頁籤選單縮合
題 名 | 運用決策樹技術探討基層診所門診量之影響因素=Analysis of Outpatient Visits to Primary Care Clinics and Its Influencing Factors Using a Decision Tree Model |
---|---|
作 者 | 許文錦; 李佳桓; 余致廷; | 書刊名 | 臺灣公共衛生雜誌 |
卷 期 | 35:3 2016.06[民105.06] |
頁 次 | 頁281-293 |
分類號 | 419.2 |
關鍵詞 | 選址分析; 基層診所; 資料探勘; 地理資訊系統; Location analysis; Primary care clinic; Data mining; Geographical information system; |
語 文 | 中文(Chinese) |
中文摘要 | 目標:開業地點對醫療院所經營具有重要影響力,便利性高的地點可提高病患就診意願,更可讓醫療院所取得競爭優勢。隨著開放資料時代來臨,如何運用開放資料協助選址成為醫療機構經營重要課題。本研究旨在運用資料探勘技術分析開放資料,探勘影響基層診所門診量的關鍵分類規則,藉以協助選址決策。方法:本研究以台北市306間高門診量基層診所為對象,蒐集相關之開放資料,包括社經資料、診所門診量、地價金額、地理空間資料等19項資料,運用CART決策樹分析影響高門診量診所之成功關鍵因素。結果:研究結果顯示決策樹分析之資訊增益順序為(可支配家庭收入>所屬行政區診所密度>女性人口數)。迴歸分析發現65歲以上人口數(β=0.836)與離最近捷運站距離(β=-0.297)顯著影響基層診所門診量(p<0.001)。結論:本研究所發現之選址分類規則可協助診所、藥局、醫療器材與政府衛生部門找出符合高門診量診所地理條件且尚未設立診所的位址,提供開業者選址參考。 |
英文摘要 | Objectives: The process of locating health facilities has been studied extensively using mathematical optimization (e.g., covering model); however, few researchers have applied the techniques of data mining to this problem. This study proposes a novel prediction model, based on the Geographic Information System and data mining to assist in the selection of optimum locations for medical clinics. Methods: This study examined 306 medical clinics in Taipei, focusing on those with a high number of outpatients, using 19 variables associated with location decisions. A CART decision tree was used in the development of the model based on the collected variables. Results: The decision tree model indicated that household disposable income has the strongest impact on the number of outpatients, followed by clinic density, and gender. Regression analysis identified age, and distance to the nearest Mass Rapid Transit station as the two factors with a significant effect on the number of outpatients at a given clinic (p<0.001). Conclusions: This study constructed a novel prediction model to aide in identifying the optimal location for a clinic and evaluating options with regard to clinic relocation. |
本系統中英文摘要資訊取自各篇刊載內容。