Spatiotemporal Analysis Of Epidemic Characteristics And Influencing Factors Of Hemorrhagic Fever With Renal Syndrome In Shandong Province

BackgroundHemorrhagic fever with renal syndrome(HFRS)is a natural focus disease caused by hantavirus(HV),mainly infected by rodents.The disease is prevalent worldwide,mainly in Asia and Europe,and poses a serious threat to human health.The data shows that in recent ten years,China has reported about 10,000 HFRS cases every year on average,which are distributed in all provinces,cities and autonomous regions.However,there are significant differences in the incidence rate of HFRS among regions,and the epidemic situation is highly sporadic and relatively concentrated.As one of the key epidemic areas of HFRS in China,Shandong Province ranks among the top five in terms of the number of reported cases of HFRS all the year round,and its annual average incidence rate and case fatality rate both exceed the national average,which has become an significant public health problem in Shandong Province.HFRS has multiple transmission routes,with animal borne transmission as the main transmission channel.Direct or indirect contact with the blood,saliva,and excreta of infected mice can cause infection.Therefore,changes in the ecological environmental conditions,animal populations,human activities,and other related factors in the epidemic focus can affect the occurrence and prevalence of HFRS.Studies have shown that meteorological factors are significantly related to the incidence of HFRS,but most of the existing studies are based on a single city,using different models to explore the effects of meteorological factors on the incidence of HFRS.The results of different studies with spatiotemporal heterogeneity are often difficult to compare.At the level of Shandong Province,there are very limited studies that comprehensively consider the nonlinear lag effect of meteorological factors,the impact of socio-economic factors,and the spatiotemporal random effect of HFRS.With the development of spatiotemporal statistical analysis methods,Bayesian spatiotemporal models have been widely used in the field of infectious diseases.The proposed integrated nested Laplace approximations(INLA)algorithm has also greatly improved the computational speed of posterior estimation of Bayesian spatiotemporal models.This model uses prior information to describe unknown parameters in the model,and fully utilizes information from both the temporal and spatial dimensions of the data by incorporating spatiotemporal independence terms and spatiotemporal interaction terms,correcting the effect relationship between the influencing factors and diseases,and improving the accuracy of parameter estimation.It has unique advantages in spatiotemporal analysis of infectious disease influencing factors.Therefore,based on the incidence data of HFRS in various cities in Shandong Province from 2011 to 2020,as well as the meteorological data and socio-economic data of the same period,this study analyzes the epidemic characteristics of HFRS in Shandong Province.By combining distributed lag nonlinear model with Bayesian spatiotemporal model,this study comprehensively explores the impact of meteorological and socio-economic factors on the incidence of HFRS,and provides reference for the prevention and control of HFRS.Materials and methodsThe data of HFRS cases,meteorological factors,and demographic and socio-economic factors from January 1,2011 to December 31,2020 in various cities of Shandong Province are collected.Firstly,descriptive analysis was conducted on the epidemiological characteristics of HFRS to analyze the temporal trend and seasonality of the disease and identify high-risk areas and populations,then screen variables through Spearman correlation analysis and multicollinearity test.The selected meteorological variables are included in the distributed lag nonlinear model in the form of cross bases.This model is used to explore the settings of maximum lag time for each meteorological factor,and the DLNM model with the relatively optimal fitting effect is selected based on the principle of minimum sum of QAICs.The nonlinear and lag effects of meteorological factors in various cities on the incidence of HFRS are quantitatively evaluated.At the same time,in order to fully reflect the distribution law of disease data in the established model,without considering covariates,the spatial independent model,the temporal independent model,the spatiotemporal independent model,and the spatiotemporal interaction model are respectively fitted.The relatively optimal Bayesian spatiotemporal model without covariates is selected from the above four models,and the prior setting of the parameters in the model is changed for sensitivity analysis.Finally,in the relatively optimal non covariate Bayesian spatiotemporal model selected,the linear terms of socio-economic variables and the cross bases of meteorological factors selected by DLNM