An empirical comparison of spatial scan statistics for outbreak detection

Background  

The spatial scan statistic is a widely used statistical method for the automatic detection of disease clusters from syndromic data. Recent work in the disease surveillance community has proposed many variants of Kulldorff's original spatial scan statistic, including expectation-based Poisson and Gaussian statistics, and incorporates a variety of time series analysis methods to obtain expected counts. We evaluate the detection performance of twelve variants of spatial scan, using synthetic outbreaks injected into four real-world public health datasets.     

A SaTScan™ macro accessory for cartography (SMAC) package implemented with SAS<Superscript>®</Superscript>software

Background  

SaTScan is a software program written to implement the scan statistic; it can be used to find clusters in space and/or time. It must often be run multiple times per day when doing disease surveillance. Running SaTScan frequently via its graphical user interface can be cumbersome, and the output can be difficult to visualize.     

Detection of arbitrarily-shaped clusters using a neighbor-expanding approach: A case study on murine typhus in South Texas

Background  

Kulldorff's spatial scan statistic has been one of the most widely used statistical methods for automatic detection of clusters in spatial data. One limitation of this method lies in the fact that it has to rely on scan windows with predefined shapes in the search process, and therefore it cannot detect cluster with arbitrary shapes. We employ a new neighbor-expanding approach and introduce two new algorithms to detect cluster with arbitrary shapes in spatial data. These two algorithms are called the maximum-likelihood-first (MLF) algorithm and non-greedy growth (NGG) algorithm. We then compare the performance of these two new algorithms with the spatial scan statistic (SaTScan), Tango's flexibly shaped spatial scan statistic (FlexScan), and Duczmal's simulated annealing (SA) method using two datasets. Furthermore, we utilize the methods to examine clusters of murine typhus cases in South Texas from 1996 to 2006.     

Enhancing spatial detection accuracy for syndromic surveillance with street level incidence data

Background  

The Department of Defense Military Health System operates a syndromic surveillance system that monitors medical records at more than 450 non-combat Military Treatment Facilities (MTF) worldwide. The Electronic Surveillance System for Early Notification of Community-based Epidemics (ESSENCE) uses both temporal and spatial algorithms to detect disease outbreaks. This study focuses on spatial detection and attempts to improve the effectiveness of the ESSENCE implementation of the spatial scan statistic by increasing the spatial resolution of incidence data from zip codes to street address level.     

A flexibly shaped spatial scan statistic for detecting clusters

Background

The spatial scan statistic proposed by Kulldorff has been applied to a wide variety of epidemiological studies for cluster detection. This scan statistic, however, uses a circular window to define the potential cluster areas and thus has difficulty in correctly detecting actual noncircular clusters. A recent proposal by Duczmal and Assunção for detecting noncircular clusters is shown to detect a cluster of very irregular shape that is much larger than the true cluster in our experiences.

Methods

We propose a flexibly shaped spatial scan statistic that can detect irregular shaped clusters within relatively small neighborhoods of each region. The performance of the proposed spatial scan statistic is compared to that of Kulldorff's circular spatial scan statistic with Monte Carlo simulation by considering several circular and noncircular hot-spot cluster models. For comparison, we also propose a new bivariate power distribution classified by the number of regions detected as the most likely cluster and the number of hot-spot regions included in the most likely cluster.

Results

The circular spatial scan statistics shows a high level of accuracy in detecting circular clusters exactly. The proposed spatial scan statistic is shown to have good usual powers plus the ability to detect the noncircular hot-spot clusters more accurately than the circular one.

Conclusion

The proposed spatial scan statistic is shown to work well for small to moderate cluster size, up to say 30. For larger cluster sizes, the method is not practically feasible and a more efficient algorithm is needed.

     

Spatial analysis of campylobacter infection in the Canadian province of Manitoba

Background  

The study describes population level variations in campylobacter incidence within the Canadian province of Manitoba, and the relationship to sociodemographic and landscape related characteristics. Using data derived from the Manitoba Health Public Health Branch communicable disease surveillance database, the study applied a number of spatial and ecological techniques to visualize, explore and model campylobacter incidence for the years 1996 to 2004. Analytical techniques used in the study included spatial smoothing, the spatial scan statistic, the Gini coefficient, and Poisson regression analysis.     

A simulation study of three methods for detecting disease clusters

Background  

Cluster detection is an important part of spatial epidemiology because it can help identifying environmental factors associated with disease and thus guide investigation of the aetiology of diseases. In this article we study three methods suitable for detecting local spatial clusters: (1) a spatial scan statistic (SaTScan), (2) generalized additive models (GAM) and (3) Bayesian disease mapping (BYM). We conducted a simulation study to compare the methods. Seven geographic clusters with different shapes were initially chosen as high-risk areas. Different scenarios for the magnitude of the relative risk of these areas as compared to the normal risk areas were considered. For each scenario the performance of the methods were assessed in terms of the sensitivity, specificity, and percentage correctly classified for each cluster.     

