2026-11-092026-11-09https://scholars.csmu.edu.tw/handle/123456789/37377In order to effectively examine possible causes and determinants of road traffic injury, reliable information on the participants, circumstances, and resultant injuries and deaths must be available. A fatal road traffic injury database, linking Police Road Traffic Accident File (RTAF) and Cause of Death File (CODF), has been established. Traditional one-to-one deterministic record linkage by ID number linked 853 records, nonetheless, the one-to-many probability record linkage procedure linked additional 2,262 records. Within the linked data set, a comparison of identical variables, such as sex, year of birth, month of birth, and ID number between the two files was made. Manually linkage confined to cause of death as motor vehicle related death data produced another 358records. Finally a total of 3486 records were linked through 3 stages of linkage process. The distribution of important variables among deterministically linked file differed significantly from those probability linked file. Nevertheless, the distribution of important variables among fatal PRTF did not differ significantly form the linked file, so the generalization based on fatal PRTF was assured. According to both data sources, head injuries were the most prominent cause of traffic accidents. The number of head injuries reported by the policemen was one fifth lower than the number reported by medical examiner. The motorcycle-related injuries sustained higher head injuries than those car-related injuries. The percentage of head injuries was significantly higher than those of United States.背景：機動車事故傷害的防制可分為初級防制與次級防制，我們必須確切瞭解車禍事故的類型、當事人特徵與受傷部位的關係，才能有效使傷害嚴重度降低。衛生署死因統計資料檔有較多受傷部位訊息（沒有全部輸入電腦檔）；警政署交通事故資料檔有較多車禍事故發生現場狀況的訊息，若能將兩者訊息結合，一定能提供相當豐富的訊息有助於交通事故防制計畫擬定之參考。 目的：本研究首先以決定連結與機率連結將死因資料與警政資料連結；進一步比較不同連結方法所產生資料之特徵差異：最後以連結資料對車禍受傷部位記錄進行分析，尤其著重車禍事故的類型、當事人特徵與受傷部位的關係。 方法：本研究以民國84年衛生署死因統計電腦資料檔與警政署交通事故電腦資料檔進行連結，分析不同連結方式所連結個案的特徵差異。經過連結的資料，再進一步以kappa值分析死因資料與警政資料記錄同一人受傷部位的一致性。最後以卡分檢定與多元邏輯回歸分析探討車禍事故的類型、當事人特徵與受傷部位的關係。 結果：以身份證字號為確認指標進行一對一決定連結獲得853筆資料，以出生年月為主發生時間為輔進行一對多機率連結經姓名確認者共2262筆資料。縮小範圍（僅針對死因為交通事故死亡者）以姓名進行比對獲得358筆資料。經過上述四步驟的連結，本研究連結3486筆資料供未來相關研究使用。不管由警政資料或衛生資料，主要受傷部位都是頭部外傷，兩資料一致性比較，警政頭部外傷報告數還是比衛生資料低五分之一。機車與腳踏車發生頭部外傷的比率較小客車或大客車高，但是比率還是遠高於美國的比率。 結論：本次研究僅以人工方式進行機率連結，未來應進一步以電腦程式進行反覆機率連結試驗，應該可以讓連結率繼續提高。本研究所整理的連結率特徵差異與連結資料與原本警政死亡者特徵差異比較，對未來使用連結資料的外推性都有重要參考。死因統計有關受傷部位的完整性與準確度還需進一步與醫院病歷資料連結才能評估。交通事故傷害資料連結死因統計Traffic AccidentRecord LinkageCause Of Death StatisticsProject