NEW YORK – A study comparing the use of a chatbot or genetic counselors to educate patients at risk for cancer about the need for genetic testing has found the two approaches are equivalent, suggesting that chatbots can be harnessed to make this education more accessible to patients in resource-limited settings.
The randomized clinical trial, called Broadening the Reach, Impact, and Delivery of Genetic Services (BRIDGE), tested the equivalence of an online chatbot versus genetic counselor appointment on the uptake of pre-test cancer genetic services and subsequent genetic testing at University of Utah Health and NYU Langone Health. The researchers found in this study that the impact of standard genetic counseling and the use of a chatbot was largely the same when they looked at whether patients completed pre-test education in both arms and if they got genetic testing to assess their cancer risk. The BRIDGE results were published in JAMA Network Open last month.
Kimberly Kaphingst, lead author of the study and the director of cancer communication research at the Huntsman Cancer Institute at the University of Utah, noted that the results suggest that chatbots could "serve this role of delivering pre-test genetic services to [cancer-]unaffected patients."
Guidelines around who should receive genetic testing for hereditary cancer risk have changed in recent years, as more is understood about the inherited genetic mutations associated with cancer. The National Comprehensive Cancer Network (NCCN) guidelines currently recommend genetic testing for hereditary cancer risk for individuals with a family history of certain high-risk or young-onset breast, ovarian, pancreatic, prostate, and colorectal cancers. Some experts have also called for universal genetic testing for hereditary cancer risk after research, such as the TAPESTRY study, has shown that these testing guidelines may miss a large proportion of patients harboring these mutations associated with cancer risk.
With the increase in testing, the demand for genetic counselors is also growing. According to the National Society for Genetic Counselors, as of March 2023, there were around 6,500 certified genetic counselors in the US. Since 2010, the profession has grown 100 percent, the group has estimated, and there is about one clinical certified genetic counselor per 100,000 people. While patients may still face long wait times to see a genetic counselor face-to-face, the field has embraced telephone and web-based counseling to extend counselors' reach, and experts have recognized that incorporating technology into the pre- and post-test setting, including chatbots, could further scale access to genetics educational services.
One study, for example, found that a chat service between patients and genetic counselors from telehealth company DNAvisit was sufficient to provide genetic information to patients about test results. Other studies, including the MAGENTA trial focusing on breast and ovarian cancer risk testing and the ProGen study on prostate cancer risk testing, found that online pre-test education led to an equal number or more patients receiving genetic testing after the technology-enabled education.
In looking at the equivalence between genetic counselors and chatbots in providing pre-test cancer risk assessment education, Kaphingst and colleagues aimed to similarly explore the utility of chatbots in supporting genetic counselors and extending access to genetics services. "We were thinking about different ways of delivering genetic services to meet increasing demand, so we were interested in using more automated digital health approaches to see whether at least some genetic services could be delivered to an increasing number of people," Kaphingst said.
In the study, researchers identified patients without cancer who were eligible for cancer risk genetic testing based on family history and NCCN guidelines and who had a primary care visit at one of the two institutions between 2020 and 2023. A random sample of 2,241 patients at University of Utah Health and 3,061 patients at NYU Langone Health were selected for the BRIDGE trial from an initial pool of more than 20,000 patients who were eligible for cancer risk genetic testing at either health system.
"At Huntsman, we have eight genetic counselors," Kaphingst said. "To deliver [genetic testing information] to thousands of patients, in addition to their actual patient workload, they recognize the need to come up with innovative ways to deliver services to hugely increasing numbers of patients."
Each health system sent the control group a message through its patient portal recommending genetic services and encouraging them to contact the genetics clinic to schedule a pre-test genetic counseling appointment. The researchers sent a reminder message one week later and made as many as two follow-up phone calls to reach out to those who didn't respond to the initial message. Those who made an appointment with a genetic counselor received standard care and most pre-test counseling appointments were done over the phone.
The chatbot group also received an initial message through their health system's patient portal recommending genetic services, but this group received a link to complete pre-test genetics education via chatbot. The researchers scripted content specifically for the BRIDGE study using an interdisciplinary team of health communications experts and genetic counselors. The chatbot was designed to cover the same information that a genetic counselor would within a standard pre-test appointment.
