應用資料探勘技術於全民健康保險研究資料庫-以慢性腎衰竭為例

Applying Data Mining Technology on National Health Insurance Research Database-For Example: Chronic Renal Failure (CRF)

林瑞益(Rui-yi Lin)

指導教授 ： 楊燕珠

大同大學/經營設計學院/資訊經營學系所/碩士(2006年)

摘要

由於國人飲食、生活及用藥習慣影響，加上高齡國民人數逐年成長，以致於有超過四萬名慢性腎衰竭(尿毒症)患者接受健保的照顧治療。根據行政院衛生署民國93年統計資料顯示，腎臟疾病佔台灣十大死因之第八位，在腎臟病中，又以慢性腎衰竭為最常見。
本文利用全民健康保險研究資料庫1997~2000年的系統抽樣檔之門診處方及治療明細檔和基本資料檔之醫事機構基本資料檔為資料來源，擷取了國際疾病代碼為585的尿毒症患者為研究對象。研究目的為分析尿毒症患者的人口學特性與就醫習慣。在敘述性統計分析結果顯示1997~2000年國人罹患尿毒症的比例以男性居多，病患大多選擇醫學中心就醫。男性和女性都集中在60~79 歲之間，罹患縣市別以新竹市居多、台南市次之。國人就醫習慣以掛號洗腎科為主，其次為腎臟內科。
為了找出尿毒症與其他疾病的關聯性，利用資料探勘之關聯法則技術，探勘出資料中的關聯性知識法則。又為了避免各醫療院所浮報費用，利用決策樹的技術找出尿毒症在各醫療院所門診申報費用規則。
藉由資料探勘(Data Mining)的技術可將資料庫中有用的知識挖掘出來，以找出具參考意義的醫學知識供各醫療機構之專業醫師在診斷時的參考，亦能提供一般民眾在自我保健的疾病防治方面有預警功效。另一方面，亦可藉此偵測異常申報案件，找出可能浮報或虛報的案件，減低健保局的支出與不必要之浪費，同時本研究希望能促進政府及國人對該病的重視，提供相關醫療保健資訊，作為政府在建立腎臟保健計劃之參考，並期望本研究在腎臟病的防治上有所貢獻。

關鍵字

關聯法則 ； 決策樹 ； 資料探勘 ； 全民健康保險研究資料庫 ； 慢性腎衰竭

並列摘要

Because of the advance of food, life, and medicine, people live longer and longer. Over 40,000 patients with chronic renal failure (uremia) accept the treatment of health insurance. According to the statistical data of Department of Health, Executive Yuan, kidney disease is ranked eighth place among ten major causes of death in Taiwan. And the most of the kidney diseases are chronic renal failure.
In this article, we use the health insurance research database during 1997 to 2000 as the source and pick up Ambulatory care expenditures and the registry for contracted medical facilities to analyze those people who take uremia which international disease code encoded as “585”. After the analysis of basic statistics, it shows that: 1. Male is more dangerous than female and most of the patients are about 60~79 year’s old. 2. Most of the patients live in the Hsinchu City, Tainan city is next. 3. Patients with kidney disease are used to taking medical treatment on National Medical Center and registering mainly at Department of Hemodialysis, secondly Department of Nephrology.
In addition, we also concern about the relations of uremia with other diseases and whether the medical institutes giving inflated expenses. The data mining technologies are adopted the association between diseases by association rules and to induce the conditions of expenses between declaration and charging of all medical institutes by decision tree.
We pick up the useful information by the technologies of data mining, in order to provide not only the meaningful medical knowledge for doctors’ reference, but come into government and social publics’ notice and alarm people about the diseases protections . On the other hand, help the health insurance bureau to detect unusual cases and fine the false cases and intent to decrease the unnecessary cost and wastes. Hope this research is a contribution to human healthy and aids the government making projects in kidney protection.

並列關鍵字

association rule ； chronic renal failure ； data mining ； decision tree ； health insurance research database

參考文獻

21.朱宗信(1997)，「慢性腎衰竭之診斷與處置」，當代醫學，24(12)，pp. 62-65。

30.曾淑芬(1999)，「從醫院管理角度論全民健保資料庫」，中華衛誌，18(5)，pp. 364。

35.鄭守夏(1999)，「全民健保學術資料庫簡介」，中華衛誌，18(3)，pp. 235-236.

9.Geisser, S. (1975), “The Predictive Sample Reuse Method with Applications”, Journal of the American Statistical Association, pp. 320-328.

10.Han, J. and Kamber, K. (2001), Data Mining: Concepts and Techniques, Morgan Kaufmann Publishers.