This article is part of the supplement: Proceedings of the Joint Environment and Human Health Programme: Annual Science Day Conference and Workshop

Research

Evidence-based selection of environmental factors and datasets for measuring multiple environmental deprivation in epidemiological research

Elizabeth A Richardson^1*, Richard J Mitchell², Niamh K Shortt¹, Jamie Pearce¹ and Terence P Dawson³

• * Corresponding author: Elizabeth A Richardson e.richardson@ed.ac.uk

Author Affiliations

¹ School of GeoSciences, The University of Edinburgh, Drummond Street, Edinburgh EH8 9XP, UK

² Section of Public Health and Health Policy, University of Glasgow, 1 Lilybank Gardens, Glasgow G12 8RZ, UK

³ School of Geography, University of Southampton, Highfield, Southampton SO17 1BJ, UK

For all author emails, please log on.

Environmental Health 2009, 8(Suppl 1):S18 doi:10.1186/1476-069X-8-S1-S18

The electronic version of this article is the complete one and can be found online at: http://www.ehjournal.net/content/8/S1/S18

Published:

21 December 2009

© 2009 Richardson et al; licensee BioMed Central Ltd.

This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

This Environment and Human Health project aims to develop a health-based summary measure of multiple physical environmental deprivation for the UK, akin to the measures of multiple socioeconomic deprivation that are widely used in epidemiology. Here we describe the first stage of the project, in which we aimed to identify health-relevant dimensions of physical environmental deprivation and acquire suitable environmental datasets to represent population exposure to these dimensions at the small-area level. We present the results of this process: an evidence-based list of environmental dimensions with population health relevance for the UK, and the spatial datasets we obtained and processed to represent these dimensions. This stage laid the foundations for the rest of the project, which will be reported elsewhere.

Introduction

The physical environment is multifactorial. Environmental factors with health relevance range from pathogenic (i.e., with potential to damage health), to salutogenic (i.e., with potential to enhance or maintain health). Constrained by data limitations and availability, studies of the health effects of the environment often focus on a subset of the environment in isolation [e.g., air pollution: [1]]. However, populations are not exposed to single environmental factors in isolation: they simultaneously experience multiple exposures. Evans and Kantrowitz [p.304; [2]] suggest that "... multiple exposures to a plethora of suboptimal environmental conditions" may help explain socioeconomic inequalities in health. Different environmental exposures may have additive, synergistic or antagonistic effects on health when experienced in combination [3], hence identifying areas experiencing multiple environmental deprivation may assist in clarifying environment and health relationships and our understanding of health inequalities.

Composite indicators, or indices, are used in other disciplines to present information from multiple variables in an understandable and usable form [4]. A UK example is the Carstairs score [5], which summarises four elements of multiple socioeconomic deprivation: material possessions, employment, living conditions and social class. Socioeconomic deprivation indices such as the Carstairs score are widely used in epidemiology, and have greatly facilitated research into the relationships between socioeconomic deprivation and health [6].

There is a growing need for measures that summarise environmental influences on health in a meaningful way, in order to inform policy-making and interventions [4]. Environmental summary measures have been trialled elsewhere (e.g., [7,8]), but none have sought to summarise multiple environmental deprivation in a specifically health-relevant way. A carefully constructed measure summarising health-relevant aspects of the physical environment could help improve our understanding of the importance of environmental determinants of health.

Our larger project aimed to quantify the overall health-related environmental burden faced by a population, by developing a health-based summary measure of multiple physical environmental deprivation for the UK and determining its utility. Here we describe the first stage of the project, in which we aimed to identify health-relevant dimensions of physical environmental deprivation and acquire suitable environmental datasets. Due to word constraints we summarise key decisions made and our justification for them. We also present the results of this process: an evidence-based list of environmental dimensions with population health relevance for the UK, and the spatial datasets we obtained and processed to represent these dimensions. Subsequent methodological steps, analysis and interpretation that builds on the work described here will be presented elsewhere.

