Carulli Christian, Rossi Stefano Marco Paolo, Magistrelli Luca, Annibaldi Alessandro, Troncone Enzo
Orthopaedic Clinic, University of Florence, Careggi University Hospital, 50121 Florence, Italy.
Department of Life Science, Health, and Health Professions, Università degli Studi Link, 00165 Rome, Italy.
J Clin Med. 2025 Jan 22;14(3):690. doi: 10.3390/jcm14030690.
Knee osteoarthritis is a prevalent condition that significantly impacts patients' quality of life. Effective management typically involves a combination of pharmacological and non-pharmacological treatments. However, establishing a consensus on the optimal treatment strategy is crucial for standardizing care. The present study is the result of a rigorous process that combines artificial intelligence with human expertise to improve the reliability of medical recommendations. A new software platform (Butterfly Decisions, 2021, Italy) was employed to leverage AI-assisted decision-making, facilitating the digitalization of the entire consensus process. The process started with data collection through an online survey including simulated clinical cases of knee osteoarthritis collected by 30 orthopedic surgeons; artificial intelligence (AI) analyzed the collected clinical data and identified the key concepts and relevant patterns. Subsequently, AI generated detailed statements summarizing key concepts extracted from the data and proposed a reformulation of the statements to be discussed during the discussion session of the advisory board. The advisory board, composed of four qualified, experienced specialists of knee osteoarthritis, evaluated statements, providing their agreement levels, confidence, and supporting evidence. The AI tools calculated the degree of certainty and contradiction for each statement based on these evaluations. The literature was critically evaluated to ensure that there was an evidence-based evaluation of the proposed treatment statements. Finally, revised versions were proposed to address the feedback, evidence was collected to refine the scientific report, and the board members evaluated the AI performance too. The consensus analysis revealed a high level of agreement in the need for a multimodal approach to treating knee osteoarthritis. The feedback highlighted the importance of integrating physical therapy and weight management, non-pharmacological methods, with Symptomatic Slow-Acting Drug for Osteoarthritis (SYSADOAs) and pharmacological treatments, such as anti-inflammatory drugs and intra-articular knee injections. The board members found that AI was easy to use and understand and each statement was structured clearly and concisely. The expert consensus about knee osteoarthritis conservative management being facilitated with AI met with unanimous agreement. AI-assisted decision-making was shown to have excellent analytical capabilities, but algorithms needs to be trained by orthopaedic experts with the correct inputs. Future additional efforts are still required to evaluate the incorporation of AI in clinical workflows.
膝关节骨关节炎是一种常见病症，严重影响患者的生活质量。有效的管理通常需要药物治疗和非药物治疗相结合。然而，就最佳治疗策略达成共识对于规范治疗至关重要。本研究是一个严格过程的成果，该过程将人工智能与人类专业知识相结合，以提高医疗建议的可靠性。采用了一个新的软件平台（Butterfly Decisions，2021年，意大利）来利用人工智能辅助决策，促进整个共识过程的数字化。该过程始于通过在线调查收集数据，该调查包括30名骨科医生收集的膝关节骨关节炎模拟临床病例；人工智能分析收集到的临床数据，识别关键概念和相关模式。随后，人工智能生成详细陈述，总结从数据中提取的关键概念，并提出在咨询委员会讨论会上待讨论陈述的重新表述。咨询委员会由四名合格的、经验丰富的膝关节骨关节炎专家组成，对陈述进行评估，给出他们的同意程度、信心和支持证据。人工智能工具根据这些评估计算每个陈述的确定程度和矛盾程度。对文献进行了严格评估，以确保对提议的治疗陈述进行基于证据的评估。最后，提出修订版本以回应反馈，收集证据以完善科学报告，委员会成员也对人工智能的表现进行了评估。共识分析显示，在采用多模式方法治疗膝关节骨关节炎方面达成了高度共识。反馈强调了将物理治疗和体重管理等非药物方法与骨关节炎症状性慢作用药物（SYSADOAs）以及抗炎药物和膝关节腔内注射等药物治疗相结合的重要性。委员会成员发现人工智能易于使用和理解，每个陈述的结构清晰简洁。关于人工智能有助于膝关节骨关节炎保守治疗的专家共识得到了一致认可。人工智能辅助决策显示出出色的分析能力，但算法需要由骨科专家用正确的输入进行训练。未来仍需要进一步努力评估人工智能在临床工作流程中的应用。
