資料連結應用於道路交通事故傷害研究：一對一確定連結與一對多人工可能連結之比較

Record linkage for Road Traffic Injury Research:One-to-One Deterministic linkage vs. One-to-Many Manual Probable linkage

呂宗學(Tsung-Hsueh Lu) ； 賴靜慧(Ching-Huei Lai) ； 林豐福(Fong-Fu Lin) ； 李孟智(Meng-Chih Lee) ； 周明智(Ming-Chih Chou)

《台灣公共衛生雜誌》 20卷2期 (2001/04) Pp. 112-120

https://doi.org/10.6288/TJPH2001-20-02-05

摘要

背景：警政單位的交通事故登錄檔(交通檔)與衛生單位的死因統計登錄檔(死因檔)是交通事故傷害研究與防制計劃的重要參考訊息，但是兩個資料庫也都有一些缺點。若能將兩資料庫連結，可以提高兩者訊息的有用性。目標：本研究希望能建立一個連結資料庫，比過去資料庫能提供更有用訊息，並提供未來電腦機率連結程式設計基本訊息。方法：一九九五年交通檔與死因檔首先以身份證字號進行一對一確定連結，未被連結到的資料再以出生年月進行一對多人工可能連結。一對多可能配對再進一步以死亡證明書記錄之訊息來確認是否為正確配對。結果：本研究共連結3,115筆資料，確定連結與人工可能連結分別可配對到853與2,262筆資料。在交通檔警察登錄死亡的3,097個案中，有2,692(87%)可以相對應的交通檔。不同資料特徵(性別、年齡、事故發生地區、當事者人數、道路使用者型態)的連結率不同，人工可能連結資料與確定連結資料在重要特徵(事故發生地區與道路使用者型態）分佈上有統計顯著差異。兩個資料檔對同一筆資料的基本資料(身份證字號、出生年月)記錄不一致的主要原因是字形相近的抄錄錯誤。結論：由於人工可能連結在確定連結後還可多配對出近三倍的資料，因此未來若要與交通檔進行資料連結時，至少要使用人工可能連結。本研究所整理的常見抄錄錯誤，是未來電腦機率連結定義權值與部分一致規則之重要參考。(台灣衛誌2001；20(2):112－120)

關鍵字

資料庫,料連結,交通事故傷害,死亡率

並列摘要

Background: The road traffic accident file (RTAF) and cause of death file (CODF) are relevant to road traffic injury research and control programs design, but both have merits as well as demerits. Linking the CODF 'with the RTAF may yield a dataset that makes full use of information from both files. Objectives: In this study we sought to establish a linked dataset which is more useful than the existing data sets and to provide information for future computerized probability record linkage (CPRL). Methods: The RT AF and CODF for the year 1995 were first linked with one-to-one deterministic record linkage (DRL), using ID as the identifier. The remaining unmatched records were then linked with one-to-many manual probable record linkage (MPRL), using year and month of birth as identifiers. The possible pairs linked by MPRL were then verified' by the information recorded on the death certificates to determine the frue matches. Results: The linkage process linked a total of 3,115 records; DRL and MPRL matched 853 and 2,262 records, respectively. Of the 3,097 police registered fatal cases in the RTAF. 2.692 (87%) were linked to matching CODF record. Of the 8,145 certifier registered traffic-related deaths in the CODF, only 2,712 (33%) had matching RTAF record. The linkage rate varied according to different characteristics. including sex, age, place of traffic accident, number of victims involved, and type of road user. The characteristics (region of accident occurred and type . of road user) of linked records obtained by MPRL differed significantly from those obtained by URL Most of the discrepancies between the two files in registering the same identifier information (lD or birth year and month) were due to transcription errors. Conc1usions: Given the high yield of MPRL compared with DRL, and the differences in the distribution of characteristics between records linked by MPRL and DRL, MPRL is a necessary procedure for future record linkage with RTAF. The common discrepancies found in this study may serve as a useful reference in defining the weights and partial disagreement rules in future CPRL efforts.(Taiwan J Public Health. 2001 ;20(2):112-120)

並列關鍵字

factual databases ； medical record linkage ； traffic accidents ； mortality.