多病共存管理中的风险预测工具:一项定性研究
该研究分析苏格兰四家整合健康与社会照护合作中30名专业人员的访谈和28名多病共存患者/照护者的六组焦点小组,探讨风险预测在多病共存管理中的益处与挑战。识别出风险预测工具合法性、工作负荷影响和工具重构三个主题;专业人员和患者均对现有工具持保留态度,但对人工智能辅助工具有条件支持。
要点
- 专业人员质疑现有风险预测工具的临床实用性,认为算法缺乏临床细微区分且过度依赖粗糙算法。
- 患者/照护者担心风险沟通可能引起不必要的焦虑、削弱自主性,且在老年阶段相关性较低。
- 对人工智能辅助风险分层有一定支持,但需无工作负荷影响、补充临床判断并考虑患者复杂性和偏好。
核心结论
- 对使用人工智能辅助工具进行风险分层有一定支持,前提是不增加工作负荷、补充临床判断并考虑患者临床复杂性和偏好。
研究设计与样本
- 对30名专业人员的访谈和28名多病共存患者/照护者的六组焦点小组进行分析,覆盖苏格兰四个整合健康与社会照护合作。
- 数据收集于2023年5月至2024年5月,采用主题分析法。
主要发现
- 识别出三个主题:风险预测工具的合法性、风险预测的工作负荷影响、风险预测工具的重构。
- 专业人员质疑现有工具的临床实用性,指出缺乏临床细微区分和过度依赖粗糙算法。
- 专业人员强调新工具增加工作负荷,需要无缝整合、明确如何回应预测的指导,并纳入心理社会因素和有意义的结局。
- 人工智能和常规数据被视为有希望提高预测准确性和实时应用,但有效使用需要更好的IT系统、培训和临床监督(人在环路中)。
- 患者/照护者观点混合,警告风险沟通可能引起不必要的焦虑、削弱自主性,且在老年阶段相关性较低;强调风险沟通应反映社会背景并与患者优先事项一致。
临床怎么用
- 使用风险预测工具时应结合临床判断,避免仅依赖粗糙算法。
- 引入新工具需评估工作负荷影响,并提供明确的风险应对指导和 seamless 整合。
- 风险沟通应结合患者社会背景和优先事项,避免引起不必要的焦虑并保护患者自主性。
- 人工智能辅助工具的应用需配备更好的IT系统、培训和临床监督(人在环路中)。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND: Risk stratification is integral to providing good care, but its value in managing patients with multiple long-term conditions (MLTC) remains uncertain. AIM: To explore the perspectives of healthcare professionals, patients, and carers on the benefits and the challenges of using risk prediction in MLTC management. DESIGN AND SETTING: Analysis of interviews with 30 professionals and six focus groups with 28 patients with MLTC/carers in four Scottish integrated health and social care partnerships was undertaken. METHOD: Data were collected between May 2023 and May 2024, and analysed thematically. RESULTS: Three themes were identified: legitimation of risk prediction tools, workload implications of risk prediction, and reconfiguration of risk prediction tools. Healthcare professionals questioned the clinical utility of existing tools, noting a lack of clinical nuance and overreliance on blunt algorithms. They stressed increased workload implications of new tools and the need for seamless integration, clearer guidance on how to respond to prediction, and inclusion of psychosocial factors and meaningful outcomes. Artificial intelligence (AI) and routine data were seen as promising in enhancing predictive accuracy and real-time application of new tools, but effective use requires better IT systems, training, and clinical oversight - 'human-in-the-loop'. Patients/carers expressed mixed views, warning that risk communication could cause unnecessary anxiety, undermine autonomy, and be less relevant in older age. They emphasised the importance of risk communication that reflects social context and aligns with patient priorities. CONCLUSION: There was some support for using AI-informed tools for risk stratification, provided they have no workload implications, complement clinical judgement, and account for patient clinical complexity and preferences.
引用
- Arakelyan S, Anand A, Mercer SW, et al. Risk prediction tools in multiple long-term conditions management: a qualitative study. Br J Gen Pract. 2026. PMID: 41344812. doi:10.3399/BJGP.2025.0371
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。