Lima Henrique A, Trocoli-Couto Pedro H F S, Zaman Marzia, Engelmann Débora C, Parkes-Ratanshi Rosalind, Junck Leah, Hendry Brenda, Taylor Amelia, El Kawak Michelle, Ravi Nirmal, Santos Henrique D P, Pawlik Timothy M, Resende Vivian
Department of Surgery, Federal University of Minas Gerais Faculty of Medicine, Belo Horizonte, Minas Gerais Brazil.
CMED Health Limited, Dhaka, Bangladesh.
Npj Health Syst. 2025;2(1):28. doi: 10.1038/s44401-025-00034-3. Epub 2025 Jul 30.
We sought to define healthcare workers' (HCW) views on the integration of generative artificial intelligence (AI) into healthcare delivery and to explore the associated challenges, opportunities, and ethical considerations in low- and middle-income countries (LMICs). We analysed unified data from selected 2023 Gates Foundation AI Grand Challenges projects using a mixed-methods, cross-sectional survey evaluated by an international panel across eight countries. Perceptions were rated on a simplified three-point Likert scale (sceptical, practical, enthusiastic). Among 191 frontline HCWs who interacted with AI tools, 617 responses were assessed by nine evaluators. Enthusiastic responses accounted for the majority (75.4%), while 21.6% were practical and only 3.0% were sceptical. The overall interclass correlation coefficient of 0.93 (95%CI: 0.91-0.94, with an average rating = 9) indicated excellent inter-rater reliability. While quantitative data underscored a generally positive attitude towards AI, qualitative findings revealed recurring cultural and linguistic barriers and ethical concerns. This is a unique study analysing data from the first applications of generative AI in health in LMICs. these findings offer early insights into generative AI implementation in LMIC healthcare settings and highlights both its transformative potential and the need for careful policy and contextual adaptation.
我们试图界定医护人员对将生成式人工智能（AI）整合到医疗服务中的看法，并探讨低收入和中等收入国家（LMICs）相关的挑战、机遇及伦理考量。我们分析了来自2023年盖茨基金会AI重大挑战项目的统一数据，采用混合方法进行横断面调查，并由一个国际专家小组在八个国家进行评估。看法通过简化的三点李克特量表（怀疑、实用、热情）进行评分。在191名与AI工具互动的一线医护人员中，九名评估人员评估了617份回复。热情的回复占多数（75.4%），实用的占21.6%，怀疑的仅占3.0%。总体组内相关系数为0.93（95%CI：0.91 - 0.94，平均评分 = 9），表明评分者间信度极佳。虽然定量数据强调了对AI的普遍积极态度，但定性研究结果揭示了反复出现的文化和语言障碍以及伦理问题。这是一项独特的研究，分析了生成式AI在LMICs卫生领域首次应用的数据。这些发现为LMICs医疗环境中生成式AI的实施提供了早期见解，并突出了其变革潜力以及精心制定政策和因地制宜的必要性。
