本研究分析了2016年湖北省高危孕产妇人群在空间上的分布规律，为优化现有医疗资源配置提供参考，同时也为探索其他相似问题的分析方法提供借鉴。本研究基于2016年湖北省孕产妇就医病历数据，以乡镇街道为研究单元，主要使用了空间自相关分析、SaTScan空间扫描统计、核密度估计三种空间分析方法，分析了湖北省孕产妇高危发生率的空间格局。研究结果发现湖北省孕产妇高危发生率最大高值聚集区位于宜昌市的西北边缘乡镇、神农架林区大部分区域和十堰市东南边缘乡镇。此外，恩施州南部、荆州市南部、黄冈市东部、襄樊市西部、省管县级市仙桃、天门市也是高危率的高聚集区。总的来看，湖北省孕产妇高危率在空间分布上呈现中西部密集，东部稀疏，边缘集聚的整体格局，且高值聚集区大多分布在市级行政单位的边缘地带，低值聚集区则多出现在市级单元的中心区。基于上述分析，本研究发现空间自相关分析能分别从全局与局部两个方向探测空间的聚集现象，SaTScan空间扫描统计能对空间内的聚集现象进行分层级描述，但两者精度皆易受边界限制；核密度估计法可以突破边界的限制，在结果中将聚类区定位得更精准，但在确定热点区边界往往比较困难。

This study analyzed the spatial clusters of pregnant women with high risks in Hubei Province in 2016, so as to provide reference for optimizing the allocation of maternal health care resources and for exploring methods for analyzing similar research questions. Three spatial analysis methods, spatial autocorrelation analysis, SaTScan space scanning statistics and kernel density estimation were used. This study used the data from the medical records of pregnant woman in Hubei Province in 2016, and analyzed at the township/street level. The results show that the highest clusters were located in the townships at the northwestern edge of Yichang, most areas of the Shennongjia Forest Region, and the townships at the southeastern edge of Shiyan. In addition, clusters of high-risk rate were also identified in the southern districts of Enshi and Jingzhou, eastern districts of Huanggang, and the western districts of Xiangfan, Xiantao, Tianmen. Overall, the high-risk rate of pregnant woman in Hubei Province was higher in the central and western regions, while lower in the east region. The clusters with high values were mostly distributed at the margins of municipal administrative units. Areas with low values were mostly located in the central area of the municipal units. Based on the above analysis, this study found that spatial autocorrelation analysis could detect spatial aggregations from both global and local directions. SaTScan spatial scanning statistics could describe the aggregation phenomena in space hierarchically. However, the accuracy of both methods is limited by the boundary. Kernel density estimation could overcome the limit of the boundary, and the clustering area is more precise in the result. It is, however, difficult to determine the boundaries of the hot zones.