Figures
Abstract
Background
Previous research has shown that environments with features that encourage walking are associated with increased physical activity. Existing methods to assess the built environment using geographical information systems (GIS) data, direct audit or large surveys of the residents face constraints, such as data availability and comparability, when used to study communities in countries in diverse parts of the world. The aim of this study was to develop a method to evaluate features of the built environment of communities using a standard set of photos. In this report we describe the method of photo collection, photo analysis instrument development and inter-rater reliability of the instrument.
Methods/Principal Findings
A minimum of 5 photos were taken per community in 86 communities in 5 countries according to a standard set of instructions from a designated central point of each community by researchers at each site. A standard pro forma derived from reviewing existing instruments to assess the built environment was developed and used to score the characteristics of each community. Photo sets from each community were assessed independently by three observers in the central research office according to the pro forma and the inter-rater reliability was compared by intra-class correlation (ICC). Overall 87% (53 of 60) items had an ICC of ≥0.70, 7% (4 of 60) had an ICC between 0.60 and 0.70 and 5% (3 of 60) items had an ICC ≤0.50.
Conclusions/Significance
Analysis of photos using a standardized protocol as described in this study offers a means to obtain reliable and reproducible information on the built environment in communities in very diverse locations around the world. The collection of the photographic data required minimal training and the analysis demonstrated high reliability for the majority of items of interest.
Citation: Chow CK, Corsi DJ, Lock K, Madhavan M, Mackie P, Li W, et al. (2014) A Novel Method to Evaluate the Community Built Environment Using Photographs – Environmental Profile of a Community Health (EPOCH) Photo Neighbourhood Evaluation Tool. PLoS ONE 9(11): e110042. https://doi.org/10.1371/journal.pone.0110042
Editor: Marc A. Adams, Arizona State University, United States of America
Received: November 6, 2013; Accepted: September 16, 2014; Published: November 4, 2014
Copyright: © 2014 Chow et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: CK Chow is supported by an Australian National Health and Medical Research Council (NHMRC) Career development Fellowship co-funded by the National Heart Foundation and a Sydney Medical Foundation Chapman Fellowship. S Yusuf is supported by an endowed chair from the Heart and Stroke Foundation of Ontario. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Previous research has shown that environments with features that encourage walking are associated with increased physical activity, with potential to lower rates of obesity. Features of the physical environment that encourage walking include presence of walking paths, traffic lights, and cross-walks that reduce the risks from vehicular traffic; green space, the presence of aesthetic architecture, structures, natural features, mixed-land use and the cleanliness and maintenance of the environment. [1]
A number of instruments now exist to capture the physical environment of communities. The methods used by these can be broadly classified into 3 categories. The first include methods which use geocoded data within Geographical Information Systems (GIS) [2] to analyse relationships between individuals and places. The second uses questionnaires administered to residents of communities to collate information on individual perceptions of the community they live in. [3] The third involves systematic observation or audit, in which a trained individual conducts an assessment of a community by observing and recording information along a pre-specified route. [4] [5]
The three methods described collect complementary data on the physical environment, but the last of these, from systematic observation, is onerous and experience so far has largely been limited to urban environments in North America, Europe and Australia. [6] [7] Consequently, it is not clear how applicable they will be when used in other parts of the world. [8]
We have developed instruments to capture information about the community environment through direct observations (EPOCH1) and surveys of perceptions among individuals residing in communities (EPOCH2). [9], [10] The capture of digital photos of the built environment and off-line analysis of data presented a potentially efficient and complementary mode of environmental assessment. Uses of photos have been embraced to promote social change in community health initiatives, e.g. Photovoice [11] and there are a number of studies that have used new forms of geospatial imaging, such as Google Streets, to capture the health-related aspects of communities, [12] but the use of photos to measure a community's built environment is still in its infancy.
