Probabilistic Data Linkageを用いた消防および医療機関データ突合の試み～検者間信頼性を用いた検証～(Probabilistic Data Linkage of local emergency medical service and hospital department data: a pilot validation study with inter–rater reliability)

要旨 【目的】Probabilistic Data Linkageを用いた消防機関および医療機関のデータ突合の精度について検者間信頼性を用いて検討した。【方法】2020年1月から4月の期間に奈良県立医科大学附属病院高度救命救急センターに搬送となった251例の奈良県救急医療情報管制支援システム（e–MATCH）データと291例の同センター外来日誌データ（転院や直接受診を含む）を対象とし，RのfastLinkを用いて次の3つの方法で突合した（①年齢，性別，病院到着日と来院日，②年齢，性別，覚知日時，③年齢，性別，病院到着日と来院日，病院到着時刻と受付時刻）。突合はe–MATCH側への突合とし，2名の検者が突合の正誤を検証した。【結果】突合されたデータセットは，①245例，②248例，③243例で，正の突合はそれぞれ198例（78.9%），209例（83.3%），215例（85.7%）で，誤突合は，①47例（18.7%），②39例（15.5%），③28例（11.1%）であった。検者間の信頼性CohenのKappa係数はいずれも0.7以上であった。【考察と結語】Probabilistic Data Linkageを用いることで消防データと医療機関の精度の高いデータの突合が期待できる。 ABSTRACT Objective: This study aimed to merge data from a local emergency medical service (EMS) with hospital data using probabilistic data linkage and to validate it in terms of inter–rater reliability. Methods: Data from the Emergency Medical Alliance for Total Coordination (e–MATCH data of 251 cases transported to the Emergency Department of Nara Medical University Hospital) and hospital department data (291 cases) between 1 January and 30 April, 2020 were linked using R software environment with fastLink package in three patterns (#1: transport date, age, and sex; #2: age, sex, date, and time for emergency calls; and #3: age, sex, hospital arrival date, hospital arrival time, and reception time). Two investigators verified true or false linkages for each dataset. Results: Cohen’s kappa coefficients for each dataset were 0.73 (#1), 0.97 (#2), and 0.78 (#3). True linkage numbers were 198 (78.9%, #1), 209 (83.3%, #2), and 215 (85.7%, #3), and false linkage numbers were 47 (18.7%), 39 (15.5%), and 28 (11.1%), respectively. Conclusions: Merging local EMS and hospital data using probabilistic data linkages with data concerning date, time, age, and sex can create satisfactory data matching.