AI-assisted preoperative communication in prostate cancer: A prospective, single-blinded, randomized phase II trial.

Abstract

5024 Background: To evaluate whether AI-assisted preoperative communication reduced patient anxiety, and physician workload compared with standard preoperative communication in patients undergoing radical prostatectomy. Methods: We conducted a prospective, randomized, single-blinded, phase II trial in newly diagnosed prostate cancer patients undergoing radical surgery, excluding those with severe psychiatric or cognitive disorders. Upon hospital admission, patients were randomized at the room level to receive either standard or AI-assisted preoperative communication. After baseline assessment, the AI-assisted group received individualized responses generated by a locally deployed large language model–based AI system prior to routine physician-led face-to-face preoperative communication, while physicians remained blinded to group allocation. The primary outcomes were patient anxiety (GAD-7) and physician workload (NASA-TLX). Comparisons were conducted using non-parametric rank-sum tests. The trial was registered at ClinicalTrials.gov (NCT07082049). Results: Between April and November 2025, 268 patients were randomized to the control group (n = 130) or the AI-assisted group (n = 138). A total of 3,507 questions were collected (avg. 13.08 per patient). After routine preoperative communication, patients in the AI-assisted group reported lower anxiety levels than those in the control group (GAD-7 median [IQR], 3.5 [2.5–5.0] vs. 7.5 [6.5–9.0]; P < 0.001). Physician workload was also reduced in the AI-assisted group, as reflected by lower NASA-TLX scores (median [IQR], 37.0 [27.0–46.0] vs. 54.0 [35.0–73.0]; P < 0.001), accompanied by a substantial reduction in preoperative communication time (12.60 vs. 21.90 minutes; P < 0.001). Patients in the AI-assisted group reported more favorable emotional status, lower negative illness perceptions, and higher satisfaction, while exploratory analyses indicated that AI-assisted communication alone improved anxiety, emotional status, and disease-related perceptions prior to physician face-to-face preoperative communication. The observed reductions in physician workload and communication time suggest potential applicability in resource-constrained clinical settings where clinical manpower and consultation time are limited. Conclusions: AI-assisted preoperative communication reduced patient anxiety, and physician workload compared with standard communication, supporting its role as a supplementary approach to routine preoperative care in prostate cancer, including settings involving older adults and limited clinical resources. Clinical trial information: NCT07082049 .