Spatial cluster detection using dynamic programming

The potential existence of spatial clusters in childhood cancer incidence is a debated topic. Identification of rare disease clusters in general may help to better understand disease etiology and develop preventive strategies against such entities.The incidence of newly diagnosed childhood malignancies under 15 years of age is 140/1,000,000. In this context, the subgroup of nephroblastoma represents an extremely rare entity with an annual incidence of 7/1,000,000. We evaluated widely used statistical approaches for spatial cluster detection in childhood cancer (Ref. [22] Schündeln et al., 2021, Cancer Epidemiology). For the simulation study, random high risk clusters of 1 to 50 adjacent districts (NUTS-level 3, nomenclature des unités territoriales statistiques) were generated on the basis of the 402 German administrative districts. Each cluster was simulated with different relative risk levels (1 to 100). For each combination of cluster size and risk level 2000 iterations were performed. Simulated data was then analyzed by three local clustering tests: Besag-Newell method, spatial scan statistic and the Bayesian Besag-York-Mollié approach (fit by Integrated Nested Laplace Approximation). The performance characteristics of all three methods were systematically documented (sensitivity, specificity, positive/negative predictive values, exact- and minimum power, correct classification, positive/negative diagnostic likelihood and false positive/negative rate).This data article links to a Mendeley online repository which includes the raw data of simulated high-risk clusters and simulated cases on the district level for an all-childhood-malignancy scenario as well as for cases of nephroblastoma. These data was used for the evaluation of the three cluster detection methods. The R code for simulation and analysis are available from GitHub.The article also includes analyzed data summarizing the performance of the cluster detection tests in very rare disease entities, using the example of simulated nephroblastoma cases.The raw data from the study can be used for benchmarking analyses applying different spatial statistical methods systematically and evaluating their performance characteristics comparatively. The analyzed data from the nephroblastoma example can be useful to interpret the performance of the three applied local cluster detection tests in the setting of extremely rare disease entities. As a practical application, data and R code can be used for performance analyses when planning to establish surveillance systems for rare disease entities.

Spatial cluster detection has been employed to identify significant connected spatial clusters in a variety of crowdsourcing applications in bioinformatics, social networks, and network security. However, despite of providing tremendous social benefits, the release of crowdsourced data will pose a considerable threat to mobile users’ privacy. Differential privacy has been adopted for privacy preservation of crowdsourced data, yet the effective mining of protected data becomes challenging. In this paper, we investigate the problem of spatial cluster detection on privacy-preserving crowdsourced data. Specifically, we propose PSCluster, a differentially private spatial cluster detection mechanism to provide services with data protected under a differential privacy model. PSCluster’s key components include a hypothesis testing framework for modeling differentially private spatial data, and a hybrid algorithm that jointly conducts expectation maximization (EM)-based parameter estimation and free-form spatial cluster detection. We evaluate PSCluster over synthetic data and real data from mobile crowdsourcing applications, and compare PSCluster with two state-of-the-art baseline methods. The experimental results show that PSCluster improves the utility without sacrificing privacy.

Regularized spatial and spatio-temporal cluster detection

The detection of spatial clusters has been suggested as a potential tool for guiding the delivery of interventions against visceral leishmaniasis; however, little attention has been given to the consistency of results by using different spatial clustering methods. The present study aimed to assess the performance of three different techniques for identifying patterns in the spatial distribution of canine leishmaniasis in the city of Teresina, Brazil. This cross-sectional study was based on a serological survey for canine leishmaniasis in which each dog domicile was georeferenced to the exact location of each animal in space. The spatial analysis was performed using three methods: the Cuzick-Edwards statistic, the Hierarchical Nearest Neighbour analysis, and the Kulldorff Scan statistic. All techniques were able to identify clusters of high prevalences of canine leishmaniasis, but results were not consistent among techniques. The feasibility of the identification and location of clusters of cases in a restricted number of villages in neighborhoods might contribute to the optimization of control measures against visceral leishmaniasis. However, given the relative inconsistency of the results provided by the different methods, protocols for assessing clusters of diseases should always include more than one method of evaluation.

