單一劑量餐包藥品核對系統成效分析：某精神專科醫院經驗分享

Assessment of Drug Recognition System Utility and Effectiveness: Experience of Mental Hospital

翁益基(Yih-Ji Wong) ； 古貞庭(Chen-Ting Ku)

《醫學與健康期刊》 8卷2期 (2019/09) Pp. 57-64

摘要

目前單一劑量餐包皆由藥師人工核對。餐包藥品核對系統（medicine detection check-out，MDC）可利用機器執行餐包內藥品比對。本報告期望瞭解此機器藥物辨識正確率及經濟效益。以本院急性病房UD餐包進行連續24工作天檢測。MDC機器檢測後再由藥師進行雙重核對。共檢測17166餐包及57516藥錠，平均每個餐包檢測時間約為3.63秒。敏感性為95.92%（47/49）、特異性96.31%（14685/17117）、儀器餐包錯誤檢測率為6.92%（47/679）、儀器餐包正確檢測率為99.99% （16485/16487）。MDC之優點：節省時間、初步篩檢、餐包資訊數據化及圖像數位化；缺點則有藥品不可重疊或站立、藥品數多於7顆，準確率下降、檢核順序與包藥機包藥順序顛倒。總結若能將MDC與藥包機整合，進行初步品管稽核及藥品數量管理，則流程品管會更加完善。

關鍵字

藥物辨識系統 ； 影像辨識 ； 用藥安全

並列摘要

It is necessary to verify pouch content even with an optimized tablet packing machine process. Hospital pharmacies are currently performing pouch verification checks manually. The medicine detection check-out system(MDC) can automatically check patient pouches. Herein we assess the system's effectiveness. Unit dose packages for the acute ward of the hospital was tested for 24 consecutive days. Pouches were first detected by the machine and then rechecked manually by the same pharmacist. A total of 17,166 pouches and 57,516 tablets (or capsules) were evaluated, and the average time spent per pouch was approximately 3.63 seconds. The sensitivity was 95.92%(47/49), specificity 96.31%(14,685/17,117), positive predictive value 6.92%(47/679), and negative predictive value 99.99%(16,485/16,487). The advantages of MDC system include saving time, assisting the pharmacists in preliminary screening, digitizing pouches information and digital images of the pouches for quality management and analysis. Disadvantages include the need to reduce the chance of pills contacting or overlapping each other. The maximum number of drugs per pouch is limited to 7, the order of inspection is in reverse of the order of the drugs in the package machine. If the MDC can be integrated into a pharmaceutical packaging machine for initial quality control, the quality management of the package manufacturing process would be more complete.

並列關鍵字

Drug recognition system ； Pattern recognition ； Medication-use safety

參考文獻

衛生福利部：台灣病人安全通報系統2016年年報。http://www.patientsafety.mohw.gov.tw/ebook/20170726/mobile/index.html#p=1。 引用2/2/2018。Ministry of Health and Welfare (Taiwan, ROC). Taiwan Patient-safty Reporting System: Annunal Report 2016. Available at: http://www.patientsafety.mohw.gov.tw/ebook/20170726/mobile/index.html#p=1. Accessed February 2, 2018. [In Chinese].

Thrall JH, Li X, Li Q, et al.: Artificial Intelligence and Machine Learning in Radiology: Opportunities, Challenges, Pitfalls, and Criteria for Success. Journal of the American College of Radiology. 2018;15:504-8.

蔡紀眉：看好中醫AI應用！念資工的他愛上傳統醫學，打造藥材辨識App。https://www.bnext.com.tw/article/48006/when-traditional-chinese-medicine-meets-artificial-intelligence。引用2/1/2018。Jimei Tsai. When traditional chinese medicine meets artificial intelligence. Available at: https://www.bnext.com.tw/article/48006/when-traditional-chinese-medicinemeets-artificial-intelligence. Accessed February 1, 2018.

Chaczko Z, Kale A: Automated Tablet Quality Assurance and Identification for Hospital Pharmacies. International Journal of Electronics and Telecommunications 2011;57: 153-8.

Špiclin Ž, Bukovec M, Pernuš F, Likar B: Image registration for visual inspection of imprinted pharmaceutical tablets. Machine Vision and Applications 2011;22:197-206.