基于Kalman融合模型的全国卫生总费用预测研究  

Forecasting of total national health expenditure based on the Kalman fusion model

作　　者：闫果 张倩[2] YAN Guo;ZHANG Qian(Hengshui City People’s Hospital,Hebei 053000,China)

机构地区：[1]衡水市人民医院,河北053000 [2]衡水市疾病预防控制中心

出　　处：《现代预防医学》2022年第15期2768-2776,共9页Modern Preventive Medicine

基　　金：国家自然科学基金(82003560)。

摘　　要：目的通过对我国卫生总费用及构成的描述与分析,采用基于Kalman融合模型对卫生总费用及构成进行分析研究,并对2021—2023年卫生总费用及构成进行预测,为卫生政策的调整及执行提供数据支撑。方法收集1990—2020年全国卫生总费用及构成数据,建立ARIMA、GM(1,1)、ELM极端学习机模型,并在Kalman框架下融合以上两种或三种模型对卫生总费用及构成进行预测,分别比较拟合和预测误差,采用平均相对误差评价模型预测能力,并对2021—2023年卫生总费用及构成进行预测。结果基于Kalman融合模型在卫生总费用、政府支出费用、社会支出费用及个人支出费用的预测中,三个预测结果优于单模型预测结果,一种预测结果与最佳单模型预测结果的平均相对误差相差0.003,对2021—2023年卫生总费用的预测结果为:84002.31、94799.95、107256.90亿元。结论基于Kalman融合模型的预测精度基本优于单一预测模型,对多种数据具有鲁棒性。卫生总费用保持了较高增速,但在国内生产总值中的占比稳步增长,其中社会支出增长较快,政府和个人支出保持同步稳定增长。Objective Through the description and analysis of the total health expenditure and composition in my country,the Kalman fusion model was used to analyze and study the total health expenditure and composition,the total health expenditure and compositions from 2021 to 2023 were forecasted to provide information for the adjustment and implementation of health policies.Methods The national total health expenditure and composition data from 1990 to 2020 was collected.ARIMA,GM(1,1),and ELM models were established,and the above two or three models were integrated under the Kalman framework to analyze the total health expenditure and composition prediction.The fitting and prediction errors were compared respectively,and the average relative error was used to evaluate the prediction ability of the model,and the total health expenditure and compositions from 2021 to 2023 were predicted.Results In the prediction of total health expenditure,government expenditure,social expenditure,and personal expenditure based on the Kalman fusion model,three prediction results were better than the single model prediction results,and the average relative error was 0.003 less than the best single model prediction result.The predicted results of the total health expenditure from 2021 to 2023 were:84002.31,94799.95,and 107256.90 billion yuan.Conclusion The prediction accuracy based on the Kalman fusion model is better than that of the single prediction model and is robust to a variety of data.The total health expenditure has maintained a relatively high growth rate,but its proportion to the GDP is steady.Among them,social expenditures have grown rapidly,and government and personal expenditures have maintained a synchronous and stable growth.

关 键 词：GM(1 1)模型 ARIMA模型 ELM极端学习机模型 Kalman框架 融合模型 

分 类 号：R1[医药卫生—公共卫生与预防医学] R197.1