基于Joinpoint回归模型的新型冠状病毒肺炎流行趋势分析  

Analysis of epidemic trend of coronavirus disease 2019 based on Joinpoint regression model

作　　者：代吉亚[1] 郭汝宁[1] 刘国恒[1] DAI Ji-ya;GUO Ru-ning;LIU Guo-heng(Guangdong Center for Disease Control and Prevention,Guangzhou,Guangdong 511430,China)

机构地区：[1]广东省疾病预防控制中心,广东广州511430

出　　处：《热带医学杂志》2020年第10期1375-1379,共5页Journal of Tropical Medicine

基　　金：广东省科技攻关课题(2019B111103001)。

摘　　要：目的探索新型冠状病毒肺炎(COVID-19)在武汉市和广东省的流行趋势和防控措施的实施效果。方法采用Joinpoint回归模型对武汉市和广东省新型冠状病毒每日新增病例数进行时间序列分段拟合,计算不同阶段新增病例数日均变化速度,推算出统计学拐点。结果武汉市新型冠状病毒肺炎每日新增病例经Joinpoint回归分析后,有4个分段点,将时间序列分为5个阶段,其中1月19日-2月6日为上升期(APC=24.3%,t=10.2,P<0.01),2月7-29日为下降期(APC=-7.9%,t=-5.2,P<0.01);广东省有5个分段点,将时间序列分为6个阶段,其中前3个阶段有统计学意义,分别是其中1月19-31日为上升期(APC=33.1%,t=7.1,P<0.01),2月1-15日为下降期(APC=-12%,t=-4.0,P<0.01),2月15-19日为快速下降期(APC=-51.2%,t=-2.1,P<0.01)。根据Joinpoint回归模型实际值和拟合值,计算均方差(MSE)分别为0.28和0.25,平均相对误差绝对值(MAPE)分别为23.73%和16.94%。结论 Joinpoint回归分析可用于时间序列分段拟合,计算疾病大流行期间有统计学意义的时间趋势变化特征。通过比对防控措施实施时间和Jionpoint模型统计学拐点的对应关系,发现减少人员流动、提高病毒检测能力、对所有病例特别是轻症病例和无症状感染者进行隔离治疗,是控制COVID-19疫情的有力措施。Objective To explore the epidemic trend of coronavirus disease 2019(COVID-19) and the effect of prevention and control measures in Wuhan and Guangdong province.Method A Joinpoint regression model was used to fit the timeseries of the daily number of new cases of COVID-19 in Wuhan and Guangdong provinces,and the daily average rate of change in the number of new cases at different stages was calculated to estimate the critical point of epidemic trend.Results New daily cases of COVID-19 in Wuhan were analyzed by joinpoint regression,and there were 4 segmentation points.The time series were divided into 5 stages,of which January 19 to February 6 was the rising period(APC=24.3%,t=10.2,P<0.01),and February 7-29 was the decline period(APC=-7.9%,t=-5.2,P<0.01).The time series were divided into 6 stages in Guangdong province,of which the first 3 stages were statistically significant,with January 19-31 the rising period(APC=33.1%,t=7.1,P<0.01),February 1-15 the decline period(APC=-12%,t=-4.0,P<0.01),and February 15-19 the rapid decline period(APC=-51.2%,t=-2.1,P<0.01).According to the actual value and fitted value of the Joinpoint regression model,the mean square error(MSE) was calculated to be 0.28 and 0.25,and the mean absolute percentage error(MAPE) was respectively 23.73% and 16.94%.Conclusions Joinpoint regression analysis can be used for time-series piecewise fitting to calculate statistically significant time trend changes during a pandemic.By comparing the correspondence between the implementation time of prevention and control measures and the statistically critical point of the Joinpoint model,it was found that reducing the flow of people,improving the ability to detect viruses,and isolating and treating all cases,especially mild cases and asymptomatic infections were the most powerful measures to control the COVID-19 epidemic.

关 键 词：新型冠状病毒肺炎 流行趋势 Joinpoint回归分析 

分 类 号：R563.1[医药卫生—呼吸系统]