P539: Chatbot-assisted informed consent in genomics research*

Introduction:The consent process for clinical research is traditionally done in-person or through phone or video call by study personnel with the potential participant.There are many inefficiencies with this process, including the need for coordination between patient and study staff availability, difficulties in assessing patient comprehension, and time pressures to complete the consent.Chatbots have been implemented in a variety of clinical settings, including study recruitment, genetic counseling, cascade genetic testing, and even cancer risk screening.We report our experience using a scripted chatbot to facilitate the complex informed consent process for enrollment in a genomics research study for undiagnosed rare disease.Methods: A HIPAA-compliant, scripted chatbot, GIA or "Genetic Information Assistant", was developed to go through an IRB-approved script for automated chat-based informed consent 6 .Participants who were interested in enrolling in the study were given the option to review consent documents using GIA or through a traditional process done with a Genetic Counselor (GC) or Clinical Research Coordinator (CRC).The GIA script included several points where recall of key information was tested; if participants were unable to do so they were reverted to a traditional consent process.Participants who opted for chatbotassisted could also choose to talk to study team members at any time.Of the 1,010 individuals that were screened, roughly 20% opted into the GIA chatbot consent process.During the enrollment process, they were posed with an opt-in question regarding secondary findings.Using two-proportion Z tests and chi-squared tests, we compared the total time to enrollment as well as the opt-in rate for secondary findings between the GIA consent group and the traditional consent group, which was further broken into CRC and GC groups.Results: On average, the group that used GIA completed enrollment 10 days faster than the non-GIA group, however this finding was not statistically significant (48 days vs 58 days, p=0.41).Opt-in rates for medically actionable secondary findings were higher with GIA compared to the traditional consent group (98% vs 92%, Z = 2.87).When the traditional consent group was broken down into GC and CRC groups, we found that opt-in rates were similar between GIA and CRC (98% vs 99%, Z = 0.55) but were different when comparing GIA and GC (98% vs 90%, Z = 3.63) as well as CRC and GC (99% vs 90%, Z = 4.12). Conclusion:We demonstrate a chatbot-assisted informed consent process may have benefits in genomics research.Potential benefits include providing flexibility to participants to complete the consent process at their own convenience, testing participants' understanding of the study, potentially reducing time to enrollment and boosting participant numbers, and allowing CRCs and clinical study personnel to focus on other aspects of the study.Incorporation of a chatbot can potentially improve equity and patient comprehension in the consent process.These advantages to using a scripted chatbot in the informed consent process for clinical studies are not just relevant in genomics or neurological research, but rather can be applied widely to any clinical trials in the field of medicine.