Differences in physical environmental characteristics between adolescents’ actual and shortest cycling routes: a study using a Google Street View-based audit

Differences in physical environmental characteristics between adolescents’ actual and shortest... Background: The objective evaluation of the physical environmental characteristics (e.g. speed limit, cycling infra- structure) along adolescents’ actual cycling routes remains understudied, although it may provide important insights into why adolescents prefer one cycling route over another. The present study aims to gain insight into the physical environmental characteristics determining the route choice of adolescent cyclists by comparing differences in physi- cal environmental characteristics between their actual cycling routes and the shortest possible cycling routes. Methods: Adolescents (n = 204; 46.5% boys; 14.4 ± 1.2 years) recruited at secondary schools in and around Ghent (city in Flanders, northern part of Belgium) were instructed to wear a Global Positioning System device in order to identify cycling trips. For all identified cycling trips, the shortest possible route that could have been taken was cal- culated. Actual cycling routes that were not the shortest possible cycling routes were divided into street segments. Segments were audited with a Google Street View-based tool to assess physical environmental characteristics along actual and shortest cycling routes. Results: Out of 160 actual cycling trips, 73.1% did not differ from the shortest possible cycling route. For actual cycling routes that were not the shortest cycling route, a speed limit of 30 km/h, roads having few buildings with windows on the street side and roads without cycle lane were more frequently present compared to the shortest pos- sible cycling routes. A mixed land use, roads with commercial destinations, arterial roads, cycle lanes separated from traffic by white lines, small cycle lanes and cycle lanes covered by lighting were less frequently present along actual cycling routes compared to the shortest possible cycling routes. Conclusions: Results showed that distance mainly determines the route along which adolescents cycle. In addi- tion, adolescents cycled more along residential streets (even if no cycle lane was present) and less along busy, arterial roads. Local authorities should provide shortcuts free from motorised traffic to meet adolescents’ preference to cycle along the shortest route and to avoid cycling along arterial roads. Keywords: Active transport, Cycling, Route choice, Physical environment, Audit, Youth *Correspondence: hannah.verhoeven@ugent.be Department of Public Health, Faculty of Medicine and Health Sciences, Ghent University, Corneel Heymanslaan 10, 9000 Ghent, Belgium Full list of author information is available at the end of the article © The Author(s) 2018. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/ publi cdoma in/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 2 of 15 Compared to the shortest routes, land use mix (i.e. the Background extent to which several types of land use, such as resi- Air pollution, which is partially caused by vehicle emis- dential and industrial areas, shops, services, are included sions, is consistently related to acute respiratory infec- in an area) was significantly higher along actual cycling tions among young children, cardiopulmonary disease routes. A study among children in the Netherlands and lung cancer [1]. By replacing private car use (passive (8–12  years) found that there were significantly fewer transport) by active modes of transport such as cycling, trees, zebra crossings and sidewalks along actual cycling carbon dioxide emissions can be reduced substan- routes compared to the shortest routes [17]. In addition, tially [2]. Although the risk of a higher intake of carbon actual cycling routes had significantly more traffic lights, dioxide can be considered as a negative aspect of active junctions and a higher chance of being on residential transport [3], a growing body of evidence emphasizes streets compared to the shortest routes. Safety showed the potential benefits of cycling for transport for public thus to be an important factor among children in this health [2, 4]. Since adolescence is characterised by a steep study. According to Dessing et al. [17], most of the zebra decrease in physical activity levels [5], increasing cycling crossings in the Netherlands are located on or near busy for transport is also a promising strategy to meet the rec- streets, that were avoided by the children. Furthermore, ommended 60 min of daily physical activity among ado- when main roads have to be crossed children preferred lescents [4, 6]. Cycling for transport has been associated signalized intersections. Because of some inconsist- with higher levels of cardiorespiratory fitness [7] and ent results across these previous studies, similar studies lower levels of overweight [8] among adolescents and it among adolescents may provide additional insights into can easily be incorporated into their daily lives once the which physical environmental