Methods

Stage 1: Identifying health-relevant dimensions of environmental deprivation

A summary measure of health-related physical environments should include only factors with a clear association with health [4], and to which substantial numbers of people are exposed. We therefore began by identifying distinct dimensions of physical environmental deprivation that were public health-relevant and quantifiable for the UK. A crucial first step was defining the 'physical environment': we decided to include external physical, chemical and biological factors and exclude social and cultural factors.

We documented the reasons for decisions made during the process in order to maximise transparency and promote replicability. We initially conducted a scoping review to guide selection of health-relevant environmental factors. To avoid overlooking less commonly-researched factors we consulted a diverse range of sources. A long list of potentially relevant environmental factors was produced using a wide range of international academic and grey literature and by browsing titles returned from general 'environment + health' searches of publication databases (PubMed, WebOfKnowledge and GeoBase). For public health relevance we required that at least 10% of the UK population was exposed to each environmental factor; the factors for which this threshold would not be met were excluded.

We then systematically searched publication databases for empirical studies that had explored the health impacts of one or more environmental factors on the long list. Search terms were derived from the long list. The WebOfKnowledge database allowed us to order search results by citation scores, ensuring that key references were not overlooked. Additionally, reference lists of papers were manually browsed to locate further studies. Non-English language studies and those pre-1980 were excluded. Searching was halted for each environmental factor when a saturation point was reached, i.e., when no novel results were being returned.

The assembled evidence was appraised by the project team during group discussions, based on prevalence of the health outcome(s), rigour of the study design, and the strength of association established. Spatial datasets were sought to enable assessment of population exposure to factors when exceedance of the 10% threshold was not apparent from the literature. The end-product of this stage was a wish-list of health-relevant environmental factors we would want to include in our summary measure.

Stage 2: Dataset acquisition and processing

To maximise future utility and reproducibility of the measure the datasets used would ideally be readily available, routinely updated, representative of the environmental factors of concern and of an acceptable and comparable quality [7,9]. For each environmental factor on our wish-list we therefore sought data that were spatially contiguous, comprehensive across the UK and centred around 2001, to correspond with the decennial census which would be our source of denominator data for subsequent testing of the summary measure's utility. The environmental data needed to be fit for the purpose of reliably representing long-term exposure to each factor.

We selected UK 2001 Census Area Statistics (CAS) wards as our geographical unit of analysis. There were 10,654 CAS wards in the UK at the 2001 census, with an average population of approximately 5,500. Using the geographical information system software ArcMap (ESRI Inc., Redlands, CA) we rendered each environmental dataset to the 2001 CAS wards.

Results

We long-listed 13 environmental factors and appraised the evidence for i) their association with health outcomes using the international literature and ii) their relevance to population health in the UK context. Consequently, seven factors were included in our wish-list for our summary measure of environmental deprivation (Table 1). Table 1 briefly outlines some key epidemiological evidence for the health associations of each wish-listed factor (from meta-analyses where available), although the full evidence review for each factor was more comprehensive than can be reported here. Six environmental factors were excluded (Table 2) as a result of this evidence appraisal process.

We attempted to obtain UK-wide datasets representative of the environment in 2001 (the date for which reliable population data are available) for the seven wish-listed factors. However, data reliability and availability issues meant that we were unable to obtain suitable information pertaining to drinking water quality and noise. For the five remaining environmental factors, the UK-wide datasets obtained and the ward-level measures derived from these data are detailed in Table 3. Mapping each measure in a geographical information system (GIS) (Figure 1) confirmed that the datasets represented expected geographical trends (e.g., higher pollution in urban areas, and higher average temperatures towards the south and east).

Discussion

We aimed to identify health-relevant dimensions of physical environmental deprivation and acquire suitable environmental datasets. Guided by accepted principles for the design of composite indicators [4], we have documented the rationale behind our decisions throughout, to ensure transparency and repeatability.

