以資料視覺化進行探索性資料分析:以探討多重用藥相關因素為例

Abstract

前言 臺灣藥物費用支出一直是總醫療費用中相當可觀的一部份，且高於其他國家的藥物支出占總醫療費用之百分比。如此大量的藥物使用，衍生出包括:藥物交互作用與重複用藥…等用藥問題，而這些問題又多與多重用藥間有著密切的關係。所謂多重用藥，是指在同一時間之內服用多類藥物，或指使用了比適應症所需更多的藥物，而本研究將利用健檢資料所提供的資訊，試圖探討多重用藥的狀況及相關因素為何。 研究目的 本研究將透過美兆診所之健康檢查與問卷調查資料，(1)了解成年人多重用藥的盛行狀況；(2)除了常見的性別、年齡因素外，探討教育程度、地區別、不良生活習慣、服用藥物種類與生化檢測值的異常與否，與多重用藥(polypharmacy)關係；(3)試圖針對不同年齡組別之多重用藥的人，藉由分群與排序的演算法，探索其用藥組合之異同。 材料與方法 資料來源為美兆健康管理顧問公司，所提供的台灣2000年到2006年的健檢資料。資料可分為兩部分:問卷資料與檢測資料。研究方法，在敘述統計部分除了傳統的統計圖表外，更運用了廣義相關圖(GAP)探索受檢者的用藥組合的異同。在推論統部分，以多元邏輯斯迴歸與二元邏輯斯迴歸，作為模型建立的方法。 研究結果 邏輯斯迴歸分析結果發現，女性、大學以下學歷者與吃飯不定時者，三者都是多重用藥的風險較高的族群。此外，心臟病與高血脂等藥物的服用者，其多重用藥的勝算，更是不服用者的10倍以上，而中藥的服用與多重用藥間，則呈現負向的關係。 而各年齡層的多重用藥受檢者，用藥的群聚分析結果發現，20-44歲與45-64歲的多重用藥受檢者，皆歸納出一群女性居多的群集，且偏好服用安眠藥、中藥、止痛藥與腸胃藥。相反地，男性居多的用藥群集之中，受檢者則偏好服用三高與心臟病用藥。而65歲以上的多重用藥受檢者，安眠藥的服用者與中藥、止痛藥的服用者分別形成了兩個明顯的用藥群集。

Introduction The drug expenditure in Taiwan has always been a significant part of the total medical costs, and the drug expenditure percentage of the total medical expenses is higher than the other countries. With such large drug usage, medication problems are derived, such as drug interactions and therapeutic duplications, etc… Besides, these problems have a close relationship with polypharmacy. The so-called polypharmacy refers to taking multiple types of drug simultaneously, or overuse drugs more than the required ones. This thesis will utilize the information provided by the health check-up database to explore the status and the related factors of polypharmacy in Taiwan. Objective This study will use the health check-up reports and questionnaire data of MJ (1) to understand the prevalence of polypharmacy in adults (2) besides the common factors, gender and age, to confirm the relationship between factors as education level, region, unhealthy habits, biochemical test, and the result, polypharmacy (3) to explore the similarity and difference of drug usage combination among people in different age groups with polypharmacy by clustering and sorting algorithms. Materials and Methods The health check-up database provided by MJ is in Taiwan from 2000 to 2006. It can be divided into two parts: questionnaire data and health check-up reports. About the research methods, in addition to the traditional statistical charts, generalized association plot(GAP) is applied to explore the similarity and difference between the subject's drug combination. As for inferential statistics, the multinomial logistic regression and binary logistic regression are the methods for model building.

Results According to the logistic regression analysis, it is found that female subjects, subjects without college degree, and subjects having meals irregularly are higher risk groups for polypharmacy. In addition, the polypharmacy odds for patients taking medicine of heart disease and hyperlipidemia is 10 times higher than that do not. Furthermore, the relationship between Chinese medicine usage and polypharmacy is negative. In the polypharmacy subjects of all age groups, drug clustering analysis shows that in 20-44- years- old and 45-64-year-old subjects with polypharmacy, a group of mostly female clusters who prefer to take sleeping pills, Chinese medicine, analgesic and gastrointestinal drugs were summarized separately. In contrast, in the male majority drug cluster, the preference for taking antihypertensive, hypolipidemic and antidiabetic agents is high. As for subjects over the age of 65, sleeping pills and Chinese medicine users as well as analgesic users form two obvious drug clusters.