Spatial Distribution Characteristics And Bayesian Spatiotemporal Modeling Of Scarlet Fever Incidence In China

Objective To understand the distribution of scarlet fever epidemic nationwide from 2004 to 2018,analyze the spatial aggregation characteristics of scarlet fever epidemic,establish a Bayesian spatio-temporal model to analyze the spatio-temporal distribution characteristics of scarlet fever morbidity risk,and further explore the effect of social and environmental factors on the risk of scarlet fever morbidity.This study provided a basis of reference and information for control and management of scarlet fever in a science-based manner.Methods The study collected information on the reported incidence of scarlet fever in 31 provinces(excluding Hong Kong,Macao,and Taiwan)from 2004 to 2018,as well as socio-environmental data from each province during the same period.The incidence of scarlet fever was first described in the temporal and spatial dimensions to provide a preliminary understanding of the spatial and temporal distribution of its incidence.To simplify the subsequent model,factor analysis was used to downscale socio-environmental indicators that may influence the onset of scarlet fever.Exploring the spatial aggregation characteristics of scarlet fever nationwide using spatial autocorrelation analysis.Four Bayesian analysis models including temporal,spatial and spatio-temporal interaction effects were constructed separately,and the optimal model was determined by evaluated indicators.The effects of temporal,spatial and spatiotemporal interaction effects on the onset of scarlet fever were quantitatively analyzed based on an optimal model.Further,the effect of socio-environmental factors on the incidence of scarlet fever was explored by incorporating the downscaled socioenvironmental factors into the optimal model,and the relative risk and 95% confidence interval of each factor were estimated.Results 1.From 2004 to 2018,a total of 655,396 cases of scarlet fever were reported in 31 provincial-level administrative regions,with an estimated average yearly incidence rate of 3.24/100,000 and a range of 1.46/100,000-5.64/100,000 in each year,with an increasing trend.The incidence peaks in May to June and November to December when the number of incidence accounts for more than 50% of the total number of cases,and troughs in February and August.In the spatial distribution of incidence,the high incidence areas were mainly in Beijing,Heilongjiang,Jilin and Liaoning provinces(cities).2.The KMO value of the factor analysis adaptation test for social-environmental factors was 0.808,and the Bartlett’s spherical test P<0.001,both of which indicated that the dataset was suitable for factor analysis,and four factors were extracted,namely,the economic development factor(EDF),the traffic and pollution factor(TPF),the medical resource allocation factor(MRAF),and the population factor(PF),which explained a total of about 84.38% of the variance.3.The spatial aggregation results showed that the global Moran’s I values for each year from 2004-2018 ranged from 0.1885-0.4168 with p-values less than 0.05,which indicated a positive spatial correlation between the study areas.The results of the local Moran’s I index analysis show that the high-high aggregation area mainly included Inner Mongolia,Liaoning,Jilin,and Tianjin,the low-low aggregation area mainly included Chongqing,Hunan,Jiangxi,Guizhou,Fujian,Guangdong,Hubei,Guangxi,and Hainan,the low-high aggregation area mainly consisted of Hebei,and the high-low aggregation was only Shanghai.Local hot spot detection showed that there were 8 hot spot areas and 12 cold spot areas for scarlet fever incidence nationwide from 2004-2018,which were roughly consistent with LISA aggregation analysis,with hot spot areas mainly distributed in the north and cold spot areas mainly distributed in the south.In addition,there was a trend to expand the scope of scarlet fever hot spot areas from2004-2014,from Inner Mongolia,Jilin,and Liaoning to gradually expand to the whole northeast and north China except Shanxi.4.Among the four Bayesian models constructed,the DIC values of the model without spatial effects were much larger than those of the model incorporating spatial effects,and the model incorporating the spatio-temporal interaction term had sharply reduced DIC values and the smallest values than the spatio-temporal independent model,which was determined to be the best-fit model.There were spatial,temporal,and spatiotemporal interaction effects on the effect of scarlet fever morbidity risk.The results of spatial effects showed that 15 regions had RR greater than 1.The regions with higher risk of morbidity were mainly located in some northern provinces,among which Beijing(RR=4.672,95%CI: 4.621-4.725)had the highest,followed by Liaoning(RR=3.433,95%CI: 3.404-3.462),Ningxia(RR=3.284,95%CI: 3.213-3.356)and Heilongjiang(RR=3.009,95% CI: 2.979-3.038).The results of the time effect showed a decreasing trend in the mean relative risk across years.The analysis of spatio-temporal interaction effects showed that the regions with higher relative risk from 2004 to 2018 were still concentrated in northern provinces and cities such as Beijing,Heilongjiang and Liaoning,and the number of regions with relative risk greater than 3 increased from2 in 2004 to 7 in 2010,and there was a gradual expansion trend of regions with relative risk greater than 1 after 2015.5.The results of the optimal model-based analysis showed that EDF(RR=1.232,95% CI: 1.147-1.339),TPF(RR=1.133,95% CI: 1.008-1.259)and PF(RR=1.219,95%CI: 1.089-1.382)were positively associated with the onset of scarlet fever.Conclusions 1.The national incidence of scarlet fever from 2004-2018 showed a clear upward trend,with incidence periods concentrated in May-June and NovemberDecember,and high incidence areas mainly in northern provinces and cities,which are key regions for scarlet fever epidemic prevention and control.2.The incidence of scarlet fever showed a spatially aggregated and nonrandom configuration,with hotspot regions mainly in northeast and north China and coldspot regions mainly in south-central China.3.The risk of scarlet fever development was significantly influenced by spatial,temporal,and spatio-temporal interaction effects.4.Economic development factors,traffic pollution factors and demographic factors all had different degrees of influence on the incidence of scarlet fever.