What Patients Want from Healthcare Chatbots: Insights from a Mixed-Methods Study

Jessica Ellis, MS – Professional Research Assistant, Center for Bioethics and Humanities, University of Colorado Anschutz Medical Campus; Annie Moore, MD – Professor, Division of General Internal Medicine, University of Colorado Anschutz Medical Campus; Marlee Akerson, BA – Professional Research Assistant, Center for Bioethics and Humanities, University of Colorado Anschutz Medical Campus; Matt Andazola, MPH – University of Colorado Hospital; Eric Campbell, PhD – Professor, Center for Bioethics and Humanities, University of Colorado Anschutz Medical Campus; Matthew DeCamp, MD, PhD – Associate Professor, Center for Bioethics and Humanities, University of Colorado Anschutz Medical Campus

Abstract: Healthcare chatbots are gaining traction as tools to assist patients with a variety of tasks, from administrative support to symptom assessment. However, the extent to which patients trust and prefer chatbots for different healthcare tasks remains underexplored. We conducted a mixed-methods study with patient users of a healthcare system chatbot integrated with an electronic health record. We purposively oversampled participants by race or ethnicity to survey 617/3089 (response rate, 20.0%) chatbot users using de novo and validated survey items. Additionally, we conducted semi-structured interviews with 46 patient users and 2 chatbot developers between November 2022 and May 2024. We used modified grounded theory to analyze interviews and descriptive statistics to analyze survey results. Patient users preferred chatbots for administrative tasks due to approachability, availability, and avoiding unpleasant human interactions. Yet they also preferred to discuss sensitive tasks with chatbots (such as sexual health or substance use) due privacy and anonymity and less embarrassment, judgment, or bias. Surveys revealed that patient users were less likely to worry about being judged based on chatbot interactions (153/608, 25.2%) compared to interactions with a doctor (219/606, 36.1%)(p < 0.001). Citing empathy, users expressed a preference for human clinicians for diagnostic tasks. Patient users appeared to simultaneously prefer chatbots for either simple tasks or sensitive ones, often motivated by a desire to avoid human judgment. Chatbots may thus fill a void or serve as a critique of healthcare’s status quo. Our findings suggest further needs to design transparent chatbots that accommodate patient preferences and desired use-cases.

Keywords: artificial intelligence, chatbots, conversational agents, digital health access