The spatial-temporal association between mrteorological factors and bacillary dysentery

Bacillary dysentery remains a worldwide public health problem, which has been found to have spatial–temporal heterogeneity, however most studies have only focused on the disease from either a time or space perspective, the spatial– temporal association between them has been still unclear. In this study, the Bayesian space–time hierarchy model was used to identify the spatial-temporal patterns of this disease in Shandong province, China. And then GeoDetector was used to quantify the determinant power of meteorological factors and their interactive effect among different regions in Shandong. The results indicated that, temporally, the incidence peaked in summer. Geographically, the hot spots were distributed discretely among three regions, among which the effect of meteorological factors on this disease exist significant discrepancy. The most important two dominant factors of eastern coastal region were wind speed and average temperature, with determinant powers of 28% and 25%, respectively. The first two dominant factors of western inland region were average temperature and precipitation, with determinant powers of 47% and 32%, respectively. The first two dominant factors of middle region were average temperature and wind speed, with determinant powers of 66% and 48%, respectively. These findings suggest that in a hot and humid environment would boost the

transmission of bacillary dysentery, which can be served as a suggestion and basis for the surveillance and will be helpful for this disease control and implementing disease-prevention policies.

Background
Bacillary dysentery is an enteric infectious disease caused by different species of Shigella (S. dysenteriae, S. flexneri, S. boydii, and S. sonnei), a group of Gramnegative, non-spore-forming, and rod-shaped bacteria [1,2]. This disease is mainly transmitted through the fecal-oral route or by contaminated food and water and contact among persons [3,4]. Typically, symptoms include diarrhea, fever, tenesmus, and mucus blood [5].
Up to now, bacillary dysentery has remained a common global public health problem for both developed and developing countries [6,7]. Annually, there are more than 160 million reported cases and 1.1 million deaths worldwide [8]. In China, the disease has caused about 300 deaths annually in the last decades and, in 2015, was perceived as one of the top five infectious diseases [9]. Due to the mechanism of bacillary dysentery is not completely clear, it may cause an epidemic at any time.
Therefore, researching the spatial-temporal heterogeneity of bacillary dysentery incidence and quantify the determinant powers of meteorological factors among different areas would provide suggestions for bacillary dysentery risk control and implementing disease-prevention policies, which were adapted to local conditions. It is widely accepted in numerous studies that meteorological factors have significant effects on the transmission of bacillary dysentery. For example, in Beijing, the capital of China, the highest incidence occurs from June to September [10]. A study in Jinan, a northern city in China, demonstrates that the risk reaches its peak in the summer and fall seasons [11]. In India, the peak of bacillary dysentery mostly occurs in hot summer [12]. In Dhaka, a city in southern Asia, from September to December, a high incidence occurs [13]. From these numerous previous studies, it can be seen that meteorological factors indeed have an impact on the temporal heterogeneity of bacillary dysentery incidence [14][15][16].
Meanwhile, the incidence of bacillary dysentery also presents obvious spatially non-  [10].
To our knowledge, few studies have concentrated on the heterogeneity of bacillary dysentery from a spatiotemporal perspective and meanwhile, quantified its spatiotemporal heterogeneity, effects of meteorological factors and their interactive effect on this disease among different regions. The aims of this study were to 1) explore the county-level spatiotemporal heterogeneity of bacillary dysentery risk, and 2) quantify the determinant powers of meteorological factors and their interactive impact, and 3) detect the hot, cold spots and its mechanisms.

Study area
Shandong, a coastal province in eastern China, in which some areas located in the east and north extend into the sea, while the central and western areas are mountainous and hilly, as a part of the North China Plain (Fig. 1

Data sources
Weekly data on bacillary dysentery cases were gained from the Chinese Centre for Disease Control and Prevention, which were from September 1, 2012, to August 31, 2013. Weekly meteorological data from the same period, including relative humidity, wind speed, and hours of sunlight, were collected from the China Meteorological Data Sharing Service System (Fig. 2).
The study design based on GeoDetector and Bayesian space-time hierarchy model to quantify the associations between meteorological factors and bacillary dysentery from spatial-temporal perspectives among different regions.

