基于时间序列模型的某三甲医院心力衰竭入院人数及治疗趋势预测

Prediction of hospitalization and treatment trend of heart failure in a tertiary hospital based on time series model

目的利用温特斯指数平滑法建立医院心力衰竭入院人数及平均住院日预测模型,探讨其应用价值,为医院管理提供科学依据。方法从某三甲医院电子病历系统收集2007—2017年心力衰竭住院患者人数及其平均住院日,通过模型诊断、参数优化等方法,构建温斯特指数平滑模型,后对该院心力衰竭入院及治疗趋势进行预测,并对预测结果进行评价。结果将该院2007年1月至2016年12月心力衰竭患者数据设定为训练样本,进行建模拟合及参数优化,以2017年1—12月数据作为测试样本,进行预测及验证。结果显示,该模型用于心力衰竭入院人数预测的平均绝对百分误差(MAPE)为7.055%,平稳R^(2)为0.738;心力衰竭平均住院日预测的MAPE为4.323%,平稳R^(2)为0.698。实际住院人数及平均住院日基本位于预测值的95%置信区间内,表明所建立的温特斯指数平滑模型能较好地用于心力衰竭入院人数及平均住院日的预测。结论温特斯指数平滑模型能较好地预测心力衰竭住院患者人数、平均住院日的季节变化趋势,能为医院合理配置医疗资源提供方法参考及科学依据。

Objective To establish a prediction model for the number of hospitalized patients with heart failure and average hopitalization days by using Winters exponential smoothing method,and to explore its application value and provide scientific basis for hospital management.Methods The number of hospitalized patients with heart failure and their average hopitalization days were collected from the electronic medical record system of a tertiary hospital,and the Winston exponential smoothing model was constructed by means of model diagnosis and parameter optimization to predict and analyze the trend of admission and treatment of heart failure in this hospital.The predicted results were evaluated.Results The data of patients with heart failure from January 2007 to December 2016 in this hospital were set as training samples for simulation construction and parameter optimization,and the data from January to December 2017 were used as test samples for prediction and verification.The results showed that the mean absolute percentage error(MAPE)of the prediction model was 7.055%,the mean stationary R 2 was 0.738,and the MAPE of the prediction model was 4.323%,the mean stationary R 2 was 0.698.The actual number of inpatients and the average length of stay were within 95%confidence interval of the predicted value,indicating that the Winters exponential smoothing model could be used to predict the number of inpatients and average length of stay of heart failure.Conclusion The Winters exponential smoothing model can predict the seasonal trend of the number and average length of stay of patients with heart failure,and provide a scientific basis for the rational allocation of medical resources of hospitals.