An Identifying Model of High Medical Utilization Among Outpatients－Using Back-Propagation Neural Network

• Main Authors: Yu-Chi Chu, 朱玉琪
• Other Authors: Peng-Teng Chang
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/98833816341646796454

碩士 === 東海大學 === 工業工程與經營資訊學系 === 97 === National Health Insurance has already run for 14 years and gradually achieves the goal which is reducing the medical burden of the public. This can be attributed to the efforts of the National Health Insurance Bureau, the cooperation of medical community and the support of the public. Particularly the operating efficiency of low-premium and high satisfaction rate has been recognized by the international community. However, a good social welfare system needs good finance to support it. Healthy finance is the key to make national health insurance sustainable. This study analysis the National Health Insurance Research Database in 2007 on the Longitudinal Health Insurance Database in three dimensions: the user characteristics, the treatment habits and the consumption of medical resources. And explore the factors impacting outpatient heavy users to use the medical resources. Use the Back-Propagation Neural Network to build a heavy user factors determining model. Detect abnormal medical object and high potential heavy user. This can be provided to National Health Insurance Bureau for heavy user countermeasures reference. This study constructs two kinds of models that use 31 and 35 variables as the input layer nodes to input 6-16 hidden layer nodes model for verification. For model 1, 14 hidden layer nodes have the highest correct testing rate. For model 2, 13 hidden layer nodes have the highest correct testing rate. These are the best testing methods of these two models. The verified results showed that 14 hidden layer nodes have 79.37% correct testing rate in model 1. 14 hidden layer nodes have 82.50% correct testing rate in model 2. Model 2 has the higher correct data-testing rate, so this study used model 2 for building factors of outpatient heavy users determining. Conclusion: This model can be used to predict in advance the potential of high users of outpatient medical care, and prevent the excessive usage of medical resources such as repeated medical treatment taking or medicine drawing. Thereby we can reducing the waste of medical resources and improve the quality of medical services.