Spatiotemporal dynamics of hemorrhagic fever with renal syndrome in Jiangxi province, China

Historically, Jiangxi province has had the largest HFRS burden in China. However, thus far, the comprehensive understanding of the spatiotemporal distributions of HFRS is limited in Jiangxi. In this study, seasonal decomposition analysis, spatial autocorrelation analysis, and space–time scan statistic analyses were performed to detect the spatiotemporal dynamics distribution of HFRS cases from 2005 to 2018 in Jiangxi at the county scale. The epidemic of HFRS showed the characteristic of bi-peak seasonality, the primary peak in winter (November to January) and the second peak in early summer (May to June), and the amplitude and the magnitude of HFRS outbreaks have been increasing. The results of global and local spatial autocorrelation analysis showed that the HFRS epidemic exhibited the characteristic of highly spatially heterogeneous, and Anyi, Fengxin, Yifeng, Shanggao, Jing’an and Gao’an county were hot spots areas. A most likely cluster, and two secondary likely clusters were detected in 14-years duration. The higher risk areas of the HFRS outbreak were mainly located in Jiangxi northern hilly state, spreading to Wuyi mountain hilly state as time advanced. This study provided valuable information for local public health authorities to design and implement effective measures for the control and prevention of HFRS.

Scientific RepoRtS | (2020) 10:14291 | https://doi.org/10.1038/s41598-020-70761-0 www.nature.com/scientificreports/ and has been also widely conformed to characterize spatial epidemiology of diseases [14][15][16][17][18][19] . However, the dynamics spatiotemporal distributions of HFRS in Jiangxi have not yet been explored systematically. So this study aims to explore the dynamics of spatiotemporal distributions based on the case surveillance data from 2005 to 2018 at the county scale in Jiangxi, and providing valuable scientific support for HFRS monitoring and control.

Materials and methods
Study areas. Jiangxi (24°29′14″-30°04′44″ N, 113°34′36″-118°28′58″ E) lies in southeastern China, with an approximate area of 16.5 thousand km 2 , and population of 44.56 million in 2010, including 11 cities and 100 counties (Fig. 1). Jiangxi belongs to a humid subtropical climate, with annual rainfall, annual average temperature, and annual average sunshine ranging from 1,341 to 1,943 mm, 16.2 to 19.7 °C, 1,473 to 2,077 h, respectively 20 . Jiangxi is surrounded by mountains on three sides and facing the Yangtze River on the other. The southern half of the province is hills with ranges and valleys interspersed; while the middle and northern half is flatter and lower in altitude. Stretching from south to north, the whole land is generally sloping towards Poyang lake, which has formed a huge basin opening to the north 20 . In historical literatures, Jiangxi was often divided into three zoogeographic regions and five zoogeographic states: Jiangxi northern plain region (including the plain state bordering on rivers and lakes and Jiangxi northern hilly state), Jiangxi central mountainous hilly region

Spatiotemporal cluster analysis.
A seasonal-trend decomposition of time series analysis was performed to explore the various characteristics of periodicity and seasonality in R software (Version 3.1 AT&T BellLaboratories, Auckland, New Zealand). A global spatial autocorrelation analysis and a Local Indicators of Spatial Association (LISA) analysis were conducted in ArcGIS software (Version 10.4, ESRI Inc., Redlands, CA, USA) to visualize the global and local spatial clustering of HFRS cases in Jiangxi from 2005 to 2018. Global Moran's I Index (ranged from − 1 to 1) was used to analyze global spatial autocorrelation. Moran's Index = 0 implied a random spatial distribution. Moran's I Index < 0 implied a dispersing spatial distribution, and Moran's I Index > 0 implied a clustering spatial distribution. Local Moran's I was calculated to explore significant hot spots (High-High), cold spots (Low-Low), and outliers (High-Low and Low-High) 21 . A Kulldorff 's spatiotemporal scan statistical analysis was used to identify the spatiotemporal clusters of HFRS cases in Jiangxi in SaTScan Software (Version 9.4, Martin Kulldorff, National Cancer Institute, Bethesda, MD, USA) 22 . The discrete Poisson probability model by a circular window with a radius was used for scanning. The maximum of the spatial and temporal size were all defined as 25%. A P < 0.05 was considered to be significant.
The maps were made in ArcGIS software (Version 10.4, ESRI Inc., Redlands, CA, USA). ethics approval. Ethical approval for the research was granted by the Chinese Center for Disease Control and Prevention Ethics Committee (No. 2012CB955504). This study didn't collect patients' samples, and the data obtained from the China Information System for Disease Control and Prevention (CISDCP) were anonymized so that subjects could not be identified. Therefore, the ethics committee agreed that no informed consent was needed from patients. All methods in our study were used in accordance with the relevant guidelines and regulations.
Seasonal decomposition analyses. The seasonal trend was explored using a season-trend decomposition in R software (Fig. 3). There was a rhythmic vibration in the raw data from 2005 to 2018. Seasonality and trend components were isolated from the raw data and also eliminated part of the random noise or reminder component. The epidemic of HFRS showed a bi-peak seasonality characteristic each year, with the primary peak occurring in winter (November to January) and the second peak happening in early summer (May to June). In addition, the epidemic of HFRS collectively showed an upward trend and the variation characteristic of periodicity with the amplitude and the magnitude of the periodical variation increasing.

