A flexibly shaped spatial scan statistic for detecting clusters

Tango, Toshiro; Takahashi, Kunihiko

doi:10.1186/1476-072x-4-11

Cited by 466 publications

(216 citation statements)

References 17 publications

(18 reference statements)

Supporting

Mentioning

211

Contrasting

Unclassified

Order By: Relevance

“…Parameter estimation was done using maximum likelihood in STATA 13. Spatial clustering of stunting was assessed using a flexible (irregular shaped) spatial clustering statistic (so-called Tango's flexibly shaped spatial scan statistic) implemented in FleXScan software assuming a binomial model, original log likelihood ratio and with a maximum spatial cluster size of 30 nodes and based on 9999 Monte Carlo replications (Takahashi, Yokoyama, & Tango, 2007;Tango & Takahashi, 2005).…”

Section: Discussionmentioning

confidence: 99%

Stunting in children (0–59 months): what is the current trend in South Africa?

Dukhi

Sartorius

Taylor

2016

Early Child Development and Care

View full text Add to dashboard Cite

Background: Stunting continues to affect young children as a global nutritional disorder. The aim of our study was to assess the prevalence, associated risk factors and spatial clustering for stunting in a disadvantaged South African District. Methods: A community-based cross-sectional weighted survey of households was conducted in the iLembe district, KwaZulu-Natal in 2014. Children aged 0-59 months were included in the study sample. Height for age Z scores were calculated for the children and the spatial clustering was assessed. Results: Of the 572 children included in the weighted analyses, 21.1% (95% confidence intervals (CI) 18.0-24.5) were classified as stunted and of these 7.5% (95% CI 5.3-10.4) were severely stunted. Conclusion: Stunting is a major public health burden in SA that requires intervention at a community level. A multi-sectoral approach is required to reduce stunting prevalence, focusing on child growth and development monitoring, nutrition education and household food security.

show abstract

Section: Discussionmentioning

confidence: 99%

Stunting in children (0–59 months): what is the current trend in South Africa?

Dukhi

Sartorius

Taylor

2016

Early Child Development and Care

View full text Add to dashboard Cite

show abstract

“…The indicator function I ( C i > Np i ) is equal to 1 when C i > Np i and 0 otherwise. A possible cluster is defined as a region with a high likelihood ratio statistic [12]. …”

Section: Methodsmentioning

confidence: 99%

“…The first method is a focused approach where regions with a high rate of disease around a possible cause (i.e., a toxic waste site) are located [9,10]. Focused cluster detection methods include the circular spatial scan statistic (CSS) [11], flexible spatial scan statistic (FSS) [12] and Bayesian disease mapping (BYM) [13]. Focused tests are used to discover potential clusters in a specific area of interest by testing the null hypothesis that there is no cluster against the alternative hypothesis that a cluster exists.…”

Section: Introductionmentioning

confidence: 99%

Application of Three Focused Cluster Detection Methods to Study Geographic Variation in the Incidence of Multiple Sclerosis in Manitoba, Canada

Torabi¹,

Green²,

Yu³

et al. 2014

Neuroepidemiology

View full text Add to dashboard Cite

Background: Macroscopic geographic variation in the incidence and prevalence of MS is well-recognized. Microscopic geographic variation in the distribution of MS is also recognized, but less well-studied. Most studies have focused on prevalent cases of MS, although studies of variation in disease incidence are more relevant for developing etiologic hypotheses. We aimed to study geographic variation in the incidence of MS using three different methods. Methods: We used population-based administrative (health claims) data to identify 2,290 incident cases of MS in the province of Manitoba, Canada from 1990 to 2006. We applied three focused cluster-detection procedures, including the circular spatial scan statistic (CSS), flexible spatial scan statistic (FSS), and Bayesian disease mapping (BYM), to the dataset. Results: The CSS and FSS methods identified 30 and 26 regions as potential clusters, respectively, although the regions identified differed slightly due to the non-circular shape of some regions in Manitoba. The BYM approach identified 37 regions as potential clusters, again with some differences as compared to the other two methods. Twelve regions were identified as potential clusters by all three methods. All methods identified the western part of the city of Winnipeg as a significant cluster. Using the BYM approach, the incidence of MS was highest among areas of higher socioeconomic status. Conclusions: Two methods CSS and FSS only capture geographical variations and are not able to control for confounders at the same time which may lead to mis-identification of clusters. However, the BYM method can simultaneously identify geographical variations and control for possible confounders.

show abstract

“…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).…”

mentioning

confidence: 99%

“…As more shapes are considered, the statistical power declines, and the computational running time may become unreasonable for typical problem sizes (11). Furthermore, if the exact case locations are available, then considering every conceivable shape is problematic; it is always possible to draw a bizarrely shaped region of infinitesimally small total area that includes every case.…”

mentioning

confidence: 99%

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

Wieland

Brownstein²,

Berger³

et al. 2007

Proc. Natl. Acad. Sci. U.S.A.

View full text Add to dashboard Cite

Existing disease cluster detection methods cannot detect clusters of all shapes and sizes or identify highly irregular sets that overestimate the true extent of the cluster. We introduce a graphtheoretical method for detecting arbitrarily shaped clusters based on the Euclidean minimum spanning tree of cartogram-transformed case locations, which overcomes these shortcomings. The method is illustrated by using several clusters, including historical data sets from West Nile virus and inhalational anthrax outbreaks. Sensitivity and accuracy comparisons with the prevailing cluster detection method show that the method performs similarly on approximately circular historical clusters and greatly improves detection for noncircular clusters.biosurveillance ͉ disease cluster detection ͉ graph theory

show abstract

A flexibly shaped spatial scan statistic for detecting clusters

Cited by 466 publications

References 17 publications

Stunting in children (0–59 months): what is the current trend in South Africa?

Stunting in children (0–59 months): what is the current trend in South Africa?

Application of Three Focused Cluster Detection Methods to Study Geographic Variation in the Incidence of Multiple Sclerosis in Manitoba, Canada

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

Contact Info

Product

Resources

About