The items included in this instrument were drawn from our earlier review of the literature which we conducted at the time we developed the family of EPOCH instruments. [13] Across the family of EPOCH instruments we aimed to measure all the constructs that we identified as environmental factors actually or potentially associated with cardiovascular disease (CVD). The EPOCH photo instrument mainly captured measures of the neighbourhood environment with respect to walkability; these constructs are summarised in the second table of our previous review. [13] Our review drew on a broad range of literature from many disciplines and included a number of instruments. Those instruments that have been tested for reliability and validity were particularly influential in the development of our method. (Table 1) [4], [14]–[17]. Four of them use a direct assessment method. [16] The fifth, NEWS (Neighborhood Environment Walkability Scale), captures individuals' perceptions of the neighborhood in which they live; in terms of sidewalks, street connectivity, safety, surroundings, and overall satisfaction. Since most of the direct measure instruments need significant human resources for data collection, whereby a researcher is required to walk individual streets to assess traffic, design, connectivity, and accessibility, our aim was to develop a method that could be potentially applicable worldwide and is less onerous with respect to data collection, thereby enabling a rapid and efficient assessment of a community environment for walkability.
The aim of this study was to develop a method to evaluate features of the built environment of communities using a standard set of photos. In this report we describe the method of photo collection, the development of an instrument for analysis, the inter-rater reliability properties of the instrument, and the creation of summary scores. Given the limited geographical scope of many previously described built environment instruments, we also wanted to assess its applicability in a much wider set of countries.
Methods
Ethics statement
The EPOCH instruments were approved by the Hamilton Health Sciences/McMaster Health Sciences Research Ethics board which operates in compliance with the ICH Good Clinical Practice Guidelines and the Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans and Division 5 Health Canada Food and Drug Regulations. No data from human participants was used in the current analyses.
Setting
The study was conducted in a convenience sample of 86 urban and rural communities (Table 2) from 5 countries (Canada, Colombia, Brazil, China and India) that were involved in the Prospective Urban and Rural Epidemiology (PURE) study of cardiovascular risk factors and disease [18] and in which the Environmental Profile of a Community's Health (EPOCH) 1 and 2 tools were developed. [9], [10] Investigators were encouraged to include urban and rural communities from a range of socioeconomic areas that had a diverse range of physical environments.
Photos
Research assistants from each country were trained using a prepared set of slides and a manual explaining the procedures to be used. Photos were taken using a standardized protocol by researchers doing the ‘Community Observation Walk’ element of the EPOCH 1 assessment. [9] In brief the community observation walk took place in the commercial or central shopping district of the community and began at a central location. From this designated ‘start-point’ photos were taken to capture a 360 degree view of the community with a minimum of 5 photos. As illustrated in Figure 1, 4 of these photos were taken from the start point in each of 4 directions and the 5th photo was of the start point, taken from across the street. From some communities we had more than this number of photos and in a few only 4 good quality photos could be assessed. Assessors were instructed as to where to stand to take photos, the views they were to obtain, and how to overlap images to achieve full coverage of a street scene. They were also given basic instructions on camera use, lighting and angle to take photos. In each community, 3 observers conducted the walk and we were sent 3 sets of photos from each community. However as these photos were found to be very similar when reviewed and our focus was on the reliability of external observers to audit the communities, we used only the first set of photos for the current analyses.
This diagram shows how the photos were taken. The cross indicates an intersection. The individual must stand at their start point and take photos of each direction from their start points. (front, 2 sides, back). Then they go across from where they were standing to take a picture of their start point. All the photos must show clear view of the street and roads in the neighbourhood without any cars, buildings or pedestrians blocking the view.
