A spatial scan statistic for nonisotropic two-level risk cluster

Authors

  • Xiao-Zhou Li,

    1. State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
  • Jin-Feng Wang,

    Corresponding author
    • State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
  • Wei-Zhong Yang,

    1. Office for Disease Control and Emergency Response, Chinese Center for Disease Control and Prevention (China CDC), Beijing, China
    Search for more papers by this author
  • Zhong-Jie Li,

    1. Office for Disease Control and Emergency Response, Chinese Center for Disease Control and Prevention (China CDC), Beijing, China
    Search for more papers by this author
  • Sheng-Jie Lai

    1. Office for Disease Control and Emergency Response, Chinese Center for Disease Control and Prevention (China CDC), Beijing, China
    Search for more papers by this author

Jin-Feng Wang, State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China.

E-mail: wangjf@lreis.ac.cn

Abstract

Spatial scan statistic methods are commonly used for geographical disease surveillance and cluster detection. The standard spatial scan statistic does not model any variability in the underlying risks of subregions belonging to a detected cluster. For a multilevel risk cluster, the isotonic spatial scan statistic could model a centralized high-risk kernel in the cluster. Because variations in disease risks are anisotropic owing to different social, economical, or transport factors, the real high-risk kernel will not necessarily take the central place in a whole cluster area. We propose a spatial scan statistic for a nonisotropic two-level risk cluster, which could be used to detect a whole cluster and a noncentralized high-risk kernel within the cluster simultaneously. The performance of the three methods was evaluated through an intensive simulation study. Our proposed nonisotropic two-level method showed better power and geographical precision with two-level risk cluster scenarios, especially for a noncentralized high-risk kernel. Our proposed method is illustrated using the hand–foot–mouth disease data in Pingdu City, Shandong, China in May 2009, compared with two other methods. In this practical study, the nonisotropic two-level method is the only way to precisely detect a high-risk area in a detected whole cluster. Copyright © 2011 John Wiley & Sons, Ltd.

Ancillary