Health & Place 33 (2015) 195–202

Contents lists available at ScienceDirect

Health & Place journal homepage: www.elsevier.com/locate/healthplace

Differences in associations between active transportation and built environmental exposures when expressed using different components of individual activity spaces Torbjorn van Heeswijck a,n, Catherine Paquet a,d, Yan Kestens b, Benoit Thierry b, Catherine Morency c, Mark Daniel a,e a

Spatial Epidemiology and Evaluation Research Group, School of Population Health, University of South Australia, Adelaide 5001, SA, Australia CRCHUM, Montreal School of Public Health, Montreal Hospital University Research Center (CRCHUM), Montreal, QC, Canada H2L 1V1 Department of Civil, Geological and Mining Engineering, École Polytechnique de Montréal, C.P. 6079, succ. Centre-ville, Montreal, QC, Canada H3C3A7 d Research Center of the Douglas Mental Health Institute, Montreal, QC, Canada H4H 1R2 e Department of Medicine, St.Vincent’s Hospital, The University of Melbourne, 41 Victoria Parade, Fitzroy 3065, VIC, Australia. b c

art ic l e i nf o

a b s t r a c t

Article history: Received 13 February 2014 Received in revised form 1 September 2014 Accepted 4 March 2015 Available online 6 April 2015

This study assessed relationships between built environmental exposures measured within components of individual activity spaces (i.e., travel origins, destinations and paths in-between), and use of active transportation in a metropolitan setting. Individuals (n ¼ 37,165) were categorised as using active or sedentary transportation based on travel survey data. Generalised Estimating Equations analysis was used to test relationships with active transportation. Strength and significance of relationships between exposures and active transportation varied for different components of the activity space. Associations were strongest when including travel paths in expression of the built environment. Land use mix and greenness were negatively related to active transportation. & 2015 Elsevier Ltd. All rights reserved.

Keywords: Activity space Walking Bicycling Geographic information systems Built environment

1. Background The low population level of engagement in regular physical activity among adults is a topic of serious concern in western societies, with implications for population levels of obesity, cardiovascular disease and general wellbeing. Research has moved beyond a focus on recreational exercise towards the more holistic concept of ‘active living’ (Sallis et al., 2006). There is considerable interest in supporting the substitution of motorised modes of daily transportation with active transport modes. Walking and cycling in particular are potential avenues for increasing population levels of physical activity. Participation in active transportation has been associated with BMI scores 0.9 points lower as well as reduced likelihood of obesity, lower rates of mortality and cardiovascular disease risk (Flint et al., 2014; Frank et al., 2004; Lopez-Zetina et al., 2006). Time spent in cars is associated with

n Correspondence to: School of Population Health, City East Campus, University of South Australia, Internal Post Code CEA-09, GPO Box 2471, Adelaide 5001, SA, , Australia. Tel.: þ61 8 830 22734; fax: þ 61 8 830 22603. E-mail addresses: [email protected] (T. van Heeswijck), [email protected] (C. Paquet), [email protected] (Y. Kestens), [email protected] (B. Thierry), [email protected] (C. Morency), [email protected] (M. Daniel).

http://dx.doi.org/10.1016/j.healthplace.2015.03.003 1353-8292/& 2015 Elsevier Ltd. All rights reserved.

increased BMI and likelihood of obesity (Furie and Desai, 2012; Hamer and Chida, 2008; Matthews et al., 2007). A body of literature investigating the role of the built environment in shaping physical activity behaviours has formed in recent decades, and numerous studies have reported associations between objectively measured built environments and adult physical activity (Committee on Physical Activity, 2005; Durand et al., 2011; Fraser and Lock, 2011; Saelens and Handy, 2008; Sugiyama et al., 2012). Built environmental features considered relevant to physical activity have been broadly characterised as belonging to categories labelled the 3 “Ds” (Sugiyama et al., 2012): density (e.g., population and destination density), diversity (e.g., mixed land use and destination types) and design, including both functional (e.g., street and public transport networks) and aesthetic (e.g., green spaces) aspects. The importance of the 3 “Ds” at both the origins and destinations of trips and their relationship to travel demand and mode choice are well established in the transportation literature (Cervero and Kockelman, 1997; Frank and Pivo, 1994) however the focus in this field, and more broadly in place and health research (Chaix et al., 2012), has largely been on environmental exposure within local residential or ‘neighbourhood’ environments (Sugiyama et al., 2012). This approach may not sufficiently capture built environment exposures relevant to behaviour occurring away from the residential space, such as daily transportation (Kestens et al., 2010; Zenk et al., 2011).

196

T. van Heeswijck et al. / Health & Place 33 (2015) 195–202

Movement through, and therefore exposure to, the built environment during daily living is not necessarily constrained to “local” areas, defined either administratively or by a uniform spatial proximity to home, but occurs in a way that is both spatially complex and unique to individuals (Kwan, 2009). Spatial models must be able to account for activity beyond the “local” context (Cummins, 2007; Kwan, 2009). Specifically, the speeds and distances typically afforded by active transportation modes also imply an intimate relationship between individuals and their surroundings during transportation movements, requiring the ability to measure environmental exposures within areas immediately proximal to individual origins, destinations and routes (Moudon and Lee, 2003). The concept of the activity space, which encompasses all the locations visited by individuals in order to undertake activities or travel between them, provides a means of representing the full extent of the environment encountered by an individual during movement within a given time period (Schönfelder and Axhausen, 2003b). One study has investigated relationships between the built environment and cycling for transport using an activity space method that accounted for environmental exposures within origin, destination, and path zones, but a separate activity space was created for each trip rather than investigating the aggregate space encountered by individuals over the course of the day (Winters et al., 2010). Other studies have found significant relationships between food environment exposures within the daily activity space and outcomes relating to health and behaviour (Lebel et al., 2012; Zenk et al., 2011). There is a need for this methodology to be applied in studies of active transportation participation to provide support for the implementation of built environmental interventions, and to lay the groundwork for research assessing the impact of these interventions on hard outcomes (e.g., reductions in population rates cardiovascular disease). This study sought to investigate relationships between built environmental exposures within individual activity spaces and use of active transport modes, and further to compare these relationships to those found when expressing the environment within the individual activity space and its component spaces.

