Department of Pediatrics, 11402 University Avenue NW, Edmonton, Alberta, Canada.
Int J Health Geogr. 2010 Jun 7;9:28. doi: 10.1186/1476-072X-9-28.
Traditional approaches to statistical disease cluster detection focus on the identification of geographic areas with high numbers of incident or prevalent cases of disease. Events related to disease may be more appropriate for analysis than disease cases in some contexts. Multiple events related to disease may be possible for each disease case and the repeated nature of events needs to be incorporated in cluster detection tests.
We provide a new approach for the detection of aggregations of events by testing individual administrative areas that may be combined with their nearest neighbours. This approach is based on the exact probabilities for the numbers of events in a tested geographic area. The test is analogous to the cluster detection test given by Besag and Newell and does not require the distributional assumptions of a similar test proposed by Rosychuk et al. Our method incorporates diverse population sizes and population distributions that can differ by important strata. Monte Carlo simulations help assess the overall number of clusters identified. The population and events for each area as well as a nearest neighbour spatial relationship are required. We also provide an alternative test applicable to situations when only the aggregate number of events, and not the number of events per individual, are known. The methodology is illustrated on administrative data of presentations to emergency departments.
We provide a new method for the detection of aggregations of events that does not rely on distributional assumptions and performs well.
传统的统计疾病聚集检测方法侧重于识别疾病的发病率或患病率较高的地理区域。在某些情况下，与疾病相关的事件可能比疾病病例更适合分析。每个疾病病例都可能与多个与疾病相关的事件相关联，并且需要在聚类检测测试中纳入事件的重复性质。
我们提供了一种新的方法来通过测试可能与最近邻合并的单个行政区域来检测事件的聚集。该方法基于测试地理区域中事件数量的精确概率。该测试类似于 Besag 和 Newell 给出的聚类检测测试，不需要 Rosychuk 等人提出的类似测试的分布假设。我们的方法结合了不同的人口规模和人口分布，可以通过重要的层次结构来区分。蒙特卡罗模拟有助于评估确定的集群总数。需要每个区域的人口和事件以及最近邻空间关系。我们还提供了一种替代测试方法，适用于仅知道事件总数而不是每个个体的事件数的情况。该方法在急诊部门就诊的行政数据上进行了说明。
我们提供了一种新的用于检测事件聚集的方法，该方法不依赖于分布假设，并且性能良好。
