Transmission of S. japonicum is influenced by many factors. Some key factors are climatic suitability (e.g. temperature and precipitation); spatial and temporal distribution of the intermediate host, O. hupensis; human activities (e.g. occupational, domestic, and recreational water-contact patterns); environmental contamination with human and animal excreta, in particular, cattle stool; and lack of a clean water supply, sanitation, and proper hygiene [12, 24–28]. As a result, S. japonicum infection exhibits marked spatial heterogeneity from the community scale even at the scale of a single administrative village) up to the regional scale, [7, 17, 28–31]. Our results from the spatial autocorrelation analysis showed that there was significant spatial cluster at the household level in 2005, when the infection prevalence was relatively high (i.e. > 9.3%). Similar results were obtained from analyses of the spatial scan statistic that was used to identify areas and population with high prevalence (a significant spatial cluster of infection was found at the household level in 2005). These findings were similar to the results reported by Peng et al. (2010) , where the human infection prevalence was 6.5%. Additionally, the results support the general view of the focal nature of S. japonicum (i.e. S. japonicum infection shows distinct spatial distribution (focality and aggregation) when the human infection prevalence is relatively high).
However, the four years’ results of both the global spatial autocorrelation analysis and the spatial scan statistical analysis all showed that no significant spatial cluster of infection was found at the household level after 2005, the results of the local spatial autocorrelation analysis also showed that the spatial distribution of S. japonicum infections became less clustered with time. These suggested that the spatial distribution pattern of S. japonicum infections might be changing over the course of control, from a heterogeneous pattern to a more homogenous pattern with a decrease in the infection prevalence at the household scale in our study area. This was different from the general phenomenon of infectious diseases (i.e. spatial heterogeneity increases with declining rates of transmission). This difference might be attributed to or related to 1) the complex life cycle of S. japonicum involving snail (Oncomelania hupensis) intermediate hosts and many vertebrate definitive hosts [24, 32] (for example, the geographic distribution of schistosomiasis japonica largely depends on the availability of the susceptible schistosome-transmitting snail O.hupensis, and the spatial distribution of snails infected with S. japonicum becomes less heterogeneous with a decrease in the infection prevalence ); 2) the effective integrated national intervention being employed during the study period ; and 3) changes in some important risk factors influencing the transmission of S. japonicum in our study area. Although some control measures of snails were carried out in our study region, the spatial distribution of snails does not obviously change, in other words, the area with snails does not change over the course of control . However, the integrated national control strategy resulted in changes in some important risk factors influencing the transmission of S. japonicum in the study area . For example, environmental contamination with cattle excreta was completely wiped out after all bovines were removed from the study region. Previously, buffaloes were responsible for approximately 75% of human transmission . Hygiene was improved, to a large extent, by constructing lavatories and latrines and by supplying tap water. Furthermore, new human infections of schistosomiaisis each year were effectively cleared by treatment with praziquantel. Also, health education programs might have changed human behavior. However, it is unclear whether there would be changes in other risk factors with the shift from clustered to random spatial distribution. For example, whether some minor and neglected risk factors (e.g. environmental contamination with wild (e.g. mouse and rabbit) and other domesticated (e.g. pig and dog) animals excreta and population movement) would become important risk factors of S. japonicum transmission when other key risk factors noted above were controlled. This is necessary to further survey and to reassess the risk factors of S. japonicum transmission. The findings of random spatial distribution patterns of S. japonicum transmission showed that it might be necessary to adjust current control measures in the community due to the changes in risk factors of S. japonicum transmission. For example, if some neglected risk factors (e.g. wild mouse, population movement) become important risk factorsfor S. japonicum transmission in a community after bovines are removed as important sources of infection, an approach for wild rat control or systematic administration of antischistosomal drugs (e.g. praziquantel) to migrantsmay need to be conducted in the community. This can help to tailor effective, locally adapted control measures .
Utzinger et al. (2003)  reported that the spatial distribution of S. mansoni infection intensity levels was random or homogenous in a single village, but the infection prevalence of S. mansoni was very high (80.4%) when compared with the infection prevalences of S. japonicum in our study. In our study, we saw a random spatial pattern of S. japonicum infection only when the human infection prevalence was relatively low (< 4%); when the infection prevalence was relatively high (e.g. > 9%), the spatial pattern was heterogeneous. These findings indicate that spatial patterns of Schistosoma transmission might be relevant to infection prevalence at a finer scale, but it is unclear whether there is a critical prevalence when the distribution shifts from heterogenous to homogenous distribution. Additionally, these results suggest that differences in spatial distribution patterns of schistosomiasis might be related to differences in the spatial distribution of risk factors of Schistosoma transmission, this needs further study.
The cohort population was a stable population except in 2008 when 8.8% of the participants were lost to follow-up. This was mainly due to population movement (e.g. mobility for paid work or school). There was no obvious difference (e.g. age, sex, occupation) between the observed population and the population lost to follow-up (data not shown). A recognized limitation of this study was related to the Kato-Katz thick smear technique used. This method has become relatively insensitive due to widespread chemotherapy that results in generally low worm burdens ; however, our previous study  showed that the sensitivity of a single stool examination with the three-slide Kato-Katz method did not change significantly when the prevalence of a single stool examination with the three-slide Kato-Katz method is between 13.0% (close to the baseline (9.3%) in 2005) and 3.9% (close to the prevalence of 3.7% in 2009). Significant spatial clusters of infection could be found by both the spatial autocorrelation analysis and the spatial scan statistical analysis in 2005; however, significant spatial clusters of infection could not be found after 2005. So, the relative insensitivity of Kato-Katz method might not be a major factor for the apparent lack of clustering. Other weaknesses of the study include a relatively small dataset and some risk factor data related to households or individuals were not collected (e.g. human activities and hygiene). This limits the ability to further analyze why the spatial distribution of S. japonicum infections changed. Given the results of the present study, it is cautiously suggested that, at the household level, the spatial distribution of S. japonicum infections become less heterogeneous as the prevalence of infection decreases.