UCSF investigators recently partnered with IIAM Corporation to determine whether a supervised machine learning (ML) model accurately predicts whether patients with head and neck cancer should be recommended for surgery based on the medical information provided at the time of referral.
New patient referrals are typically triaged and processed by practice coordinators with limited medical backgrounds who may be unable to distinguish between benign and potentially serious findings. This research shows that using AI to help process referrals could significantly decrease time to treatment for patients with cancer and improve patient experiences and outcomes.
“This technology has remarkable potential to transform the workflow of referrals, solve issues with patient access and optimize patient care,” said UCSF neurotology and skull base surgeon-scientist Nicole Jiam, MD, who led the study. She is also CEO of IIAM.
Optimizing processes improves patient outcomes
The researchers conducted a retrospective cohort study using 64,222 records of patients with head and neck cancer from the Surveillance, Epidemiology and End Results (SEER) database, and used 70% of the records to train the algorithm and 30% to test it. The team identified and isolated 21 from a total of 263 patient variables that would be pertinent and plausibly available at the time of a new referral. The ML model they developed correctly classified patients who were offered head and neck surgery with an 81% accuracy rate.
These results help identify the specific medical information most likely to affect cancer treatment decisions. The variable that carried the most weight in the decision to recommend surgery was tumor site. Other key variables included diagnostic confirmation, histology, and T and N stages.
“This machine learning algorithm is able to recognize patterns, synthesize data and allow us to understand how a physician might assess the urgency of a referral,” Jiam said. “It can help support the team that’s triaging and processing referrals so that patients can be seen by the right person at the right time.”
Patients and providers often encounter difficulties when navigating referral processes and scheduling appointments with specialists. “I think that’s something health care systems all around the country are dealing with,” Jiam said. “When there are concerns about potential cancer, the stress from sitting with the unknown, from a patient's perspective, can be very daunting. We’re trying to make this process better for everyone.”
The new frontier of health care technology
Through innovative work being done by the Bakar Computational Health Sciences Institute and the Division of Clinical Informatics and Digital Transformation, as well as other programs, centers and individuals, UCSF is committed to leveraging technology to advance the field of precision medicine and improve patient outcomes.
“UCSF is at the forefront of understanding the digitalization of health care and defining this new frontier,” Jiam said. “I'm very excited for this new frontier.”
Cancer research and treatment take place within the UCSF Helen Diller Family Comprehensive Cancer Center.
To learn more
UCSF Head and Neck Surgery Cancer Program
Phone:(415) 885-7528 | Fax: (415) 885-7711
Refer a patient
Clinical trials