Photo analysis
We scored each set of photos centrally using a standard form developed for this study. We identified the components to include in this form from reviewing literature on existing instruments as described above. The photos were assessed for ease of walking and biking, street type, pedestrian density, traffic, safety features such as the presence and quality of cross walks and median strips, aesthetics of the neighborhood (Appendix S1). The EPOCH Photo Neighbourhood Evaluation Tool (EP-NET) and accompanying manual was used by trained assessors to evaluate photos systematically. The first 32 items of the EP-NET instrument are objective measures, asking about the presence and absence of items and the number of items. Items 33 to 37 are items that ask the photo assessor to give an overall measure with respect to a construct. All items included in the instrument are listed in Table 3. All of the constructs included in the EP-NET instrument are drawn from the literature; however we adapted the exact means of measurement to be appropriate when applied to the medium of photography. For example a number of instruments have items that capture features of neighbourhood beautification. Our literature review indicated that the presence of grass/flowers/trees/public art/interesting features were consistent with the construct of neighbourhood beautification. These could have been captured by self-report by residents or by direct counting from neighbourhood audits. However in EP-NET we included the presence/absence of these items and a semi-quantitative measure of how many of these items were present e.g. none, 1 or 2, some or many. For items 33 to 37, the response scales were based on the NEWS questionnaire. [14]
Statistical analysis
We describe the characteristics of the communities overall and by sub-group (rural/urban and by country/regional grouping) using descriptive statistics. To assess the reliability of extraction of data from photos, each set of photos was scored independently by 3 individual observers based in Hamilton, Canada and trained using a study manual. The inter-rater reliability was assessed using the intra-class correlation coefficient (ICC) statistic, derived from a two-way random effects model. [9] The range of the ICC is between 0.0 and 1.0 and will be higher when less variation is present between observers. We defined a high ICC (greater than 0.75) to indicate good agreement and a low ICC (less than 0.4) to indicate poor agreement. All analyses were performed in Stata version 12.1.
Results
Descriptive characteristics
The three observers' responses on each characteristic are in detailed in Appendix S2. Observers reported the time taken to code a set of photos from a community to be on average 10 to 15 minutes and a maximum of 20 minutes per community. Observers reported higher scores for pedestrian facilities and safety characteristics in Canada compared to other countries and in urban compared to rural communities. For example, observer 1 noted 97.3% of communities in Canada had sidewalks, but this was 33.3% of communities in India. No cross-walks were reported in 10.8% of communities in Canada, but this was 86.7% of communities in India. Scores for beautification/aesthetic features were generally higher in Canada, though there was some variation. For example there were many planted trees in 32.4% of communities in Canada, but this was 15% in Brazil/Colombia, 6.7% in India and 42.9% in China. No man-made landscapes were observed to be present in 21.6% of communities in Canada compared to 60% in Brazil/Colombia, 93.3% of India and 21.4% of China. Scores for community disorder were higher in rural communities. Urban communities scored higher on overall appeal in all countries. (Appendix S2).
Reliability
Table 3 summarizes the inter-rater reliability for each community characteristic evaluated and Table 4 summarizes the reliability measures by regions. The reliability of each item's measure by region is in the ‘Combined table’ in Appendix S2. Overall 77% (46 of 60) items had an ICC of ≥0.75, 17% (10 of 60) had an ICC between 0.60 and 0.74, 0% had an ICC between 0.4 and 0.59 and 5% (3 of 60) items had an ICC ≤0.40. One item, the presence of desert, could not be assessed for reliability as it did not occur in any of the communities included in the sample. Variables with a score of 1.0 were those exhibiting no variation, such as the presence of buildings (Q27), which was universal. The three items with low ICCs were: 1) the number of buses seen in the photos, where some assessors seemed to miss some of them; 2) the presence of greenbelt/forest, which highlighted differing interpretations of what constituted a greenbelt/forest and whether to include it if it was only in the distant background of the photo, and 3) the observer's opinion about suitability for biking, which again revealed different interpretations.
Creation of summary scores
To enable some qualitative examination of face validity, we have created a means of summarizing items into simple scores across 5 domains. For each domain, points were allocated to the main characteristics contributing to components of the domain and the total points summed within domains (Table 5). We described the community as high scoring if they had a score in the top tertile of scores and a low scoring community if they were in the bottom tertile of scores. The domains and their components were: 1) Pedestrian facilities and safety features – presence of sidewalks, completeness and quality of sidewalks, cross walks, and cross walk safety features – Traffic signals/signs, white/coloured lines, different road surface/pacing, and presence of a median and/or grass strip. We separated out Bicycle lanes and quality (presence/absence and quality); 2) Aesthetics/Beatification – natural features, man-made landscaping, street trees, street furniture/public art; 3) Community disorder - presence of litter/garbage, vacant/derelict buildings, buildings poorly maintained; 4) Urbanization/density – street density, vehicle density, density of parked cars; and 5) Overall appeal – assessed using a sum of four Likert scales by the central observers indicating level of agreement for each of the statements ‘neighbourhood is safe/pedestrian friendly,’ ‘streets and sidewalks in the neighbourhood are suitable for walking’, ‘buildings, homes and structures in this neightbourhood are very attractive’ and ‘neighbourhood as a whole is aesthetically appealing’.