model were incorporated to fit the final multivariate Bayesian spatiotemporal model,and the spatiotemporal effects,meteorological factors,and socio-economic factors affecting the incidence of HFRS in Shandong Province were comprehensively analyzed.Results1.There were 11,753 HFRS cases in Shandong Province from 2011 to 2020,with an average annual incidence rate of 1.19/100,000,and the highest case fatality rate in 2018 was 2.97%.HFRS occurs throughout the year,with two peaks each year,namely,the autumn and winter peaks from October to November and the spring and summer peaks from March to June.The incidence is mainly concentrated in the central and eastern regions of Shandong Province,with sporadic occurrence in other regions.The gender ratio of male to female cases during the study period was 2.61:1,with onset occurring at all ages.The age range of 30 to 70 years was the highest,accounting for 82%of the total number of cases during the study period.The occupation with the highest incidence was farmers,accounting for 85%.2.The relationship between various meteorological factors and the incidence of HFRS is non-linear,and the rise of RR of average temperature and average relative humidity has a certain hysteresis phenomenon.Taking the mean value of each variable as a reference,the impact of weekly average temperature and weekly average relative humidity on the incidence of HFRS generally presents an "S" pattern.When the temperature is between-0.8℃ and 13.9℃ and above 24.9℃,the RR decreases,and when the temperature is between 14.1℃ and 20.6℃,the RR increases.When the relative humidity is between 41.6%and 65.9%and above 96.9%,the RR decreases,and when the relative humidity is between 66.1%and 84.2%,the RR increases.The effect curves of cumulative weekly sunshine hours and average weekly wind speed with the onset of HFRS showed an inverse"J" shape and a monotonic upward trend when the cumulative delay was 1 week.3.There are 5 socio-economic variables in the model with statistically significant posterior estimates of relative risk(P≤0.05),among which urbanization rate and per capita park green space area are negatively correlated with the incidence of HFRS,with RR is 0.9996(95%CI:0.9994～0.9997)and 0.9452(95%CI:0.9277～0.9631),respectively;The per capita urban road area,green coverage rate in built-up areas,and the number of hospital beds per 10,000 people are positively correlated with the incidence of HFRS,with RR is 1.0305(95%CI:1.0190～1.0421),1.0204(95%CI:1.0043～10368),and 1.0233(95%CI:1.0117～1.0350),respectively.4.In the non covariate Bayesian spatiotemporal model established in this study,when the spatial and temporal terms are separately included,the DIC values of the model are relatively large,with values of 21,217.16 and 28,310.85,respectively;After adding the spatiotemporal interaction term,a relatively optimal model was obtained with DIC=20,527.85,which significantly improved the fitting effect of the model.Change the prior setting of hyperparameter in the spatiotemporal interaction model for sensitivity analysis,the relative change in DIC value of the model is less than 0.01%,and the model is relatively stable.After incorporating the screened cross basis of meteorological factors and the linear term of socio-economic factors,the DIC value of the multivariate Bayesian spatiotemporal model was 20,171.54,indicating an improvement in the fitting effect of the model.Conclusions1.The incidence rate and mortality of HFRS in Shandong Province showed a fluctuating downward trend,and the monthly incidence had obvious periodicity and seasonality.The central and eastern regions of Shandong Province had relatively high incidence.Men,middle-aged and elderly people,and farmers are the high risk groups.2.There are nonlinear and hysteretic effects between meteorological factors and the risk of HFRS in various cities in Shandong Province.The impact of weekly average temperature and relative humidity on the incidence of HFRS at the provincial level generally present an "S" pattern;The correlation between wind speed and the incidence of HFRS presents an inverted "J" pattern;The correlation between sunshine hours and the incidence of HFRS showed a monotonic increasing trend.3.Among the socioeconomic factors,urbanization rate and per capita park green space area were negatively correlated with the incidence of HFRS,while per capita urban road area,green coverage rate in built-up areas,and the number of hospital beds per 10,000 people were positively correlated with the incidence of HFRS.4.In the Bayesian spatiotemporal model established in this study,when spatial,temporal,and spatiotemporal interaction terms are included simultaneously,the model fitting effect is the best.The sensitivity of the model to a priori distribution of hyperparameters is low,and the model is relatively robust.After incorporating covariates,the DIC value of the model significantly decreased,and the fitting effect significantly improved.