GeoDetector
The GeoDetector q statistics was used to quantify the determinant power of impact factors from spatial-temporal perspectives and the stratified heterogeneity of a responding variable (the temporal and spatial variations of bacillary dysentery risk)

Bayesian space-time hierarchy model
The Bayesian space-time hierarchy model (BSTHM) was contributed to reveal the spatiotemporal patterns of this disease. Exactly, a BSTHM with Poisson distribution was used in modeling the number of cases y it and the risk population n i , as follows: where u it describes the spatiotemporal risk of bacillary dysentery in county i ( i = 1, …,103) and week t (t = 1,…,53). The term α is the fixed effect. The index s i indicates the spatial disease risks in county i, affected by some relative stable factors in the study period, such as local geographic environment, economic conditions, and medical resources. The temporal term b 0 t * + v t expresses the overall time trend with v t ∼ N (0, σ v midpoint t m id over the study period. And b 1i captures the departure extent from b 0 for county i, such as, if b 1i ≥ 0, the local variation intensity is higher than the overall variation trend [23]. The term ε 1i ~ N (0, σ ε 2 ) represents the Gaussian noise random variable, according to Gelman [24].
Then, the study area was further classified into hot, cold, and other spots in accordance with the following criteria [25]. A county was defined as a hot spot if the posterior probability p (exp (s i ) >1 | data) ≥ 0.975. Conversely, if that value less than 0.025, the county was defined as a cold spot. The remaining counties were considered as neither hot nor cold spots, in which all processes were implemented in the WinBUGS software [26].

Spatiotemporal Heterogeneity Detection
Between September 1, 2012, and August 31, 2013, a total of 8,014 cases in 53 weeks of bacillary dysentery in 103 counties were reported in Shandong Province.
The highest number of cases occurred in summer (June to August), with a monthly incidence of 1.22 per 10,000 people. The lowest number of cases appeared in winter (December to February), with a monthly incidence of 0.35 per 10,000 people.
Geographically, the relative risks (RRs) differed dramatically, with the q statistic value of 0.51, indicating that there exists apparent spatial heterogeneity. Figure 3 presents the spatial RRs of bacillary dysentery by county level from 2012 to 2013.
The spatial RRs of counties in eastern and northern Shandong mostly located in coast areas were higher, denoting that these counties have relatively higher bacillary dysentery risk. Conversely, counties in western and southern Shandong mainly belong to inland regions, presented relatively lower disease risk.
Additionally, the overall temporal trend presents an increase (Fig. 4), and meanwhile there remains seasonality, in which the highest disease risk occurred in summer (June to August) and the lowest disease risk occurred in winter (December to February), indicating that the risks of bacillary dysentery have obvious temporal heterogeneity, demonstrated by the q statistic value of 0.51.
Among the 103 counties in Shandong, 13 (12.62%) and 12 (11.65%) counties were perceived as hot and cold spots, respectively. Another 78 (75.73%) counties were considered to be neither hot nor cold spots. Figure 5 presents that hotspot areas were mainly distributed in the eastern and northern coast areas.

Risk Factor Analysis
The bacillary dysentery risk is obviously related to seasonal changes ( Fig. 2 and 4), Specifically, in middle mountainous areas, average temperature presents the most dramatically relationship with bacillary dysentery, with a q value of 0.66, it was the dominant factor explaining the temporal variation of the bacillary dysentery incidence ( Table 1).
The other selected potential meteorological risk factors also have a non-negligible impact, such as, wind speed and precipitation had significant association with a higher extent of deviations, with q values of 0.48 and 0.22, respectively ( Table 1).
The results of interactive impacts in GeoDetector denote that combined effect between randomly two meteorological factors also play an important role in the transmission of bacillary dysentery. Taking an example, in central regions, the determinant power of average temperature and relative humidity is 0.82, the determinant power of average temperature and sun hour is 0.82, and the determinant power of average temperature and precipitation is 0.81 (Table 1).
Among these meteorological factors, comparing with their independent influence, all shows "bivariate enhance" effect.
Meanwhile, in western hilly areas, there also was a strong relationship between bacillary dysentery and average temperature, with the value of q being 0.47 (Table   2). Precipitation and wind speed have a similar determinant power with a higher extent of deviations, and q values were 0.32 and 0.32, respectively ( Table 2).
The results of interaction in GeoDetector indicate that interactive impact between randomly two meteorological factors also play an important role in the transmission of bacillary dysentery. For example, in western regions, the determinant power of average temperature and wind speed is 0.64, the determinant power of wind speed and sun hour is 0.62, and the determinant power of sun hour and precipitation is 0.58 (Table 2). Comparing with their independent influence among these meteorological factors, all shows "bivariate enhance" effect.
Additionally, in eastern coastal areas, wind speed presents the most significant impact on the bacillary dysentery, with the value of q was 0.28. And then average temperature and sun hour also show obvious association with bacillary dysentery, with the q value was 0.25 and 0.22, respectively (Table 3).
And the results of interaction in GeoDetector present that interactive effect between randomly two meteorological factors play an important role in the transmission of bacillary dysentery. For example, in eastern regions, the determinant power of sun hour and wind speed is 0.71, the determinant power of wind speed and precipitation is 0.69, and the determinant power of sun hour and average temperature is 0.59 (Table 3). Comparing with their independent influence among these meteorological factors, also shows "bivariate enhance" effect. These results indicated that a hot and moist environment is more likely to promote the transmission of bacillary dysentery.