Spatio-temporal cluster analysis of county-based human West Nile virus incidence in the continental United States

Background  

West Nile virus (WNV) is a vector-borne illness that can severely affect human health. After introduction on the East Coast in 1999, the virus quickly spread and became established across the continental United States. However, there have been significant variations in levels of human WNV incidence spatially and temporally. In order to quantify these variations, we used Kulldorff's spatial scan statistic and Anselin's Local Moran's I statistic to uncover spatial clustering of human WNV incidence at the county level in the continental United States from 2002–2008. These two methods were applied with varying analysis thresholds in order to evaluate sensitivity of clusters identified.     

Spatial and Temporal Algorithm Evaluation for Detecting Over-The-Counter Thermometer Sale Increases during 2009 H1N1 Pandemic

Background

Spatial outbreak detection algorithms using routinely collected healthcare data have been developed since the late 90s to identify and locate disease outbreaks. However, current well-received spatial algorithms assume only one outbreak cluster present at the same point of time which may not be valid during a pandemic when several clusters of geographic areas concurrently occur. Based on a retrospective evaluation on time-series and spatial algorithms, this paper suggests that time series analysis in detection of pandemics is still a desirable process, which may achieve more sensitive performance with better timeliness.

Methods

In this paper, we first prove in theory that two existing spatial models, the likelihood ratio and the Bayesian spatial scan statistics, are not useful if multiple clusters occur at the same point of time in different geographic regions. Then we conduct a comparison between a spatial algorithm, the Bayesian Spatial Scan Statistic (BSS), and a time series algorithm, the wavelet anomaly detector (WAD), on the performance of detecting the increase of the over-the-counter (OTC) medicine sales during 2009 H1N1 pandemic.

Results

The experiments demonstrated that the Bayesian spatial algorithm responded to the increase of thermometer sales about 3 days later than the time series algorithm.

Conclusion

Time-series algorithms demonstrated an advantage for early outbreak detection, especially when multiple clusters occur at the same time in different geographic regions. Given spatial-temporal algorithms for outbreak detection are widely used, this paper suggests that epidemiologists or public health officials would benefit by applying time series algorithms as a complement to spatial algorithms for public health surveillance.     

Spatial and temporal patterns of enzootic raccoon rabies adjusted for multiple covariates

Background  

With the objective of identifying spatial and temporal patterns of enzootic raccoon variant rabies, a spatial scan statistic was utilized to search for significant terrestrial rabies clusters by year in New York State in 1997–2003. Cluster analyses were unadjusted for other factors, adjusted for covariates, and adjusted for covariates and large scale geographic variation (LSGV). Adjustments were intended to identify the unusual aggregations of cases given the expected distribution based on the observed locations.     

Tuberculosis incidence in Portugal: spatiotemporal clustering

Background  

The statistics of disease clustering is one of the most important tools for epidemiologists to detect and monitor public health disease patterns. Nowadays, tuberculosis (TB) – an infectious disease caused by the Mycobacterium tuberculosis – presents different (development in populations and antibiotics resistance) patterns and specialists are very concerned with it and its association to several other diseases and factors. Each year, tuberculosis kills about three million people in the world. In particular, it is responsible for the death of more than one-third of HIV-infected people, who prove particularly susceptible due to a decline in their immune defences. The purpose of this study is to determine if there are spatiotemporal tuberculosis incidence clusters in continental Portugal. The presented case study is based on the notification of new tuberculosis cases (disease incidence), between 2000 and 2004. In methodological terms, the spatial scan statistic, used to identify spatiotemporal clusters, was improved by including two new approaches: definition of window sizes in the cluster scanning processes considering empirical mean spatial semivariograms and an independent and posterior validation of identified clusters (based on geostatistical simulations).     

Penalized likelihood and multi-objective spatial scans for the detection and inference of irregular clusters

Background  

Irregularly shaped spatial clusters are difficult to delineate. A cluster found by an algorithm often spreads through large portions of the map, impacting its geographical meaning. Penalized likelihood methods for Kulldorff's spatial scan statistics have been used to control the excessive freedom of the shape of clusters. Penalty functions based on cluster geometry and non-connectivity have been proposed recently. Another approach involves the use of a multi-objective algorithm to maximize two objectives: the spatial scan statistics and the geometric penalty function.     

Investigation of geo-spatial hotspots for the occurrence of tuberculosis in Almora district, India, using GIS and spatial scan statistic

Background  

The World Health Organization has declared tuberculosis a global emergency in 1993. It has been estimated that one third of the world population is infected with Mycobacterium tuberculosis, the causative agent of tuberculosis. The emergence of TB/HIV co-infection poses an additional challenge for the control of tuberculosis throughout the world. The World Health Organization is supporting many developing countries to eradicate tuberculosis. It is an agony that one fifth of the tuberculosis patients worldwide are in India. The eradication of tuberculosis is the greatest public health challenge for this developing country. The aim of the present population based study on Mycobacterium tuberculosis is to test a large set of tuberculosis cases for the presence of statistically significant geographical clusters. A spatial scan statistic is used to identify purely spatial and space-time clusters of tuberculosis.     