Bell Canada White Pages Residential Telephone registry (Agence Métropolitaine de Transport, 2012). Interviewers spoke to the adult (18 years or older) from each household most knowledgeable about the transportation activities of the household’s members, requesting self-report data about the origin, destination and mode of all transportation activities (trips) undertaken by all members of that household over the 24-h period preceding the interview. Age, sex, employment status and driver’s license ownership were reported for each household member. Home location, car ownership and household type were reported for the household. All reported home locations, trip origins and destinations were geocoded as part of the MO-D survey. 2.2.2. Environmental data Data pertaining to spatial features of the CMMA region were drawn from a comprehensive Geographic Information System (GIS) named the Montreal Epidemiological and Geographic Analysis of Population Health Outcomes and Neighbourhood Effects (MEGAPHONE) (Daniel and Kestens, 2007). This GIS includes data describing land use patterns (residential, industrial, commercial, government and institutional, open space), road networks, public transit networks, census tract (CT) level census data and public, private and institutional businesses and services. 2.3. Sample

2. Methods

The MO-D survey included responses from 120,511 participants between the ages of 20 and 89. A minimum age restriction was employed as this study is part of a larger project linking these measures to mortality data available only for individuals aged 20 years and older. Of these participants 67,654 reported residence locations within the Island of Montreal, and 40,699 also reported trip origins and destinations within the Island of Montreal only. Individuals with trips or residences outside the Island of Montreal were excluded as environmental data sources of equivalent quality were not consistently available in those areas. Finally an additional 3534 individuals for whom complete data were not available for all variables were excluded to produce a sample of 37,165 individuals.

2.1. Population and setting

2.4. Measures

This study was carried out using travel survey data collected by the Quebec Ministry of Transportation from a sample of the population of the Island of Montreal, Canada. The Island of Montreal covers an area of 500 km2 and contained 1.8 million people at the time of the survey (Statistics Canada, 2002).

2.4.1. Outcome variable The outcome variable for this study was physical activity accrued during daily transportation behaviour, estimated as total daily metabolic equivalent-minutes (METS) of physical activity and categorised as having reported sedentary or physically active transportation behaviour. The total METS for each individual was calculated using estimated trip distances, walking and cycling speeds, and METS conversion factors as described below. Geocoded trip origins and destinations were mapped to the Montreal Island street network using ArcGIS 9.3 (ESRI, 2008). The shortest (by travel time) transportation network (including street, train and metro networks) constrained path between origin and destination was calculated for each trip using ArcGIS Network Analyst, taking into account the reported transportation mode(s). Multi-modal trips were analysed as single trips with the locations of mode change between origin and destination used as anchors for route choice. Average walking speeds were estimated for each individual using data for comfortable gait speeds (Bohannon, 1997), assuming an inverse linear relationship between age and speed. Walking speeds calculated this way ranged from 4.5 to 5.0 km/h. Average commuter cycling speed was estimated at 18 km/h (de Geus et al., 2007; Oja et al., 1998). For each trip with walking or cycling as the reported transportation mode an estimated travel time was calculated from the trip distance

2.2. Data sources 2.2.1. 1998 Montreal origin-destination survey Data on individual and household transportation behaviour were drawn from the 1998 Montreal Origin-Destination (MO-D) survey. This computer-assisted telephone interview (CATI) survey is conducted every 5 years by the Technical Committee on Travel Surveys of the Greater Montreal Area on a representative sample of the Census Montreal Metropolitan Area (CMMA) population. The 1998 MO-D survey was undertaken from August 25th to December 18th and contains data from 164,076 individuals in 65,227 households. The sample population was drawn from 73 sampling areas containing all 762 census tracts (CTs) of the CMMA, with sampling quotas of 3–9% of the total number of households in each area. Quotas were based on analysis of the previous MO-D survey results and the desire to adequately represent variables of interest. Households were randomly sampled by telephone number from the July 1998

T. van Heeswijck et al. / Health & Place 33 (2015) 195–202

197

and the individual’s estimated speed. METS were estimated for each walking and cycling trip using conversion factors taken from the International Physical Activity Questionnaire (IPAQ Group, 2005). Total METS for all trips reported for each individual were summed to give individual daily METS accrued during transportation and categorised as having reported sedentary (o7 daily METS) or active (7rdaily METS) transportation behaviour based on the criteria used for health surveys by the Australian Bureau of Statistics (Australian Bureau of Statistics, 2009). A binary outcome was used as the majority of subjects (82%) were classified as reporting sedentary transportation. Ninety-eight walking and cycling trips with an estimated travel time of two hours or greater were excluded as likely errors in the selfreported data.

2.4.2. Environmental variables Environmental variables were selected to represent built environmental constructs of potential importance. Land use mix was represented by an entropy measure (Feng et al., 2010) incorporating five categories of land use (residential, industrial, commercial, government/institutional and open space, 2002 DMTI data, values between 0 for homogenous land use and 1 for perfectly mixed land use). Access to public transportation was represented by the presence of bus lines (kilometres of road being serviced by bus(es) per square kilometre, 2005 data) and distance to the nearest metro station (street network aligned kilometres to the nearest metro station, 2004 DMTI data). Street network connectivity was represented by intersection density (count of street intersections per square kilometre, 2000 DMTI data) and barriers by the presence of highways (kilometres of highwaydesignated road per square kilometre, 2001 data). Greenness was represented by normalised difference vegetation index averaged across buffer areas (NDVI, June 2001 Landsat image, 30 m pixel resolution). Availability of destinations of interest was represented by the density of destinations (kernel density of commercial, food and activity destinations, 2005 Zipcom database).

