Use of a patient-facing digital tool for glioblastoma outside scheduled clinic visits.

Abstract

1514 Background: Care for patients with glioblastoma is largely organized around scheduled clinic encounters, yet symptom management, decision-making, and caregiver responsibility extend beyond office visits. While patient portals, phone triage, and after-hours call coverage exist, these pathways are often resource-intensive, variably available, and insufficient to meet ongoing informational and emotional needs. Scalable, low-cost tools that extend access to vetted support outside clinic hours without increasing clinician workload may help address gaps in care delivery. We evaluated real-world use of a freely accessible, patient- and caregiver-facing digital support tool in glioblastoma. Methods: A freely accessible AI chatbot supporting informational, emotional, and care-navigation needs was deployed online without registration or recruitment. The tool excluded medical decision-making and clinical recommendations. De-identified conversation logs generated during routine use were retrospectively analyzed over a 4-month period (Sep 15, 2025 – Jan 14, 2026). Analyses included descriptive characterization of conversation volume, geographic distribution, timing of use, engagement depth, user-reported feedback, and primary topics addressed. All analyses used aggregate, de-identified data with no interaction or intervention involving human subjects. The protocol is IRB exempt per 45 CFR 46.104(d)(4) (WCG IRB 20260304). Results: During the study period, the chatbot supported 1,661 conversation threads comprising 15,197 messages across 54 countries. Most conversations originated from the United States (59.9%), followed by the United Kingdom (7.3%), Australia (5.2%), Canada (4.1%), and Greece (3.8%). Conversations were predominantly multi-turn and included use outside clinic hours. The tool was used primarily in English; however, ~12% of messages occurred in other languages, most commonly Spanish and French with additional use of Arabic, German, Italian, Portuguese, Hebrew, Hindi, and several Central and Eastern European languages. Conversations most frequently addressed imaging interpretation and scan-related uncertainty, treatment decision-making, symptom management and treatment-related side effects, and caregiver concerns. Additional topics included clinical trial exploration and prognosis clarification. When clinical trial–related questions arose, users were directed to a brain tumor trial finder. Across the study period, 374 users engaged in trial searches across 22 countries; among respondents to a trial-interest query (n = 251), 92% expressed interest. Voluntary user feedback (n = 97) demonstrated high perceived usefulness and trust (net promoter score, 80). Conclusions: A freely accessible AI chatbot demonstrated feasibility as a scalable approach to extending access to care for patients with glioblastoma and their caregivers outside office hours.