Using chatbots to screen for heritable cancer syndromes in patients undergoing routine colonoscopy

Hereditary colorectal cancer (HCRC) syndromes account for 10% of colorectal cancers but remain underdiagnosed. This feasibility project tested the utility of an artificial intelligence-based chatbot deployed to patients scheduled for colonoscopy to identify HCRC risk factors, educate participants about HCRC and obtain consent to genetic testing as an extension of genetic counselling of appropriate subjects. Genetic counsellor (GC) and genetic counselling assistant (GCA) time spent per subject was also measured.

Patients scheduled for colonoscopy at Cleveland Clinic were invited via electronic medical record patient portal or letter prior to colonoscopy with a link to a chatbot administering the Colon Cancer Risk Assessment Tool (CCRAT) to screen for HCRC syndromes. Those with ≥1 positive response to a CCRAT question received chatbot-deployed genetic education and the option to receive genetic testing. An order for a 55-gene pan-cancer panel was placed for those consenting, and the subject had blood drawn on the day of colonoscopy. Results were disclosed by a GC or GCA by telephone. Subject demographics, progression through the chat, responses to CCRAT, personal and family history, genetic test results and communication with the subject were recorded. Descriptive statistics and two-tailed unpaired t-test and Fisher's exact test were used.

506/4254 (11.9%) initiated and 487 (96.2%) completed the chat with the chatbot. 215 (44.1%) answered 'yes' to ≥1 CCRAT question and all completed pretest education. 129/181 (71.3%) subjects who consented completed testing, and 12 (9.3%) were found to have a germline pathogenic variant. Per subject, the GC spent a mean of 14.3 (SD 7.3) and the GCA a mean of 19.2 (SD 9.8) minutes.

The use of a chatbot in this setting was a novel and feasible method, with the potential of increasing genetic screening and testing in individuals at risk of HCRC syndromes.