Barry Rowlingson

Spatial point pattern analysis and its application in geographical epidemiology

This paper reviews a number of methods for the exploration and modelling of spatial point patterns with particular reference to geographical epidemiology (the geographical incidence of disease). Such methods go well beyond the conventional ‘nearest-neighbour’ and ‘quadrat’ analyses which have little to offer in an epidemiological context because they fail to allow for spatial variation in population density. Correction for this is essential if the aim is to assess the evidence for ‘clustering’ of cases of disease. We examine methods for exploring spatial variation in disease risk, spatial and space-time clustering, and we consider methods for modelling the raised incidence of disease around suspected point sources of pollution. All methods are illustrated by reference to recent case studies including child cancer incidence, Burkitt’s lymphoma, cancer of the larynx and childhood asthma. An Appendix considers a range of possible software environments within which to apply these methods. The links to modern geographical information systems are discussed.

SPLANCS: spatial point pattern analysis code in S-Plus

In recent years, Geographical Information Systems have provided researchers in many fields with facilities for mapping and analyzing spatially referenced data. Commercial systems have excellent facilities for database handling and a range of spatial operations. However, none can claim to be a rich environment for statistical analysis of spatial data. We have made some powerful enhancements to the S-Plus system to produce a tool for display and analysis of spatial point pattern data. In this paper we give a brief introduction to the S-Plus system and a detailed description of the S-Plus enhancements. We then present three worked examples: two from geomorphology and one from epidemiology.

A Conditional Approach to Point Process Modelling of Elevated Risk

SUMMARY We consider the problem of investigating the elevation in risk for a specified disease in relation to possible environmental factors. Our starting point is an inhomogeneous Poisson point process model for the spatial variation in the incidence of cases and controls in a designated geographic region, as proposed by Diggle. We develop a conditional approach to inference which converts the point process model to a non-linear binary regression model for the spatial variation in risk. Simulations suggest that the usual asymptotic approximations for likelihood-based inference are more reliable in this conditional setting than in the original point process setting. We present an application to some data on the spatial distribution of asthma in relation to three industrial locations.

Childhood malaria in the Gambia: a case‐study in model‐based geostatistics

The paper develops a spatial generalized linear mixed model to describe the variation in the prevalence of malaria among a sample of village resident children in the Gambia. The response from each child is a binary indicator of the presence of malarial parasites in a blood sample. The model includes terms for the effects of child level covariates (age and bed net use), village level covariates (inclusion or exclusion from the primary health care system and greenness of surrounding vegetation as derived from satellite information) and separate components for residual spatial and non-spatial extrabinomial variation. The results confirm and quantify the progressive increase in prevalence with age, and the protective effects of bed nets. They also show that the extrabinomial variation is spatially structured, suggesting an environmental effect rather than variation in familial susceptibility. Neither inclusion in the primary health care system nor the greenness of the surrounding vegetation appeared to affect the prevalence of malaria. The method of inference was Bayesian using vague priors and a Markov chain Monte Carlo implementation.

Spatial modelling and the prediction of Loa loa risk: decision making under uncertainty.

Abstract Health decision-makers working in Africa often need to act for millions of people over large geographical areas on little and uncertain information. Spatial statistical modelling and Bayesian inference have now been used to quantify the uncertainty in the predictions of a regional, environmental risk map for Loa loa (a map that is currently being used as an essential decision tool by the African Programme for Onchocerciasis Control). The methodology allows the expression of the probability that, given the data, a particular location does or does not exceed a predefined high-risk threshold for which a change in strategy for the delivery of the antihelmintic ivermectin is required.

Spatial and spatio-temporal log-gaussian cox processes: Extending the geostatistical paradigm

In this paper we first describe the class of log-Gaussian Cox processes (LGCPs) as models for spatial and spatio-temporal point process data. We discuss inference, with a particular focus on the computational challenges of likelihood-based inference. We then demonstrate the usefulness of the LGCP by describing four applications: estimating the intensity surface of a spatial point process; investigating spatial segregation in a multi-type process; constructing spatially continuous maps of disease risk from spatially discrete data; and real-time health surveillance. We argue that problems of this kind fit naturally into the realm of geostatistics, which traditionally is defined as the study of spatially continuous processes using spatially discrete observations at a finite number of locations. We suggest that a more useful definition of geostatistics is by the class of scientific problems that it addresses, rather than by particular models or data formats.

Bivariate Binomial Spatial Modeling of Loa loa Prevalence in Tropical Africa

We present a state-of-the-art application of smoothing for dependent bivariate binomial spatial data to Loa loa prevalence mapping in West Africa. This application starts with the nonspatial calibration of survey instruments, continues with the spatial model building and assessment, and ends with robust, tested software intended for use by field workers for online prevalence map updating. From a statistical perspective, we address several important methodological issues: building spatial models that are sufficiently complex to capture the structure of the data but remain computationally usable, reducing the computational burden in the handling of very large covariate data sets, and devising methods for comparing spatial prediction methods for a given exceedance policy threshold.

Short communication: Negative spatial association between lymphatic filariasis and malaria in West Africa

Objective To determine the relationship between human lymphatic filariasis, caused by Wuchereria bancrofti, and falciparum malaria, which are co‐endemic throughout West Africa.

Testing for clustering of health events within a geographical information system framework.

An approach to analysing data for spatial clustering is outlined, with special reference to environmental epidemiology. The method is based on recent developments in spatial point-process modelling; specifically, on the use of so-called ‘second-order’ analysis of bivariate point patterns. This continuous-space approach offers some advantages over analytical methods that aggregate health events to areal units. The method is implemented within the framework of a proprietary geographical information system, ARC/INFO, and is illustrated with reference to health data from a questionnaire survey of children in Preston (Lancashire). The nature of the data gained from the questionnaire means that variables which may affect the health of the children studied can be accounted for within the analysis.

Seasonal Influenza Vaccination amongst Medical Students: A Social Network Analysis Based on a Cross-Sectional Study

Introduction The Chief Medical Officer for England recommends that healthcare workers have a seasonal influenza vaccination in an attempt to protect both patients and NHS staff. Despite this, many healthcare workers do not have a seasonal influenza vaccination. Social network analysis is a well-established research approach that looks at individuals in the context of their social connections. We examine the effects of social networks on influenza vaccination decision and disease dynamics. Methods We used a social network analysis approach to look at vaccination distribution within the network of the Lancaster Medical School students and combined these data with the students’ beliefs about vaccination behaviours. We then developed a model which simulated influenza outbreaks to study the effects of preferentially vaccinating individuals within this network. Results Of the 253 eligible students, 217 (86%) provided relational data, and 65% of responders had received a seasonal influenza vaccination. Students who were vaccinated were more likely to think other medical students were vaccinated. However, there was no clustering of vaccinated individuals within the medical student social network. The influenza simulation model demonstrated that vaccination of well-connected individuals may have a disproportional effect on disease dynamics. Conclusions This medical student population exhibited vaccination coverage levels similar to those seen in other healthcare groups but below recommendations. However, in this population, a lack of vaccination clustering might provide natural protection from influenza outbreaks. An individual student’s perception of the vaccination coverage amongst their peers appears to correlate with their own decision to vaccinate, but the directionality of this relationship is not clear. When looking at the spread of disease within a population it is important to include social structures alongside vaccination data. Social networks influence disease epidemiology and vaccination campaigns designed with information from social networks could be a future target for policy makers.