PO.CL11.02 · 临床研究
AI驱动的聊天机器人应用用于癌症患者症状管理的早期结局：一项前瞻性数字健康研究的中期分析
Early outcomes of an AI-driven chatbot application for symptom management in patients with cancer: Interim analysis of a prospective digital health study
该海报暂无可下载的资料
AACR 官方页面
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
背景：癌症患者经常经历持续性症状，如疼痛、疲乏和治疗相关的痛苦。尽管有基于指南的建议，但实时症状监测和支持性护理的提供仍不理想。我们开发了一款基于AI的聊天机器人应用，旨在提供个性化的症状管理教育、自我管理指导和互动支持。本研究评估用户体验、自我效能、心理痛苦和早期临床影响。
方法：开展了一项前瞻性数字健康干预研究，采用一款依据NCCN支持性护理指南提供教育的AI聊天机器人。共纳入192名参与者，包括实体瘤和血液系统恶性肿瘤患者。所用的经过验证的工具包括用于可用性的UMUX-Lite、用于自我管理自我效能的SEMCD-6、用于心理痛苦的DT、感知有用性（PU）、参与频率和电子健康素养测量（工具详情见幻灯片）。在基线和随访时评估结局。采用混合效应模型分析组别×时间效应。
结果：在整个研究期间用户体验始终保持较高水平，UMUX和PU评分良好。虽然短期和中期客观改善幅度不大，但干预组在关键的患者报告结局方面显示出显著更大的改善，包括：心理痛苦减轻（P = 0.013），自我效能各维度改善（多个SEMCD条目显示显著变化；主要维度P = 0.008）。较高的电子健康素养与自我效能和感知有用性更大的改善相关。亚组分析提示存在异质性的治疗效应，男性、肺癌患者以及受教育程度较低的个体表现出更大的未满足需求，需要量身定制的支持。参与度（每周使用聊天机器人的天数）与结局改善呈现正向的剂量-反应趋势。
结论：该AI聊天机器人应用在癌症患者中显示出持续的高用户满意度，以及在减轻心理痛苦和增强自我管理信心方面的早期获益信号。虽然客观临床指标在短期内变化有限，但不同人口学和诊断亚组之间的差异效应凸显了针对性个性化的必要性。正在进行的分析将评估中介因素，如数字素养、治疗联盟、干预可用性和长期临床结局。
查看英文原文 English abstract
Background:Cancer patients frequently experience persistent symptoms such as pain, fatigue, and treatment-related distress. Despite guideline-based recommendations, real-time symptom monitoring and supportive care delivery remain suboptimal. We developed an AI-based chatbot application designed to provide personalized symptom management education, self-management guidance, and interactive support. This study evaluates user experience, self-efficacy, psychological distress, and early clinical impact.
Methods:A prospective digital-health intervention study was conducted using an AI chatbot delivering education aligned with NCCN supportive-care guidance. A total of 192 participants were enrolled, including both solid tumor and hematologic malignancy patients. Validated instruments included UMUX-Lite for usability, SEMCD-6 for self-management self-efficacy, DT for psychological distress, perceived usefulness (PU), engagement frequency, and e-health literacy measures (instrument details in slide deck). Outcomes were assessed at baseline and follow-up. Group × time effects were analyzed using mixed-effects models.
Results: User experience remained consistently high throughout the study period, with favorable UMUX and PU scores. Although short-term and mid-term objective improvements were modest, the intervention group demonstrated significantly greater improvements in key patient-reported outcomes, including:- Reduced psychological distress (P = 0.013)- Improved self-efficacy domains (multiple SEMCD items showing significant change; P = 0.008 in primary domains) Higher e-health literacy was associated with larger improvements in self-efficacy and perceived usefulness. Subgroup analyses suggested heterogeneous treatment effects, with males, lung-cancer patients, and individuals with lower educational attainment showing greater unmet needs and requiring tailored support. Engagement (days of chatbot use per week) showed a positive dose-response trend with outcome improvement.
Conclusions: The AI chatbot application demonstrated high sustained user satisfaction and early signals of benefit in psychological distress reduction and self-management confidence among cancer patients. While objective clinical markers showed limited short-term changes, differential effects across demographic and diagnostic subgroups highlight the need for targeted personalization. Ongoing analyses will evaluate mediators such as digital literacy, therapeutic alliance, intervention usability, and long-term clinical outcomes.
利益披露 Disclosure
H. Lee, None.