Automated health chats for symptom management of head and neck cancer patients undergoing radiation therapy

Objective: To report the early experience using an automated chatbot (Chats) for patient-reported outcomes (PRO) and symptom self-management in head and neck cancer (HNC) patients undergoing radiation treatment (RT). Methods: Patients aged >= 18 years diagnosed with HNC who were scheduled to begin RT were given the option to use Chats from June 2018 to June 2019. Enrolled patients received chat notifications two days before weekly ontreatment visits and every 1-4 weeks after RT for an additional 4 months. After the first in-person follow-up visit, participants completed an electronic usability and satisfaction questionnaire. Results: Of 95 patients who agreed to participate, 84 were eligible for analysis. Participants were significantly younger than patients who declined participation (mean age 61.3 vs 68.3 years; p-value < 0.001). Patient engagement with Chats was highest at 67% during the first month and declined over time (p-value = 0.004). Concordance between PRO and clinician-reported outcomes (CRO) was fair, ranging from 0.10 to 0.43 (Cohen kappa statistics). The most commonly under-reported symptoms were salivary duct inflammation (53%), xerostomia (41%), and mucositis (37%). 89% (39 of 44) of patients who completed surveys found Chats easy to use, and 61% reported that Chats helped with symptom self-management and reduced the need to call the care team. Conclusions: These early results suggest that an interactive chatbot is feasible and provides support for HNC patients during and after RT. Chats identified discordance between PRO and CRO. Further study is required to measure benefits of Chats in a larger population.