Conversations with patients about cancer surgery need to include a discussion about the short- and long-term risks and benefits of the procedure, yet surgeons often do a poor job of explaining how patients’ lives could be impacted. Not only can ineffective conversations about surgery impact patient quality of life, it can also increase healthcare costs due to potentially avoidable postoperative complications. With a new two-year, $69,000 grant from the Palliative and Supportive Care Disease-oriented Team at the UW Carbone Cancer Center, Division of Endocrine Surgery Associate Professor Courtney Balentine, MD, MPH and Department of Radiology Professor Alan McMillan, PhD will be exploring how large language models (LLMs) can be used to improve patient-surgeon communication. LLMs are advanced artificial intelligence systems that are trained on massive datasets to recognize, summarize, translate, predict, and generate content.
“Research shows that when patients with cancer are adequately prepared for how treatment can affect their lives, it profoundly reduces both the impact and incidence of adverse treatment effects, thereby enhancing patient outcomes and quality of life,” explained Balentine. “Using thyroid cancer as a pilot case, we plan to develop and test an LLM chatbot that can serve as a companion to surgeon-patient discussions. Patients can use the chatbot to discuss how thyroid cancer treatment typically affects patients and learn about strategies they can use to minimize these effects based on the lived experiences of other thyroid cancer patients.”
To train the LLM chatbot, the research team will use an existing qualitative dataset that explores the short- and long-term effects of thyroid cancer treatment on patient quality of life, physical function, and cognitive function. They will then validate the chatbot by asking 20 patients who have completed thyroid cancer treatment to engage with the chatbot to assess the accuracy of the information it shares, the acceptability of this technology, and the feasibility of its use. Balentine and McMillan will then apply for funding to conduct a randomized trial to test the effectiveness of the chatbot in the clinical setting.
“There is an urgent need to improve the quality of surgeon-patient communication to enhance both the quality of decision-making and the quality of recovery from cancer surgery,” said Balentine. “If we’re successful, this chatbot could serve as a model for developing LLMs to aid pre-operative conversations about other cancer-related surgeries, ultimately improving the quality of postoperative recovery and survivorship for patients with cancer.”