基于百度指数的京津冀地区COVID-19高峰预测与空间分布特征

Peak prediction and spatial distribution characteristics of COVID-19 in Beijing-Tianjin-Hebei region based on baidu index

目的分析2022年11月11日―12月22日京津冀地区各市COVID-19的进展周期和空间聚集情况。方法基于各市每日“发烧”关键词百度指数搜索数据,使用logistic回归分析模型模拟并预测此轮COVID-19发展进程,对感染进展周期划分为渐增期、快增期及缓增期,预测感染高峰日,同时对“发烧”百度指数搜索率进行全局和局部空间自相关分析。结果从logistic回归分析模型模拟及预测结果来看,模型拟合效果较好,各市COVID-19流行进展速度及流行阶段各不相同,石家庄市、保定市及邢台市最早进入感染快增期阶段;空间自相关分析显示仅有5 d京津冀地区存在全局空间正相关性(Moran’s I:0.314~0.491,均P<0.05),其他时间均呈随机分布。结论京津冀地区此轮COVID-19流行均呈暴发趋势,各市疫情进展阶段及感染高峰有较大差异,且大部分时间不存在显著的空间自相关,为卫生医疗配置提供参考依据。

Objective To analyze the progression cycle and spatial clustering of the coronavirus disease 2019(COVID-19)in Beijing-Tianjin-Hebei region from November 11 to December 22,2022.Methods Based on the daily fever keyword Baidu index search data in each city,the logistic model was used to simulate and predict the development process of this round of COVID-19 infection.The infection progression cycle was divided into an increasing period,a rapid increase period and a slow increase period,and the peak day of infection growth was predicted.At the same time,the global and local spatial autocorrelation analysis of the fever Baidu index search rate was conducted.Results The logistic model simulation and prediction yielded a robust model fitting effect.The epidemic progression speed and stage of COVID-19 infection varied across cities.Shijiazhuang City,Baoding City and Xingtai City were the first to enter the rapid increase stage of infection.Spatial autocorrelation analysis showed that there was a significant global spatial positive correlation in Beijing-Tianjin-Hebei region only for 5 days(Moran′s I:0.314~0.491,P value was<0.05),whereas other times exhibited random distribution.Conclusions This round of COVID-19 infection epidemic in Beijing-Tianjin-Hebei region shows an outbreak trend.The epidemic progress stage and infection peak in each city are significantly different,and most of the time there is no significant spatial autocorrelation,which provides a reference for health care configuration.