factors are related to an skills for cycling have been acquired [9]. individuals’ route choice. The role of the physical environment for health behav - Methodologies to assess the physical environment iours such as cycling for transport has been acknowl- include both subjective and objective measurements. edged by socio-ecological models and previous research Subjective measurements, such as self-reported ques- [10–12]. However, the majority of previous studies inves- tionnaires, encounter limitations such as recall bias [18] tigating physical environmental correlates of cycling for and may not accurately assess the effect of the actual transport focused on the neighbourhood environment physical environmental factors on cycling for transport close to home, although cycling for transport does not [11]. Therefore, observational field audits are frequently necessarily take place in the immediate neighbourhood applied as an objective tool for measuring the physical environment. Nevertheless, the evaluation of physi- environment related to physical activity [19–21]. Vanwol- cal environmental characteristics along adolescents’ leghem et al. [22] developed EGA-Cycling (Environmen- actual cycling routes remains understudied, although tal Google Street View Based Audit-Cycling) to virtually it is important to find out why individuals chose a spe - assess physical micro- and macro-environmental charac- cific cycling route. In addition, although previous studies teristics along cycling routes using Google Street View. emphasized the importance of distance for adolescents’ EGA-Cycling was based on existing audit instruments cycling for transport [12–14], it is likely that adoles- (e.g. Pikora-SPACES instruments [20], Audit Tool Check- cents do not always take the shortest cycling route. By list version [21], Irvine-Minnesota Inventory [23]), but comparing adolescents’ actual cycling routes with the was adapted to the Flemish street infrastructure. In the shortest possible cycling routes, important information last decade, using virtual technologies, such as Google regarding which physical environmental characteristics Street View, to assess the physical environment is gain- determine the route choice of adolescent cyclists may be ing attention [24–29]. Auditors are able to virtually walk obtained. Among adults, two recent studies compared through a street which is time- and cost-saving [24, 28] physical environmental characteristics of actual and and they are not exposed to unsafe (traffic) situations shortest cycling routes [15, 16]. Winters et al. [16] found compared to field audits. Previous studies showed good that actual cycling routes of Canadian adults had signifi - agreement between virtual and field audit tools [24, cantly more traffic calming facilities (e.g. traffic circles or 26, 29]. However, virtual audit tools showed to be less median barriers to slow or block motorized traffic) and accurate when measuring micro-environmental charac- participants cycled less along arterial (busy) roads and teristics (e.g. litter, sidewalk condition) [24, 26, 28]. Nev- more along local roads, off-street paths and roads with ertheless, Ben-Joseph et  al. [28] concluded that Google cycling facilities. Krenn et  al. [15] also found that Aus- Street View was more accurate in measuring small fea- trian cyclists avoid busy roads and prefer roads with cycle tures compared to Google Maps and MS Visual Oblique. lanes. Actual cycling routes included more green and The aim of the present study is to gain insight into the aquatic areas and had fewer traffic lights, fewer cross - physical environmental characteristics determining the ings and less hilly roads compared to the shortest routes. Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 3 of 15 route choice of adolescent cyclists by comparing differ - adolescents (response rate = 84.1%) participating in the ences in physical environmental characteristics between study. their actual cycling routes and the shortest possible cycling routes using a Google Street View-based audit Study protocol (EGA-Cycling). The study protocol consisted of two parts (see Fig.  1 for a flow chart). In the first part of the study, each participat - ing school was visited three times by the research team Methods between September and December 2015. During a first Participants visit, the purpose of the study was explained to the ado- A convenience sample of 12 secondary schools in and lescents and informed consent was obtained. Each par- around Ghent was contacted to participate in the study. ticipant received a unique ID number in order to be able Ghent is a city in Flanders, northern part of Belgium, to link data of all measurements. Participants completed that has 253,266 inhabitants and comprises an area of a questionnaire assessing socio-demographics. Further- 2 2 156.2  km (population density: 1622  h/km ) [30, 31]. In more, participants received a Global Positioning System the six schools that agreed to participate, school prin- (GPS) device and a charger for the device together with cipals or staff members randomly selected at least two verbal