As indicated above, the scores were calculated by summing items in each domain, giving roughly equal weighting to each feature. When creating these scores we chose to exclude some items that we had collected in the evaluation tool. We excluded the following items as they did not fit well with existing constructs for all countries studied: Q27 - presence of buildings, Q28 – number of buildings, Q29 - awnings, Q8 - parking lots, Q9- street width. Further, we excluded Q13 – number and types of vehicles, as there were few data points and some of the measures were unreliable or irrelevant for some regions within countries. We excluded the following items because they had poor or borderline measures of reliability, few counts, fitted more than one domain or were interpreted differently in different settings, these were: Q11 obstacles to pedestrian walking, Q32 - building design. We separated bike lanes from the pedestrian safety domain due to few counts in some regions. We also removed the item on bike lanes from the ‘overall appeal’ domain as it was not reliable and there were few counts of bike lanes.
Summary scores results
The mean of 3 observers' summary scores for each domain are tabulated overall and by geographical region in Table 6. In Table 7, the mean scores for all communities are given across the three observers. The urbanization/density score and community disorder score were higher in Brazil/Colombia and the aesthetics/beatification score and pedestrian safety scores were higher in Canada. A sample of photos is shown in Figure 2 and Figure 3 for communities that scored high or low on multiple domains from Canada, Colombia and India.
In these communities from urban Canada and rural Colombia the common high-scoring characteristics are complete sidewalks, several planted trees, traffic signals, and pedestrian traffic signs, well maintained buildings and roads and the presence of street furniture such as benches and street lamps.
These are pictures of 2 low scoring communities overall. You can see in communities from urban Canada and rural India there is partial or no sidewalks, not many crosswalks, not many planted trees or aesthetically pleasing features. In addition the buildings are not well maintained.
Discussion
This report describes a novel method to evaluate the built environment of communities by means of a set of photos taken according to a standardized method and applying a standard data extraction form to analyse each photo set for features of the community's physical environment. This method of community assessment is relatively rapid and involves minimal training of the on-site assessor (photographer). This could be important for studies that are conducted across multiple countries, particularly where there is not yet research expertise with regards to environmental assessment. The high reliability of the majority of measures indicates that measurement analysis is feasible and repeatable. The EPOCH photo instrument complements our previously reported instruments developed to collect environmental data relevant to cardiovascular health in large scale international studies - EPOCH 1 and EPOCH 2. EPOCH 1 is an objective environmental audit tool, in which a trained researcher directly observes and systematically records physical aspects of the environment and includes 5 sections, 1) community characteristics – a checklist of infrastructure and services, 2) Community observation walk – this walk starts at the ‘Start point’ described in this paper and is where the photos are taken 3) tobacco store assessment, 4) grocery store assessment and 5) local restaurant assessment. EPOCH 2 is a questionnaire to capture what participants observe in the community, their awareness of laws, regulations and their opinions about behaviours and laws.