Objective to analyze the spatial-temporal distribution and factors associated with HIV/AIDS mortality among young people in the Northeast from 2001 to 2020. Method ecological study with 2,509 deaths from HIV/AIDS from northeastern residents aged between 10 and 24 years of age, reported in the Mortality Information System. Temporal analysis techniques (Joinpoint) and detection of spatial clusters (Spatial Autocorrelation, Gets-Ord Gi* and Scan were used. Three spatial error and spatial lag (Spatial Error and Spatial Lag) and non-spatial regression models (Ordinary Least Squares-OLS) were used to identify the factors associated with mortality in northeastern municipalities, considering p<0.05. Results the HIV/AIDS mortality rate among young people in the northeast was 0.4 deaths per 100,000 inhabitants. Maranhão (APC:7.1; CI95%:2.3-12.1), Sergipe (PcA:6.9; CI95%:1.8-12.2), Rio Grande do Norte (PcA:6.4; CI95%:1.8-11.2), Ceará (PcA:4.2; CI95%:1.5-7.0) and Alagoas (APC:3.2; CI95%:0.1-6.4) showed a significant increasing trend of deaths. Cluster detection techniques indicated clusters of deaths mainly on the coast of Pernambuco and north-central Maranhão. The indicators proportion of the population in households with density >2 (β=0.012; p<0.001) and per capita transfer of the continued benefit (β=0.000; p<0.001) showed a positive relationship with the outcome. On the other hand, the proportion of extremely poor people (β=-0.011; p=0.029) and the Brazilian Deprivation Index (β=-0.195; p=0.009) were negatively associated with mortality. Conclusion there was a trend of increased mortality in five of the nine northeastern states. The spatial clusters were located mainly in Pernambuco and North-Central Maranhão. Interventions aimed at socioeconomic factors should be established to prevent HIV infections and deaths among young people.

Spatial cluster detection is an important problem in a variety of scientific disciplines such as environmental sciences, epidemiology and sociology. However, there appears to be very limited statistical methodology for quantifying the uncertainty of a detected cluster. In this paper, we develop a new method for the quantification and visualization of uncertainty associated with a detected cluster. Our approach is defining a confidence set for the true cluster and visualizing the confidence set, based on the maximum likelihood, in time or in one‐dimensional space. We evaluate the pivotal property of the statistic used to construct the confidence set and the coverage rate for the true cluster via empirical distributions. For illustration, our methodology is applied to both simulated data and an Alaska boreal forest dataset. Copyright © 2017 John Wiley & Sons, Ltd.

Stepwise and stagewise approaches for spatial cluster detection

Hotspots of change in use of public transport to work: A geospatial mixed method study

Zero-inflated Bell scan: A more flexible spatial scan statistic

This study investigates the relationship between digital inclusive finance (DIF) and foreign direct investment (FDI) in China, using panel data from 31 provinces, municipalities, and autonomous regions from 2011 to 2022. The research employs a dual fixed spatial Durbin model to empirically analyze the impact of DIF on FDI and its spatial effects. The primary issue addressed in the study is how DIF influences FDI, both within a region and across neighboring areas. The main objective of the research is to evaluate the spatial clustering characteristics between DIF and FDI and to determine whether DIF significantly enhances FDI levels in a region. Furthermore, the study aims to understand the "siphon effect," where DIF may attract FDI away from surrounding areas. Methodologically, the dual fixed spatial Durbin model is utilized to capture both the direct and spillover effects of DIF on FDI across regions. The study examines regional heterogeneity, particularly how the impact of DIF varies across eastern, central, and western China. The results reveal that DIF indeed has a spatial clustering distribution with FDI, which can significantly boost FDI within a region. However, it also generates a siphon effect, attracting FDI from surrounding areas. Additionally, the study highlights regional differences, with the positive effect of DIF on FDI diminishing as one moves from the eastern to the western regions. Meanwhile, the negative spatial spillover effect of DIF development on neighboring areas is significant across all regions. Based on these findings, the study recommends tailored policy interventions to address regional disparities and mitigate adverse spillover effects, ensuring balanced development across regions

Spatial clustering has been widely used for spatial data mining and knowledge discovery. An ideal multivariate spatial clustering should consider both spatial contiguity and aspatial attributes. Existing spatial clustering approaches may face challenges for discovering repeated geographic patterns with spatial contiguity maintained. In this paper, we propose a Spatial Toeplitz Inverse Covariance-Based Clustering (STICC) method that considers both attributes and spatial relationships of geographic objects for multivariate spatial clustering. A subregion is created for each geographic object serving as the basic unit when performing clustering. A Markov random field is then constructed to characterize the attribute dependencies of subregions. Using a spatial consistency strategy, nearby objects are encouraged to belong to the same cluster. To test the performance of the proposed STICC algorithm, we apply it in two use cases. The comparison results with several baseline methods show that the STICC outperforms others significantly in terms of adjusted rand index and macro-F1 score. Join count statistics is also calculated and shows that the spatial contiguity is well preserved by STICC. Such a spatial clustering method may benefit various applications in the fields of geography, remote sensing, transportation, and urban planning, etc.

