| Objective: Hemorrhagic fever with renal syndrome(HFRS),as a notifiable communicable disease under national Class B management,currently there is no effective drug treatment,and their prevention and control task are relatively heavy.With the advent of the era of big data,the data about HFRS,such as spatial data,are becoming more and more detailed,and the statistical algorithm is becoming more and more perfect.The Bayesian spatial-temporal modeling with more accurate fitting results can comprehensively analyze the influence of host factors,geographical environmental factors,socio-economic factors and meteorological factors on the incidence of HFRS.In addition,considering the characteristics of various factor data,the continuous and discrete data were included into the model for comparative analysis results,which was expected to provide auxiliary suggestions for the prevention and control of HFRS,as well as developing research ideas and provide reference for the analysis of other infectious diseases.Methods: The number of reported cases of HFRS in each district and county of Huludao City,Liaoning Province from 2012 to 2021 was collected,and the incidence of HFRS in each district and county was described from the aspects of time trend and spatial trend.The annual host data,annual geographic environmental data,annual socio-economic data and annual meteorological data were collected from 2012 to 2021 and the number of reported cases of HFRS in each district and county in Huludao City for correlation analysis,generalize linear model(GLM)analysis and Bayesian spatial-temporal modeling analysis.The analysis results of various models were compared by model evaluation index.The hysteresis of monthly meteorological data should be considered first,and then various monthly meteorological indicators that are most correlated with the number of reported cases of HFRS in each district and county should be included in the model for analysis.In addition,considering the characteristics of the analysis results of different data types,the annual geographic environment data,annual meteorological data and monthly meteorological data were discretized to compare whether there are differences between the data analysis results of different types.Results: 1.In terms of annual time trend,the incidence of diseases in Lianshan District,Nanpiao District,Suizhong County,Jianchang County and Xingcheng City remained at a high level during 2012-2015 and 2016-2018,and showed a small decreasing trend during2015-2016,while the incidence of diseases in the above five areas showed an obvious steep decline during 2019-2021,and maintained a low level of incidence overall.In contrast,the incidence in Longgang District remained at a low level in the past ten years,and the fluctuation range was not obvious.In terms of monthly time trend,the incidence of HFRS was at a high level in the whole season of spring and summer,from March to August,accounting for more than 70% of the total numbers of HFRS in this period of the year,and the peak of HFRS occurred from March to May.In some years,there will be a small climax in late autumn and early winter,which was from October to December every year,and the peak of the incidence is mostly from October to November every year.In terms of spatial trend,Suizhong County and Xingcheng City,located in the southwest of Heludao City,had the highest proportion of morbidity about 25% or even more than 35%,but Suizhong County had the incidence ratio of less than 20% since 2019.In Jianchang County,located in the northwest of Huludao City,the proportion of morbidity was about 10% for most of the time,and the number of reported cases increased to about 15% during 2016-2019.Among the three districts in the northeast of Huludao City,the proportion of morbidity of Longgang District was below 7%,that of Nanpiao District was between 10% and 22%,and that of Lianshan District fluctuated between 15% and 27%.2.The results of spatial autocorrelation analysis showed that,except 2012 and 2015,the occurrence of HFRS in all districts and counties of Huludao City in other years had spatial autocorrelation.In 2013,2014 and 2018-2021,the direction of spatial autocorrelation was negative,while the Global Moran’s Index was positive in 2016 and 2017.It indicated that there was a positive spatial autocorrelation in the space.3.Among host factors,there was a positive correlation between the number of HFRS in Huludao City and the host factors,such as rat density and rat viral carriage rate,with correlation coefficients of 0.615 and 0.461(both P values were less than 0.05).4.Among geographical environmental factors,the posterior mean and 95%confidence interval(CI)of normalized difference vegetation index(NDVI)were 6.68(3.680,9.721),indicating that there was a positive correlation between NDVI and the incidence of HFRS in Huludao City.The posterior mean and 95%CI of altitude were-0.003(-0.005,-0.001),indicating a negative correlation between the two.5.Among socioeconomic factors,the posterior mean and 95%CI of gross domestic product(GDP)were-0.002(-0.004,-0.001).The posterior mean and 95%CI of the population density were0.032(0.002,0.062).The posterior mean and 95%CI of the number of beds in medical institutions were-0.001(-0.001,-0.001).The posterior mean and 95%CI of the area of administrative divisions were-0.002(-0.003,-0.001).The posterior mean and 95%CI of the total power of agricultural machinery were-0.003(-0.019,0.014).6.Among the annual meteorological factors,the posterior mean and 95%CI of mean air temperature were 0.074(-0.127,0.275);The posterior mean and 95%CI of the mean relative humidity were-0.027(-0.039,-0.016).The posterior mean and 95%CI of average wind speed were-0.174(-0.263,-0.085).The posterior mean and 95%CI of the mean air pressure were 0.016(-0.012,0.044).The posterior mean and 95%CI of total precipitation were-0.001(-0.002,-0.001).The posterior mean and 95%CI of mean sunshine duration were-0.015(-0.027,-0.004).7.Among the monthly meteorological factors,after considering the lag of each index,the analysis results showed that the posterior mean and 95%CI of the monthly minimum air temperature were 0.017(-0.007,0.041);The posterior mean and 95%CI of monthly maximum air temperature were-0.029(-0.044,-0.014).The posterior mean and 95%CI of monthly mean air temperature were 0.010(-0.019,0.038).The posterior mean and 95%CI of monthly mean relative humidity were 0.020(0.014,0.027).The posterior mean and95%CI of monthly mean wind speed were-0.172(-0.259,-0.088).The posterior mean and95%CI of monthly mean pressure were-0.002(-0.021,0.016).The posterior mean and95%CI of total monthly precipitation were-0.005(-0.008,-0.003).The posterior mean and95%CI of total monthly sunshine duration were-0.006(-0.008,-0.004).8.In the above analysis results,continuous data had a good fit in the independent and identically distributed(IID)model.Discrete data were well fitted in the stochastic partial differential equations(SPDE)model of Bayesian spatial-temporal modeling.Conclusions: 1.From 2012 to 2021,the incidence of HFRS in all districts and counties in Huludao City showed an overall fluctuation trend.The incidence of HFRS in each district and county was higher in spring and summer and in late autumn and early winter,and the peak occurred from March to May and from October to November.2.Among the annual factors,NDVI,altitude,GDP,population density,number of beds in medical institutions,area of administrative divisions,average annual relative humidity,average annual wind speed,total annual precipitation and average annual sunshine duration had an impact on the incidence of HFRS in each district and county in Huludao City.Among the monthly factors,the monthly maximum temperature,the monthly average temperature,the monthly average relative humidity,the monthly average wind speed,the monthly total precipitation and the monthly total sunshine duration had an impact on the incidence of HFRS in each district and county in Huludao City.3.Compared with GLM,Bayesian spatial-temporal modeling considered the interaction effect of time and space as a random variable for analysis and calculation,and the fitting results were better,suggesting that it had a certain application prospect.The IID model of Bayesian spatial-temporal modeling was suitable for continuous data analysis,and the SPDE model was suitable for discrete data analysis,which can provide methods for developing ideas for the spatial-temporal analysis of infectious diseases in the future,and also provide auxiliary guidance for disease prevention and control departments. |
PDF Full Text Request