建立急診留觀病人留置時間之預測模式

鄭孜歆(Cheng Tzu Hsin)

指導教授 ： 胡雅涵

國立中正大學/管理學院/資訊管理學系/碩士(2013年)

摘要

急診部門為緊急的醫療處置場所，在現今大幅增加的急診需求下，有效控制與預測急診病人於急診留置時間的要求也隨之增加。過去對於急診相關研究，均針對品質、急診相關因素、72小時再返方面進行探討，對於急診留觀病人留置時間方面研究較少。相關文獻指出，急診病人留置時間過長是導致急診壅塞原因之一。因此為減少急診壅塞，適切管控急診留觀病人留置時間為第一個需解決的問題。考量兒科急診就診病人疾病複雜因素，若以此類急診留觀病人的留置時間建立預測模式，進一步提供醫療管理者與急診醫師對於病人留置時間的有效管控，並延伸至各年齡層之急診病人，不僅可以提昇急診救護品質，更利未來國家衛生政策其區域性急診監控系統之建立。
本研究以2005-2007年北部某區域醫院兒科病患就診資料為研究對象，並彙整其相關變項，利用SPSS統計軟體進行敍述性統計、雙變項分析等資料分析，進而使用WEKA應用軟體之分類技術(M5P和Linear Regression)建立急診病人留置時間的預測模式。
研究結果發現單一分類器M5P預測數值較精確，其MAE值均提升10%；.且依據實驗進一步發現，包括：急診重疊時間病人數量、醫師是否具專科資格、到診班別區間看診醫師人數、抵達班別與是否假日等五項變項，可做為預測模型之基礎，更有助於建立急診留觀病人留置時間的有效監控標準。

關鍵字

急診留觀 ； 急診留置時間 ； 急診壅塞

並列摘要

Under the increasing need for emergency services, how to effectively control and predict length of stay (LOS) in the Emergency Department (ED) becomes a critical issue. However, recent studies are most on quality issue or unplanned ED revisits within 72 hours; the topic of ED observation length of stay is seldom discussed. Recent researches suggest that ED length of stay is a key factor contributing to ED overcrowding. Therefore, the concern of adequate monitoring and control of ED observation length of stay arises. Considering the disease complexity of pediatric patients in ED, the establishment of prediction model of ED length of stay among pediatric patients may extent to different age groups. This prediction model not only assists ED physicians and health managers to effectively manage the ED length of stay, but also provide essential reference to National Policy for future establishment of the regional emergency monitoring system.
These study objects are pediatric patients from a regional hospital with 6 branches in northern Taiwan from 2005 to 2007. The related variables were collected and the descriptive statistics and inferential statistics analysis were conducted by SPSS. The prediction model of ED length of stay was established by data mining method (M5P and Liner regression).
The result shows that the value of MAE in M5P is significantly superior to those of Linear Regression over 10% in predicting the ED length of stay. This study also demonstrates significant variables in predicting ED length of stay, such as the number of patients and doctors during the same working shift, percentage of ED physicians with specialist certificate, working shift, and visiting during a non-holiday or holiday.

並列關鍵字

ED observation unit ； ED Length of stay (LOS) ； ED overcrowding

參考文獻

沈希哲（民91）。瞭解並分析停留時間長短對急診醫療品質良窳的影響。財團法人醫院評鑑暨醫療品質策進會。台北市：行政院衛生署。

陳錦杏、蔡玲真、金榮義、呂彥陞（民100）。應用迴歸樹於急診室衛材存量之預測。健康管理學刊，9，85-102。

顏瑞昇、石富元、蔡光超、陳石池、陳文鍾（民92）。回診的急診病患:分析危險因子和回診原因。中華民國急救加護醫學會，14，93-98。

翁紹仁、蔡柏祥、王立敏、張群岳（民101）。資料包絡法評量及提昇急診作業績效。品質月刊，48，12-14。

紀志賢、曾詠淑、林錫璋、陳澤生、蔡良敏、駱聰成（民88）。以資料挖掘模式實施緊急救護量預測。慈濟醫學，11，337-341。