| 摘要: |
| 目的 通过对糖尿病住院费用的主要影响因素进行分析，探寻有效分析方法，为合理控制糖尿病住院费用提供理论依据。方法 研究对象为北京市2011年13家三级甲等医院的全部2型糖尿病住院患者病案首页数据，诊断依据为出院主诊断（ICD编码为E11）。采用反向传播神经网络建模，并与多元线性回归模型相比较，分析糖尿病住院费用影响因素。结果 (1)反向传播神经网络预测误差比多元线性回归低，在数据拟合方面优于多元线性回归模型。(2)两模型结果显示，糖尿病住院费用影响因素中排前两位的是住院日和出院科别，单项费用中药品费、化验费、检查费占据前三位。结论 在对糖尿病住院费用分析时反向传播神经网络优于多元线性回归。 |
| 关键词: 住院费用 糖尿病 人工神经网络 多元线性回归 |
| DOI： |
|
| 基金项目: |
|
| Application of Artificial Neural Network in the Study of Factors Affecting Hospitalization Expenses on Diabetes |
|
| () |
| Abstract: |
|
Objective To analyze the main influencing factors of hospitalization expenses on diabetes，to explore effective analytical methods and to provide a theoretical basis for reasonable control of hospitalization expenses on diabetes. Method Inpatient medical record summaries of patients with type 2 diabetes in 2011 from 13 top hospitals in Beijing were used as data source. We employed Back-propagation (BP) Artificial Neural Network (ANN) model to do analysis, and we compared it with the Multiple Liner Regression (MLR) model. Result （1）The performance of ANN is superior than MLR in identifying main factors affecting hospitalization expenses on diabetes. （2）Medicine expenses are the largest proportion of hospitalization expenses on diabetes. Top 2 major factors affecting hospitalization expenses are hospitalization days and discharge department. Conclusion ANN performs better than MLR in identifying the main factors affecting hospitalization expenses on diabetes. |
| Key words: hospitalization expense,diabetes,Artificial Neural Networks,multiple liner regression |
