pone.0067575

Neighbourhood, Route and Workplace-Related Environmental Characteristics Predict Adults’ Mode of Travel to Work Alice M. Dalton1,3*, Andrew P. Jones2,3, Jenna R. Panter3,4, David Ogilvie3,4 1 School of Environmental Sciences, University of East Anglia, Norwich, United Kingdom, 2Norwich Medical School, University of East Anglia, Norwich, United Kingdom, 3UKCRC Centre for Diet and Activity Research (CEDAR), Institute of Public Health, Cambridge, United Kingdom, 4Medical Research Council Epidemiology Unit, Institute of Public Health, Cambridge, United Kingdom Abstract Objective: Commuting provides opportunities for regular physical activity which can reduce the risk of chronic disease. Commuters’ mode of travel may be shaped by their environment, but understanding of which specific environmental characteristics are most important and might form targets for intervention is limited. This study investigated associations between mode choice and a range of objectively assessed environmental characteristics. Methods: Participants in the Commuting and Health in Cambridge study reported where they lived and worked, their usual mode of travel to work and a variety of socio-demographic characteristics. Using geographic information system (GIS) software, 30 exposure variables were produced capturing characteristics of areas around participants’ homes and workplaces and their shortest modelled routes to work. Associations between usual mode of travel to work and personal and environmental characteristics were investigated using multinomial logistic regression. Results: Of the 1124 respondents, 50% reported cycling or walking as their usual mode of travel to work. In adjusted analyses, home-work distance was strongly associated with mode choice, particularly for walking. Lower odds of walking or cycling rather than driving were associated with a less frequent bus service (highest versus lowest tertile: walking OR 0.61 [95% CI 0.20–1.85]; cycling OR 0.43 [95% CI 0.23–0.83]), low street connectivity (OR 0.22, [0.07–0.67]; OR 0.48 [0.26–0.90]) and free car parking at work (OR 0.24 [0.10–0.59]; OR 0.55 [0.32–0.95]). Participants were less likely to cycle if they had access to fewer destinations (leisure facilities, shops and schools) close to work (OR 0.36 [0.21–0.62]) and a railway station further from home (OR 0.53 [0.30–0.93]). Covariates strongly predicted travel mode (pseudo r-squared 0.74). Conclusions: Potentially modifiable environmental characteristics, including workplace car parking, street connectivity and access to public transport, are associated with travel mode choice, and could be addressed as part of transport policy and infrastructural interventions to promote active commuting. Citation: Dalton AM, Jones AP, Panter JR, Ogilvie D (2013) Neighbourhood, Route and Workplace-Related Environmental Characteristics Predict Adults’ Mode of Travel to Work. PLoS ONE 8(6): e67575. doi:10.1371/journal.pone.0067575 Editor: Harry Zhang, Old Dominion University, United States of America Received November 19, 2012; Accepted May 22, 2013; Published June 19, 2013 Copyright: � 2013 Dalton 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: The Commuting and Health in Cambridge study was developed by David Ogilvie, Simon Griffin, Andy Jones and Roger Mackett, and initially funded under the auspices of the Centre for Diet and Activity Research (CEDAR), a UKCRC Public Health Research Centre of Excellence. Funding from the British Heart Foundation, Economic and Social Research Council, Medical Research Council, National Institute for Health Research and the Wellcome Trust, under the auspices of the UK Clinical Research Collaboration, is gratefully acknowledged. The study is now funded by the National Institute for Health Research Public Health Research programme (project number 09/3001/06: see http://www.phr.nihr.ac.uk/funded_projects). David Ogilvie is also supported by the Medical Research Council (Unit Programme number MC_U106179474). The views and opinions expressed herein are those of the authors and do not necessarily reflect those of the NIHR PHR programme, NHS or the Department of Health. 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. * E-mail: a.dalton@uea.ac.uk Introduction People who engage in physical activity are less likely to suffer from a range of preventable chronic diseases [1]. However, most people do not meet government guidelines for physical activity [2], which contributes to the burden of disease in the UK [3] and globally [4]. One way of increasing population activity levels may be to build it into daily routines, for example in the form of active travel (cycling and walking) for part or all of the journey to work [5]. This has the potential to make a significant contribution to physical activity, as commuting to and from work constitutes 15% of the journeys made in the UK and the United States [6,7] and 19% of the distance travelled [7]. There has been a steady decline in cycling and walking to work during the last few decades as the car has become a more popular choice [8] and people have tended to live further from work [9]. Nevertheless, there is evidence that regular active travellers are physically fitter and have higher levels of mental wellbeing and lower sickness absence from work [10,11]. In addition, public transport may provide an alternative to the car for those living further from work and offer health as well as PLOS ONE | www.plosone.org 1 June 2013 | Volume 8 | Issue 6 | e67575 environmental benefits, with US research finding that people walking to transit stops can accrue around 30% of their recommended daily activity levels by doing so [12]. A recent modelling study has suggested that a step-change in the prevalence of walking and cycling in England and Wales could save the National Health Service approximately £17bn in healthcare costs associated with chronic conditions such as type-2 diabetes, cancer, heart disease and depression over the next 20 years [13]. Features of the built and natural environments within which people live and work may influence the mode of travel used [1], such as for journeys to work, by either supporting or acting as a barrier against walking and cycling. In order to develop environments more supportive of a modal shift towards more active travel at the population level, we therefore need to better understand the importance of characteristics of the home, work and journey