2.4.3. Individual activity spaces Two types of spatial buffers were used to construct representations of individual activity spaces: Network-constrained buffers (1.6 km distance) were created around each location of residence, trip origin and trip destination, representing an area accessible by approximately 20 min walking at average speed. Straight-line buffers (200 m distance) were created around calculated paths between each origindestination pair to represent the area proximal to the calculated travel path and nearby alternative paths. Four representations of the individual activity space were constructed using the following buffer type combinations (Fig. 1): (1) place of residence buffers, (2) origin and destination (including place of residence) buffers, (3) path buffers only, and (4) the combined activity space incorporating all buffers together (origins, destinations and paths). In all representations buffers based on origins and destinations located less than 1 km straightline distance from a previously reported origin or destination for that individual were excluded to minimise overlap of buffers where individuals visited locations repeatedly or distances between visited locations were short (totalling 141,843 out of 224,510 reported origins and destinations). Built environmental attributes were calculated within each buffer using data extracted from the MEGAPHONE GIS. Where buffers covered only part of a built environmental feature (e.g. a particular land use parcel) only the area included within the buffer was considered for the purpose of measurement. For each of the four activity space representations, exposure to an aspect of the built environment was expressed as the average of that built environmental feature across the buffers used to represent the activity space.

Fig. 1. Example of activity space components—two routes between origin and destination with associated buffers along routes and at each location.

2.4.4. Covariates Previous work has found individual access to automobiles and employment status to be associated with active transportation participation (Adams, 2010). Accordingly employment status, driver’s license status, and number of cars per license holder in the household were extracted from the MO-D survey for inclusion as individual-level covariates along with participant age and gender. Household type (based upon age of members and categorised as single, couple, couple with children, lone parent or other households) was also included as a covariate because living with other individuals, particularly dependents, was expected to present barriers to engaging in non-motorised transportation. The total distance travelled by individuals was included in all models because trips requiring long distance travel were expected to discourage the use of non-motorised transport. The distance from place of residence to nearest Metro station was included as access to public transportation was expected to impact individual’s preference for non-motorised transportation. Population density, the proportion of people holding university degrees, the proportion of households below the low-income threshold and crime rates were included in all models as area-level covariates because previous work has demonstrated significant relationships between area-level socioeconomic status and walking rates (Riva et al., 2009). These sociodemographic data were extracted at the CT level from the 1996 Canadian Census for the CT where participants resided. 2.5. Data analysis Relationships between all built environmental variables and active transportation were tested using binary logistic Generalised Estimating Equation (GEE) models. GEE was used to account for spatial clustering at the Census Tract level (Genmod procedure in SAS software, version 9.2; SAS Institute Inc., Cary, North Carolina). This method was selected over multilevel models due to convergence problems. Separate models were tested for each representation of the activity space. Models were adjusted for all individual- and area-level covariates simultaneously. SES and environmental predictors were standardised. Statistical significance was set at alpha¼ 0.01 due to the large size of the sample increasing the likelihood of type I error. Odds ratios and their 99% confidence intervals were obtained by exponentiation of regression parameter estimates. Differences between odds ratios were tested for statistical significance at the 99% confidence level (‘Comparison of two ratios’ procedure in WINPEPI software, version 11.39, Abramson, 2011).

198

T. van Heeswijck et al. / Health & Place 33 (2015) 195–202

Table 1 Descriptive statistics for study population drawn from 1998 Montreal origin-destination survey (n ¼37,165). Data sources: 1998 Montreal origin-destination survey excluding non-residents of Montreal Island and those under the age of 20, 1996 Canadian Census, and 2000–2005 MEGAPHONE GIS data. All individuals

Male

Female

Characteristic

Mean/n

SD/%

Mean/n

SD/%

Mean/n

SD/%

Age (mean, SD) Sex Male (n, %) Employment Full time (n, %) Part time (n, %) Student (n, %) Retired (n, %) Other (n, %) Driver’s license Yes (n, %) Car/license ratio o0.5 (n, %) 0.5–0.99 (n, %) 4 ¼ 1 (n, %) Household type Single person, no child (n, %) Couple, no child (n, %) Couple with children (n, %) Lone parent (n, %) Other (n, %) SES variables Proportion holding university degree (mean, SD) Proportion low income households (mean, SD) Population density (median, IQR) Other covariates Total distance travelled (median, IQR) Crime rate (median, IQR) Distance to nearest metro station (median, IQR) Built environment Bus line density (mean, SD) Highway density (median, IQR) Land use mix (mean, SD) Greenness (mean, SD) Destination density (median, IQR) Intersection density (mean, SD)