and written instructions on how and when to wear classes from the first to fourth grade (12–16  years). A the device. All participants were instructed to wear the total of 18 classes was selected and 283 adolescents were GPS device, which was attached to their waist with an invited to participate in the study. Only participants who elastic belt, during waking hours until the research team were present at school when measurement materials returned to the school to collect the devices (4–5  days were handed out, could be included in the study. Pas- later). During activities that could damage the GPS sive informed consent was obtained from adolescents’ device or during which it could be uncomfortable to parents. If parents did not agree to let their child partici- wear it (e.g. showering, swimming or rugby), the adoles- pate in the study, they had to sign a form. Furthermore, cents were asked to temporarily remove the GPS device. researchers also obtained active informed consent from They were also instructed not to turn off the GPS device adolescents. This procedure resulted in a group of 238 during data collection. Participants were asked for their Fig. 1 Flow chart Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 4 of 15 Structured one‑on‑one interview with personal travel maps mobile phone number. Two text messages per day (in the GPS data were stored in a PostgreSQL database with morning and evening) were sent to the participants will- PostGIS in order to generate a personal travel map ing to give their number in order to remind them to wear per day in the web application. This web application the GPS devices and to charge it. During a second visit, showed the geographical position of participants for researchers returned to the schools to collect the devices. every 30-second-interval. Figure  2 shows an example Afterwards, the GPS data were downloaded and a web of a personal travel map. These personal travel maps application was created in order to visualize the data on were used as a guide to conduct a structured one-on- a personal travel map. During the last visit, which took one interview discussing routes on two selected days. place within the first week after collection of the GPS The first week- and weekend day with complete data devices, researchers conducted a structured one-on-one (excluding the day the devices were handed out) were interview (Additional file  1) during which a researcher selected for the interview. When no weekdays with chronologically discussed the personal travel maps. Per complete data were available, two weekend days were trip travelled, participants were asked about their trans- selected and vice versa. When only 1 day with complete port mode and why they took a particular route. Partici- data was available, the structured interview was com- pants who completed all measurements and returned the pleted for 1  day. During these interviews, a researcher GPS device received an incentive (i.e. movie ticket). chronologically identified, together with the partici - In the second part of the study, adolescents’ cycling pants, the trips they made during a day. For each iden- routes were selected, and for each actual cycling route tified trip, the participant was asked which transport the shortest cycling route was calculated using Google mode was used. For active trips (walking or cycling/ Maps. For each cycling route which was not the short- skateboard/…) the participant was also asked why est cycling route, an adapted version of EGA-Cycling he/she chose that particular route to reach his/her was used to obtain information about physical environ- destination. mental characteristics along adolescents’ actual cycling routes and along the corresponding shortest routes using Google Street View. The study protocol was approved by the Ethics Com - GPS data processing mittee of the University Hospital of Ghent University (EC Data processing was executed using the Personal Activ- 2015/0317). ity and Location Measurement System (PALMS©) [34, 35]. PALMS filtered invalid GPS data when extreme speed (> 150  km/h) or extreme changes in distance Measurements and data processing (> 1000 m) or elevation (> 100 m) between two consecu- Questionnaire tive data points were identified. The programming soft - Participants completed a paper-and-pencil questionnaire ware Python was used to combine the PALMS dataset assessing following socio-demographics: home address, with information on school schedules of each participat- gender, date of birth, grade (first to fourth year), educa - ing class, school addresses and home addresses of par- tional type (general, technical or vocational) and high- ticipants. PALMS categorised data into location (home, est education of parents (primary education, secondary school, leisure) or transport. Data were categorised in the education, tertiary education-non university, tertiary domain ‘transport’ when a trip was detected. A trip was education-university, I don’t know). Education of parents defined as a period of at least 3  min of movement with was used to as a proxy for socio-economic status (SES). the same