environments [14,15]. Whilst there is good evidence that modal choice is associated with travel distance, evidence is more equivocal regarding the importance of many other characteristics of the physical environment including street network connectivity, urban design, walkable destinations (includ- ing schools, shops, leisure and cultural facilities), land use, infrastructure for walking or cycling, and the availability of or access to public transport [1,16,17]. It is also noteworthy that despite the constraints of distance, some people do use active modes of travel even if they live a relatively long way from work. In part, our limited understanding of the correlates of active travel may reflect methodological limitations of some previous studies. A recent systematic review of the environmental correlates of cycling noted that few studies had measured the environment objectively and none of the 21 studies identified could be regarded as methodologically strong [16]. Reasons included failing to control for confounders, such as gender and age; failing to carry out analysis at the individual level; and using inappropriate statistical methods. Many studies have examined environmental attributes, public transport (transit) accessibility, or restrictions on car use singly but not in combination, meaning that the relative importance of these factors is not known. For these reasons, a recent review concluded that empirical research either does not include or provides inconclusive evidence on the influence of the transport environment on travel behaviour [1]. In addition, research has tended to treat walking and cycling as a singular behaviour, whereas they are distinct behaviours likely to have different correlates [18,19]. A further limitation is that character- istics of the home neighbourhood have often been the focus for analysis, yet the workplace environment and that of the route between home and work may also be important [20,21,22]. Although some research has considered distances to public transport (transit) stops [23], overall accessibility and the ease by which that transport may be used to reach work has rarely been considered [16,24]. In general, previous studies on active travel suggest that existing models fail to account for a significant proportion of the variance in travel behaviour. Environmental attributes have been shown to explain some of this variance, but prior empirical research has typically tested a limited range of independent variables in rather environmentally homogeneous, often urban, settings. Using a much wider range of objective environmental measures than have commonly been tested, the aim of this study is to investigate the extent to which features of the environment are associated with modal choice amongst a sample of commuters in Cambridge, UK. We focus on the individual level and assess correlates for travel by all modes. In doing so, we aim to identify modifiable environmental characteristics that might form the targets of future interventions to increase the prevalence of active travel. Methods Study design and setting This research analysed cross-sectional data obtained from a sample of commuters taking part in the Commuting and Health in Cambridge study in Cambridge, UK. The details of this study have been outlined previously [25]. Participants were aged 16 and over, working in Cambridge and living within 30 km of the city. They were sampled using a workplace recruitment strategy that targeted a variety of workplaces and employers in a range of geographical locations across the city centre and urban fringe. The data used for this analysis were collected between May and November 2009 using postal questionnaires [26]. Participants reported their recent physical activity (at home, at work and for recreation), general health, and travel to and from work and for other purposes. Personal characteristics such as age, gender, weight, height and highest educational attainment were also reported. 1168 respondents returned questionnaires, 1155 of whom provided valid postcodes which could be used to identify their home and work locations. Ethics statement Ethical approval for the study was obtained from the Hertfordshire Research Ethics Committee (reference number 08/H0311/208) and written informed consent was provided by each participant. No minors/children participated in this study. Determining modal choice Two sections of the survey questionnaire were relevant to this analysis. The first comprised the Recent Physical Activity Questionnaire (RPAQ), a validated instrument that measures physical activity in the previous four weeks [27]. RPAQ was used to classify participants according to their usual mode of travel to work. The survey listed four modes (car/motor vehicle, works or public transport, bicycle, or walking) and participants were asked to specify if they ‘always’, ‘usually’, ‘occasionally’ or ‘never or rarely’ travelled to work by each mode. Most participants selected the option of ‘always’ for one mode of transport, and their usual mode was classified accordingly. For the 178 cases in which more than one mode was identified, rules were established to classify participants according to their predominant modal choice based on the most frequently reported travel mode: if participants stated that they always or usually walked and/or cycled but that they also always or usually used the car, they were coded as using the car as it was presumed this would constitute the main component of the journey (n = 92); if they stated they always or usually walked and/ or cycled but also always or usually used public transport, they were coded as using public transport (n = 39); if they reported always or usually walking and cycling they were coded as cycling (n = 9); and if they stated that they sometimes used the car and sometimes the bicycle, they were coded as using the car (n = 3). For 35 participants, no predominant mode could be determined at this stage. In these cases, a question from the second relevant section of the questionnaire, ‘About your travel to and from work in the last seven days’, was used to identify the predominant mode or modes used in the previous seven days. Of those always reporting more than one mode per journey, 19 using the car and bus were classified as car users, one reporting walking and using the train was classified as using the train, and one using the train and the car was classified as using the