43.3

15.7

43.0

15.5

43.6

15.8

17,910

48









20,782 2,468 3,792 5,689 4,434

56 7 10 15 12

11,453 810 183 2,548 1,269

64 5 10 14 7

9329 1658 1962 3141 3165

48 9 10 16 16

28,820

78

15,459

86

13,361

69

9,992 10,850 16,323

27 29 44

4,417 5,448 8,045

25 30 45

5575 5402 8278

29 28 43

6,309 10,582 11,481 1,836 6,957

17 29 31 5 19

2,883 5,368 5704 439 3,516

16 30 32 2 20

3426 5214 5777 1397 3441

18 27 30 7 18

19.9 33.0 6,316

13.2 16.0 7540

19.9 33.3 7,970

13.1 16.1 6414

19.8 32.8 7800

13.2 15.9 6156

16.4 9.9 2.1

22.8 10.9 5.9

18.3 10.3 2.0

24.0 11.2 5.9

14.7 9.8 2.1

21.4 9.7 6.0

5.0 0.6 0.7  0.12 218.4 80.0

1.1 1.5 0.1 0.07 420.1 17.5

5.1 0.7 0.7  0.12 227.2 80.3

1.1 1.7 0.1 0.06 444.3 17.9

4.9 0.6 0.7  0.11 211.0 80.0

1.1 1.3 0.1 0.07 393.0 17.2

3. Results 3.1. Sample characteristics Table 1 presents descriptive statistics for the sample population. The final study population included 37,165 individuals from 24,707 households in 500 CTs. Participants were largely employed (62.5% in full or part-time employment). The majority were licensed to drive (77.6%) and had reasonable access to motor vehicles (73.1% with at least 1 car for every 2 licensed drivers in the household). A higher proportion of men than women were employed full time and held a driver’s license but otherwise there was little difference between genders. Table 2 presents descriptive statistics for survey questions for all island-resident participants of the 1998 MO-D survey. Additional categories for employment and driver’s license are due to data reported for individuals below 18 years of age. The sample population for this study was on average older, more likely to be employed full time and less likely to be single. Based on census data the sample population was similar to that of the CMMA in terms of age and sex distribution, number of people per household and proportion of two-parent families (Statistics Canada, 1996). Participants were more likely to be employed full-time and less likely to be single or a lone parent compared to the CMMA population. Table 3 reports the odds ratio and 99% confidence interval of participants being classified as physically active across all built

Table 2 Descriptive statistics for all Montreal Island residents participating in 1998 Montreal origin-destination survey (n ¼67,654) Data sources: 1998 Montreal origin-destination survey excluding non-residents of Montreal Island. Characteristic

Mean/n

SD/%

Age (mean, SD) Sex Male (n, %) Employment Full time (n, %) Part time (n, %) Student (n, %) Retired (n, %) Other (n, %) Child/refused (n, %) Driver’s license Yes (n, %) No (n, %) Child/refused (n, %) Car/license ratio o 0.5 (n, %) 0.5–0.99 (n, %) 4 ¼1 (n, %) Household type Single person, no child (n, %) Couple, no child (n, %) Couple with children (n, %) Lone parent (n, %) Other (n, %)

35.5

20.3

33,493

50

28,256 3,354 16,566 8,366 7,074 4,038

42 5 25 12 11 6

40,997 13,759 12,898

61 20 19

15,605 20,585 31,464

23 30 47

7,557 15,141 28,392 4,918 11,646

11 22 42 7 17

T. van Heeswijck et al. / Health & Place 33 (2015) 195–202

Table 3 Odds ratios of reporting active transportation in relation to one unit increase in standardised built environment variables expressed for full activity space. Predictor

Odds ratio

CI (99%)

p

Bus line density Highway density Land use mix Greenness Destination density Intersection density

0.903 0.903 0.448 0.591 1.256 1.044

0.777–1.049 0.802–1.016 0.404–0.496 0.530–0.659 1.109–1.423 0.953–1.144

0.079 0.026 o 0.001 o 0.001 o 0.001 0.226

environmental variables when measured using the combined activity space. Greater density of destinations of interest was associated with an elevated odds of using active transport, while greater land use mix and greenness were unexpectedly associated with a lower odds of using active transport. Significant associations with use of active transport were not found for bus line density, highway density and intersection density. The odds ratio and 99% confidence interval for associations between built environmental exposures and reported active transportation for each activity space representation method are shown in Fig. 2. An elevated odds of reported active transportation was observed for greater intersection density and lower highway density within the origin/destination based activity space. Greater odds of reporting active transportation were found for lower greenness and lower highway density within the residence based activity space. Active transportation was also associated with greater density of destinations, lower land use mix, lower greenness and lower bus line density for the path-based activity space. Greater odds of active transportation were found for greater density of destinations, lower land use mix and lower greenness for the path buffer and combined activity space. Lower greenness was also positively associated with odds of active transportation within the residential space. Results for analyses that included path buffers when expressing environmental exposures were statistically different from those obtained when not including path buffers for greenness (OR¼ 0.68 and 0.59 for expressions including paths, 0.85 and 0.90 without paths) and land use mix (OR¼0.49 and 0.45 including paths, 1.21 and 1.01 without paths). Differences in statistical significance were also observed for density of destinations (significant including paths, not significant without paths) and highway density (not significant including paths, significant without paths), though the confidence intervals overlapped for these variables.

4. Discussion In this study, built environmental variables expressed for individual daily activity spaces were found to be associated with the odds of engaging in active transportation. The density of destinations, land use mix, and greenness of individual activity spaces were significantly associated with active transportation after accounting for individual and SES factors as well as trip distance and access to public transport. These associations were weaker or non-significant when measuring built environment exposures using residence-only or origin/destination-based spatial expressions. An extensive body of literature now exists on the relationships between the objectively measured built environment and physical activity (including active transportation) expressed using locally-based expressions of environmental exposure, in which the environment is spatially defined using either administrative areas (e.g., census tract) or geometrically defined buffers tied to the location of residence (Cummins, 2007; Sugiyama et al., 2012). This body of literature has led to promotion of the built environment as an important target for intervention in efforts to increase population levels of physical activity

