Originally published by our sister publication Gastroenterology & Endoscopy News

WASHINGTON—The use of a conversational artificial intelligence–based patient navigator doubled the colonoscopy completion rate in patients from underserved communities who had previously missed or avoided an appointment, according to researchers from Montefiore Einstein Comprehensive Cancer Center.

“This is outreach and engagement in action,” said Alyson Moadel-Robblee, PhD, the deputy director of community engagement and cancer health equity at Montefiore Einstein, in Bronx, N.Y.

Colorectal cancer disparities among people of color are well documented. Compared with white individuals, Black individuals have a 20% higher incidence of CRC and are 40% more likely to die from the disease, Dr. Moadel-Robblee noted.

Located in one of the most underserved urban areas in the United States, Montefiore Einstein primarily serves people from communities of color and low-income households. Despite active outreach by the center’s professional patient navigators, 59% of more than 6,600 patients either canceled or did not show up for their colonoscopy in 2022.

This low completion rate spurred Dr. Moadel-Robblee and her co-investigators to conduct a quality improvement project to test the ability of the MyEleanor AI patient navigator to “reengage” patients. The population included 2,400 English- and Spanish-speaking patients who were nonadherent with previous colonoscopy appointments. The study’s cohort was 44% Hispanic and 39% Black; 73% spoke English and 25% Spanish; 32% were unemployed.

image

How It Works

The MyEleanor AI navigator was developed by the digital health technology company MyndYou (myndyou.com). The program was trained on thousands of clinical cases and can detect nuances in conversation, she said.

MyEleanor was used to discuss rescheduling, assess barriers to colonoscopy uptake, offer live transfers to staff for rescheduling of appointments and provide reminders about procedure preparations (including instructions for medication pickup and bowel prep).

After determining the patient’s preferred language, the AI navigator followed this patient-friendly script: “I’m Eleanor, the automated care assistant for your team at Montefiore. I am calling today because we noticed that you missed your most recent colonoscopy appointment, and we would like to help you reschedule it at the end of this call.”

During the call, patients were also asked about their personal barriers to colonoscopy, reasons for missing their last appointment and other issues that could affect colonoscopy adherence.

Doubling Completion Rates

Presenting the results of the study at the 2024 annual meeting of the American Society of Clinical Oncology (abstract 100), Dr. Moadel-Robblee reported that more than half (57%) of the participants fully engaged with MyEleanor, and one-third of the study population (58% of those who engaged) accepted the phone transfer to a human navigator for rescheduling. “We were super excited about this. … It tells us this is a feasible way to intervene with a community that might have barriers to following up with their cancer screening,” Dr. Moadel-Robblee said.

The completion rate for prior “no shows” nearly doubled, from 10% to 19%. The number of patients completing colonoscopies rose from 3,898 in 2022 to 5,621 in 2023. Furthermore, patient volume at the center increased by 1,363 patients, or 36%.

“We also determined that our seven patient navigators gained 52 hours that were freed up for them to field the live transfer calls and schedule new patients,” she added.

Barriers Reported by Patients

“We found a host of barriers that were intervenable,” she said. More than 50% of participants reported at least two barriers to screening, most citing transportation (38%), lack of perceived need (36%), time constraints (36%), absence of physician prompting (33%), medical mistrust (32%), concerns about findings (28%) and cost (27%).

Patients were more likely to engage with the AI program if they reported that they were unemployed, on disability or expressed a greater number of barriers. Spanish-dominant patients and those who declined to identify their race reported up to twice the number of barriers. Participants most likely to not complete colonoscopy were most likely to report cost and fear as barriers.

The researchers plan to measure MyEleanor’s effects on patient navigator burden and acceptance, patient satisfaction and cost of care and are examining the potential for MyEleanor to assist with screening for other cancers.

Commenting on the findings, Cameron Carlin, MS, the manager of clinical decision support data science at City of Hope in California, said, “AI-based patient navigators, such as the work done with colonoscopy outreach at Montefiore, have the ability to assist with cancer patients in underserved communities, so long as they are implemented with proper healthcare system integration, data security, and community engagement and education.”

He cautioned that—as with human-initiated phone call follow-up—AI-based patient navigation must be as “streamlined as existing processes” because additional barriers “could cause fatigue with end users and lead to negative perceptions and outcomes with the [AI] system.”

Patients within the community need to understand how the AI tool works and be confident that their data are secure, he said. “This includes a realistic perspective of what the system can do when speaking to it, clearly defined limitations of the system, information of how to loop back with live transfers and reassurance that their healthcare experience will improve from these methods.”

—Caroline Helwick

Mr. Carlin and Dr. Moadel-Robblee reported no relevant financial disclosures.