Our findings have face validity because comparisons of measures and summary scores across regions fit directionally with expected constructs. The pedestrian facilities and safety characteristics observed were higher in the urban compared to the rural communities of Canada and there were relatively fewer pedestrian facilities in communities in India, which matches with expectations of individuals (many of the authors of this paper) that have travelled and observed both settings. While higher pedestrian and safety characteristics have been related to increased physical activity, [19] it is hard to ascertain whether the higher pedestrian facility and safety characteristics in Canada compared to other countries is more a function of the construct used to gain an objective measure of these characteristics being derived in countries with similar built environments such as the United States, Canada and Australia. [20] [21] It will require further investigation as to whether similar constructs about the built environment relate to physical activity in diverse countries. Similarly the higher measure of ‘overall appeal’ in Canada may be due to this measure being dominated by characteristics considered to be appealing in countries more similar to Canada or the assessment through a ‘westernised’ lens. The utility of the overall appeal measures also need to be examined in analyses of predictive validity. Our analyses here did find that some variables requiring the observer to express an opinion had high reliability. However in the case of judgments about suitability for biking there were specific inter-observer differences reflecting their own experiences. Our findings with respect to the three items with low ICCs described at the end of the results, has led us to remove the first and third of these and to clarify the definition for the second.
The study has some limitations: It was conducted in a convenience sample of communities in a limited number of countries, although compared with other studies to evaluate environmental assessment tools, this was an unusually diverse sample of communities. The photos were assessed only on constructs that have been previously reported to be associated with physical activity but the majority of these instruments have been developed in the United States. Consequently, there may be other aspects of walkability we did not capture. However, we intend to store the photos indefinitely which will permit re-analyses for additional domains when these are suggested by other research. The summary scores were created simply with equal or near equal weight put on each characteristic within domains and may require refinement in examining their predictive validity against outcomes such as physical activity. Finally, photos were only taken of commercial streets, unlike other built environment scales which included coverage of residential and commercial streets.
In conclusion, we describe an approach to environmental assessment that is relatively rapid, low-cost and simple with respect to the data collected, thus offering a means of obtaining data on the built environment as experienced by communities in diverse locations. Our team and our collaborators have now used this method to collect and analyse photographic data on the built environments from a large number of communities in the Prospective Urban Rural Epidemiology (PURE) study (conducted in 17 high, middle and low income countries) [18] and the Health In Times of Transition (HITT) study conducted across former Soviet Union countries. [22]
Supporting Information
Appendix S3.
Instruction manual for evaluating communities using photos with EP-NET EPOCH Photos - Neighbourhood Evaluation Tool.
https://doi.org/10.1371/journal.pone.0110042.s003
(PDF)
Acknowledgments
We would like to acknowledge the work of Allison Zaruk who was a Summer student at the Population Health Research Institute and assisted with sorting and analysing photo data.
We would like to thank the following people for administrative support and local centre management: PHRI Project office: Pam Mackie, Sumathy Rangarajan; China: Sun Yi; India: Jayachitra KG, Mario Vaz; Colombia: Ronald Garcia; Brazil: Cláudia Stéfani Marcílio, Antonio Cordeiro Mattos; Canada (Quebec): Samantha Kellet, Ginette Turbide, Nadine Bonneville; Canada (BC) : Samantha Hage-Moussa, Danijela Gasevic
Author Contributions
Conceived and designed the experiments: CC DC KL. Performed the experiments: CC DC M. Madhavan PM LW YS YW SS PL DG AA SL GD KT. Analyzed the data: CC DC. Wrote the paper: CC DC KL M. McKee SY.
References
- 1. Sallis JF, Bowles HR, Bauman A, Ainsworth BE, Bull FC, et al. (2009) Neighborhood environments and physical activity among adults in 11 countries. Am J Prev Med 36: 484–490.
- 2. Foster C, Hillsdon M, Jones A, Grundy C, Wilkinson P, et al. (2009) Objective measures of the environment and physical activity—results of the environment and physical activity study in English adults. J Phys Act Health 6 Suppl 1: S70–80.
- 3. Cerin E, Conway TL, Cain KL, Kerr J, De Bourdeaudhuij I, et al. (2013) Sharing good NEWS across the world: developing comparable scores across 12 countries for the Neighborhood Environment Walkability Scale (NEWS). BMC Public Health 13: 309.
- 4. Day K, Boarnet M, Alfonzo M, Forsyth A (2006) The Irvine-Minnesota inventory to measure built environments: development. Am J Prev Med 30: 144–152.