199

as well as creating safer and more sustainable transportation environments (Killingsworth et al., 2003). Reports and policy documents typically suggest that environmental interventions to promote active transportation should cover the full range of spaces within which individuals carry out the activities of daily living (Global Advocacy Council for Physical Activity, 2010; National Public Health Partnership, 2001). There remains, however, a gap in the literature to support proposed measures beyond residential areas. Sugiyama et al. (2012) reviewed the literature analysing relationships between built environment features associated with routes or destinations (e.g., street connectivity, sidewalk presence) and physical activity, concluding that there was some evidence for relationships between these features and walking for transport, but noted that correspondence between environmental measures and the places where walking behaviours took place was poor. Reviews have called for greater attention to environmental exposures specific to the context of physical activity behaviours, particularly in contexts beyond the residential environment (Ding and Gebel, 2012; Sugiyama et al., 2012). The activity space concept offers a potential method for better representing the environmental exposure of individuals during daily travel, and several methods have been explored as tools for describing the spaces individuals are exposed to from behavioural data (Rai et al., 2007; Schönfelder and Axhausen, 2003a). Nevertheless, few studies relating the built environment to physical activity have utilised this concept (Winters et al., 2010; Zenk et al., 2011) and to our knowledge only one such study has investigated utilitarian physical activity specifically (Winters et al., 2010). One recent study used very similar methods to the present work to investigate levels of food environment exposure at work and along commuting paths in addition to the home environment, concluding that exposures were greatly underestimated when measurement was restricted to the residential neighbourhood (Burgoine and Monsivais, 2013). Another study used data from the Montreal origin-destination survey to investigate relationships between the built environment and walking behaviours (Wasfi et al., 2013), but street intersection density was the only built environmental variable measured beyond the home location of individuals. Relationships between the built environment and active transportation observed in this study were similar when measured using residence and origin/destination based spatial expressions of environment, and most were not statistically significant. Negative associations with active transportation behaviour were found for greenness (OR¼0.85) and highway density (OR¼0.83) measured within residential environments. Markedly different results were obtained for the path and activity space based expressions across several variables including a positive association with active transport behaviour for density of destinations, and strongly negative associations for greenness and land use mix measured within individual activity spaces. The similarity between relationships found for the path-based and combined activity space environmental expressions in terms of both significance and effect size across all variables suggests that active transport participation is more strongly related to the nature of the environment experienced along the path between origin and destination than the environment surrounding those locations, a conclusion consistent with that reported for relationships between built environments and commuter cycling behaviour (Winters et al., 2010). Intersection density, Highway density and bus service line density all produced results that differed between buffer combinations in terms of statistical significance, however the difference in effect sizes were generally small and confidence intervals overlapped the point estimates obtained using different spatial representations. A generally positive relationship between physical activity and connectivity, represented by intersection density, is in agreement with previously reported results (Frank et al., 2005; Troped et al., 2010) though evidence for associations with walking and cycling for transport is inconclusive (Committee on Physical Activity, 2005; Heinen et al.,

200

T. van Heeswijck et al. / Health & Place 33 (2015) 195–202

Fig. 2. Odds ratio of reporting active transportation in relation to one unit increase in standardised built environment variables measured using different components of Activity Space.

2010; Sugiyama et al., 2012). A previous study found that highway density was correlated with lower cycling participation in all zones, but unlike the results from this work the relationship is strongest so along the trip path (Winters et al., 2010). Similar point estimates for the different spatial representations suggest that these relationships may not be uniquely influenced by the context of exposure. Greenness may represent the presence and quality of a variety of built environmental features including open public space, street foliage and shade or aesthetic qualities of the built environment. Exposure to green spaces is expected to encourage physically active behaviour due to preference for ‘natural’ environments (Tilt et al., 2007) but evidence for the relationship between greenness and active transportation is mixed. Subjective greenness and associated aesthetic qualities have been associated positively with walking and cycling (Tilt et al., 2007; Titze et al., 2010), and local park area was found to be related to walking and cycling for transport (Wendel-Vos et al., 2004; Zlot and Schmid, 2005). However Tilt et al. (2007) reported no association between objectively measured greenness and walking, and negative associations with walking and cycling have also been described recently (Troped et al., 2010; Winters et al., 2010). Winters et al. suggest that this result may reflect active transport users being forced to travel through less green areas that offer more direct routes, as active transportation participation is strongly negatively associated with greater trip distance. Alternatively it is possible that greener areas were more likely to be encountered at greater distances from the urban core, and were thus more likely to be encountered by individuals using motorised transportation. Total distance travelled and odds of active transportation were strongly associated in this work but stratification by distance travelled did not affect the relationship between greenness and active transportation for the combined

activity space (data not shown). Another complication is that the survey period for the origin-destination survey data used in this work ends at a time when the climate in the Montreal region is becoming colder as winter approaches. It is possible that the desirability of exposure to ‘green’ spaces changed significantly during this period, and that typically active respondents were misidentified as inactive due to less hospitable weather discouraging them from engaging in active transportation. Land Use Mix (LUM) is often highlighted as an important component of the built environment for promotion of physical activity, and is a component of the popular ‘walkability’ construct (Frank et al., 2006; Owen et al., 2007). LUM scores are typically calculated using an entropy measure similar to that employed by Frank et al., however there is some variety in the number and type of land use categories included. The strong negative correlation between LUM and physical activity found in this work is at odds with previously reported results of positive or non-significant relationships (Grasser et al., 2013; Saelens and Handy, 2008). It is possible that the inclusion of a category for industrial land use in this study’s LUM measure influenced the direction of association, as a land use type that might not be expected to promote physically active behaviour it is often omitted from LUM measures. However, a recent study demonstrated a positive relationship between a LUM measure with the same categories used in this work and walking for transport (Duncan et al., 2010). Changing the categories included in a LUM measure has previously been shown to impact the strength of relationships between LUM and walking behaviours (Christian et al., 2011), though not to the extent of reversing relationships. It would be expected for a single undesirable category to bias relationships towards the null rather than cause a reversal of direction, and this result therefore bears further

