Using Data Mining to Forecast Medical Resource Consumption of Diabetic Nephropathy Patients

• Main Authors: HUANG,BO-LIN, 黃柏霖
• Other Authors: LEE,TIAN-SHYUG
• Format: Others
• Language: en_US
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/88249752083947701286

碩士 === 輔仁大學 === 企業管理學系管理學碩士班 === 105 === Diabetes has become an important public health issue in the twenty-first century and dialysis treatment has become a large burden on the National Health Insurance (NHI) system of Taiwan, and diabetic nephropathy (DN) is the leading factor that affects whether diabetic patients need dialysis treatment. The rate of end-stage renal disease (ESRD) in Taiwan is the highest in the world. Statistics produced by the Ministry of Health and Welfare in 2015 indicate that chronic kidney failure (uremia) was the second highest cause of visits to primary outpatient clinics in 2015, second only to cancer. According to the National Health Insurance Administration at the Ministry of Health and Welfare, 6% of the health insurance budget is spent on dialysis treatment for ESRD patients. Since there have been few studies on the medical resources consumed by diabetic nephropathy in Taiwan. Therefore, this study proposes a forecasting model for DN patients. In this study, we used five techniques including multiple regression, stepwise regression, multivariate adaptive regression splines (MARS), support vector regression (SVR) and a two stage model (T-SVR), to establish a model for predicting the medical resources consumption of diabetic nephropathy patients. To construct the T-SVR model, the input variables we used are the union of variables that are identified as important variables by stepwise regression and MARS for constructing T-SVR model. The results of comparing these models show that the best performance was odtained usuing SVR, followed by the T-SVR model. In addition, we found five important variables, namely hypertension disease, dyslipidemia disease, cardiovascular disease, cerebrovascular disease and kidney disease excluding DN. The results in this paper identify the important factors that have a significant impact on medical resources consumption, as well as the model with the best forecasting performance of all the data mining techniques. This study can provide suggestions of medical institutions for allocating medical resources and controlling medical resources consumption, so that the medical resources can be allocated more suitably and effectively.