應用迴歸樹於急診室衛材存量之預測

Application of Regression Tree in Emergency Room to Predictthe Stock

陳錦杏(Chin-Hsing Chen) ； 蔡玲真(Ling-Jen Tsai) ； 金榮義(Jung-I Chin) ； 呂彥陞(Yan-Shang Liu) ； 陳冬蓮(Tung-Lien Chen)

《健康管理學刊》 9卷2期 (2011/12) Pp. 218-235

https://doi.org/10.29805/JHM.201112.0007

摘要

不足的衛材庫存量將會導致醫療處置的不完善，降低病患就診的品質，更有引起醫療糾紛的可能性；但過多的庫存量，除造成醫院的成本負擔外，也容易造成衛材的閒置品（呆料）及過期品（廢料）等問題出現。本研究以中部某地區醫院急診室之衛材存量為研究材料，分析急診病患的就醫紀錄，且針對ICD-9、檢傷級數、氣溫、濕度等因素，找出其衛材使用法則，並進行衛材存量預測，以達到醫院成本之有效管控。因此，我們透過歷史之資料分析，加入就醫特性（ICD-9主次診斷碼、檢傷級數、就診科別）及環境干擾（氣溫、雨量、濕度、風速、能見度）等進行資料探勘（Data Mining），以找出影響急診室衛材使用量之影響因素。在考量整體模型的易用性、易懂性及易調整性下，我們選定迴歸二進樹演算法，建立每季各類別的衛材項目的預測模型樹，以進行衛材庫存量的預測。讓使用者參考所在地之歷史氣候變化，進而從各類別的衛材項目之預測模型樹中預測每季的各類別衛材庫存準備量。研究發現探勘結果確實可幫助醫院對急診室衛材存量有效預測與管理。

關鍵字

衛材存量 ； 資料探勘 ； 決策樹 ； 迴歸二進樹

並列摘要

Insufficient stock of medical material will reduce the quality of medical, and too much material stock will induce the cost burden of hospital. In this study, we investigated the storages of sterilized material in emergency room for a central section hospital in Taiwan. We analyzed the medical records of emergency patients in this hospital, and hoped to find a material usage rule to predict sterilized material storage by the fields of ICD-9, injured progression, air temperature, humidity, etc. and then to achieve effective material management in emergency room of hospital. Therefore, in this research, we addressed critical factors ICD-9, injured progression, holiday/non holiday, air temperature, humidity and applied algorithm of data mining technology to analyze and find the factors which affected the sterilized materials consumption. Cosidering the ease of use, ease of understanding, and ease of adapting for the whole model, we chose classification and regression tree (CART) to construct predicting model tree for every material item by seasons to precede the stock of sterilized material prediction. The explored results are much critical and helpful to predict and manage the storage of sterilized materials for the hospital in emergency room.

並列關鍵字

sterilized material storages ； data mining ； decision tree ； classification and regression tree(CART)