The scan statistic has been a methodology widely applied for spatial and space-time cluster detection. This is because, by identifying areas with a significant concentration of points, we gain evidence for the analysis of underlying phenomena. This article aims to conduct a bibliometric study on spatial scan regression. Spatial scan regression includes correlation between the outcome and potential regressor variables into the cluster detection. The bibliometric study applies a robust literature search methodology, using Web of Science and Scopus as the main databases. Thus, the most relevant articles on the topic (n=458) were selected and analyzed using Bibliometrix, a package developed for the R® statistical programming language, and the VOSViewer software, employing text mining tools. Subsequently, we proceeded with: (i) matrix construction of attributes and (ii) presentation of descriptive analysis of data and network relationships. The sample covers the years from 1993 to 2023 and represents an annual growth rate of 12.36% with an average of 14.77 articles published each year. Despite the Increasing studies on spatial scan regression, having 82% of the publications made since 2013, with emphasis on the period from 2018 to 2023, comprising 255 publications (56% of the total), it is noteworthy that most of them are produced by relatively few countries, with a significant concentration on health-related themes. It is observed that this method has broad potential for application in other areas. Collaboration networks in research still exhibit a structure to be developed, emphasizing the essential need for creating new partnerships between countries and research institutions on this specific topic.

This article provides a review of spatial analysis methods for use in health promotion and education research and practice. Spatial analysis seeks to describe or make inference about variables with respect to the places they occur. This includes geographic differences, proximity issues, and access to resources. This is important for understanding how health outcomes differ from place to place; and in terms of understanding some of the environmental underpinnings of health outcomes data by placing it in context of geographic location. This article seeks to promote spatial analysis as a viable tool for health promotion and education research and practice. Four more commonly used spatial analysis techniques are described in-text. An illustrative example of motor vehicle collisions in a large metropolitan city is presented using these techniques. The techniques discussed are as follows: descriptive mapping, global spatial autocorrelation, cluster detection, and identification and spatial regression analysis. This article provides useful information for health promotion and education researchers and practitioners seeking to examine research questions from a spatial perspective.

Residential address errors in public health surveillance data: A description and analysis of the impact on geocoding

The similarity between maximum entropy (MaxEnt) and minimum relative entropy (MRE) allows recent advances in probabilistic inversion to obviate some of the shortcomings in the former method. The purpose of this paper is to review and extend the theory and practice of minimum relative entropy. In this regard, we illustrate important philosophies on inversion and the similarly and differences between maximum entropy, minimum relative entropy, classical smallest model (SVD) and Bayesian solutions for inverse problems. MaxEnt is applicable when we are determining a function that can be regarded as a probability distribution. The approach can be extended to the case of the general linear problem and is interpreted as the model which fits all the constraints and is the one model which has the greatest multiplicity or “spreadout” that can be realized in the greatest number of ways. The MRE solution to the inverse problem differs from the maximum entropy viewpoint as noted above. The relative entropy formulation provides the advantage of allowing for non-positive models, a prior bias in the estimated pdf and `hard' bounds if desired. We outline how MRE can be used as a measure of resolution in linear inversion and show that MRE provides us with a method to explore the limits of model space. The Bayesian methodology readily lends itself to the problem of updating prior probabilities based on uncertain field measurements, and whose truth follows from the theorems of total and compound probabilities. In the Bayesian approach information is complete and Bayes' theorem gives a unique posterior pdf. In comparing the results of the classical, MaxEnt, MRE and Bayesian approaches we notice that the approaches produce different results. In␣comparing MaxEnt with MRE for Jayne's die problem we see excellent comparisons between the results. We compare MaxEnt, smallest model and MRE approaches for the density distribution of an equivalent spherically-symmetric earth and for the contaminant plume-source problem. Theoretical comparisons between MRE and Bayesian solutions for the case of the linear model and Gaussian priors may show different results. The Bayesian expected-value solution approaches that of MRE and that of the smallest model as the prior distribution becomes uniform, but the Bayesian maximum aposteriori (MAP) solution may not exist for an underdetermined case with a uniform prior.