| Background:Japanese encephalitis (JE), a mosquito-borne disease caused by the Japaneseencephalitis virus (JEV) from the Flaviviridaefamily, is clinically characterized byfever, nausea and vomiting, headache, lowered level of consciousness, seizures,movement disorders, and acute flaccid paralysis. JEV is transmitted in an enzooticcycle among mosquitoes and vertebrate hosts, primarily domestic pigs and ardeidbirds. The primary vector species Culex tritaeniorhynchus is abundant in rural areas,where it breeds in low lying flooded areas containing grasses or flooded rice paddies.Humans are a dead-end host and get infected when bitten by infected mosquitoes.WHO studies showed that24countries and regions in Asia and the Western Pacificregion had JE prevalence, about67,900cases occurred each year, and50%occurredin China.After the1970s, JE incidence decreased remarkablely in Chinafrom20.92/100000in1971to0.23/100000in2008due to the nationwide vaccinationprogram. Although China has made remarkable achievements in JE control, moreafforts are needed to meet the WHO’s target of the JE control (JE should be controlledwith the incidence of children aged less than15years lower than0.5/100000). Thedistribution of JE is very heterogenous in the national level in China with a lowerincidence in coastal areas and higher incidence inYunnan, Guizhou, Sichuan, andGuangxi. At the same time, JEV endemic regions have expanded. Tibet, once thoughtto be free of JE because of its high elevation, has isolated genotype I virus from Cx.Tritaeniorhynchus and found its neutralizing antibodies in human serum andimmunoglobulin M antibodies in swine serum. It also should be noted that JEoutbreaks in recent years have presented new features. A JE outbreak in July-August2006, Yuncheng, Shanxi Provincepresented with high mortality (up to28.8%),anincrease in the proportion of adult JE cases (86%cases were adults above30years ofage),and simultaneously isolation of genotype I and genotype III, which were not appeared in previous outbreaks. Based on those evidences previously noted, it can beconcluded that JE still remains a serious public health problem in China.In recent years, the rapid development of spatial technology has been widelyapplied to the study of schistosomiasis, malaria, hemorrhagic fever with renalsyndrome, tick-borne diseases and other natural foci of disease. However, studies inChina were mainly concentrated in analysis the epidemiological characteristics of JEin local regions for a long monitoring time, or in the whole country for a short-termmonitoring time, which lack an overall understanding the longterm epidemiccharacteristics. Also, spatial analysis was only in Guangxi Province. So, there is a biggap in the study concering the spatiotemporal pattern and hotspots of JE. Predictionanalysis is in the initial stage, which was researche at large scale and only provided anoverall trend. Therefor, this method, lack of accuracy, can provide limited informationfor target JE control and prevention. In summary, it is necessary to understand thespatiotemporal pattern of JE at small scale, to identify high risk clusters, to analyzethe main factors influenced JE occurrence, and predict the risk of JE occurrence,which will provide the scientific basis for the prevention and control of JE.Objective:To understand the trends and changes in the epidemic characteristics of JE and toinvestigate the temporal and spatial distribution patterns of JE;to illustrate therelationship between JE and environmental factors in different highly endemic areas,to identify the main drivers of JE occurrence, and to predict the short-term JEoccurrence; to analyze the impact of meteorological, geographical and environmentalfactors on the disease and perform higher solution prediction model to predict the riskof JE occurrence, to assess spatial differences of JE occurrence, and to provide ascientific basis for the prevention and control of JE.Methods:1. Monthly JE monitoring data from2002-2010were collected at countylevel,and then used descriptive epidemiology to analyze the epidemiologicalcharacteristics of both JE cases and deaths by using SAS9.2and SPSS16.0 programmers. The county was considered as the spatial unit of analysis,which wasjoined with JE cases and demographic information to draw JE annual incidencedynamically by using ArcGIS programmer. Global spatial autocorrelation analysis,partial autocorrelation analysis (Local Indicator of Spatial Association, LISA) andspatiotemporal cluster analysis (spatial scan statistics) were performed to assess if JEoccurrence existed autocorrelation and detecte high risk spatiotemporal clusters byusing GeoDaTM0.95i、ArcGIS9.3) and SaTScan9.1.1programmers.2.After adjustment of autocorrelation, seasonality and long-term trends of JEoccurrence, used Zero-inflated negative binomial (ZINB) to detect the relationshipsbetween environmental factors and JE occurrence in Zunyi and Bijie (both located inthe primary cluster, but with variations in elevation and geographical features), andcompared with negative binomial regression modeland zero-inflated Poisson model(ZIP). These analyses were performed by using Stata programmer (version11).3. After eliminating duplicate records in the same village per year and removingrecords of unavailable addresses, the remaining cases were geocoded in theadministrative village centeraccording to the patient’s residential address. Using GIStechnology, meteorological factors (average temperature, minimum temperature,maximum temperature, relative humidity and rainfall), elevation, land use, normalizeddifference vegetation index, pig density and population density were projected into1×1km