Article Text

Download PDFPDF

Location, location, location: what can geographic information science (GIS) offer sexual health research?
  1. I Simms1,
  2. M Gibin2,
  3. J Petersen3
  1. 1HIV&STI Department, Health Protection Services, Public Health England, UK
  2. 2Department of Geography, Environment and Development Studies, Birkbeck College, University of London, London, UK
  3. 3Institute for Social & Economic Research, University of Essex, London, UK
  1. Correspondence to Dr I Simms, HIV&STI Department, Health Protection Services, Public Health England, NW9 5EQ, UK; ian.simms{at}

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Person, time, place is a mantra recited often when describing infectious disease epidemiology. To date sexual health research has been largely focused on individual demographic characteristics, sexual networks and behaviours. However, location is increasingly recognised as an important determinant of incidence and studies that include geographic information systems (GIS) techniques are being published in Sexually Transmitted Infections and elsewhere.1 Blanchard and Aral suggested that an individual's sexual health outcomes are highly dependent on those of others and the local environment, an influence exerted through sexual networks, risk behaviours and systems based around social interaction, education, microbiological factors, as well as healthcare provision and access behaviour.2 Spatial variation in the epidemiology of sexual health has been widely described and the success of intervention strategies is seen as being dependent on ‘knowing your epidemic, globally and locally’.3 GIS techniques allow researchers to explore characteristics that make locations inherently similar or unique. Here we focus on the local level as we consider how geospatial analysis can enrich our knowledge of the epidemiology and public health impact of sexual health, and consider the ethical, governance and technical challenges associated with this rapidly emerging field.

GIS integrate software and hardware to digitally capture, manage, analyse and visualise geographically referenced data but the term is also used to describe any analysis which includes location. Such analyses are not new: one of the iconic examples of spatial analysis is John Snow's map of a London cholera outbreak (1854).4 In 2014, the visual communication of geospatial information through interactive web based resources, infographics or traditional maps are commonplace and the technical resources that underpin them are within the reach of most researchers. Introductory texts and a wealth of high quality, detailed surveillance data sets that capture location specific information are available, such as the genitourinary medicine clinic activity data set, together with open access geodemographic data.5 Specialist open source GIS software that combine georeferenced outcome and attribute (covariate) data, such as Quantum GIS (OSGeo Project) and Kulldorf's software for spatial, temporal and space-time scan statistics (SaTScan), are also available.6 Sexual health researchers have used geospatial analytical techniques for a variety of purposes: to map and assess the geographical distribution of infection; to plan, improve and evaluate sexual health services; and to identify clusters of sexually transmitted infections (STI) in space and time within the context of covariates. Nakaya et al7 visualised spatiotemporal models of trends in the Japanese HIV epidemic using circular cartograms. Location has a variety of implied geospatial features such as place of residence, services visited, educational context, social and sexual networks, and opportunities for clinical care and Huntington et al8 provided insight into the complex, context driven decision making process that patients infected with HIV negotiate when selecting HIV clinical services in England. Le Polain de Waroux et al9 used genitourinary medicine clinic activity data set data to analyse factors associated with travelling to non-local genitourinary medicine clinics for gonorrhoea care in London, information that could be used to help design services to meet local population needs and facilitate prompt, equitable access to care. In Baltimore, core groups of gonococcal infection were identified through geodemographic profiling of geocoded residential addresses of gonococcal cases.10 Tanser et al11 investigated fine scale geographical variations in HIV prevalence and clustering of infections in a high prevalence, rural population in KwaZulu-Natal, geolocating the 12 221 study participants to homestead residence (small rural settlement). Although the study area population is characterised by a late stage, generalised HIV epidemic, working at high geographical resolution revealed heterogeneity not seen at higher geographies. This allowed investigators to identify localised HIV epidemics of varying intensities contained within geographically defined communities bordering a national road. The space-time scan statistics geospatial analysis also highlighted a need for interventions that target treatment and care programmes to local communities at greatest risk in addition to the measures aimed at the general population. In addition, the study shows that choice of the optimum scale is crucial to analyses that investigate geographical processes. Administrative boundaries are frequently used within analyses but as these reflect physical and geopolitical structures their use can result in a loss of analytical discrimination and bias at low levels of geography. Postcode level data or global position may appear to be a solution but their use has to be weighed against risk of disclosing the identity of individuals. Nevertheless, the Southwark Health Atlas study of teenage pregnancy showed that detailed locally focused analyses which maintain confidentiality can be produced.12 Here health service data sets for the London Borough of Southwark (population approximately 270 000) were pseudonymised and geocoded to higher geographies and linked to small area geographical information. Highest densities of teenage conceptions adjusted for population size together with local schools and general practices were identified using a cluster detection algorithm and percentage volume contours (Kernel Density Estimation). This information formed the basis of a local evidence-based strategy which was successful in engaging local schools and communities within high-risk areas.

Interpreting data at high resolution can be problematic and contrasting analyses undertaken by geopolitical unit and geospatial location as part of the Southwark Health Atlas illustrates these problems. Clusters were identified through geospatial analysis not seen at ward level, an administrative unit of around 5500 people. This is an example of the ecological fallacy: the assumption that an individual's characteristics are the same as those for a group to which the individual belongs. The modifiable areal unit problem can occur when data are aggregated to different areal units, such as government administrative areas or institutional boundaries.13 Distinctly different measures can be produced due to variations in denominator populations, risk groups and the influence of surrounding areas. The modifiable areal unit problem needs to be considered when investigating local data but can be minimised by working at high resolution. Maps and other outputs would then need to mask an individual's exact location using techniques such as smoothing or cluster detection. The balance between the need to analyse and present small area level health and social data while retaining individual confidentiality is the subject of ongoing debate within information governance standards.14 Data custodians need to use their organisations’ existing disclosure rules as the basis for deciding on the level of geographical identifiers that can be released.

Spatial-temporal analyses are becoming increasingly feasible as software and hardware are developed but remain resource intensive and time consuming when detailed data sets are used. For conventional data sets time of event could be captured through administrative processes or survey questionnaire but the adoption of Web 2.0 internet platforms, in which the user is the provider of website content, has opened the opportunity for novel geospatial analysis. Since most internet users collect and provide data through smartphones, tablet computers and other global positioning system capable mobile devices the volunteered geographical information that is captured is accurately georeferenced allowing spatial-temporal analysis. For example, mobile applications such as GrindR, a geosocial networking application geared towards men who have sex with men, allow users to locate other men who have sex with men within close proximity in time and space. The person/time/location information available from such internet platforms represents a new source of data for the investigation of social and sexual interaction. However, since volunteered geographical information is not created by authoritative sources methodologies need to be developed which allow data extraction, analysis and interpretation that are consistent with ethical standards.

The relationship between socioeconomic status, health inequality and location over time emphasised in ‘Fair society, healthy lives’ has been taken up by Public Health England in its aim of improving health by influencing people at the local level.15 GIS techniques complement these aspirations, providing a new dimension with which to develop our understanding of the interplay between individual behaviours, and social and sexual networks that shape the epidemiology and control of STIs and HIV.2



  • Contributors This editorial was cowritten and all drafts reviewed by all authors.

  • Competing interests None.

  • Provenance and peer review Commissioned; internally peer reviewed.