Evaluating a hybrid AI-powered digital health guide to improve informed decision making and streamline access to genetic testing for hereditary cancer.

319

Background: Despite expanded testing guidelines, lower costs, and clear clinical utility, uptake of genetic testing remains low. Traditional referral-based models that rely on separate clinician visits, including telehealth, do not meet the needs of many patients. Chatbots represent a promising approach to support informed decision making, supplement or replace genetic counseling, improve access to genetic care delivery, and enhance health outcomes across diverse populations. The study’s goal was to develop and evaluate the feasibility, acceptability, and preliminary efficacy of a hybrid AI-powered digital health guide (DHG) designed to promote informed decision making and improve guideline-based genetic testing for hereditary cancer predisposition. Methods: The DHG prototype was developed following community-engaged formative research (focus groups (n = 51), and semi-structured interviews (n = 14)) with cancer patients and at-risk relatives and guided by Ottawa Decision Support Framework. Cancer patients meeting germline testing criteria were ascertained from the electronic health record and recruited through proactive outreach. The DHG was refined through user (content-focused) and usability (functionality-focused) testing. In the pilot trial (n = 28), probands were randomized to the DHG or enhanced usual care (EUC) arms. Qualitative data were analyzed using the Framework Method and quantitative data were summarized with descriptive statistics and t-tests. Methods for minimizing hallucinations from the large language model DHG chat function were used. Results: In response to patient and relative input, we incorporated features to enhance trust, accessibility, and cultural relevance, including an interculturally competent, multimedia DHG with text and voice options; clear and actionable genetic information; patient testimonials; cascade testing; and flexible options for result delivery. Users’ feedback also led to clearer language, improved navigation, and stronger messaging regarding privacy and data security. The Chatbot Usability Questionnaire score for the DHG was 70.3 (IQR = 12.5), indicating good acceptability. The trial showed genetic testing uptake was substantially higher in the DHG arm (73.3%) compared with enhanced usual care (EUC; 7.7%). Pretest genetic counseling was requested by 1 of 13 EUC participants and none of 15 DHG participants. DHG participants reported significantly lower decisional conflict (33.37 ± 21.09) and decision regret (17.5±16.50) compared with EUC (53.25±22.66 and 37.08±17.38, respectively; p < 0.05). Conclusions: Our approach has strong potential to increase access, remove barriers, streamline care, guide decision making, and inform prevention and surveillance practices for high-risk patients. Clinical trial information: NCT06184867 .