Discussion
Bacillary dysentery has remained a worldwide public health threat in recent years [6,7]. In this study, the spatiotemporal heterogeneity of the disease risk was The spatial distribution of bacillary dysentery risk was non-homogeneous in Shandong. Counties with the highest risk (hot spots) of this disease were discretely distributed in the eastern, north central coastal areas and southwestern, where the economic level has significant difference, such as, higher in eastern or lower in western areas, which were consent to previous studies [10,18]. Although the exact mechanism cannot be explained completely, several reasons include but are not limited to the followings. On the one hand, in these regions, a moist environment would promote the virus to breed and spread. On the other hand, economies in these areas are almost higher than other parts of Shandong, contributing to a high population density and attracting a large migrant population from other counties; while, economies in western areas are almost lower than other parts of Shandong, poor medical conditions, imperfect public facilities and lack of personal hygiene awareness, which explain the high risk of bacillary dysentery in these areas.
Meanwhile, the risk of bacillary dysentery presents obvious seasonality.
Meteorological factors are perceived as crucial environmental factors, which are widely accepted, influencing the breeding and spreading of viruses that cause this disease [14,[27][28][29]. This study found that average temperature takes on the Changsha, the incidence of bacillary dysentery would rise by 14.8% as 1°C increase in average temperature [30]. A study, conducted in Jinan, also found that, as each 1°C increase in average temperature, the incidence of bacillary dysentery is responded to increase by 11% [15]. The conduction is similar to that in Peru, where it was found that bacillary dysentery risk increased by 8% with a rise of 1°C in temperature [31]. In the United Kingdom, a rise in average temperature was associated with a 5% increase in the bacillary dysentery risk [32]. The potential mechanism may be that higher temperatures enhance exposure to pathogens, promote the breeding of the bacteria, and prolong the survival of bacteria in the moderate environment and contaminated food or water [31]. Another mechanism could be that moderate and warm temperatures in the environment may promote specific behavioral patterns in the population, such as more outdoor activities, which can increase contact among people, thereby facilitating the spread of bacillary dysentery infection.
Precipitation was also found to have strong association with bacillary dysentery, especially in western areas with the q value was 0.32. This result was consistent with some of previous studies, such as, Ma et al. denoted that there was a strong correlation between the incidence of bacillary dysentery and precipitation [33], whereas Li et al. indicated that precipitation have a significant impact on the incidence of bacillary dysentery [34]. The potential impact mechanism may be that precipitation affects water and food, which affects the growth and reproduction of bacteria, and therefore has a certain role in promoting the transmission of this disease.
Furthermore, in this study, wind speed also was found to have significant association with bacillary dysentery, especially to eastern regions. Notably, the effect of wind speed on bacillary dysentery in past research was not always consistent. A previous study conducted by Li [15]. Reasons for these phenomena could be that regional characteristics in different study areas may have different climate conditions, thus affecting the epidemiology of bacillary dysentery.
Additionally, the interactive effect of these meteorological factors was non- This study has some limitations that should be mentioned. Firstly, spatial data at the county level was used, introducing an inevitably ecological fallacy [35].

Availability of data and materials
The datasets generated and/or analyzed during the current study are not publicly available due to policy of Chinese Disease Control and Prevention.

Competing interests
The all authors declare that they have no competing interests.

Funding
The study was supported by the following grants: National Natural Science   Figure 1 Geographic location of the Shandong Province in China and its monthly incidence of bacillary Temporal relative risks of weekly bacillary dysentery from September 1, 2012, to August 31, Figure 5 Distribution of the hot and cold spots of bacillary dysentery in Shandong Province. Note: The