Cluster Detection Methodology for Small Area Cancer Data

描述(由申请人提供)： 这个项目的总体目标是开发新的和灵活的统计方法，用于检测空间癌症疾病发病率数据中的聚集性。在这方面，我们有兴趣开发和评价集群建模方法的使用，以解决集群在哪里以及集群规模有多大的确定问题。目前这一领域公认的方法论受到假设检验基础的限制(例如SATScan[1])，为此开发新的基于模型的方法有相当大的余地。 主要的具体目标是： 1)发展贝叶斯模型，允许依赖于病例的局部密度(数据依赖模型)。 2)不仅在空间环境中而且在时空环境中的方法的发展。这在公共卫生应用中很重要，因为时间的变化可能具有特别重要的意义。我们建议通过使用分层贝叶斯方法来建模特定地区和年份的集群分布，特别是估计时空连续体中的疾病风险，从而将依赖于数据的空间集群建模扩展到时间设置。 3)对已开发方法的评价。我们建议通过与一小部分替代比较方法的比较来评估数据依赖集群模型的使用。 4)软件开发。需要提供灵活的软件，使负责癌症集群检测的研究人员和公共卫生工作者能够使用建模方法，使他们能够灵活地建立对观察到的疾病数据的适当描述。

DESCRIPTION (provided by applicant): The general aim of this project is to develop new and flexible statistical methods for the detection of clustering within spatial cancer disease incidence data. In this connection we are interested in developing and evaluating the use of cluster modeling methods to address the issue of the identification of where clusters are and how large the clusters are. Current accepted methodology in this area is limited by its basis on hypothesis testing (e.g. SaTScan [1]), and there is considerable scope to develop new model-based methods for this purpose. The major specific aims are: 1) The development of Bayesian models that allow dependence on the local density of cases (data-dependent models). 2) The development of methods in not only spatial settings but also within space-time situations. This is important in public health applications where changes in time can have particular importance. We propose to extend the data-dependent spatial cluster modeling to the temporal setting by using a hierarchical Bayesian approach to modeling the region and year specific cluster distributions, in particular, to estimate the excess of disease risk within a space-time continuum. 3) The evaluation of the developed methods. We propose to evaluate the use of data-dependent cluster models in comparison to a small set of alternative comparison methods. 4) The development of software. There is a need for flexible software to be made available that can allow researchers and public health workers tasked with cancer cluster detection to be able to use modeling approaches, with their ability to flexibly build appropriate descriptions of the observed disease data.