T. van Heeswijck et al. / Health & Place 33 (2015) 195–202

investigation. Since significant relationships were only found for activity space representations that included travel paths, and since these relationships were so strong, there might be something important about land use in the immediate environment during individual travel behaviour that previous studies have not been able to investigate. The cross-sectional nature of the present analysis limits our ability to make causal inferences, as existing preferences of individuals for environmental features or travel modes may cause them to move to or visit environments that support their preferred lifestyle—processes known as residential self-selection and selective daily mobility bias. (Frank et al. (2007) found that neighbourhood type preference and neighbourhood walkability score had a combined effect on walking trips undertaken, and a review of studies investigating the influence of self-selection on relationships between the built environment and travel behaviour (Cao et al., 2009) found that the extent of confounding attributed to residential self-selection varied, with the evidence supporting a significant impact of the built environment on travel behaviour remaining after analyses had accounted for self-selection. The activity spaces, active transport measures and individual covariates in this study were derived from self-reported data that are open to errors – inaccurate or under/over reporting of trips or trip segments (e.g., walking to a metro station), recall and interview bias— particularly when interviewees were responding on behalf of other household members. It is unlikely that any bias stemming from these limitations would be systematic. Most current literature faces similar issues. Another limitation of our method is the assumption that individuals travelled along the lowest travel-time route between two points. If individuals consistently avoid the fastest route due to a negative quality of the environment along it then our method may report associations between behaviour and environment that are in the opposite direction to the true association. It has also been shown that the optimal method and accuracy of estimating route choice from travel survey data varies with trip distance and mode choice—lowest travel-time is very effective for longer motorised trips but not as accurate as crow-flies estimates for walking or shortest-trip estimates for cycling (Wolf et al., 2007). Furthermore all methods lose accuracy as trip distances decrease, potentially introducing a systematic loss of accuracy in those trips for which active transportion modes are most likely to be chosen. GPS has become an attractive option for work in this area as it allows for more accurate tracking of individual movements across a survey period, but the cost and planning required to use the technology presents barriers for use with large sample populations and historical data such as that used in this work. With increasing use in the field and technological advancement GPS may become increasingly appropriate as a measurement tool for large-scale work (Matthews et al., 2009). Change in the built environment occurring between collection of travel survey data and measurement of the built environment may also lead to errors in environmental exposure measurement. Analysis of land use data from the MEGAPHONE GIS showed that up to 20% of the greater metropolitan Montreal area changed land use categories between the years 2002 and 2007 (data not shown). The majority of this change resulted from transfers between Residential and Open Space categorisations in both directions and it is unlikely that a systematic bias was introduced as a result. Access to public transport increased slightly over a similar time period with the number of bus routes serving the island increasing by 7% for the years 1998–2003 and 5% for the years 2003–2008 (Grimsrud and El-Geneidy, 2013). The sample population used in this work was an order of magnitude larger than those in the few other studies we are aware of that assess built environment/active transport relationships specifically within route and activity space contexts. The increased likelihood of type I error was compensated for by reducing α to 0.01 and

201

we were able to report the remaining statistically significant results with increased confidence. Active transportation was measured indirectly in this study based on participant’s reported travel modes. As participants were not being directly questioned about their physical activity or exercise habits any systematic misreporting of physical activity was unlikely. Walking trips constituted the majority of reported active transport so the use of a combined walking and cycling measure for active transport potentially masked specific relationships between the built environment and cycling behaviours.

5. Conclusion In this study specific to the densely populated Island of Montreal, several features of the built environment were found to be significantly related to the odds of reporting active transport use when measured using an activity space expression of environmental exposure. These relationships, though weaker, were also observed when expressing environmental exposure only within paths taken between origins and destinations. Relationships between the built environment and active transport use were generally far weaker or non-significant when environmental exposure measurement was limited to areas surrounding residences or trip origins and destinations. This study is the first to compare relationships between the built environment and both walking and cycling behaviours within residential and path based buffers, and our results strongly indicate that future work must incorporate travel paths when defining the spaces within which environmental exposures are measured.

Acknowledgements The authors express appreciation to the Ministère des Transports du Québec, Direction générale de Montréal, for enabling this research through sharing data from l’EnquêteOrigine-Destination. Data integration and spatial data were supported by a grant to Mark Daniel from the Canada Foundation for Innovation (Project no. 201252). Catherine Paquet was funded by a National Health and Medical Research Council (Australia) Post-doctoral Training Research Fellowship (#570139). Yan Kestens holds a Canadian Institutes of Health Research Applied Public Health Chair in Urban Interventions and Population Health. References Abramson, J.H., 2011. WINPEPI updated: computer programs for epidemiologists, and their teaching potential. Epidemiol. Perspect. Innov 8, 1. Adams, J., 2010. Prevalence and socio-demographic correlates of “active transport” in the UK: analysis of the UK time use survey 2005. Prev. Med. 50 (4), 199–203. Agence Métropolitaine de Transport. (2012). 1998 Montreal Origin-Destination Survey Sampling Plan. Australian Bureau of Statistics. (2009). National Health Survey: Users’ Guide. Bohannon, R.W., 1997. Comfortable and maximum walking speed of adults aged 20–79 years: reference values and determinants. Age Ageing 26 (1), 15–19. Burgoine, T., Monsivais, P., 2013. Characterising food environment exposure at home, at work, and along commuting journeys using data on adults in the UK. Int. J. Behav. Nutr. Phys. Act. 10, 85. Cao, X. (Jason), Mokhtarian, P.L., Handy, S.L., 2009. Examining the impacts of residential self selection on travel behaviour: a focus on empirical findings. Transp. Rev. 29, 359–395. Cervero, R., Kockelman, K., 1997. Travel demand and the 3Ds: density, diversity, and design. Transp. Res. Part D: Transp. Environ. 2 (3), 199–219. Chaix, B., Kestens, Y., Perchoux, C., Karusisi, N., Merlo, J., Labadi, K., 2012. An interactive mapping tool to assess individual mobility patterns in neighborhood studies. Am. J. Prev. Med. 43 (4), 440–450. Christian, H.E., Bull, F.C., Middleton, N.J., Knuiman, M.W., Divitini, M.L., Hooper, P., Giles-Corti, B., 2011. How important is the land use mix measure in

