Large language model–assisted preoperative communication reduces patient anxiety and physician workload in prostate cancer: a prospective randomized phase II trial

Abstract

Abstract This prospective, randomized, single-blinded, phase II trial evaluated whether large language model (LLM)-assisted preoperative communication, compared with standard communication, reduces patient anxiety and improves illness perception and satisfaction while decreasing physician workload. Newly diagnosed prostate cancer patients scheduled for radical prostatectomy were randomized by room to the control or AI-assisted groups (based on LLM). Baseline assessments of anxiety, emotion, and illness perception were performed, and patient questions were collected. In the AI-assisted group, patients received locally deployed LLM-generated personalized responses before routine face-to-face communication, with physicians blinded to allocation. Primary endpoints were patient anxiety and physician workload; others included emotion, illness perception, satisfaction, and communication time, with outcomes reported as mean [SD]. A total of 268 patients were randomized (control, n = 130; AI-assisted, n = 138), and 3507 questions were collected (mean, 13.1 per patient). AI-assisted patients reported significantly lower post-communication anxiety (GAD-7, 3.2 [2.7] vs. 5.7 [3.1]; P = 0.001) and better emotion, illness perceptions, and satisfaction. Physician workload was reduced, with lower NASA-TLX (39.9 [16.9] vs. 56.8 [19.5]; P < 0.001) and shorter communication time (11.3 [4.5] vs. 19.9 [6.1] min; P < 0.001). AI-assisted preoperative communication significantly reduces patient anxiety and improves psychological and cognitive outcomes while decreasing physician workload. ClinicalTrials.gov registration: NCT07082049, 07/2025.