train. For those reporting a mixture of different modes which varied throughout the week, six used the car Environment, Travel Mode and the Commute to Work PLOS ONE | www.plosone.org 2 June 2013 | Volume 8 | Issue 6 | e67575 most frequently and one used the train most frequently. One participant did not answer the RPAQ question but responded as only using the car to the ‘last seven days’ question. Six participants could not be attributed to one predominant mode because they reported using a mixture of different modes of transport throughout the week with equal frequency, and were therefore not included in subsequent analyses. Environmental measures, neighbourhood delineation and route identification For each participant, environmental measures were calculated for the home neighbourhood, the work neighbourhood, and the route between home and work. Table 1 presents a total of 30 different measures together with the data sources that were used to calculate them, grouped into themes of ‘roads and routes’, ‘public transport’ and ‘land use’. In order to generate neighbourhoods, participants’ home and work postcodes were georeferenced using the Ordnance Survey (OS) Address Layer 2Hdatabase [28]. A pedestrian route network dataset was constructed in the ArcGIS 9.3 [29] geographic information system (GIS) software package by combining road data, excluding motorways, from the OS MasterMapH Integrated Transport NetworkTM (ITN) database with local authority rights- of-way data (public footpaths, bridleways and byways), cycle route information from the charity Sustrans [30], and other informal pathways recorded on OpenStreetMap.com. Using this informa- tion, home neighbourhoods were delineated to represent areas that could be accessed within an approximate ten-minute walking time (equating to 800 m along the pedestrian network). This distance has been used in previous accessibility analysis to represent a practically walkable neighbourhood area [19,31,32]. To delineate the route between home and work for each participant, ArcGIS was used to identify sections of the pedestrian network that comprised the shortest route between the two locations. Subsequently, the characteristics of the environment within a 100 m distance of the route were quantified. The ‘roads and routes’ environmental variables that were computed included a measure of distance to work (length of shortest route between home and work), route directness (ratio of route network to straight line distance), whether participants were travelling into or out of the city centre on their journey from home to work (a participant was defined as working in the city if they worked within 2 km of Cambridge central bus station, and defined as living in the city if they lived in the Cambridge urban area according to the 2001 census [33]) and the proportion of route length that was along A or B (major) roads. In addition, five measures of the walkability of the streets in the home neighbour- hood were calculated, including road density, junction density, road connectivity, existence of A class (major) roads, the proportion of foot/cycle paths, and the general ‘walkable area’ (defined as the area walkable within 800 m along the route network buffer from participants’ home location divided by the area within an 800 m straight line distance). Public transport variables were derived using the route network dataset in combination with the National Public Transport Access Nodes (NAPTAN) and Data Repository (NPTDR) datasets [34,35]. Firstly, it was established whether or not there was a bus service passing through the neighbourhood that would take the participant to work, either directly or with one or more changes. The other variables calculated for both the home and work locations were the distance to the nearest bus stop, bus service frequency, the number of bus stops, the number of bus routes served and the distance to the nearest railway station. These indicators used the pedestrian route network, the frequency of bus services from the nearest bus stop on a typical weekday, the number of serviced bus stops present, and the number of bus routes available in the neighbourhood based on a count of unique service numbers. Land use variables included an indicator of land use mix - the Herfindahl-Hirschmann Index (HHI) [36] - and building density at the home location and on the route to work. A count of the number of destinations (schools, eating and drinking establish- ments, stage and screen venues, sports complexes, and retail units) was taken after they were mapped along the route to work and in the home and work neighbourhoods. A measure of local area deprivation was calculated for the home neighbourhood, based on the proportion of people in lower socioeconomic classes (semi- routine occupations, routine occupations, never-worked and long- term unemployed categories) living within the relevant Census Output Area. In addition, the availability of workplace car parking (none, charged or free) reported by each participant in their questionnaire was used. Data analysis Unadjusted associations between usual travel mode (car, public transport, cycling or walking) and home, work, and route characteristics were examined using chi-square tests and analyses of variance. Continuous variables were categorised as tertiles or using other appropriate groupings. Prior to model fitting, multi- collinearity was managed using a pair-wise correlation matrix to identify variables that were highly associated, defined as having a Pearson’s correlation coefficient of .0.55 based on previous empirical research [37]. Only one variable from each correlated pair was added into each model, the chosen one being that with the strongest association in the expected direction with the outcome variable. Multinomial regression models were then fitted to examine adjusted associations, using commuting by car as the reference category for the outcome variable. The models were fitted in a number of stages. First, a personal model was created by entering all potential individual and household-level covariates into a multinomial logistic regression. Prior empirical evidence has shown mixed associations between individual factors and active travel [17], therefore all such variables were included at this stage. After all these potential predictors were added, those for which p.0.1 were removed in a backwards stepwise manner leaving only those which were statis