以健保資料庫與癌症登記檔建構糖尿病確診後罹患為肝癌之預測模型

Using NHIRDB and Taiwan Cancer Registry to Construct a Predict Model for Diagnosed Type 2 Diabetes and Liver Cancer.

林宥安(Yu-An Lin)

指導教授 ： 徐建業

臺北醫學大學/醫學科技學院/醫學資訊研究所/碩士(2014年)

https://doi.org/10.6831/TMU.2014.00105

摘要

動機：自民國71年以來我國十大死因排名第一名即為惡性腫瘤，糖尿病和惡性腫瘤在歷年來均出現在國人十大死因排行之中，對於國人的健康具有重大的影響，以致醫療資源的耗用甚鉅。且根據行政院衛生福利部的2012年統計資料顯示，肝癌是台灣癌症死亡的第二大原因，每年約有7,000名病患死於肝癌。主要的原因是B型肝炎及C型肝炎的高盛行率，使慢性肝病得以進一步發展成肝硬化及肝癌。目的：利用類神經網路建構第二型糖尿病確診後罹患為肝癌之預測模型。方法：藉由全民健康保險研究資料庫找出第二型糖尿病罹患為肝癌之研究對象，並利用類神經網路建構預測模型，預測模型所使用之因素則利用卡方檢定來了解其因素與肝癌之間關聯分析之檢定結果。研究結果：研究樣本共65,871人，其中共有515人得到癌症；男性共31947人，其中358得到癌症；女性共33924人，其中157人得到癌症。而疾病風險預測模型方面，類神經網路所建立出之模型其Sensitivity可達0.802、Specificity可達0.773、ROC取線下之面積可達0.873。結論：酒精性預測因子所建立出的預測模型準確性均較為低落，其原因可能為糖尿病與酒精性肝臟疾病因子間很有關聯且酒精性肝臟疾病因子與肝癌也很有關係，但糖尿病卻無直接與肝癌間有直接影響所以導致酒精性肝臟疾病因子進行預測較為低落，在未來的研究應結合更多不同的資料庫，如個人生活習慣資料庫與健康檢查資料庫，進而再尋找可能引發糖尿病確診後罹患為肝癌之風險因子。

關鍵字

全民健康保險研究資料庫、糖尿病、肝癌、類神經網路、預測模型

並列摘要

Background: Since 1982 the first one causes of death is cancer. And diabetes is also the top 10 of death in Taiwan, it severely impact people`s health, and cause a lot medical resources in Taiwan. And according to statistics in 2012 by Ministry of Health and Welfare (Taiwan) show that liver cancer is the second leading cause of cancer death in Taiwan, about 7,000 patients died of liver cancer. The main reason is the high prevalence of hepatitis B and hepatitis C, lead chronic liver disease develops into liver cancer. Objective: Using neural networks to construct a predict model that Type 2 diabetes after diagnosed conversion to Liver cancer. Methods: By Using National Health Insurance Research Database to identify Liver cancer complicated by diabetes and use of neural network to construct prediction model. Predictive models factors is use of the chi-square test and T test to test the relation between factors and Lever cancer. Results: In this study we include 31,953 males, and 364 people have liver cancer; include 33,767 females, and 157 have liver cancer; totally include 65,356 people, and 521 people have cancer. The disease risk prediction models, was created by neural network, Sensitivity is 0.802, Specificity is 0.773, and Area Under ROC is 0.873. Conclusion: The model create by Alcohol factor are lower accuracy, the reason maybe is that diabetes and Alcoholic liver disease have relation, and alcohol-related factor have relation with Liver cancer, but dabetes didn’t has a direct impact to liver cancer, so cause that alcohol-related factor can’t to predict liver cancer, In future we should be combined with other database, such as personal health database and health record database, which can find more risk factors.

並列關鍵字

National Health Insurance Research Database, Diabetes, Liver, Neural Networks, Predictive Models.

參考文獻

[20]. 謝桂鈴、郭乃文，胰島素幫浦對糖尿病患者生活品質及血糖控制的成效探討，臺北醫學大學，碩士在職專班碩士論文，2011。

[45]. 黃琮榆、姚立德、陳文輝，類神經網路於肝癌與卵巢癌質譜資料分類之應用，國立台灣科技大學自動化科技研究所，碩士論文2007。

[39]. 湯宗泰、簡守維、賴仲亮、陳永福、高琳詠，利用決策樹方法分析探討糖尿病人血糖控制之成效，2008 資訊科技國際研討會論文集，2008。

[29]. 汪忠川，肝臟轉胺酶 Alanine Aminotransferase的上升與糖尿病之發生，(台灣老年醫學雜誌 2007；3(1)：54-63。

[21]. 侯穎蕙、林士弼、邱于容、許雅蓉、鄭俐玲，糖尿病門診照護品質與就醫機構相關性探討，澄清醫護管理雜誌，2011，7(1)：32-41。