Adaptations for finding irregularly shaped disease clusters

Background  

Recent adaptations of the spatial scan approach to detecting disease clusters have addressed the problem of finding clusters that occur in non-compact and non-circular shapes – such as along roads or river networks. Some of these approaches may have difficulty defining cluster boundaries precisely, and tend to over-fit data with very irregular (and implausible) clusters shapes.     

A space-time analysis of the proportion of late stage breast cancer in Massachusetts, 1988 to 1997

Background  

Spatial variation in patterns of disease outcomes is often explored with techniques such as cluster detection analysis. In other types of investigations, geographically varying individual or community level characteristics are often used as independent predictors in statistical models which also attempt to explain variation in disease outcomes. However, there is a lack of research which combines geographically referenced exploratory analysis with multilevel models. We used a spatial scan statistic approach, in combination with predicted block group-level disease patterns from multilevel models, to examine geographic variation in prostate cancer grade and stage at diagnosis.     

A CUSUM framework for detection of space-time disease clusters using scan statistics

Several methods for timely detection of emerging clusters of diseases have recently been proposed. We focus our attention on one of the most popular types of method; a scan statistic. Different ways of constructing space-time scan statistics based on surveillance theory are presented. We bridge the ideas from space-time disease surveillance, public health surveillance and industrial quality control and show that previously suggested space-time scan statistics methods can be fitted into a general CUSUM framework. Crucial differences between the methods studied are due to different assumptions about the spatial process. An example is the specification of the spatial regions of interest for a possible cluster, another is the increased rate to be detected within a cluster. We evaluate the detection ability of the methods considering the possibility of a cluster emerging at any time during the surveillance period. The methods are applied to the detection of an increased incidence of Tularemia in Sweden.     

Small-area distribution of multiple sclerosis incidence in western France: in search of environmental triggers

Background

Despite intensive research over several decades, the etiology of multiple sclerosis (MS) remains poorly understood, although environmental factors are supposedly implicated. Our goal was to identify spatial clusters of MS incident cases at the small-area level to provide clues to local environmental risk factors that might cause or trigger the disease.

Methods

A population-based and multi-stage study was performed in the French Brittany region to accurately ascertain the clinical onset of disease during the 2000–2004 period. The municipality of residence at the time of clinical onset was geocoded. To test for the presence of MS incidence clusters and to identify their approximate locations, we used a spatial scan statistic. We adjusted for socioeconomic deprivation, known to be strongly associated with increased MS incident rates, and scanned simultaneously for areas with either high or low rates. Sensitivity analyses (focusing on relapsing-remitting forms and/or places of residence available within the year following clinical onset) were performed.

Results

A total of 848 incident cases of MS were registered in Brittany, corresponding to a crude annual incidence rate of 5.8 per 100,000 inhabitants. The spatial scan statistic did not find a significant cluster of MS incidence in either the primary analysis (p value ≥ 0.56) or in the sensitivity analyses (p value ≥ 0.16).

Conclusion

The findings of this study indicate that MS incidence is not markedly affected across space, suggesting that in the years preceding the first clinical expression of the disease, no environmental trigger is operative at the small-area population level in the French Brittany region.

     

Power evaluation of disease clustering tests

Background  

Many different test statistics have been proposed to test for spatial clustering. Some of these statistics have been widely used in various applications. In this paper, we use an existing collection of 1,220,000 simulated benchmark data, generated under 51 different clustering models, to compare the statistical power of several disease clustering tests. These tests are Besag-Newell's R, Cuzick-Edwards' k-Nearest Neighbors (k-NN), the spatial scan statistic, Tango's Maximized Excess Events Test (MEET), Swartz' entropy test, Whittemore's test, Moran's I and a modification of Moran's I.     

Geovisual analytics to enhance spatial scan statistic interpretation: an analysis of U.S. cervical cancer mortality

Background  

Kulldorff's spatial scan statistic and its software implementation – SaTScan – are widely used for detecting and evaluating geographic clusters. However, two issues make using the method and interpreting its results non-trivial: (1) the method lacks cartographic support for understanding the clusters in geographic context and (2) results from the method are sensitive to parameter choices related to cluster scaling (abbreviated as scaling parameters), but the system provides no direct support for making these choices. We employ both established and novel geovisual analytics methods to address these issues and to enhance the interpretation of SaTScan results. We demonstrate our geovisual analytics approach in a case study analysis of cervical cancer mortality in the U.S.     

Geographic distribution modeling and spatial cluster analysis for equine piroplasms in Greece

Maximum entropy ecological niche modeling and spatial scan statistic were utilized to predict the geographic range and to investigate clusters of infections for equine piroplasms in Greece, using the Maxent and SaTScan programs, respectively. The eastern half of the country represented the culminating area with high probabilities (p > 0.75) of presence of equine piroplasms and encompassed most regions with high concentration of equid host populations. The most important environmental factor that contributed to the ecological niche modeling was land cover followed by temperature. Significant clusters (p < 0.0001) were detected for Babesia caballi and Theileria equi infections in North and Central regions of Greece, respectively, which have significant equine populations. Maximum entropy ecological niche modeling and spatial scan statistic have proved to be useful tools for the surveillance of animal diseases.