202

T. van Heeswijck et al. / Health & Place 33 (2015) 195–202

understanding walking behaviour? Results from the RESIDE study. Int. J. Behav. Nutr. Phys. Act. 8 (1), 55. Committee on Physical Activity. (2005). Does the Built Environment Influence Physical Activity?. Cummins, S., 2007. Commentary: investigating neighbourhood effects on health— avoiding the “local trap”. Int. J. Epidemiol. 36 (2), 355–357. Daniel, M., Kestens, Y., 2007. MEGAPHONE: Montreal Epidemiological and Geographic Analysis of Population Health Outcomes and Neighbourhood Effects. Canada Registered Copyright 2007 (no. 1046898). Canada. De Geus, B., De Smet, S., Nijs, J., Meeusen, R., 2007. Determining the intensity and energy expenditure during commuter cycling. Br. J. Sports Med. 41 (1), 8–12. Ding, D., Gebel, K., 2012. Built environment, physical activity, and obesity: what have we learned from reviewing the literature? Health Place 18 (1), 100–105. Duncan, M.J., Winkler, E., Sugiyama, T., Cerin, E., duToit, L., Leslie, E., Owen, N., 2010. Relationships of land use mix with walking for transport: do land uses and geographical scale matter? J. Urban Health: Bull. N.Y. Acad. Med. 87 (5), 782–795. Durand, C.P., Andalib, M., Dunton, G.F., Wolch, J., Pentz, M.a., 2011. A systematic review of built environment factors related to physical activity and obesity risk: implications for smart growth urban planning. An Official Journal of the International Association for the Study of Obesity. Obes. Rev. 12 (5), e173–e182. ESRI. (2008). ArcGIS 9.3. Redlands: ESRI. Feng, J., Glass, T.A., Curriero, F.C., Stewart, W.F., Schwartz, B.S., 2010. The built environment and obesity: a systematic review of the epidemiologic evidence. Health Place 16 (2), 175–190. Flint, E., Cummins, S., Sacker, a., 2014. Associations between active commuting, body fat, and body mass index: population based, cross sectional study in the United Kingdom. BMJ 349 (aug19 13), g4887–g4887. Frank, L.D., Andresen, M.A., Schmid, T.L., 2004. Obesity relationships with community design, physical activity, and time spent in cars. Am. J. Prev. Med. 27 (2), 87–96. Frank, L.D., Pivo, G., 1994. Impacts of mixed use and density on utilization of three modes of travel: single-occupant vehicle, transit, and walking. Transp. Res. Rec.. Frank, L.D., Saelens, B.E., Powell, K.E., Chapman, J.E., 2007. Stepping towards causation: do built environments or neighborhood and travel preferences explain physical activity, driving, and obesity? Soc. Sci. Med. 65 (9), 1898–1914. Frank, L.D., Sallis, J.F., Conway, T.L., Chapman, J.E., Saelens, B.E., Bachman, W., 2006. Many pathways from land use to health: associations between neighborhood walkability and active transportation, body mass index, and air quality. J. Am. Plann. Assoc. 72 (1), 75–87. Frank, L.D., Schmid, T.L., Sallis, J.F., Chapman, J., Saelens, B.E., 2005. Linking objectively measured physical activity with objectively measured urban form: findings from SMARTRAQ. Am. J. Prev. Med. 28 (2 Suppl. 2), 117–125. Fraser, S.D.S., Lock, K., 2011. Cycling for transport and public health: a systematic review of the effect of the environment on cycling. Eur. J. Public Health 21 (6), 738–743. Furie, G.L., Desai, M.M., 2012. Active transportation and cardiovascular disease risk factors in U.S. adults. Am. J. Prev. Med. 43 (6), 621–628. Global Advocacy Council for Physical Activity, 2010. The Toronto charter for physical activity: a global call for action. J. Phys. Act. Health 7 (Suppl. 3), S370–S385. Grasser, G., Van Dyck, D., Titze, S., Stronegger, W., 2013. Objectively measured walkability and active transport and weight-related outcomes in adults: a systematic review. Int. J. Public Health 58 (4), 615–625. Grimsrud, M., El-Geneidy, A., 2013. Driving transit retention to renaissance: trends in Montreal commute public transport mode share and factors by age group and birth cohort. Public Transp. 5 (3), 219–241. Hamer, M., Chida, Y., 2008. Active commuting and cardiovascular risk: a metaanalytic review. Prev. Med. 46 (1), 9–13. Heinen, E., van Wee, B., Maat, K., 2010. Commuting by bicycle: an overview of the literature. Transp. Rev. 30 (1), 59–96. IPAQ Group, 2005. Guidelines for Data Processing and Analysis of the International Physical Activity Questionnaire (IPAQ)—Short and Long Forms (pp. 1–15). Kestens, Y., Lebel, A., Daniel, M., Thériault, M., Pampalon, R., 2010. Using experienced activity spaces to measure foodscape exposure. Health Place 16 (6), 1094–1103. Killingsworth, R.E., De Nazelle, A., Bell, R.H., 2003. Building a new paradigm: improving public health through transportation. ITE J.. Kwan, M.-P., 2009. From place-based to people-based exposure measures. (1982). Soc. Sci. Med. 69 (9), 1311–1313. Lebel, A., Kestens, Y., Pampalon, R., Thériault, M., Daniel, M., Subramanian, S.V., 2012. Local context influence, activity space, and foodscape exposure in two

