Impact Of Meteorological Factors On Hand Foot And Mouth Disease And Forecast And Early Warning

BackgroundClimate change has been considered as one of the greatest challenges threatening population health in the 21st century.Controlling the health threat caused by climate change has become a common goal in the world.In recent years,the incidence rate of infectious diseases closely related to meteorological factors is still high.The health impact of meteorological factors on infectious diseases has received increasing attention due to its increased burden on our society and economy.Hand,foot,and mouth disease(HFMD)is one of the notifiable Category-C infectious disease,which is caused by a variety of enteroviruses.It predominantly affects infants and children aged 0-4 years old.It is highly contagious and transmitted through direct contact,respiratory droplets or fecal-oral route.The national annual incidence of HFMD ranked the first among all notifiable category-C infectious diseases.Given its high prevalence and morbidity,HFMD has become one of the public health problems that our country focuses on.Previous studies have shown that temperature,relative humidity and other meteorological factors can affect the occurrence and prevalence of HFMD by affecting pathogens,hosts and human behavior.However,the research results in different regions are controversial,even contradictory.In addition,most research only explore the single meteorological factor effects,and there is little research on the complex interaction between temperature and humidity.At the same time,most studies reported the effects of meteorological factors on HFMD via relative risk(RR),and lack of the attributable numbers or attributable fractions of HFMD due to meteorological factors.The time-dependent reproduction number(Rt)represents the expected number of secondary cases arising from a primary case infected at time t.Different from the case number,it can reflect the change of disease transmission force in real time.To better understand the epidemic trend of HFMD,various predictive models have been used so far.Timely and effective prediction can minimize the epidemic risk and adverse effects of HFMD.However,due to the collinearity of independent variables and fitting degree,the practicability of model is poor and operation is complex.It lacks a simple and effective high-precision prediction model.At the same time,in China,the existing outbreak detection system is based on routinely reported cases data.When the early warning signal is issued,the infectious disease has occurred,the early warning preparation time is relatively short,and the epidemic situation cannot be early warning before it occurred.In this study,based on the monitoring data of HFMD in China,the spatiotemporal distribution characteristics of HFMD were analyzed by global autocorrelation(Moran I)and spatiotemporal scanning.Geodetector was used to explore which meteorological factors affecting HFMD distribution on spatial scale.We systematically estimate the effect of meteorological factors on HFMD,calculate attributable risk and explore the heterogeneous sources.Meanwhile,we also inference of Rt and establish forecast and early warning model in high incidence area.Our study will help to understand the impact of meteorological factors on population health,formulate more reasonable prevention and control measures to effectively curb the epidemic of HFMD.Objectives1.To understand the distributional characteristics of HFMD in China from 2014 to 2016.2.To study the spatiotemporal distribution characteristics of HFMD and to explore which meteorological factors affecting the spatial distribution variation of HFMD.3.To evaluate the regional differences,attributable risks and heterogeneity sources,and to explore the influence of meteorological factors on its transmission capacity in high incidence area.4.To establish a weather-based forecasting and early warning model for HFMD in the high incidence area.Methods1.Data collection:Daily data of reported HFMD cases from 2014 to 2016 were obtained from the Chinese Center for Disease Control and Prevention through the China National Notifiable Disease Surveillance System.Meteorological data during the same period were downloaded from the China Meteorological Data Sharing Service System,and the humidex was calculated by the average temperature and relative humidity.City-specific social characteristics were collected from China city statistical yearbook.As HFMD mainly affects infants and children under 5 years old,we focused on the incidence of HFMD among children aged 0-4 years in this study.To reduce the uncertainty,324 cities with number of HFMD cases greater than P5 were selected from 334 prefecture level cities in China.2.Statistical analysis:(1)Descriptive analysis was conducted to describe the distribution characteries of disease and meteorological factors on the dimension of time,space and population.