台灣政府衛生福利開放資料現況及利用率分析

An analysis of the current status and utilization of government health and welfare open data in Taiwan

目標：檢視台灣政府衛生福利開放資料品質、內容與應用狀況。方法：擷取衛生福利部暨所屬附屬機構釋出在台灣政府資料開放平臺(data.gov.tw)之資料集屬性特徵進行描述性分析。結果：截至2015年12月31日衛生機構釋出890個資料集，中央健康保險署，疾病管制署，國民健康署與食品藥物管理署分別釋出249，193，178與158個資料集。約七成資料集是屬於全球資訊網協會資料格式分類三顆星或以上較佳品質等級。瀏覽次數前十名資料集分別是：死因統計，醫院病床統計，藥局基本資料，登革熱每日確定病例統計，國際醫藥品稽查協約組織有關藥品優良製造指引藥廠名單資料集，災民收容所，不符合食品資訊資料集，各醫院癌症資源中心，可供食品使用原料彙整一覽表，健保用藥品項壓縮總檔。平臺收錄52個活化應用程式，其中有25個有使用到政府衛生福利開放資料。結論：台灣衛生福利機構釋出政府衛生福利開放資料集數目雖然比其他部門機構少，但是被活化應用的比例相對較高。但是，有些衛生機構將同一主題資料集切割成太多小資料集，不符合資料完整原則：資料應該以其來源最完整型式與最精細的尺度收錄，不使用加總或修改後的格式。此外，未來應可加強政府衛生福利開放資料的加值分析質與量，主動回饋釋出機構改善釋出資料品質與內容，朝建立良好資料加值分析生態系統的方向努力。

Objectives: To examine the current status and utilization of government health and welfare open data in Taiwan. Methods: Characteristic information from datasets released by the Ministry of Health and Welfare and related Institutions available at data.gov.tw were extracted for descriptive analyses. Results: On December 31 2015, 890 health and welfare datasets were available at data.gov.tw. The numbers of datasets released by the National Health Insurance Administration, Center for Disease Control, Health Promotion Administration, and Food and Drug Administration were 249, 193, 178, and 158, respectively. More than seventy percent of the datasets were ranked as third star or above (good quality) according to the W3C (World Wide Web Consortium) classification. The ten most frequently browsed datasets were cause of death statistics, hospital beds statistics, basic information about pharmacies, the daily number of confirmed dengue fever cases, the PIC/S GMP (Pharmaceutical Inspection Convention and Pharmaceutical Inspection Co-operation Scheme Guide to Good Manufacturing Practice for Medicinal Products) pharmaceutical manufactures, the list of shelters for disaster victims, the list of foods which did not meet government standards, the list of cancer treatment centers, the list of food materials for human use, and the list of drugs reimbursed by the National Health Insurance. Fifty-two applications were registered at data.gov.tw and 25 of these used government health and welfare open data. Conclusions: Although the number of datasets released by health institutions was lower than that of other sectors, the number of applications using government health and welfare open data was relatively high. Some of the datasets were divided into too many small datasets by some health institutions however, and this did not follow the primacy principle. We suggest that the quality and quantity of value-added analyses should be improved and that feedback be provided to the institutions releasing the datasets in order to establish a good valueadded analyses ecosystem.