Spatial autocorrelation analysis.
The global spatial autocorrelation analysis was performed based on county-level HFRS cases in Jiangxi ( Table 2). The annual global Moran's I indexes, which ranged from 0.14 to 0.38, all passed the significance level test (p < 0.05), indicating that the spatial distribution of the HFRS epidemic was not random from 2005 to 2018, with t-shaped spatial autocorrelation characteristics at county-level scale. Moran's I indexes showed an uptrend from 2006 to 2010, and then stayed at a relatively stable level after 2011, suggesting that the spatial aggregation increased first and then stabilized.
LISA was performed to detect local spatial clusters of the HFRS epidemic (Fig. 4). Three counties, Yifeng, Shanggao and Gao'an, were hot spots (High-High cluster area) in the whole study period and An'yi was a Spatiotemporal clusters analysis. The incidence of HFRS was aggregated through space and time using A most likely cluster, and two secondary likely clusters were detected in a 14-year duration ( Table 3). The most likely cluster included Jiujiang city (Wuning county), Nanchang city (An'yi and Wanli county), and Yichun city (Jing'an, Yifeng, Shanggao, Gao'an and Fengxin county). Wanli was in the plain states bordering on rivers and lakes, and the other counties attributed to Jiangxi northern hilly state. The expected case number was 85.05,

Discussion
In recent years, though an obvious decline in HFRS cases followed intensified control campaigns in China, HFRS would persist in China under the current control and prevention measures 23 . The results of this study also revealed that the amplitude and the magnitude of HFRS outbreaks in Jiangxi increased from 2005 to 2018. The susceptible population remained mainly in male farmers in the age groups of 16 to 60 years old, but the proportion of female HFRS cases was increasing. Consistent with several studies 9,11,24 , the results of this study showed that the incidence of HFRS in Jiangxi had a major epidemic season (winter, November-January) and a minor seasonal peak (early summer, May-June) each year, which is corresponded to the epidemic peak of HTNV-type HFRS and SEOV-type HFRS, respectively 15 . Hereby, this study indicated that Jiangxi was a mixed epidemic area dominated by HTNV. The result differed from that of a previous virology study in 2002, which indicated that Jiangxi was a mixed epidemic area dominated by SEOV-type based on the amplification and sequencing results of hantavirus-positive mice 25  www.nature.com/scientificreports/ explanation was that environmental change, with the rapid urbanization process, could alter the composition of rodent species and increase the number of striped field mice in urban area 26 , causing that Jiangxi developed into a HTNV-type dominated mixed epidemic area. In addition, SEOV infections were more likely to be neglected because of the mild symptoms causing low rates of visiting doctors 27 . However, SEOV infection should not be ignored even though HTNV infection was dominated at present. Because unlike other hantaviruses, SEOV could easily spread throughout the entire colony when they were introduced 28,29 . Moreover, the host of SEOV, R. norvegicus, whose lifestyle differs greatly from that of A. agrarius, feeding on various food items from humans instead of field crops, are widely distributed in Jiangxi 30,31 , which is more likely to transmit the virus to humans compared with striped field mice. Unluckily, we are lack of the virus surveillance data until now. Therefore, more research efforts should be undertaken in animal hosts surveillance, causative agents detection, and latent infection investigation in the future in Jiangxi to implement targeted strategies of HFRS prevention and control. The global model revealed that the distribution of HFRS cases in Jiangxi exhibited the characteristics of spatial aggregation. Subsequently, the LISA model and spatiotemporal scan analysis showed that the higher risk areas of the HFRS outbreak were mainly located in Jiangxi northern hilly state, spreading to Wuyi mountain hilly state as time advanced, but no HFRS cases were reported in Jiangxi southern mountainous state. This phenomenon might be affected by natural environmental factors, such as altitude. Some studies have revealed that HFRS occurrence was restricted to areas of altitude of below 200 m 15,32,33 . The average altitude ranges from 300 to 500 m in Jiangxi southern mountainous state, while it is flatter and lower in the middle and northern half. Moreover, though it was difficult to collect available data of the rodent host for our study, a past study in 1985 in Jiangxi indicated that A. agrarius was dominant rodent species in Jiangxi northern plain region, with no distribution in Jiangxi southern mountainous region 25 . This suggested that the distribution of HFRS cases was consistent with that of the rodent host. However, the higher incidence areas of HFRS were in Jiangxi northern hilly state instead of the low-lying plain state of bordering rivers and lakes, which was also different from the results of neighboring provinces 34,35 . This phenomenon might be explained by several factors. Firstly, the low-lying plain state was not only the Poyang Lake eco-economic zone, but also the political, cultural, and economic center of Jiangxi province with an urbanization process at an unprecedented pace over the past 15 years, which has reduced the risk of human-rodent contact because of the better environment [36][37][38][39] . Secondly, in 1998, the Yangtze river flooding occurred in the whole region, and then the project of "returning the farmland to the lake" was implemented in 2003, which caused the water level of Poyang lake to rise. The expansion of the wetland area might have induced that the rodent host shifted to Jiangxi northern hilly state with higher altitude 40,41 . Finally, to protect the ecological environment of Poyang lake, the implementation of "Mountain-River-Lake Program", with the ecological concept of "mountain as source, river as connection flow, and lake as storage" 20 years ago has accelerated the local industrial transformation, and more and more local farmers worked in enterprises or engaged in technical production 42 . All the above factors in the region provided human beings with fewer opportunities for exposure to rodent hosts. Some limitations should be considered when interpreting our findings. Firstly, the bias could exist in this study because the HFRS cases did not differentiate HTNV from SEOV infections, and came from the passive surveillance system called CISDCP. Secondly, social factors and urbanization level might have an important impact on the epidemic process of HFRS but were not included in the analysis due to data unavailable. Finally, specific species of rodent populations were not characterized in our study area.
In conclusion, this study has analyzed the spatiotemporal characteristics of HFRS comprehensively in Jiangxi from 2005 to 2018. We found that HFRS cases were aggregated in Jiangxi northern hilly state, spreading to Wuyi mountain hilly state as time advanced. Therefore, more interventions should be implemented for the prevention and control of HFRS in the future, including animal hosts surveillance, causative agents detection, medical