資料探勘技術應用於健保資料庫探討重大疾病患者就診特性分析-以雲林縣居民為例

Use of Data Mining Approach for Digging the Characteristics of Major Disease Patients-A Case Study of Yunlin Residents

張淑芬(Shu-Fen Chang)

指導教授 ： 陳大正

國立虎尾科技大學/管理學院/資訊管理研究所/碩士(2010年)

摘要

隨著醫療水準提升、衛生改善及民眾知識水準提高，已造成民眾對過去由被動疾病治療觀念轉為主動的疾病的預防，因此本研究藉由全民健康保險研究資料庫資料，以雲林縣重大疾病患者為研究對象，運用資料探勘技術來探索疾病患者就診特性，希望可以挖掘出病患潛在的就診特性並加以分析做為疾病預防的參考。
本研究利用C5.0決策樹演算法來進行資料探勘，以八項重大疾病做為分類的標準，藉由帕列扥 ( Pareto Principle ) 法則篩檢最重要的20%法則，來顯現80%就診資料之特性。根據實驗結果，以年齡、性別、就診費用可看出較大關聯性。在年齡部份：發現重大疾病大都發生在中老年人身上，性別部份：發現男性罹患重大疾病的機率略高於女性，就診費用部份：發現沿海地區鄉鎮患者、糖尿病患者及中老人花費較高醫療資源。可將這些現象提供給相關醫療單位做為疾病預防與衛教政策參考的方向，更希望政府可以規劃完善的社會福利資源及健康照護體系，將可提升醫療品質及有效運用健保資源。

關鍵字

資料探勘 ； 決策樹 ； 全民健康保險

並列摘要

Upon the promotion of medical care, environment improvement and people with higher education, people are with more concern about the disease prevention instead of the treatment of diseases. Therefore, the disease treatment can be decreased by disease prevention in advance. According to the data from National Health Insurance Research Database, data mining approach has been applied to discover the characteristics of major diseases patients, the study is to discover the explicit rules so as to find the related knowledge in medical records .
Since March 1, 1995, National Health Insurance was executed, but many problems exists in the insurance system. For example, abuse of medical resources, unequal distribution of funds, in particular, very serious differences of medical care between urban and rural areas. Also, the remote villages and towns are with fewer medical resources which cause to considerable insfufficient medical care for these residents.
In this study, the C5.0 decision tree algorithm has been applied for mining the interesting hidden pattern and expressed by if-then rules. The experiental results show that sight classification of major diseases can be successfully represent by these rules. Based on these rules, the significant patterns for each disease can be observed. According to the results, the most serious diseases found in the old people, with the probability of major diseases in male patients were higher than women, and the high cost in coastal towns patients, diabetes patients, and the elderly people.We hope this research results will help the Government to improve social welfare resources planning.

並列關鍵字

Data Mining ； Decision tree ； National Health Insurance

參考文獻

1. 陳美霞，「南臺灣的公共衛生問題：與北臺灣的比較分析」，臺灣公共衛生雜誌，504-518頁，第24卷，2005年。

4. 張鴻仁、黃信忠、蔣翠蘋，「全民健保醫療利用集中狀況及高、低使用者特性之探討」，台灣衛誌，第21卷，第3期， 207-213頁，2002年。

20. 王慧齡，台灣地區民眾高門診就醫族群家庭特性的研究，暨南國際大學，碩士論文，2009年。

13. 周歆凱，利用資料探勘技術探討急診高資源秏用者之特性，台灣大學，碩士論文，2004年。

14. 邱涵芬，應用資料探勘與文件分類建構個人化知識查詢網站－以全民健康保險研究資料庫與醫藥新聞為例之研究，大同大學，碩士論文，2006年。