目的:在實務界,資料探勘中之叢集分析技術早在市場行銷上蓬勃發展,透過大量收集顧客資料,根據客戶的屬性加以分群,挖掘出具同樣特性及特徵之消費族群。然而,本研究應用該工具來探討72小時內再返急診之病患特性,探勘出病患屬性與其返診行爲之潛在關係。方法:本研究收集某醫學中心急診室一年之72小時內再返診病患資料共2,516筆,並利用資料探勘技術中之叢集分析找出具有相同特徵或特性之群集,以便求得72小時內再返診病患之共同屬性組合。結果:叢集分析結果發現,兒科因呼吸系統疾病來急診,及產科因生產及產後合併症且檢傷一二級的病患屬於72小時內再返急診的高危險族群;若以疾病別來看,腫瘤疾病、神經系統疾病、泌尿系統疾病及呼吸系統疾病爲高危險群;若以年齡層來看,75歲以上的老人及4歲以下的小孩爲高危險群;此外,檢傷四級的假急診病患也是72小時內再返診的潛在高危險群。結論:在群集分析結果中,本研究發現許多叢集組合均與一般統計分析所歸納出之高危險群相類似;因此本研究證實資料探勘技術於醫療領域之實用性,可挖掘出72小時內再返急診之高危險群病人,並監測異常之叢集組合。
Objective: In practice, the technology of cluster analysis in data mining was developed primarily for marketing. This study applies cluster analysis to explore the properties of patient re-visits to the Emergency Department during a 72 hour period after the initial visit. Method: This study collected 2,516 patient records with re-visits to the Emergency Department of a medical center during a one year period[Please consider inserting the actual year of data collection.]; and used cluster analysis as a data mining tool to discover the shared properties of patients with re-visits. Result: The result of cluster analysis revealed that patients with characteristics that placed them in a high risk group were more likely to re-visit the Emergency Department during the 72 hours following an initial visit. High risk groups included patients with (1) oncology, neurology, urology or respiratory disease, (2) Pediatric patients with respiratory disease, or (3) Obstetric patients with pregnancy complications. Conclusion: As a result of cluster analysis, this study found that many cluster components were the same as that of high risk groups generated from general statistics; therefore, this study proved the usefulness of data mining in the field of medicine. Cluster analysis mines high risk patient re-visits to the Emergency Department during a 72 hour period, but will also monitor abnormal cluster components.