Patient-Reported Outcomes Using Automated Chatbot for Breast Cancer Patients Receiving Radiation Therapy

Purpose/Objective(s) We previously reported our experience using innovative and automated chatbots (“Health Chats”) in head and neck cancer patients during radiation treatment (RT). We created a breast cancer (BC) Health Chat. The aims of this pilot study are to 1) assess the usability of the Health Chats, 2) engage patients by providing frequent individualized education, and 3) identify areas of unmet needs specific to BC patients via collection of patient-reported outcomes (PROs). Materials/Methods Between January 2020 and December 2020, adults with BC aged ≥ 18 years who were English literate and were scheduled for RT at one of four academic hospitals were screened. Participants were sent chats 2 days before on-treatment visits during the duration of RT, then weekly for 4 weeks post-treatment, and monthly for an additional 3 months (“long-term follow-up”). Health Chats used PRO-CTCAE questions customized for BC patients. Clinicians were notified if any “red flag” symptoms were reported (PRO-CTCAE ≥ 3 or pain score ≥8). At the end of each chat, patients received customized educational material specific to symptoms reported. Results 20 breast cancer patients were enrolled, and 16 patients completed at least one chat. These 16 patients completed a median of 9 chats, out of a maximum of 10-12 chats. 11 patients continued to chat during the immediate post-treatment phase, and 10 patients continued to chat during long-term follow-up. Patient engagement with Chats is high, with 6 patients providing written feedback about the details of their symptoms, and 3 patients using unscheduled chats to contact the care team. The average System Usability Scale score (n = 9) was 82.77. A total of 811 pain or PRO-CTCAE scores were collected. There were 2 counts of pain scores ≥ 8, and no reports of grade 3+ skin toxicity or breast swelling. 5 out of 11 patients reported that their pain continued to worsen after finishing their last RT treatment, and 10 out of 11 patients reported that their breast pain, swelling, or skin toxicities peaked 2 weeks after RT. Patients reported severe and prolonged fatigue or insomnia: there were 4 counts of grade 3 fatigue and 3 counts of grade 4 fatigue, and 2 counts of grade 3 insomnia and 1 count of grade 4 insomnia. 4 out of 10 patients reported continued fatigue or insomnia during long-term follow-up. Of the 5 patients who screened positive for depression or anxiety using PHQ2 or GAD2 questionnaire, 3 patients continued to report persistent symptoms during long-term follow-up. Conclusion To our knowledge, this is the first study using chatbots to collect PROs for BC patients receiving RT. Our results show that BC patients are highly engaged and satisfied with Health Chats. As expected, patients report that pain and skin toxicities peak 2 weeks after RT. We also identify unexpectedly high rates of severe fatigue, insomnia, and persistent anxiety and depression. These toxicities are difficult to observe by clinicians and may represent unmet patient needs that should be addressed by future interventions.