient population to parse out and reduce the level of variability in data, providing a more accurate representation of the population.

Through more accurate measurements, predictive AI could be used to the betterment of disabled people by optimizing suggested intervention to improve quality of life and more favorable outcomes all along the process.

“Most importantly,” noted Dr. Binkley, “Tailoring AI models to the specific patient base would enhance the democratization of healthcare in shifting to a holistic, human-centric approach in the proactive, long-term care in improving lifelong trajectories of populations facing such disparities.”

We use cookies to improve your site experience. By using this site,
you agree to our Terms & Conditions. Also, please read our Privacy Policy. Accept All CookiesLearn More
X