運用決策樹技術探討急診病患醫療費用之消耗

Utilization of a Decision Tree for High Expenditure Patients in the Emergency Department

周歆凱(Hsin-Kai Chou) ； 蘇喜(Syi Su) ； 黃興進(Hsin-Ginn Hwang) ； 蔡明足(Ming-Tsu Tsai) ； 翁林仲(Lin-Chung Woung)

《台灣公共衛生雜誌》 25卷6期 (2006/12) Pp. 430-439

https://doi.org/10.6288/TJPH2006-25-06-03

摘要

目標：傳統上，決策樹之分類技術在市場上多運用於顧客資料的區隔分析。然而，本研究也應用該工具來探討急診病患醫療費用之耗用，期望從大量之費用資料庫中，探勘出病患屬性與其醫療費用消耗之潛在關係。方法：本研究收集某醫學中心急診室一年之病患就診資料，並利用資料探勘技術中之決策樹工具來觀察各醫療費用群（低費用組、一般費用組、高費用組）間之病患特質（人口學特質、就醫屬性）的分類；藉由分類規則的建立，可預測病人於就診時可能消耗之醫療費用多寡。結果：決策樹以多層次之樹枝分佈及顏色區塊等視覺化方式呈現研究結果；其中資訊增益順序為（滯留時間＞疾病分類＞離院後動向＞檢傷分級＞科別），該資訊增益之順序也代表屬性影響醫療費用分佈之程度，意即滯留時間為決定急診病色醫療費用多寡之首要因素。結論：本研究建議個案醫院能針對不同類型之病人，給予個人化的照護服務，期望改善病人再回診的情形、降低滯留急診的時間，同時也能降低病人於急診發生之醫療費用。

關鍵字

急診醫學 ； 資料探勘 ； 決策樹

並列摘要

Objective: Traditionally, classification via a decision tree has been primarily used to distinguish between types of customers. In the current study, however, a decision tree was used to track a patient's medical expenditures in the emergency department and determine the potential relationship between patient attributes and expenses, as derived from a large database maintained in the emergency department of the hospital. Method: Patient records were collected inform the emergency department of a medical center over the course of approximately one year and a decision tree was used to classify patient data based on the magnitude of medical expenses incurred (i.e., lower, average, or higher); in the future, we will be able to predict the potential medical expenditures of emergency department patients according to such a classification. Result: The decision tree consisted of multiple levels of branches and color blocks to present the output and the sequence of information gathered (e.g., length of stay＞disease classification＞mode of departure from the hospital＞triage＞medical specific) and reflected the degree to which the distribution of medical expenses were influenced. Conclusion: This research suggests that the hospital can supply professional and personal services to various patients who have some special needs; at the same time, the hospital also can reduce the number of patients that re-visit the emergency department within 72 hours or remain in the emergency department ＞24 hours, thereby decreasing the expenditures within the emergency department.

並列關鍵字

Emergency Medicine ； Data Mining ； Decision Tree

參考文獻

Craven MW, Shavlik JW(1997).Understanding time series networks: a case study in rule extraction.(Int J Neural Syst).

Mustard CA, Kozyrskyj AL, Barer ML(1998).Emergency department use as a component of total ambulatory care: a population perspective.(Canadian J Med Association).

Rund DA, Rausch TS(1981).Triage.(St.Louis: CV Mosby Co).

Tsien CL, Fraser HS, Long WI, Kennedy RL(1998).Using classification tree and logistic regression methods to diagnose myocardial infarction.(Medinfo).

Wailer AE, Hohcnhaus SM, Shah PJ, Stern EA(1996).Development and validation of an emergency department screening and referral protocol for victims of domestic violence.(Ann Emerg Med).