Jefferson Health turns to AI to address a persistent gap in cancer care

Advertisement

Changing diet while undergoing cancer treatment is incredibly challenging. That’s why Jefferson Health, based in Philadelphia, is testing whether large language models such as ChatGPT can help cancer patients receive more tailored nutrition guidance.

“It’s difficult to advise patients on what to eat. While we provide support from oncology dietitians, that adds another appointment, and patients often have questions when they’re cooking at home and can’t easily reach a dietitian,” Nicole Simone, MD, vice chair for research and faculty development in the Department of Radiation Oncology at Jefferson Health, told Becker’s.

Those challenges prompted Jefferson Health to explore whether AI could help fill the gap.

“We wanted to see if thoughtful prompt design could generate accurate, supportive diet plans tailored to cancer patients, including considerations like calorie targets and socioeconomic factors,” Dr. Simone said. “It was important to us that the tool could suggest options across different budgets, whether that’s $10 a day or $100 a day.”

Dr. Simone, who is leading the trial, said Jefferson Health tested and refined the prompts, then had oncology dietitians develop their own plans and score the AI-generated versions.

“Once we incorporated the right variables, the AI performed quite well in terms of macronutrients and micronutrients,” she said.

In addition to cost, Jefferson Health explored whether AI could generate diets tailored to specific cultural cuisines, including Asian, Hispanic and Caribbean foods. Patients’ eating habits often reflect cultural norms, and one-size-fits-all dietary guidance may not be practical or relevant.

“It actually performed very well across those categories,” Dr. Simone said. “That reinforced for me how much unmet need exists when it comes to providing dietary advice that’s truly useful and relevant for all patients, and that realization was a major driver of this project.”

Still, she emphasized the importance of recognizing AI’s limitations. While the technology can adjust recommendations for common comorbidities such as diabetes or heart disease, it lacks the clinical nuance physicians would expect.

“That’s why I see this as a supportive tool, not a replacement for clinical judgment — something that works best in tandem with an initial consultation to help tailor guidance to a patient’s specific situation,” Dr. Simone said. “Balance is key.”

Jefferson Health’s goal is to make tools like this accessible to cancer patients nationwide. While diet plays an important role in cancer care, nutrition counseling is not reimbursed by insurance for most outpatient cancer patients. Coverage exists for conditions such as diabetes and heart disease, but not cancer.

“Larger cancer centers often absorb that cost and provide dietary support as a service, but many other centers simply don’t have the resources to do so, creating a real gap in care,” Dr. Simone said. “If this tool proves useful, it could help fill that gap by giving patients accessible, ongoing support. One of the things I value most about diet-focused work in cancer care is that it gives patients back a sense of control at a time when so much feels overwhelming.”

Advertisement

Next Up in Oncology

Advertisement

Leave a Reply

Your email address will not be published. Required fields are marked *