Abstract OncoEducate is a clinician-supervised generative artificial intelligence (AI) application designed to deliver standardized, regimen-level, plain-language education at the point of care. We developed a templated handout generator using a locked prompt and predefined content structure that incorporates diagnosis, treatment regimen, and treatment intent, with clinician review prior to patient distribution. We then conducted a prospective, two-phase pilot at a single academic cancer center. In Phase I, six clinicians evaluated AI-generated handouts for nine commonly used palliative-intent genitourinary (GU) oncology regimens using 7-point Likert-scale items. In Phase II, patients with advanced kidney, prostate, or urothelial cancers initiating palliative-intent therapy received clinician-reviewed handouts and completed follow-up surveys at the next office or infusion visit or by telephone when in-person completion was not feasible. Clinicians rated handouts as useful (median 6/7) and accessible (median 6.5/7), with median accuracy ratings of 6–7/7 across domains. During an 8-week period, 20 of 21 approached patients enrolled (95%). Patients rated handouts as informative and readable (median 7/7 for both), and 65% correctly identified treatment intent as palliative. These findings support the feasibility and acceptability of clinician-reviewed, templated AI-generated handouts in routine GU oncology care, justifying larger randomized studies to evaluate clinical impact.
Litt et al. (Tue,) studied this question.