從美國經驗探討：台灣生成式AI於醫療病歷應用之發展與挑戰

Abstract

隨著生成式人工智慧（Generative AI）技術的快速發展，其於醫療領域中的應用日益受到關注，特別是在電子病歷（Electronic Medical Record, EMR）撰寫、資料整理與臨床決策支援等面向展現出高度潛力。然生成式AI的臨床落地仍面臨法規限制、資料整合困難與臨床信任不足等多重挑戰。 本研究旨在探討生成式人工智慧（Generative AI）於醫療病歷應用的發展現況，並結合PESTEL分析法與個案研究法進行質性研究，從政治、經濟、社會、科技、環境與法律六大構面，深入分析美國與台灣在生成式AI醫療應用上的產業環境差異。藉由比較兩國具代表性的企業—美國EPIC Systems與台灣ASUS華碩之導入經驗，歸納其推動歷程中的成功作法與挑戰，進而提出具體且可行的策略與政策建議，作為台灣未來發展生成式AI醫療應用之參考依據。 本研究結果針對政府、醫療機構與企業三個關鍵角色分別提出具體建議。 在政府層面，應加速建立相關法規與責任歸屬制度，推動醫療AI監理沙盒與試點制度，並整合全國EHR平台與資料標準化政策，以奠定制度基礎。同時，政府亦應主導語料庫建置與本土語言模型開發，推動公私協作的資料共享機制，並整合跨部會資源，制定中長期政策藍圖。 對醫療機構而言，則須強化病歷書寫標準化與術語規範，推動AI應用試點與效益評估，促進臨床醫師參與模型導入與驗證流程，並培育跨域應用人才與建立品質追蹤機制。 至於企業，則應聚焦於特定醫療場域的垂直整合應用，深化模組化工具開發，並透過跨業合作與平台聯盟擴展生態系，同時積極參與語料與模型訓練計畫，並建立臨床驗證與產品認證機制，以提升產品實用性與合規性。 本研究期望能為台灣在推動生成式AI於醫療病歷應用之發展提供策略性參考，並為未來政策規劃與學術研究提供理論基礎與實務指引。

With the rapid advancement of Generative Artificial Intelligence (Generative AI), its applications in the healthcare sector have garnered increasing attention - particularly in electronic medical record (EMR) documentation, data summarization, and clinical decision support. Despite its promising potential, the clinical adoption of generative AI still faces multiple challenges, including regulatory restrictions, data integration issues, and lack of trust among healthcare professionals. This study aims to explore the current development of Generative Artificial Intelligence (Generative AI) in the application of electronic medical records. It adopts a qualitative research approach by integrating the PESTEL analysis and case study methodology to examine the industrial environment differences between the United States and Taiwan across six dimensions: Political, Economic, Social, Technological, Environmental, and Legal. By comparing the implementation experiences of two representative companies - EPIC Systems in the U.S. and ASUS in Taiwan - this research identifies key success factors and challenges encountered during the adoption process. Based on these findings, the study proposes concrete and actionable strategies and policy recommendations to serve as a reference for the future development of generative AI in Taiwan's healthcare sector. The findings of this study offer concrete recommendations for three key stakeholders: the government, healthcare institutions, and enterprises. At the government level, efforts should be made to accelerate the establishment of relevant regulations and accountability frameworks, promote AI regulatory sandboxes and pilot programs for healthcare, and integrate the national Electronic Health Record (EHR) platform with data standardization policies to lay a solid institutional foundation. Furthermore, the government should take the lead in developing medical corpora and local language models, foster data-sharing mechanisms through public-private collaboration, and coordinate inter-ministerial resources to formulate a mid- to long-term policy roadmap. For healthcare institutions, it is essential to enhance the standardization of medical record documentation and terminology, promote AI application pilots and benefit evaluations, and encourage clinical physicians to participate in model implementation and validation processes. Institutions should also cultivate interdisciplinary talents and establish mechanisms for quality monitoring. As for enterprises, the focus should be on vertically integrated applications in specific medical domains, advancing the development of modular tools, and expanding ecosystems through cross-industry collaborations and platform alliances. Additionally, companies should actively engage in corpus construction and model training initiatives, and establish clinical validation and product certification mechanisms to enhance the practicality and regulatory compliance of their offerings. This study aims to provide strategic insights for Taiwan’s future advancement of generative AI in EMR applications and to offer theoretical and practical guidance for policy formulation and academic research.