Fast detection of arbitrarily shaped disease clusters

Abstract

Disease cluster detection and evaluation have commonly used spatial statistics methods that scan the map with a fixed circular window to locate candidate clusters. Recently, there has been interest in searching for clusters with arbitrary shape. The circular scan test retains high power of detecting a cluster, but does not necessarily identify the exact regions contained in a non-circular cluster particularly well. We propose, implement and evaluate a new procedure that is fast and produces clusters estimates of arbitrary shape in a rich class of possible cluster candidates. We showed that our methods contain the so-called upper level set method as a particular case. We present a power study of our method and, among other results, the main conclusion is that the likelihood-based arbitrarily shaped scan method is not appropriate to find a cluster estimate. When the parameter space includes the set of all possible spatial clusters in a map, a large and discrete parameter space, maximum likely cluster estimates tend to overestimate the true cluster by a large extent. This calls for a new approach different from the maximum likelihood method for this important public health problem.

A Spatial Scan Statistic with a Restricted Likelihood Ratio

“…Kulldorff et al (2006b) explored an elliptic version of the spatial scan statistic by introducing the eccentricity penalty that discourages eccentric clusters. Regarding the penalized likelihood approach, Assunção et al (2006) made an important comment that this approach is a possible solution but certainly plagued with a large dose of subjectivity in the penalty parameters and Jpn J Biomet Vol. 29, No.…”

“…However, since it uses a circular window to scan the potential cluster areas, it has difficulty in correctly detecting actual non-circular clusters. To detect arbitrarily shaped clusters which cannot be detected by the circular spatial scan statistic, Duczmal and Tango Assunção (2004), Patil and Taillie (2004), Tango and Takahashi (2005) and Assunção et al (2006) have proposed different spatial scan statistics. It should be noted that all of these scan statistics are based on maximizing the likelihood ratio.…”

Density-equalizing Euclidean minimum spanning trees for the detection of all disease cluster shapes

“…Assunção et al (13) used minimum spanning tree of graphs with different vertices, edges, and edge weights to consider contiguous administrative regions having similar disease rates, whether high or low. By contrast, we locate sets of individual cases corresponding to a mathematical formalization of a cluster, using specific subsets of the EMST.…”

“…One class of methods based on graph theory has recently emerged to address this problem (11)(12)(13)(14). However, these have several limitations: they are restricted to clusters that fit inside a circular region of fixed size (11), they attempt to examine a set of potential clusters too large to exhaustively search (12), they have poor specificity (13), or they have yet to be implemented or evaluated (14).…”

“…This problem surfaces when data are aggregated into small regions. Indeed, one study identified excessively large clusters with highly irregular shapes having greater likelihood ratios than the inserted clusters that were the detection targets (13).…”