A Quantitative Analysis of Global Artificial Intelligence-related Medical Studies: An Urgent Call for More, and Better Randomized Controlled Trials

This study-level dataset underpins the scoping review “A Quantitative Analysis of Global Artificial Intelligence-related Medical Studies: An Urgent Call for More, and Better Randomized Controlled Trials.” It includes all 4,667 primary studies with a substantial AI component that were extracted from 218 systematic reviews of medical AI published between 2012 and 2024. Each row corresponds to one primary study and contains standardized variables describing: bibliographic information (review ID, first author, publication year), research stage along the AI development pathway (preclinical development/offline validation, early live clinical evaluation, comparative prospective evaluation/RCT), medical specialty (mapped to first-level ICD-11 categories), country/region, center type (single center, single-country multicenter, multinational multicenter), sample size, and the AI role in clinical practice (diagnostic yield/performance, clinical decision-making, care management, patient behavior/symptoms). For studies identified as randomized controlled trials, additional fields capture design characteristics (e.g., comparator type, randomization unit), primary outcome type and direction of effect, and self-reported adherence to reporting guidelines. The dataset contains no patient-level data and only includes information available from published articles. It is intended to facilitate secondary analyses of temporal trends, geographic and specialty concentration, study design features, and reporting practices in medical AI research, and to support methodological, policy, and implementation work aimed at improving the evidence base for AI in healthcare.本研究级数据集为范围综述《全球人工智能相关医学研究的定量分析：亟需开展更多高质量随机对照试验》提供支撑。该数据集提取自2012年至2024年间发表的218篇医学人工智能系统综述，共纳入4667项具备核心人工智能组件的原始研究。每一行对应一项原始研究，包含标准化变量，具体涵盖：文献书目信息（综述ID、第一作者、发表年份）、人工智能研发路径中的研究阶段（临床前研发/离线验证、早期真实世界临床评估、前瞻性对照评估/随机对照试验（Randomized Controlled Trial, RCT））、医学专科（映射至ICD-11（国际疾病分类第11版）一级分类）、国家/地区、中心类型（单中心、单国家多中心、跨国多中心）、样本量，以及人工智能在临床实践中的角色（诊断效能/性能、临床决策、诊疗管理、患者行为/症状分析）。对于被归类为随机对照试验的研究，额外字段记录了其设计特征（如对照类型、随机化单元）、主要结局类型与效应方向，以及作者自评的报告指南依从性。本数据集不包含任何患者级数据，仅纳入已发表文献中的公开信息。其构建目的是助力开展二次分析，涵盖医学人工智能研究的时间趋势、地域与专科分布集中度、研究设计特点及报告实践现状，同时可为旨在完善医疗人工智能领域循证证据基础的方法学研究、政策制定与落地实践提供支持。