Spatial, temporal and space-time scan statistics are new tools to detect aggregation of disease cases, to test the occurrence of any statistically significant clusters, and ultimately to find evidence of risk factors. Scan statistics also identify whether cases of disease in space or time can be explained by chance or are statistically significant. The spatial scan statistic method is a useful instrument for HFRS analysis of cluster patterns. The space-time scan statistic method is proposed as a dynamic supplement to purely spatial statistical methods for outbreak detection and prediction. Aiming prevention strategies at areas of highest risk can potentially increase the public health intervention's effectiveness. People at highest risk should be informed of the high risk and the possibilities for risk reduction. The result of this study will assist health departments to develop a better prevention strategy. Further epidemiological studies will be conducted in populations defined by the study to gain insights into specific risk factors.
We investigated the spatial distribution of HFRS cases and identified areas with high endemicity of HFRS and clustering patterns with spatial scan statistics. We showed that in the period from 1988 to 2001 as a whole, the geographic distribution patterns of HFRS cases in Liaoning Province were not random. Spatial cluster analysis identified one most likely cluster and four secondary clusters based on a maximum spatial cluster size of 50% of total population.
We identified the eastern areas as having one most likely cluster in Liaoning Province which was also reported by Hualiang Lin . It is possible to think of Liaoning Province as three approximate geographical regions: the highlands in the west, plains in the middle, and hills in the east. The eastern part of Liaoning Province is dominated by the Changbai Shan and Qianshan ranges, which extends into the sea to form the Liaodong Peninsula. The most likely clusters of HFRS in Liaoning Province were mainly located in the hills. As for the secondary cluster, our result was different from that of Hualiang Lin's although with the same parameters. We found other three secondary clusters (Yuhong district, Tai'an county and Xinchengzi district) besides the western mountainous regions. The reasons could be as follows. First, the time we studied was different. The period was during 2000 to 2005 in the other study, whereas it was during 1988-2001 in our study. Second, the other study considered the city-governed region and others at a county level; therefore there were 58 county-level study areas, whereas in our study, Liaoning Province was divided into 100 study areas. In this study, we also conduct the temporal and space-time analyses besides spatial analyses.
After calculating the Moran's I statistic annually, it was significant from 1990 to 2001 at a significance level of 0.05, while it was not significant in 1988 and 1989. This implied that the distribution of HFRS in Liaoning Province from 1990 to 2001 was not distributed randomly in space, while in 1988 and 1989, it was distributed randomly. It is likely that environmental factors contribute to the epidemic and development of HFRS. Some studies have shown that some factors are related to the high incidence of HFRS such as environmental and climatic factors [21, 22]. Clearly, these factors could also affect the geographical clustering. It is necessary for us to study the changes in these factors between 1988 and 1989 and 1990 and 2001 in a future study.
We also conducted a space-time cluster analysis besides the purely spatial cluster analysis. Using the maximum spatial cluster size of 50% of the total population, and the maximum temporal cluster size of 50% of the total population, we identified one most likely cluster and two secondary clusters. When we compared the most likely and secondary clusters of the purely spatial cluster analysis with those of the space-time cluster analysis, we found an interesting phenomenon. The areas of the most likely cluster in the spatial cluster analysis became the areas of the secondary cluster in the space-time cluster analysis. Vice versa, the areas of the secondary cluster in the spatial cluster analysis became the areas of the most likely cluster in the space-time cluster analysis. That is to say, if we regarded the period 1988-2001 as a whole, the 10 areas, Hengren County, Xinbin County, Kuandian County, Benxi County, Qingyuan County, Nanfen District, Fushun County, Xihu District, Wanghua District, Fengcheng City had a higher RR than the 8 areas, Longgang District, Lianshan District, Xingcheng City, Taihe District, Nanpiao District, Linghe District, Guta District, Linghai City. The RR was 5.679 and 5.541 respectively. However, during the period 1995-2001, the 8 areas had a higher RR than the 10 areas if we considered both space and time. The RR was 9.065 and 8.560 respectively. Therefore, space-time disease-surveillance methods should be proposed as a dynamic supplement to purely spatial statistical methods for outbreak detection to detect and predict localized outbreaks before they spread to larger regions.
Temporal cluster analysis identified one most likely cluster based on a maximum temporal cluster size of 50% of the total population which had a significantly (P
< 0.001) increased HFRS risk, and the time frame was 1998-2001. In this analysis, we considered Liaoning Province as a whole place, and only considered the influence of time for HFRS.