Spatial Analysis On The High Risk Regions For Schistosomiasis Japonica And Identification Of Active Transmission Sites

The epidemic situation of schistosomiasis japonica has rebound rapidly in China recently and showed a tendency of expanding.We cannot but reconsider the strategy of sustainable schistosomiasis control due to the decreased financial supports,extensive snail habitats, declined compliance rate due to repeated drug treatment,and bad effects for chemotherapy to interrupt schistosomiasis transmission.In this study,we selected Guichi region of Chizhou city,Anhui province as our study area.We precisely identified the high risk regions of schistosomiasis step by step and located the active transmission sites of snail habitats finally, which provided more practical directions for schistosomiasis control.At the same time,the systematical approaches for spatial analysis were established.It added new contents for spatial epidemiology.The whole study was divided into six parts.PartⅠSpatial descriptive analysis on schistosomiasis casesObjective To conduct descriptive analysis on acute schistosomiasis cases from the spatial point of view and establish the corresponding approaches.Methods Acute schistosomiasis cases were collected through retrospective survey method in Guichi region from 2001 to 2006 and their spatial positions were recorded by GPS machines.Borrowing ideas from descriptive analysis in classical statistics and methods to describe crime incidents,we put forward the approaches of describing spatial central and dispersion tendency.Weighted mean center and weighted standard deviance ellipse were used to analyze the data of schistosomiasis cases, and their results were compared with the conventional descriptive analysis to explore their merits.Results The computational methods,application conditions and potential merits on using these statistical indices to perform spatial descriptive analysis were systematically established and introduced.The demographic characteristics of these acute cases were not significantly different across the six years,such as gender(p=0.42),age(p=0.08) and occupation(p=0.08).The time of disease occurrence was mainly from July to October,while no cases existed between December and March.The results from spatial descriptive analysis showed that Qiupu River was the emphasis for schistosomiasis control in Guichi region and the spatial center of acute cases had a tendency to move southward.Conclusion Spatial descriptive analysis was valuable supplements for descriptive analysis in classical statistics, and their combination would give a more comprehensive description on acute cases in Guichi region. PartⅡStudy on the spatial pattern of schistosomiasis casesObjective To explore the spatial distribution status for acute schistosomiasis cases,and establish the corresponding techniques of quantitative analysis.Methods The theories of G,F, J and K functions were introduced on the basis of inter-case spatial distances,and they were used to analyze the data of acute schistosomiasis cases in Guichi region.In the course of analysis,the study distance was from 0 to 3000m and the interval was 50m.Results The computational methods for the four quantified indices were obtained and the concept of edge effect was simultaneously put forward.For the acute schistosomiasis cases,G and K functions were above,while F and J functions were below the simulated envelopes of spatial random distribution.They showed the same result that these cases were a clustered spatial pattern. Conclusion The spatial pattern of acute cases in Guichi region was clustered based on the assumption that the distribution of background population at risk was homogeneous.PartⅢSpatial cluster analysis on the schistosomiasis casesObjective Spatial cluster analysis was discussed from different aspects with the heterogeneous distribution of background population at risk adjusted.Methods The controls with the same sample sizes as cases were obtained by the sampling method of probability proportionate to size with the village population as the weights to represent the heterogeneous distribution of background population at risk.Cuzick-Edwards test,the methods based on first- and second- order attributes of point process,and spatial scan statistic approach through moving window were used to study the spatial clustering of acute schistosomiasis from different point of view and their results were compared with each other for a validation. Results Cuzick-Edwards test showed that the cases had a significant global clustering (p＜0.01).The result from the approaches based on the second-order attribute of point process was that the acute cases showed a significant clustering when the study scale is less than 13000m(p＜0.05),and the degree of clustering increased first and then decreased.For the results of the methods based on the first-order attribute of point process and spatial scan statistic approach,they detected the two coincident clusters.The most likely cluster was situated where the Qiupu River feeds into the Yangtze River.Its center was(117.43°longitude, 30.67°latitude) and its radius was 5.63km(LLR=19.56,p=0.001).The secondary likely cluster was located in the southeastern part of Guichi region.Its center was(117.71°longitude, 30.36°latitude) and its radius was 9.74 km(LLR=7.25,p=0.07).Conclusion The acute schistosomiasis cases in Guichi region had a significant spatial clustering when adjusting the heterogeneous distribution of background population at risk and two significant high risk clusters were detected.PartⅣIdentifying high risk regions of schistosomiasisSection one Study on the impact factors of schistosomiasis and Oncomelania hupensis1.Large-scale impacts of air temperature Objective To study the impacts of air temperature on the O.hupensis at a large scale,and explore the sensitive temperature index. Methods Weather stations in the neighboring provinces where snails exist and do not exist were selected as study objects with north latitude 340 as the boundary of snail distribution. The stations in the south and north of the boundary were coded as regions that the air temperature was and was not suitable for snails to live,respectively.The differences of yearly extreme low temperature