基于倾向值匹配研究低温对流行性感冒的影响

Effect of low temperature on influenza based on propensity score analysis

目的分析重庆市流感发病趋势,探讨低温对流感发病率的独立影响,为制定流感的防控措施和策略提供参考。方法基于重庆市2010-2015年流感报告发病数据与同期气象资料,以温度为自变量、日流感发病率为因变量,均衡气压、气流等混杂因素,进行卡钳值为0.1的倾向值匹配,建立均衡混杂因素后的负二项回归模型。结果 2010-2015年间重庆市流感流行强度总体有上升趋势(χ2=38.61,P<0.001);混杂气象因素在经倾向值匹配后的模型中得到平衡,负二项回归模型在χ检验、偏差、调整偏差、AIC、AICC上均优于匹配前的对应模型;且日最低气温越低,流感日发病率越高。结论气象因素对流感发病影响有滞后性;匹配前采用负二项回归控制混杂因素可能会低估温度对流感发病的影响;倾向值匹配后的负二项回归模型预测流感发病率,其结果更可靠;此外,低温可能是导致流感发病率增加的一个重要气象因子。

We analyzed the trend of flu incidence in Chongqing,explored the independent influence of low temperature on the incidence of influenza and provided reference for the development of influenza prevention and control measures and strategies.Based on the epidemiological data of influenza from 2010-2015 in Chongqing and the meteorological data in the same period of time,using the temperature as the independent variable and the morbidity rate of daily flu as the dependent variable,we use the propensity value to match the balance of pressure,air flow and other confounding factors,establishing four negative binomial regression models.Confounding meteorological factors were balanced by the propensity score.The negative binomial regression model was better than the corresponding model before matching in Chi-square test,Deviance,Scale deviance,AIC and AICC.The lower the daily minimum temperature,the higher the incidence of influenza.The influence of meteorological factors on the incidence of influenza is lagging,and the use of negative binomial regression to control confounding factors before matching may underestimate the impact of temperature on the incidence of influenza.In addition,low temperature may be an important meteorological factor that leads to an increased incidence of influenza.