- 5. Pomerleau J, Knai C, Foster C, Rutter H, Darmon N, et al. (2013) Measuring the food and built environments in urban centres: reliability and validity of the EURO-PREVOB Community Questionnaire. Public Health 127: 259–267.
- 6. Schaefer-McDaniel N, Caughy MO, O'Campo P, Gearey W (2010) Examining methodological details of neighbourhood observations and the relationship to health: a literature review. Soc Sci Med 70: 277–292.
- 7. Schaefer-McDaniel N, Dunn JR, Minian N, Katz D (2010) Rethinking measurement of neighborhood in the context of health research. Soc Sci Med 71: 651–656.
- 8. Gasevic D, Vukmirovich I, Yusuf S, Teo K, Chow C, et al. (2011) A direct assessment of "obesogenic" built environments: challenges and recommendations. J Environ Public Health 2011: 161574.
- 9. Chow CK, Lock K, Madhavan M, Corsi DJ, Gilmore AB, et al. (2010) Environmental Profile of a Community's Health (EPOCH): an instrument to measure environmental determinants of cardiovascular health in five countries. PLoS One 5: e14294.
- 10. Corsi DJ, Subramanian SV, McKee M, Li W, Swaminathan S, et al. (2012) Environmental Profile of a Community's Health (EPOCH): an ecometric assessment of measures of the community environment based on individual perception. PLoS One 7: e44410.
- 11. Wang C, Burris MA (1997) Photovoice: concept, methodology, and use for participatory needs assessment. Health Educ Behav 24: 369–387.
- 12. Charreire H, Mackenbach JD, Ouasti M, Lakerveld J, Compernolle S, et al. (2014) Using remote sensing to define environmental characteristics related to physical activity and dietary behaviours: a systematic review (the SPOTLIGHT project). Health Place 25: 1–9.
- 13. Chow CK, Lock K, Teo K, Subramanian SV, McKee M, et al. (2009) Environmental and societal influences acting on cardiovascular risk factors and disease at a population level: a review. Int J Epidemiol 38: 1580–1594.
- 14. Cerin E, Saelens BE, Sallis JF, Frank LD (2006) Neighborhood Environment Walkability Scale: validity and development of a short form. Med Sci Sports Exerc 38: 1682–1691.
- 15. Frank LD, Sallis JF, Saelens BE, Leary L, Cain K, et al. (2010) The development of a walkability index: application to the Neighborhood Quality of Life Study. Br J Sports Med 44: 924–933.
- 16. Pikora TJ, Bull FC, Jamrozik K, Knuiman M, Giles-Corti B, et al. (2002) Developing a reliable audit instrument to measure the physical environment for physical activity. Am J Prev Med 23: 187–194.
- 17. Weich S, Burton E, Blanchard M, Prince M, Sproston K, et al. (2001) Measuring the built environment: validity of a site survey instrument for use in urban settings. Health Place 7: 283–292.
- 18. Teo K, Chow CK, Vaz M, Rangarajan S, Yusuf S (2009) The Prospective Urban Rural Epidemiology (PURE) study: examining the impact of societal influences on chronic noncommunicable diseases in low-, middle-, and high-income countries. Am Heart J 158: 1–7 e1..
- 19. Lovasi GS, Schwartz-Soicher O, Neckerman KM, Konty K, Kerker B, et al. (2013) Aesthetic amenities and safety hazards associated with walking and bicycling for transportation in New York City. Ann Behav Med 45 Suppl 1: S76–85.
- 20. Parsons JA, Singh G, Scott AN, Nisenbaum R, Balasubramaniam P, et al. (2010) Standardized observation of neighbourhood disorder: does it work in Canada? Int J Health Geogr 9: 6.
- 21. Foster S, Giles-Corti B (2008) The built environment, neighborhood crime and constrained physical activity: an exploration of inconsistent findings. Prev Med 47: 241–251.
- 22. Watson K, Roberts B, Chow C, Goryakin Y, Rotman D, et al. (2013) Micro- and meso-level influences on obesity in the former Soviet Union: a multi-level analysis. Eur J Public Health 23: 291–298.