An Initial Study of Using Big Data Analysis to Explore Intradialytic Hypotension and Ultrafiltration

• Main Authors: Dong-ShengCai, 蔡東陞
• Other Authors: Jr-Jung Lyu
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/yxk7w6

碩士 === 國立成功大學 === 工業與資訊管理學系 === 107 === When hemodialysis is used to treat end-stage kidney disease, a common risk is intradialytic hypotension. During hemodialysis, different clinical decisions will further affect the risks generated in hemodialysis. Applying big data analysis will improve clinical decision-making, because there is a great correlation between clinical decision-making and medical quality and patient safety. Therefore, many researches have shown that data analysis can improve medical quality and patient safety, and can assist new medical staff in making decisions. In medical-related data research, the more commonly used machine learning method is the decision tree, because each decision stage of the decision tree is very clear and powerful. Therefore, this study will use decision trees to build predictive models. Finally, combined with the professional knowledge of medical staff to explore the feasibility of applying big data in the healthcare. The data set was collected from October 1, 2016 to December 14, 2017 where there were 131 patients with end-stage kidney disease in the case hospital. There are 10 dependent variables that can be adjusted manually by the dialysis machine. The response variable is the next systolic blood pressure of the subjects. The final data set has 936,700 rows and 42 variables. The data set is further divided into training and test sets through data partition and predictive models are established by machine learning methods, namely decision tree. Finally, we obtained some research results through the decision tree model: the five key variables affecting systolic blood pressure in dialysis treatment. In addition, we also obtained five rules for patients with intradialytic hypotension risk in the future. Finally, based on the research results, the feasibility of applying big data in the medical field is listed. In addition to providing early warning indicators, it can also provide relevant rules to improve the dialysis treatment process. For junior staff, it can accelerate the learning curve, shorten the training schedule, reduce the burden of overall medical care, and then improve the quality of patient care and overall medical quality.