雙重資料輸入品質的比較—以Oracle Clinical 為例

中央研究院生物醫學科學研究所資料協調組(Data Coordinating Group)實驗室，以下簡稱DCG，利用Oracle Clinical這套資料庫軟體收集生醫所基因體計畫下各研究主題的資料，其目的是把傳統紙本的問卷答案經由人員上網連線，將資料輸入到資料庫以電子問卷的型態儲存。本篇採用的分析資料是由DCG實驗室針對某研究計畫採用雙重資料輸入的方式得到，也就是同一紙本問卷經由兩批不同的人員將資料輸入到資料庫中，其中第一次輸入資料(1 key)的人員是各計畫的護士人員，而第二次輸入資料(2 key)的人員是DCG實驗室內部的人員，而後由另一個資料管理人員來交互比對資料，希望藉由兩次不同人，不同時間的輸入資料，和第三者的比對來提高資料的正確性。本篇研究的主要目的在於探討兩次資料輸入時，分別產生非預期性錯誤(discrepancy)的資料品質比較，研究進行的假設如下: (1)1 key 和 2 key產生非預期性錯誤的情形有所不同。(2)不同醫院的1 key資料輸入品質有所不同。(3) 不同問卷題型發生的非預期性錯誤的情形有所不同。研究結果顯示發生錯誤情形普遍偏低，表示確實可將紙本資料轉成電子資料儲存。在上網輸入資料前，DCG人員須確實做好訓練工作以增加醫院護士操作OC資料庫的能力。雙重資料輸入的作業方式確實可以廢除，改成單一資料輸入即可，以降低人力成本。在醫院資料輸入品質方面，比較結果顯示，大型醫院整體比起小型醫院素質來的優良，計畫管理人員在尋找合作醫院時，可以列入考慮。在錯誤題型方面，統計結果顯示錯誤型態以應達而未答和資料寫錯居多數，顯示未來在設計問卷時，對於跳答和複雜性的題目，盡量設計成可簡單回答的單選題題型。

關鍵字

資料庫；電子問卷；非預期性錯誤； Oracle Clinical ；雙重資料輸入

並列摘要

Data Coordinating Group, Institute of Biomedical Sciences, Academia Sinica, hereafter referred to as DCG, utilize the database software --Oracle Clinical to collect the study material of the National Clinical Core for Genomic Medicine. The purpose is to access the questionnaire data through the net, which transfers the results of the paper-based questionnaire into those of electronic-based questionnaire to the database as storage. The analysis material is from DCG laboratory, which adopt double data entry in some plan. That is, the identical paper questionnaire answers are inputted to the database by two different groups. The first entry of data (1 key) is by nurses in each site and the second entry of data (2 key) is by DCG, and then another data manager staffs alternately collate the data, wishing that different groups input the data at different time can enhance the accuracy of data. The objective of this research is to discuss the discrepancy and the data quality comparison while entering the two data. The following three hypotheses are (1)The discrepancy appears in 1 key is different from the discrepancy appears in 2 key. (2)The quality of 1 key data varies from hospital and hospital. (3) The discrepancy of different questionnaire structure is not the same. The results showed that the discrepancy ratio is generally low, which tells that it’s practical to transferring paper data to the electronic data. Before entering the data, the DCG must complete the training to increase staff nurses’ ability to deal with OC database. To reduce the manpower cost, sole data entry can certainly take the place of double data entry. In the aspect of hospital data entry, compared with the results, the large-scale hospitals are better than small hospitals. While looking for the cooperation hospitals, project manages may put this reason into consideration. In the aspect of wrong structure, the statistical result showed that majority forms of the wrong state are not answering the question and filling in the wrong data. Thus, when designing questionnaire in the future, the questions need skipping to answer and the complex questions should be designed as multiple-choice questions which are easy to answer

並列關鍵字

Database ； eCRF ； Discrepancy ； Oracle Clinical ； Double data entry

參考文獻

1. Codd E.F. (1970), A Relational Model of Data for Large Shared Data , Communications of the ACM , 13:377-387.

3. Lehmann EL (1933) The Fisher,Neyman-Pearson Thories of Testing Hypotheses :One theory or two? Journal of the American Statistical Association 88:1242-1248.

14. 劉鵬程 (2005) ，心導管檢查或治療在不同醫院屬性下之醫療資源耗用、死亡率及重返心導管室率之分析、探討與比較，國立台灣大學醫療機構管理研究所。pp.57-63。

英文部分