and mean temperature in a year between snail areas and no snail areas were tested using t test to show their significance on the impacts of snail distribution, respectively.And unconditional logistic regression analysis was used to determine the sensitive index of air temperature to predict the snail distribution.Then the histogram for the sensitive temperature index was generated and the suspicious temperature range for snails to live was calculated through the overlapped parts between the histograms.Results The weather stations for snail areas and no snail areas were 37 and 24,respectively.The yearly extreme low temperature and mean temperature in a year were significantly different between snail areas and no snail areas(t=-6.49,p＜0.01;t=-3.93,p＜0.01),and they were both lower in the no snail areas and their differences were 6.72℃and 3.02℃,respectively.Logistic regression analysis showed that only the annual extreme low temperature was significant for indicating the snail distribution(χ~2=15.69,p＜0.01).The temperature range for the suspicious snail areas was(-7.6℃,1.5℃).Conclusion Annual extreme low temperature was the sensitive temperature index for the living of O.hupensis at a large scale,and the snails were not suitable to live in the places where the temperature was lower than -7.6℃.2.Small-scale impacts of vegetation Objective To study the impacts of vegetation changes on the distribution of O.hupensis in the lake and marshland regions.Methods A bottomland along the Qiupu River in Guichi region was randomly selected as the study field.Low-grass group,boundary-grass group,hay group and control group were designed to represent four different types of vegetation's status.The snail density,soil temperature and soil moisture were surveyed half a month after the study began.The movement of snails due to vegetation changes was analyzed and inferred.Results The snail densities of the hay group,low-grass group,control group,boundary-grass group were 0.13/0.11m~2,32.1/0.11m~2,49.07/0.11m~2,and 53.6/0.11m~2,respectively.There was no significant difference for the snail density between the boundary-grass group and control group(p＞0.05),but the differences were significant among the other groups(p＜0.05).The environments of high soil moisture and low soil temperature in the hay group and low soil moisture and high soil temperature in the low-grass group were both unfavorable for the living of O.hupensis.Conclusion Vegetation was an important factor for the living of O.hupensis,and it would lead the snails to move from low-grass group to boundary-grass group if the vegetation was cut down.3.Study on the factors of affecting snail density at a small scale Objective To study the important factors of affecting snail density at a small scale.Methods Study area was randomly selected from the bottomlands along Qiupu River in Guichi region.The sampling method was stratified random sampling according to the types of vegetation;the frame size of snail survey was 0.11m~2 and snails were collected by crosscheck-random sampling inspection survey.Elevation,soil and air temperature,height of vegetation,soil humidity and types of vegetation were simultaneously collected,and the data were collected in April and September, 2006,respectively.Generalized linear models were used to develop the prediction model for snail density,and the statistics of deviance and AIC were used to determine the best model structure.Model diagnostics and model evaluation of efficiency were performed on the chosen best model.Then the impacts of different factors at a small scale were discussed following the results of the fitting model.Results The sample size was 162.There were 6 explanatory variables,including 2 categorical variables and 4 quantitative variables. Complicated relationships existed between variables,snail data was positively related with the height of vegetation(r=0.36),while negatively related with soil humidity(r=-0.22);air temperature had a strong positive relationship with soil temperature(r=0.59);soil temperature was negatively related with height of vegetation(r=-0.36);soil humidity had negative relationships with both soil and air temperature(r=-0.34 and -0.12).The best model structure was determined with gamma distribution as error distribution,inverse as link function,and mean^2 as variance function.The results showed that different vegetation had different impacts on the snail density.Height of vegetation(t=-2.371,p=0.01897) and low elevation (t=-3.202,p=0.00166) were positive impacts,while soil humidity(t=3.124,p=0.00214) was negative impacts for snail density.Soil temperature showed an ambiguous effect(t=-1.989, p=0.04849),while air temperature was not significant for predicting snail density.Conclusion Soil temperature,vegetation,elevation and soil humidity were important factors for snails at a small scale and generalized linear models were promising to establish a prediction model of snail density.Section two Detecting high-risk regions for schistosomiasis using GAMsObjective To identify high-risk regions for schistosomiasis with the impacts of covariates adjusted.Methods The environmental variables(NDVI and LST) from remote sensing images,topographical variables(elevation and slope) from digital elevation model,nearest distance from cases and controls to the rivers,and spatial positions(X/Y) were used as independent variables to predict the disease risk for schistosomiasis from the spatial point of view using generalized additive models(GAMs).The goodness of model fitting and the prediction ability were evaluated by the partial residual plots and the area under the receiver operating characteristic curve(AUC),respectively,and piecewise 'p value surface' approach was adopted to conduct the statistical test in space.And the results were also compared with that of PartⅢto explore their differences and reasons.Results The final fitting model had a high prediction ability(AUC=0.911).Distance(χ~2=19.6879,p=0.0002),spatial positions X (χ~2=11.7625,p=0.3809),Y(χ~2=26.3038,p=0.0009) and elevation(χ~2=13.4844,p=0.0037) were significant at a small scale,and they were piecewise linear,quadratic curve,linear and piecewise linear relationships for predicting the disease risk of schistosomiasis.The