2025/07/09 []

人工智慧輔助診斷糖尿病視網膜病變之成本效用分析

目標：糖尿病視網膜病變可能影響視力及降低生活品質，眼底檢查是診斷糖尿病視網膜病變的重要工具，眼底檢查率也是糖尿病照護品質的重要指標，隨著醫療科技的進步，人工智慧於糖尿病視網膜病變的影像判讀是最受重視的應用之一，也被視為一項執行眼底檢查的重要策略。因此，本研究以中央健康保險署觀點，探討以人工智慧輔助診斷糖尿病視網膜病變的成本效用。方法：本研究以決策樹及馬可夫的混合模型，比較兩種眼底檢查策略之成本及健康生活品質校正生命年：人工智慧(VeriSee DR) 輔助眼底檢查、傳統非眼科醫師眼底檢查。以衛生福利部衛生福利資料科學中心資料庫與文獻回顧，估計模型參數，並透過單維及機率敏感度分析評估結果的穩健度。結果：人工智慧輔助眼底檢查較傳統非眼科醫師檢查具備成本效用，遞增成本效用比為每增加一個健康生活品質校正生命年需增加9,555元，敏感度分析結果亦同。結論：人工智慧輔助眼底檢查係具成本效用的檢查策略，並能提升糖尿病病人的照護結果。

  • 預定刊載卷期：台灣衛誌 2025；44(3)
  • 原著 Original Article
  •  
  • 郭玟妤、連加恩、蕭鈺靜、滕欣、謝佩芬、楊惇筑、董鈺琪
    Wen-Yu Kuo, Chia-En Lien, Amy Y. Hsiao, Shin Teng, Pei-Fen Hsieh, Dun-Jhu Yang, Yu-Chi Tung
  • 人工智慧、糖尿病視網膜病變、成本效用分析、遞增成本效用比
    artificial intelligence, diabetic retinopathy, cost-utility analysis, incremental cost-utility ratio
  • 目標：糖尿病視網膜病變可能影響視力及降低生活品質，眼底檢查是診斷糖尿病視網膜病變的重要工具，眼底檢查率也是糖尿病照護品質的重要指標，隨著醫療科技的進步，人工智慧於糖尿病視網膜病變的影像判讀是最受重視的應用之一，也被視為一項執行眼底檢查的重要策略。因此，本研究以中央健康保險署觀點，探討以人工智慧輔助診斷糖尿病視網膜病變的成本效用。方法：本研究以決策樹及馬可夫的混合模型，比較兩種眼底檢查策略之成本及健康生活品質校正生命年：人工智慧(VeriSee DR) 輔助眼底檢查、傳統非眼科醫師眼底檢查。以衛生福利部衛生福利資料科學中心資料庫與文獻回顧，估計模型參數，並透過單維及機率敏感度分析評估結果的穩健度。結果：人工智慧輔助眼底檢查較傳統非眼科醫師檢查具備成本效用，遞增成本效用比為每增加一個健康生活品質校正生命年需增加9,555元，敏感度分析結果亦同。結論：人工智慧輔助眼底檢查係具成本效用的檢查策略，並能提升糖尿病病人的照護結果。
  • Objectives: Diabetic retinopathy (DR) can impair vision and reduce quality of life. It is primarily diagnosed through fundus screening, and the rate of fundus screening is among the most crucial health-care quality indicators for patients with diabetes. Innovations in medical technology have led to artificial intelligence (AI) being applied to assist with DR screening, and it is considered a crucial strategy. Accordingly, the objective of this study was to conduct a cost-utility analysis of AI-based DR screening from the perspective of Taiwan's National Health Insurance Administration. Methods: Using a decision tree-Markov hybrid model, we compared the costs and quality-adjusted life years (QALYs) for AI-based (VeriSee DR) fundus screening and traditional nonophthalmologist fundus screening. We analyzed the Health and Welfare Data Science Center Database and conducted a literature review to estimate the model parameters. Additionally, one-way sensitivity analysis and probabilistic sensitivity analysis were performed to evaluate the robustness of the assessment. Results: Compared with traditional nonophthalmologist fundus screening, AI-based fundus screening was more cost-effective, with an incremental cost-utility ratio (ICUR) of $9,555 per QALY gained. Similar results were also obtained from the sensitivity analysis. Conclusions: AI-based fundus screening is a cost-effective means of conducting DR screening, and therefore, it can lead to improved health-care outcomes for patients with diabetes.
  • 257-268
  • http://bit.ly/3r4HS9R