Effect of chatbot utilization on genetic testing uptake in diverse clinic settings.

Abstract

e13710 Background: Chatbots are a digital health tool used to identify patients who may benefit from genetic testing, provide pre-test educational content, and streamline genetic test ordering, thus allowing clinical teams to provide scalable patient risk assessments. The Invitae Gia Hereditary Cancer chat is one such tool that has demonstrated high patient engagement and satisfaction, in which about one quarter of patients meet National Comprehensive Cancer Network (NCCN) criteria for germline genetic testing (GGT) (PMID: 34735417). In this study, we demonstrate the utilization of Gia across a variety of clinical settings. Methods: A retrospective analysis of data was performed from the Gia Hereditary Cancer chat that was deployed between February 2020 - August 2024 in clinics categorized as high-risk (focused on the treatment of cancer or cancer risk assessment) or average-risk (focused on cancer surveillance in the general population). NCCN and American Society of Breast Surgeons criteria were used to determine if patients met eligibility for GGT. Single and multivariate analysis was performed to determine the impact of patient characteristics on the uptake of GGT. Clinically actionable pathogenic germline variants (PGV) were defined as those with potential eligibility for published management recommendations, targeted therapy, and/or clinical trials. Results: 143,208 patients were invited to Gia (20% high-risk and 80% average-risk, Table): 75% White, 96% female, mean (SD) age of 53.9 (14.5), 12% reporting a personal history of cancer and 72% reporting a family history of cancer. >70% of patients from both high- and average-risk clinics completed the chat (Table). 31% (43,661/117,876) of patients met criteria for GGT and of those, 19% (8,149/43,661) completed GGT. Additionally, 3% (3,038/99,547) of patients did not meet criteria but still completed GGT. Patients from high-risk clinics were more likely to complete GGT (OR: 2.53, Cl: 2.4 - 2.7, p<0.001) compared to patients from average risk clinics. Of all patients who completed GGT (those who met criteria and those who didn’t), 11% (1249/11,187) were identified with a PGV, 91% (1131/1249) of which were potentially clinically actionable. There was no significant difference in the rate of any PGV between patients undergoing GGT at average-risk clinics and high-risk clinics (11% vs. 12%). Conclusions: Chatbot utilization resulted in identification of potentially clinically actionable PGVs in patients across a variety of clinic settings, including those from average-risk clinics who may not have otherwise been offered GGT through standard workflows. However, only 1 in 5 patients meeting criteria completed GGT, suggesting other barriers to testing that should be studied further.