disease risk began to decrease when the elevation was higher than 86m in the local area.The disease risk was constant and high when the distance was less than 1500m apart from the high risk rivers,and the risk became very low for the places where distance was larger than 18000m. The important factors for schistosomiasis risk at a small scale were social,geographical and environmental factors in sequence.Four significant high risk regions for schistosomiasis were detected.Two true high risk clusters were in agreement with the results of PartⅢ,but here they were more precise.The other two new clusters had similar environments as the two true high risk clusters and were regarded as potential high risk clusters of schistosomiasis. Conclusion There were two true and two potential high risk clusters for schistosomiasis in Guichi region.PartⅤRecognition of high risk snail habitatsObjective To locate the high risk snail habitats in Guichi region.Methods Two remote sensing images were obtained,one representing the "wet season" and the other denoting the "dry season" in Guichi region.The two indices of NDWI and NDVI were jointly applied to extract snail habitats.The accuracy of this method was evaluated through field survey.Then, the results from the previous PartⅣwere overlaid with extracted snail habitats to locate the high risk snail habitats.The detailed information for these high risk habitats were obtained by geographic information system(GIS) and field investigation was conducted to validate these analysis results following the navigation of global positioning system(GPS).Results The sensitivity and specificity of the above method to extract snail habitats were 90%and 100%, respectively.There were 349 places of snail habitats in Guichi region,and the total area was about 107km~2.Six high risk snail habitats were finally located.Conclusion Centered on the six detected high risk snail habitats,their surrounding regions of 1500m were the emphasis of schistosomiasis control in Guichi region.PartⅥStudy on the statistical distribution of Oncomelania hupensisSection one Study on the survey method to collect snails in the fieldObjective To find a better method for collecting snails in the field.Methods A bottomland was randomly selected as the study field along the Qiupu River in Guichi region.The conventional survey,individually repeated survey,crossed-repeated survey and crosscheck-random sampling inspection survey were designed to collect snails in the same area.Snail density and omission rate were used to evaluate the data quality of different survey methods.Results There were no statistical significant differences for snail density among the following three survey methods,individually repeated survey,crossed-repeated survey and crosscheck-random sampling inspection survey(χ~2=3.873,p=0.144),which were significantly different from that of conventional survey(U=309,p＜0.01).Omission rates for individually repeated survey,crossed-repeated survey and crosscheck-random sampling inspection survey were significantly different(χ~2=37.44,p＜0.01) and they were 10.26%,5.24%and 0.57%, respectively.Conclusion Crosscheck-random sampling inspection survey was the best method in the field to collect snails accurately.Section two Study on the statistical distribution of Oncomelania hupensisObjective To study the statistical distribution of O.hupensis.Methods 4 bottomlands were randomly selected as the study fields along the Qiupu River in Guichi region to collect snails in October,2005,and 2 bottomlands from the 4 selected bottomlands were randomly chosen again to survey snails in April,2006.Maximum likelihood estimation was used to fit generalized negative binomial distribution(GNBD) and negative binomial distribution(NBD). Results from different bottomlands in the same season and the same bottomlands in different seasons were compared with each other to discuss the statistical distribution of O.hupensis. Results Snail density in different seasons and different bottomlands were different,but they had the similar positive skewness distribution.GNBD was successfully fitted to all the snail data,but NBD was not.And the goodness of fitting for GNBD was better than NBD for all the collected snail data.The fitted parameters of GNBD from the same bottomlands in different seasons changed in different direction,and the parameters of GNBD from different bottomlands in the same season were not similar even if their snail density were close.Tiny changes in different snail habitats could be sensitively reflected by the parameters of GNBD. Conclusion GNBD could reflect the complexity of snail distribution than NBD very well, which was more promising for the statistical distribution of O.hupensis.Section three Relative impacts of biological behavior and environmental factors on the snail distributionObjective To study the relative impacts of biological behavior and environmental factors on the snail distribution for correctly understanding its distribution.Methods A bottomland was randomly selected along the Qiupu River in Guichi region,and the survey area of 200cm×200cm was also randomly determined.All the snails were collected,marked and measured with the corresponding Cartesian coordinates.G and F functions were first applied to conduct the exploratory data analysis on the distribution pattem of snails.Then,the transformed reduced second moment function,L-function,was used to do the multi-scale analysis and the relative impacts of biological behavior and environmental factors on the snail distribution were tried to be separated according to the results from the small and large scales,respectively. Finally,a point process model was fitted to the snail data and Monte Carlo simulation was used to test the goodness of fitting.Results An area of 122cm*172cm was surveyed finally and 528 live snails were found.The distribution pattern caused by biological behavior was random in the survey area.It was the environmental factors that led it to be an aggregated distribution and its impacts were 5.96 times as that of biological behavior.Conclusion Multi-scale analysis could be used to separate the relative impacts of biological behavior and environmental factors on the snail distribution,and the latter had larger impacts.