"[The genetic counselors on the team] are invested in this question of how to come up with innovative service delivery models where some of the education can be taken by automated digital health tools and then genetic counseling time can be used for other types of tasks, like the return of more complicated findings," Kaphingst said.
After watching an initial video message from the lead genetic counselor at the patient's site introducing the chatbot, individuals in this arm moved through the scripted information. They could request more information on any scripted topic or ask open-ended questions to the chatbot. If the researchers had developed responses to patients' questions, the chatbot answered with those scripted responses, but if patients asked questions for which there were no scripted responses, the chatbot was programmed to email those to the genetic counseling team. The chatbot collected basic family history information, shared educational information about genetics and how cancer develops, and explained what genetic testing looks for and how results are reported. Patients could also receive specific information about how the risk of individual cancers like breast, ovarian, colorectal, and pancreatic cancer increases if an individual has a family history of these tumors or high-risk genetic mutation.
Patients could also request an appointment with a genetic counselor instead of or in addition to the chatbot. In both health systems, only about 2 percent of patients in the chatbot group requested an appointment. At the end of the chatbot script, patients were offered the option to move forward with genetic testing. If they agreed, a genetic counseling assistant called them to confirm, answer any remaining questions, and collect more family history information, if needed.
Patients from both groups who agreed to genetic testing after the pre-test education session received saliva collection kits for genetic testing at their homes. Their samples were then tested on pan-cancer, multigene panels for germline mutations in 34 to 36 genes associated with heightened risk of cancer. Patients were not billed for pre-test genetics services in either group, but genetic testing was billed as a clinical test in both groups. The patients' primary care physicians were also notified that the patient would be contacted about genetic testing.
The researchers measured whether patients completed their pre-test education through the chatbot or with a genetic counselor, and if afterwards, they went on to receive genetic testing.
They found there was equivalence between the two methods for beginning pre-test services and completing genetic testing across both health systems, with 25.7 percent in the chatbot group completing pre-test services compared to 23.8 percent in the control group. Completion of genetic testing overall was also similar between the two arms, with 12.3 percent in the chatbot group completing testing versus 13.6 percent in the control arm. The researchers also determined there was equivalence between the two arms in patients' willingness to begin pre-test services using a chatbot or a genetic counselor.
While there was equivalence between the two arms in the study overall, there were some differences when researchers looked at outcomes in specific healthcare systems. They noted, for example, that at NYU, there was a larger difference in the proportion of patients who completed genetic testing between the chatbot group, 10.6 percent, and the control group, 13 percent.
Kaphingst and her team hope to continue developing the genetic testing chatbot and possibly make it available to other healthcare systems.
"The biggest challenge is often how quickly technology moves," she said. "We had what's called a rule-based chatbot that used natural language processing to answer open-ended questions and trying to keep the content up to date. As chatbots themselves become more sophisticated, we're thinking now in its second generation about how to use other approaches, potentially generative AI, in our chatbot."
She also hopes to explore using the chatbot to reach different patient populations, such as people in rural areas. Her team is exploring another potential clinical trial to evaluate the use of the chatbot in different communities and as a cost-effective option for some patients.
"Digital health approaches could potentially overcome some of the barriers, either insurance or financial barriers, or geographical barriers for people having to travel long distances to get genetic counselors," Kaphingst said. "There is potential that this approach could overcome a number of different types of access barriers that patients might face to genetic testing."
Ultimately, this study and deployment of a chatbot in the pre-test cancer risk assessment setting at these two healthcare systems required buy-in from hospital leadership, primary care physicians, and the information technology team to work with genetic counselors. Kaphingst reiterated that genetic counselors play an important role in a healthcare system, and these findings suggest that digital health tools, like the chatbot, may also have a secondary role in delivering these services.
"Utilizing chatbot tools strategically within the genetic counseling workflow can increase efficiency and access to services," Kaphingst said. "By supplementing genetic counselors with chatbot tools for tasks such as the delivery of pre-test genetics education, the time and availability of certified genetic counselors can be utilized for downstream tasks that require their unique and important clinical expertise."