目的 运用自回归积分滑动平均模型(Autoregressive Intergrated Moving Average,ARIMA)建立月平均住院费用和住院日的医学经济学模型,为医院精细化管理提供依据.方法 利用R4.0.2软件对2017年1月—2021年12月四川大学华西医院宜宾医院(宜宾市第二人民医院)的平均住院费用和住院日数据建立时间序列ARIMA预测模型.结果 住院费用最优模型为ARIMA(0,1,1),赤池信息准则(Akaike information criterion,AIC)=924.35,贝叶斯信息准则(Bayesian Information Criterion,BIC)=928.51,残差Ljung-Box Q=12.51(P=0.768),可认为残差序列为白噪声.平均住院日的最优模型为ARIMA(5,1,1),AIC=87.49,BIC=104.11,残差Ljung-Box Q=10.05(P=0.612),可认为残差序列为白噪声.2022年1-12月实际值与预测值基本吻合,月人均住院费用和人均住院日的平均相对误差为0.55%、0.29%.结论 建立基于时间序列ARIMA模型能够为合理配置卫生资源提供强有力的数据支撑.
ARIMA Model Based on R Language Time Series Used to Predict the Av-erage Monthly Hospitalization Expenses and Hospitalization Days of a Ter-tiary General Hospital
Objective To establish the medical economics model of monthly average hospitalization expenses and hos-pitalization days using Autoregressive Intergrated Moving Average(ARIMA)model,which provided the basis for the refined management of hospitals.Methods R4.0.2 software was used to establish a time series ARIMA prediction model for the average hospitalization expenses and hospitalization days data of Yibin Hospital of West China Hospital of Sichuan University(Yibin Second People's Hospital)from January 2017 to December 2021.Results The optimal model of hospitalization expenses was ARIMA(0,1,1),akaike information criterion(AIC)=924.35,bayesian informa-tion criterion(BIC)=928.51,residual Ljung-Box Q=12.51(P=0.768),which can be considered as white noise.The op-timal model of average length of stay is ARIMA(5,1,1),AIC=87.49,BIC=104.11,residual Ljung-Box Q=10.05(P=0.612),which can be considered as white noise.The actual value from January to December 2022 was basically consis-tent with the predicted value,and the average relative error of monthly average hospitalization expenses and hospital-ization days is 0.55%and 0.29%.Conclusion The establishment of ARIMA model based on time series can provide strong data support for the rational allocation of health resources.
Autoregressive intergrated moving average modelAverage hospitalization expensesAverage hospitaliza-tion daysForecast
万方数据
维普
