基于sARIMA模型的某三甲医院神经外科手术工作量预测分析

Prediction and Analysis of Neurosurgery Workload in a Tertiary Hospital Based on sARIMA Model

目的应用sARIMA模型分析和预测某三甲医院神经外科手术工作量,为资源配置和科学管理提供参考。方法采用Python软件建立手术量的ARIMA乘积季节模型,收集2017年1月1日-2019年12月31日、2021年1月1日-2021年12月31日各月神经外科手术量数据,共48个月。以2017年1月1日-2019年6月30日数据对模型进行检验和拟合,并对2019年下半年和2021年的手术量进行预测,与实际值进行比较并观测预测效果。结果s ARIMA(2,1,2)x(2,1,0)12是某院某科室手术工作量最优拟合预测模型,2019年下半年实际值与预测值的拟合情况较好,预测的手术量与实际序列趋势基本相同,2019年下半年平均相对误差为1.30%;疫情后的预测误差变大,2021年平均相对误差为7.89%。2019年8月和10月误差较大(超过10%),2021年1月、7月、10月(超过30%)、11月(近20%)误差相对较大。结论时间序列分析中的s ARIMA模型对于手术工作量有一定预测价值,但受疫情影响数据无法平稳化也会降低预测精度,使得2021年及之后数据预测有一定局限性。然而,仍可对有参考性的信息进行利用。节假日(如春节、国庆节)会推迟手术使得手术量下降,酷暑7月-8月和入冬(11月)手术量会激增,疫情限制、复工复产等因素会影响手术量。应提前做好资源配置和充分准备,及时调整人员与物资,应对手术工作量的变化。

Objectives This study aims to use the sARIMA model to analyze and predict the workload of neurosurgery in a tertiary hospital,so as to provide a reference for resource allocation and scientific management.Methods Python software was used to establish an ARIMA product seasonal model of surgical volume.Collect data on the volume of neurosurgery operations from January 1,2017 to December 31,2019,and from January 1,2021 to December 31,2021,for a total of 48 months.The data from January 1,2017 to June 30,2019 were used to test and fit the model and predict the operation volume in the second half of 2019 and 2021.The result was compared with actual values and was used to observe predicted performance.Results sARIMA(2,1,2)x(2,1,0)12 was the optimal fitting prediction model for the workload of a department in a certain hospital.In the second half of 2019,the actual value and the predicted value fitted well.The predicted operation volume was basically the same as the actual serial trend,with an average relative error of 1.30%in the second half of 2019.After the epidemic,the prediction error became larger,and the average relative error in 2021 was 7.89%.In August and October 2019,the error was large(more than 10%).The errors in January,July,October(over 30%),and November(nearly 20%)in 2021 were relatively large.Conclusions The sARIMA model in time series analysis has certain predictive value for the surgical workload.However,the inability to stabilize the data affected by the epidemic also reduces the prediction accuracy,which makes the data prediction for 2021 and beyond have certain limitations.But informative information can still be utilized.Holidays(such as Spring Festival and National Day)will postpone the operation,which will reduce the operation volume,and the operation volume will surge in the hot summer(July-August)and winter(November).Factors such as epidemic restrictions and resumption of work and production will affect the number of surgeries.Resource allocation and full preparation should be done in advance,and personnel and materials should be adjusted in time to cope with changes in surgical workload.