canadian metropolitan settings: is daily mobility exposure associated with overweight?. 912645. J. Obes. 2012. Lopez-Zetina, J., Lee, H., Friis, R., 2006. The link between obesity and the built environment. Evidence from an ecological analysis of obesity and vehicle miles of travel in California. Health Place 12 (4), 656–664. Matthews, C.E., Jurj, A.L., Shu, X.-O., Li, H.-L., Yang, G., Li, Q., Zheng, W., 2007. Influence of exercise, walking, cycling, and overall nonexercise physical activity on mortality in Chinese women. Am. J. Epidemiol. 165 (12), 1343–1350. Matthews, S., Moudon, A.V., Daniel, M., 2009. Work group II: using geographic information systems for enhancing research relevant to policy on diet, physical activity, and weight. Am. J. Prev. Med. 36 (4 Suppl.), S171–S176. Moudon, A.V., Lee, C., 2003. Walking and Biking: An Evaluation of Environmental Audit Instruments: , (February), 1–42. National Public Health Partnership, 2001. Promoting Active Transport: An Intervention Portfolio to Increase Physical Activity as a Means of Transport. Oja, P., Vuori, I., Paronen, O., 1998. Daily walking and cycling to work: their utility as health-enhancing physical activity. Patient Educ. Counsel. 33 (1 Suppl.), S87–S94. Owen, N., Cerin, E., Leslie, E., duToit, L., Coffee, N., Frank, L.D., Sallis, J.F., 2007. Neighborhood walkability and the walking behavior of Australian adults. Am. J. Prev. Med. 33 (5), 387–395. Rai, R.K., Balmer, M., Rieser, M., Vaze, V.S., Schönfelder, S., Axhausen, K.W., 2007. Capturing human activity spaces: new geometries. Transp. Res. Rec. 2021 (-1), 70–80. Riva, M., Gauvin, L., Apparicio, P., Brodeur, J.-M., 2009. Disentangling the relative influence of built and socioeconomic environments on walking: the contribution of areas homogenous along exposures of interest. (1982). Soc. Sci. Med. 69 (9), 1296–1305. Saelens, B.E., Handy, S.L., 2008. Built environment correlates of walking: a review. Med. Sci. Sports Exerc. 40 (7 Suppl.), S550–S566. Sallis, J.F., Cervero, R.B., Ascher, W., Henderson, K.a., Kraft, M.K., Kerr, J., 2006. An ecological approach to creating active living communities. Annu. Rev. Public Health 27, 297–322. Schönfelder, S., Axhausen, K.W., 2003a. Activity spaces: measures of social exclusion? Transp. Policy 10 (4), 273–286. Schönfelder, S., Axhausen, K.W., 2003b. On the Variability of Human Activity Spaces. Arbeitsbericht Verkehrs- Und Raumplanung, 149. Statistics Canada. (1996). Coverage: 1996 Census Technical Reports. Statistics Canada. (2002). A National Overview, Population and Dwelling Counts. Sugiyama, T., Neuhaus, M., Cole, R., Giles-Corti, B., Owen, N., 2012. Destination and route attributes associated with adults’ walking: a review. Med. Sci. Sports Exerc. 44 (7), 1275–1286. Tilt, J.H., Unfried, T.M., Roca, B., 2007. Using objective and subjective measures of neighborhood greenness and accessible destinations for understanding walking trips and BMI in Seattle, Washington. Am. J. Health Promot.: AJHP 21 (4 Suppl.), 371–379. Titze, S., Giles-Corti, B., Knuiman, M.W., Pikora, T.J., Timperio, A., Bull, F.C., van Niel, K., 2010. Associations between intrapersonal and neighborhood environmental characteristics and cycling for transport and recreation in adults: baseline results from the RESIDE study. J. Phys. Act. Health 7 (4), 423–431. Troped, P.J., Wilson, J.S., Matthews, C.E., Cromley, E.K., Melly, S.J., 2010. The built environment and location-based physical activity. Am. J. Prev. Med. 38 (4), 429–438. Wasfi, R.A., Ross, N.A., El-geneidy, A.M., 2013. Achieving recommended daily physical activity levels through commuting by public transportation: unpacking individual and contextual influences. Health Place 23 (23), 18–25. Wendel-Vos, G.C.W., Schuit, a.J., De Niet, R., Boshuizen, H.C., Saris, W.H.M., Kromhout, D., 2004. Factors of the physical environment associated with walking and bicycling. Med. Sci. Sports Exerc. 36 (4), 725–730. Winters, M., Brauer, M., Setton, E.M., Teschke, K., 2010. Built environment influences on healthy transportation choices: bicycling versus driving. J. Urban Health: Bull. N.Y. Acad. Med. 87 (6), 969–993. Wolf, J., Bonsall, P., Oliveira, M., Leary, L., Lee, M., 2007. Review of the Potential Role of ‘New Technologies’ in the National Travel Survey (p. 94). Zenk, S.N., Schulz, A.J., Matthews, S.a., Odoms-Young, A., Wilbur, J., Wegrzyn, L., Stokes, C., 2011. Activity space environment and dietary and physical activity behaviors: a pilot study. Health Place 17 (5), 1150–1161. Zlot, A.I., Schmid, T.L., 2005. Relationships among community characteristics and walking and bicycling for transportation or recreation. Am. J. Health Promot. 19 (4), 314–317.

Differences in associations between active transportation and built environmental exposures when expressed using different components of individual activity spaces.

This study assessed relationships between built environmental exposures measured within components of individual activity spaces (i.e., travel origins...
717KB Sizes 0 Downloads 5 Views