(2)The high incidence spatial cluster were determined through spatial autocorrelation and spatiotemporal scanning analysis.Geodetector model was used to explore which meteorological factors affecting the spatial distribution variation of HFMD.(3)A two-stage model was used to explore the relationship between average temperature and HFMD at cities,regions and the whole country level.In the first stage,distribution lag nonlinear model was conducted to obtain the estimated value of average temperature,relative humidity,humidex on HFMD at city level.In the second stage,multivariate meta-analysis was used to combine the cumulative exposure effects of different cities to evaluate the relationship between temperature,relative humidity and humidex on HFMD at regional and national levels.Quantitatively analyze the attributable risk of HFMD due to high temperature,use the meta regression model to explore the heterogeneity sources.Inference of Rt to assess the impact of average temperature on the transmission power of HFMD in the highest-incidence spatial cluster.(4)The city of highest-incidence spatial cluster was select as the research site to develop the forecast and early warning model.The data from 2014 to 2015 were used to build random forest model and generalized additive model(GAM)to predict the incidence trend of HFMD in 2016.The RMSE,MAPE and R2 were used to evaluate the prediction effect of the model.Warning thresholds were selected by ROC curve method.The sensitivity,specificity and AUC index were used to evaluate the early warning model effect.Results1.A total of 7 213 021 HFMD cases were reported during 2014-2016.The number of HFMD showed a distinct seasonal pattern with two peaks.The epidemic peak in summer was obvious,accompanied by a small peak in winter.The incidence of HFMD in male was significantly higher than that in female(60.14%vs 30.86%).The number of children under 5 years old was the largest,accounting for 90.08%of the total cases.In the occupational distribution,the scattered children account for the highest proportion(73.96%)2.The results of spatial autocorrelation and spatiotemporal scanning showed that the incidence of HFMD in China was heterogeneous and spatially aggregated.The most likely aggregation area was in Southern China,and Shenzhen city in Guangdong province was the strongest.The results of Geodetector showed that the annual average temperature was the greatest impact factors on the spatial distribution of HFMD.The interaction between annual mean temperature and relative humidity is the strongest,which is a nonlinear enhancement.3.Effect of average temperature on HFMD at national level indicated an approximately "S" type.High temperature can increase the risk of HFMD with the highest cumulative risk at 30.5℃(RR=1.20,95%CI:1.15-1.25).The effect of relative humidity and humidex on HFMD was similar to "S" type.Expect for northwest,Tibet and Jianghuai areas,high temperature can increase the risk of HFMD in other areas.The number of HFMD cases attributed to high temperature in China was 778 282,accounting for 10.80%(95%CI:10.00%～11.16%)of the total cases.The temporal fluctuation pattern was similar between the HFMD cases and Rt,but the fluctuation of Rt was earlier than the HFMD cases.The cumulative effect showed that high temperature could significantly increase the risk of Rt.When the temperature exceeded 25℃,the risk of Rt increased by 13.85%(95%CI:12.70%～15.11%)for each degree of temperature rise.The sources of heterogeneity identified by meta-regression model included population growth rate,number of students,longitude,latitude,annual average temperature,annual average humidity and region.4.The fitting effect of random forest model is good,and the R2 of Shenzhen is 0.89,which is better than the GAM(R2=0.82).The results of early warning model show that the thresholds we identified were reasonable and can correctly identify the possible epidemic.Under different threshold standards,the results of early warning model are stable,the sensitivity and the specificity are higher than 0.90,and the AUC is above 0.95.Conclusions1.The proportion of children under 5 years old is the highest.There are two epidemic peaks of HFMD in China.The epidemic characteristics of HFMD in the north and south are different.2.The distribution of HFMD cases is heterogeneous and spatial aggregation.Temperature and relative humidity are important factors that affect the spatial variation of HFMD incidence rate.3.Effect of average temperature,relative humidity,humidex on HFMD indicate an approximately "S" type,with high temperature,high humidity,high humidex could increase the risk of HFMD.There are obvious differences among different regions and cities.High temperature could increase the transmission of HFMD.4.The early warning and forecasting model based on meteorological factors and historical incidence using random forest model has good performance and high practicability.