resolution raster images. Using ecological niche models based on maximumentropy, the relationship between environmental variables and JE occurrence wereanalyzed, and the contribution of each variable to predict the risk of JE occurrencewas evaluated by using MaxEnt programmer (version3.3.3k).Results:1.A total of48,892JE cases were reported from2002-2010, with male to femaleratio of1.57:1.Children younger than15years of age accounted for87.40%of thetotal cases, but in recent years the proportion of adult JE (age above40years) had anincreasing trend (Cochran-Armitage trend test, Z=9.60, p <0.001). High incidencewas concentrated in younger age group, with children younger than10years had JEincidence higher than1/100000. A total of1,531deaths were reported from2005-2011, with male to female ratio of1.29:1. However, mortality of females was significantly higher than that of males (5.83%Vs4.79%, χ2=15.30, p <0.001.).Being grouped by10-year intervals showed that mortality increased with age(Cochran-Armitage trend test, Z=13.06, p<0.001), with persons above70yearsshowing a rate of14.67%. Foreigners mortality was higher than that of residents (12%Vs.5.77%, χ2=4.71, p=0.03). According to epidemic characteristics (adecreasingtrend from2002to2005, peaking in2006and thenfluctuating), we divided2002-2010into four periods:2002,2003-2005,2006, and2007-2010to perform spatialautocorrelation analysis. The occurrence of JE exist spatial autocorrelation. LISAanalysis found high-high areas were mainly concentrated in southwest China, with anexpanding trend to central China. Identified high risk primary clusters were in Julyand August in southwest China, with geographic extent of119,125,133, and144counties respectively. Each primary cluster had a higher incidence levels (ranged from1.79/100000to3.27/100000) with30.17%-77.59%cases concentrated in3.88%-4.97%population.2. After adjustmentof spatial autocorrelation, seasonality, and long-term trends ofJE, we performed ZINB models in Zunyi and Bijie, both located in the primary cluster.According to AIC and BIC (the lower the better), we selected the final models, whichperformed better than negative binomial models and ZIP models. The occurrence ofJE in Zunyi was negatively associated with relative humidity and rainfall at a lag of3and4months respectively, and accordingly1%increase in relative humidity at a lagof3months and1mm increase in rainfall at a lag of4months maybe associated with5.66%(95%CI:1.67%-9.45%) and0.05%(95%CI:0.01%-0.08%) decreases in JEcases respectively. The occurrence of JE in Bijie was positively associated withmaximum temperature at a lag of1month, and accordingly1oC increase in maximumtemperature at a lag of1month maybe associated with16.47%(95%CI:5.52%-28.55%) increase in JE cases. Residuals of predictions were white noise sequenceand found no significant autocorrelation and partial autocorrelation between residualsat different lags in the final models.3. Combination of ecological niche models based on maximum entropy,geographic information system (GIS) and remote sensing (RS) to detect relationshipbetween JE occurrence and environment factors and performing predictions inadvance of1year showed that minimum temperature, population density, average temperature, and elevation were the most important factors affecting the occurrence ofJE. Minimum temperatures contributed the most (17.94%-38.37%). Populationdensity, average temperature and elevation contributed second to fourth, with15.47%-21.82%,3.86%-21.22%, and12.05%-16.02%countribution, respectively.Response curve reflected more information on the relationship between JE occurrenceand environmental factors. Areas with monthly average temperature of12oC orminimum temperature higher than-8oC had JE occurrence. Areas with monthlyaverage rainfall in80-120mm or NDVI value up to150had higher risk of JEoccurrence. When the relative humidity higher than65%, risk of JE occurrenceincreased with relative humidity; after peaking at85%, risk of JE occurrence showeda small decline or remain stable. Elevation was negatively associated with JEoccurrence. Irrigated land, rain-fed land and building land had a strong correlationwith JE occurrence. The risk of JE occurrence increased with pig density and humandensity, and remained stable when pig density and human density reached to400and2500per sq.km. Duplication10times for each model to calculate average ROC curvefound that high AUC (ranged from0.82to0.91) and low extrinsic omission rate(ranged from5.44%to7.42%), indicating that the prediction models performed well.Our model predicted high risk areas of JE occurrence was mainly concentrated inSouthwest and Central China. Approximately60%of JE cases occurred in predictedhigh risk areas, which covered less than6%of areas in mainland China.Conclusions:This study described the prevalence and epidemic characteristics of JE inmainland China. GIS and spatial scan statistics were imployed to analysis thespatiotemporal pattern of JE in small scale. The spatial autocorrelation, high riskprimary and secondary clusters of JE were identified. ZIBN models were built toevaluatethe relationships between JE occurrence and environmental factors indifferent areas located in the primary cluster. Quantitative analyses the main factorsand time lag effects which influence JE distribution with variations in areas werecarried out.A nationwide high resolution (1×1km) prediction model was got. Themajor environmental factors that influence the occurrence of JE were assessed andhigh-risk and potential risk areas were found. The results of the work would bebeneficial to the target of control and prevention of JE occurrence. |
PDF Full Text Request