The summary measures we will create and test during subsequent phases of the project will only be as good as the datasets we have used, hence we sought the most reliable data available. We were only able to obtain reliable and contiguous data for five of the seven environmental factors on our wish-list, but the list remains as a useful by-product of the process. The excluded wish-list factors (drinking water quality and noise) could be included in future attempts at summarising multiple environmental deprivation should suitable datasets become available. Furthermore, based on evidence of effect size we judged that the health impacts of the two excluded factors were unlikely to be as substantial as, say, air pollution, and hence anticipated that a useful composite indicator could be constructed using the remaining five factors.

Additionally, the acquired datasets had limitations. To ensure the future utility and replicability of the work we prioritised datasets that were readily available and likely to be routinely updated in the future. The pollutant and meteorological datasets selected for meeting these criteria were relatively coarse grids (1 and 5 km), therefore they may not adequately represent individual-level exposure in each ward. Although finer resolution data may have helped alleviate these exposure misclassification issues it is less likely that any such data, if available, would be readily available for the whole UK and/or routinely updated.

The preparation of separate datasets by the four countries of the UK, sometimes to different standards, also proved problematic. In particular the detailed land use mapping product (Generalised Land Use Database, GLUD) was only available for England, whereas the coarser resolution dataset (Coordination of Information on the Environment, CORINE) was available for the whole UK. For industrial facility locations the readily available EPER database (EU-wide) usefully combined all UK data. Whilst this circumvented the need for us to acquire different datasets from separate organisations, and would aid future reproducibility of the work, EPER data from the first reporting year (2001) are known not to be fully comprehensive for all facilities in all countries [10]. Future work may therefore have more comprehensive data to incorporate.

Where necessary our environmental datasets were averaged to give an indication of long-term exposure in each ward. However, short-term variations in some environmental factors have health-relevance (e.g., air pollutant peaks, extreme low temperature events). Implicit in our approach, therefore, is the assumption that the severity of the short-term extremes correlates well with longer-term averages.

Conclusion

We have demonstrated that it is possible to identify health-relevant dimensions of the physical environment and acquire suitable datasets to represent variation in these dimensions across the UK. In doing so we have laid the foundations for the rest of the project: constructing a summary measure of multiple environmental deprivation for the UK and determining its utility in researching spatial inequalities in health. The process and its limitations have been transparently described, to aid further work and the informed use of its outputs.

List of abbreviations used

AEA: Atomic Energy Authority; CA: California State; CAS: Census Area Statistics; CI: Confidence interval; CORINE: Coordination of Information on the Environment; CVD: Cardiovascular disease; DBP: Disinfection By-Products; ELF: Extremely Low Frequency radiation; EPER: European Pollutant Emission Register; ESRI: Environmental Systems Research Institute; EU: European Union; GIS: Geographical Information System; GLUD: Generalised Land Use Database; HR: Hazards ratio; NAEI: National Atmospheric Emissions Inventory; OR: Odds ratio; RD: Respiratory disease; RF: Radio Frequency radiation; RR: Relative risk; UK: United Kingdom; UV: Ultraviolet.

Competing interests

The authors declare that they have no competing interests.

Authors' contributions

ER prepared the reviews and drafted the manuscript. ER acquired and processed datasets, assisted by TD. RM conceived of the study. All authors helped collate literature. All authors participated in design and coordination of the study, and helped review the empirical evidence. All authors read and approved the final manuscript.

Note

The peer review to this article can be found in Additional file 1.

Acknowledgements

We would like to thank two anonymous reviewers for their constructive comments. The work was supported by the Joint Environment & Human Health Programme (NERC, EA, Defra, MOD, MRC, The Wellcome Trust, ESRC, BBSRC, EPSRC and HPA), grant number NE/E008720/2.

This article has been published as part of Environmental Health Volume 8 Supplement 1, 2009: Proceedings of the Joint Environment and Human Health Programme: Annual Science Day Conference and Workshop. The full contents of the supplement are available online at http://www.ehjournal.net/supplements/8/S1.

References