transport mode, allowing for stationary periods Adolescents were identified as being ‘of a higher SES fam - of maximum 3 min. PALMS classified all trips into walk - ily’ when at least one parent completed tertiary education ing, cycling or motorised transport based on speed. A [32]. trip was classified as walking when speed was between 1 and 9 km/h, cycling between 10 and 24 km/h and motor- GPS device ised transport starting from 25 km/h [33, 36]. The geographical position of participants was recorded Subsequently, all data from the structured one-on-one by the QStarz BT-Q1000X GPS device. In addition, the interviews were inserted into the database. For trips or GPS device recorded participants’ speed which was used locations that were misclassified by PALMS [e.g. when a to define their transport mode [33]. The GPS devices car trip was classified as a bicycle trip due to traffic con - were set to collect data every 30  s using Q-travel soft- gestion (speed < 25 km/h)], corrections were made based ware. Furthermore, the devices were set to stop logging on the data of the structured interviews. The number when the memory was full (this did not occur during of corrections due to misclassification by PALMS was data collection). Q-travel software was used to download rather limited. the collected GPS data. Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 5 of 15 Fig. 2 Example of a personal travel map. Every 30 s a dot was placed on the map (temporal resolution: 30 s). The green arrow represents the first data point of the day registered by the GPS and the ‘finish flag’ represents the last registered data point of the day by the GPS EGA‑cycling shortest possible cycling routes were selected to be used EGA-Cycling (Additional file  2) consists of five subscales in subsequent analyses. All routes for which the actual and includes 37 items: (1) land use (8 items; e.g. commer- cycling route was not the shortest possible cycling route cial destinations, heavy industry and public destinations), were included, even if only one segment differed between (2) general characteristics of the street segment (12 the actual and the shortest cycling route. Differences in items; e.g. road type and speed limit), (3) cycling facilities distance between actual cycling routes and the short- (7 items; e.g. type and width of cycle lane), (4) pedestrian est possible cycling routes were calculated absolutely in facilities (3 items; e.g. presence and maintenance of the meters as well as relatively in percentage of the shortest sidewalk) and (5) aesthetics (7 items; e.g. trees and front cycling route (reported as ‘detour’). Google My Maps (a yards). EGA-Cycling shows acceptable reliability and Google Maps application) was used to visualize actual validity [22]. However, since measures about (safety at) cycling routes and the corresponding shortest cycling intersections are very limited from this tool, three addi- routes. Each cycling route was manually divided into tional items regarding this topic were added (i.e. amount several street segments (average distance: 342 ± 468  m), of side streets, amount of intersections and visibility at a new street segment started when participants turned the corners). The item regarding visibility at the corners into another street or when the street name changed. For is part of the Microscale Audit of Pedestrian Streetscape each street segment, EGA-Cycling was filled out by one (MAPS) Global tool [37]. Furthermore, another item was out of three trained observers (the first author and two included that assessed whether or not the street segment independent observers). Google Street View was used to concerned a walking/cycling road (i.e. a separate road perform the audits, which took approximately 2 weeks per only accessible for non-motorised traffic). Data on dif - observer (6  weeks in total). Google Street View images ferences in pedestrian facilities between actual cycling ranged from March 2009 till April 2015. The majority routes and the shortest possible cycling routes are not (53.0%) of images were taken between August 2014 and shown since they are not relevant for cycling. October 2014. Prior to auditing the pre-defined routes, two independent observers were trained by the first author. The training included specific instructions; all items of the EGA-Cycling tool were explained and illus- Auditing of actual and shortest cycling routes trated with photographs if necessary. Subsequently, the For all cycling trips that could be identified in the previ - observers audited three random street segments which ous steps, the shortest cycling route was calculated using enabled them to raise questions. Thereafter, five test Google Maps. Only actual cycling routes that were not the Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 6 of 15 routes (i.e. no routes that were part of the study) were the dataset as were participants who were absent when rated by the first author and the two independent observ - the structured interviews were completed (n = 17). A ers. Prior to auditing the pre-defined routes, 95% agree - final sample of 204 adolescents (85.7%) was used for ment with the first author’s scores was required. For the data analyses (46.5% boys, 14.4 ± 1.2  years). Table  1 pre- actual audits, only street segments for which there was no sents descriptive characteristics of the sample. Within overlap between the actual cycling route and the short- this sample, a total of 1126 trips was identified. Passive est cycling route were audited (516 segments). Distances transport (car, as a passenger) was used most frequently of segments were measured in Google My Maps. Figure 3 (34.6% of trips), followed by public transport (33.9% of shows examples of actual versus shortest cycling routes. trips). Active transport such as walking and cycling was used for 17.2 and 14.2% of trips, respectively. The pur - Data analyses pose of a trip and the transport mode used showed to be Data were analysed using IBM SPSS Statistics 24. A related to each other (Chi = 257.1; p < 0.001). For trips paired samples t-test was used to calculate the difference to and/or from school, the majority (57.2%) was done by in distance between actual and shortest cycling routes. public transport, 18.4% was done by bicycle, 12.6% by Because EGA-Cycling was developed to assess physical foot and 11.8% by passive transport. For leisure-related environmental characteristics along entire cycling routes instead of individual segments [22], a total score per cycling route was calculated for each item. Per item, the Table 1 Descriptive characteristics of the sample (n = 204) score for a particular segment was multiplied by the dis- Socio-demographic characteristics tance of that segment. These weighted item scores of sev - eral segments of a route were summed to obtain one total Gender (% boys) 46.5 score per route for that item. Subsequently, item scores Age (years; mean ± SD) 14.4 ± 1.2 were expressed in m/km in order to be able to compare Socio-economic status (SES) parents (%) the actual cycling route with the shortest cycling route Lower SES (% no parent completed tertiary education) 28.4 (for which the route length differed). Univariate multi - Higher SES (% at least one parent completed tertiary educa- 71.6 level logistic regression analyses were used to investigate tion) differences in physical environmental characteristics Grade (%) between actual and shortest cycling routes (three levels: 1st year of secondary school 8.3 participant, route and street segment). Statistical signifi - 2nd year of secondary school 7.4 cance was set at p < 0.05. 3rd year of secondary school 46.1 4th year of secondary school 38.2 Results Educational type (%) Sample characteristics General education 65.2 From the 238 adolescents participating, adolescents older Technical education 10.3 than 17 years (n = 4) and participants who did not wear/ Vocational education 24.5 charge the material properly (n = 13) were removed from Fig. 3 Examples of actual versus shortest cycling routes Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 7 of 15 trips, nearly half (49.6%) was done by passive transport, and the shortest possible cycling routes (detour) was 20.1% was done by foot, 17.8% by public transport and 516 ± 369 m (med = 400 m; min = 100 m; max = 1600 m). 12.5% by bicycle. The median distance for car trips was The average detour was 15.6% (med = 12.1%; min = 2.0%; 6312  m and for public transport a median distance of max = 45.7%) in comparison to the shortest possible 4934 m was found. Walking trips had a median distance cycling route. of 710 m, whereas for cycling trips a median distance of Table 2 presents results on differences in physical envi - 2633 m was found. ronmental characteristics concerning land use between Out of 160 actual cycling trips, 73.1% did not differ actual cycling routes and the shortest possible cycling from the shortest possible cycling routes. Thirty-eight routes. An increase in 100 m/km of mixed land use along unique cycling routes for which the actual route dif- the actual cycling route, resulted in 16% lower odds fered from the shortest possible cycling route could that the actual cycling route was chosen over the short- be identified (see Fig.  4). The 38 routes were spread est cycling route. In addition, an increase in 100  m/km over 22 adolescents, with a range of 1 to maximum 4 where commercial destinations are present along the routes per person. A significant difference in distance actual cycling route, resulted in 17% lower odds that the between actual and shortest possible cycling routes actual cycling route was chosen over the shortest cycling was found (t = 8.606; p < 0.001). Actual cycling routes route. had a mean distance of 4505 ± 2201  m (med = 4100  m; Table  3 presents results on differences in general char - min = 1000  m; max = 8800  m), whereas for the shortest acteristics between actual cycling routes and the short- possible cycling routes a mean distance of 3989 ± 2048 m est possible cycling routes. An increase in 100  m/km of (med = 3600  m; min = 700  m; max = 8100 m) was found. a road type which consists of two roads divided in two The mean difference between actual cycling routes lanes each direction (i.e. arterial road) along the actual Fig. 4 Overview of the 38 actual cycling routes that differed from the shortest possible cycling route Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 8 of 15 Table 2 Presence of items on land use along actual cycling routes compared to shortest cycling routes Item Actual cycling route Shortest cycling route OR (95% CI) (m/km; M ± SD) (m/km; M ± SD) Mixed land use 256 ± 226 386 ± 317 0.84 (0.71; 1.00)* Types of buildings Single buildings 155 ± 247 153 ± 256 1.00 (0.83; 1.21) Closed/semi-detached buildings 225 ± 190 139 ± 182 1.30 (0.99; 1.70) Apartment buildings 111 ± 200 161 ± 244 0.90 (0.73; (1.12) Commercial destinations 233 ± 229 367 ± 304 0.83 (0.69; 0.99)* Heavy industry 9 ± 39 17 ± 63 0.74 (0.29; 1.91) Public destinations 248 ± 196 355 ± 300 0.84 (0.70; 1.02) Recreational destinations 85 ± 105 121 ± 212 0.87 (0.65; 1.17) Natural features 315 ± 314 257 ± 286 1.07 (0.91; 1.25) Openness view Open view 65 ± 174 24 ± 81 1.29 (0.85; 1.96) Not open/closed view 354 ± 229 370 ± 271 0.98 (0.81; 1.17) Closed view 166 ± 168 146 ± 185 1.07 (0.82; 1.39) Reference = shortest cycling route. For ease of interpretation of OR, distances were converted to hectometres (100 m/km) OR odds ratio, CI confidence interval *p ≤ 0.05; **p ≤ 0.01; p ≤ 0.1 cycling route, resulted in 47% lower odds that the actual cycling route was chosen over the shortest cycling cycling route was chosen over the shortest cycling route. route. Finally, for an increase in 100  m/km of a cycle An increase in 100 m/km with a speed limit of 30  km/h lane that is covered by lighting along the actual cycling along the actual cycling route, resulted in 50% higher route, 25% lower odds that the actual cycling route was odds that the actual cycling route was chosen over the chosen over the shortest cycling route was found. shortest cycling route. Furthermore, an increase in Table  5 presents results on differences in aesthetics 100  m/km for roads where few buildings with windows between actual cycling routes and the shortest possible on the street are present along the actual cycling route, cycling routes. For none of the included variables, a sig- resulted in 192% higher odds that the actual cycling route nificant result was found. was chosen over the shortest cycling route. This last item Subjective results of the structured one-on-one inter- refers to crime safety/social control (i.e. if few buildings views showed that, for the 38 actual cycling routes that with windows on the street (or few buildings in general) differed from the shortest possible cycling route, ado - are present, few people have a clear view on the street lescents still indicated for 35.1% (n = 13) of the trips and there is thus less social control) [22]. that they chose that route because it was the shortest/ Table  4 presents results on differences in cycling fastest route. For 16.2% (n = 6) of these cycling trips, facilities between actual cycling routes and the short- participants indicated they chose that particular route est possible cycling routes. For an increase in 100  m/ to cycle together with friends/siblings/…. For another km of a cycle lane which is part of the road (cycle lane 16.2% (n = 6) of the trips, they chose that particular separated from traffic by white lines) along the actual route because of lower traffic density. Furthermore, for cycling route, 36% lower odds that the actual cycling 13.5% (n = 5) of the trips participants indicated that route was chosen over the shortest cycling route was route choice was determined by their parents and for found. For an increase in 100 m/km road with no cycle 5.4% (n = 2) of the trips they indicated to choose that lane along the actual cycling route, 25% higher odds particular route because of the presence of few/safe that the actual cycling route was chosen over the short- crossings. For another 5.4% (n = 2) of the trips, par- est cycling route was found. In addition, an increase in ticipants indicated they chose that particular route 100 m/km of a small cycle lane along the actual cycling because of a commercial destination they wanted to route, resulted in 32% lower odds that the actual visit. Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 9 of 15 Table 3 Presence of items on general characteristics along actual cycling routes compared to shortest cycling routes Item Actual cycling route Shortest cycling route OR (95% CI) (m/km; M ± SD) (m/km; M ± SD) Road type Walking/cycling road 89 ± 99 62 ± 111 1.29 (0.82; 2.03) One road for one-direction traffic 120 ± 168 56 ± 107 1.43 (0.97; 2.11) One road not divided into lanes 259 ± 232 160 ± 195 1.25 (0.99; 1.57) One road divided in one lane each direction 82 ± 125 123 ± 199 0.86 (0.64; 1.15) Two roads divided in one lane each direction 21 ± 58 7 ± 24 2.44 (0.63; 9.41) Two roads divided in two lanes each direction 14 ± 46 130 ± 245 0.53 (0.28; 0.99)* Speed limit 30 km/h 145 ± 185 58 ± 131 1.50 (1.02; 2.21)* 50 km/h 309 ± 261 316 ± 255 0.99 (0.83; 1.18) 70 km/h or more 43 ± 87 106 ± 225 0.78 (0.57; 1.08) Traffic calming measures 248 ± 222 344 ± 304 0.87 (0.73; 1.04) Amount of side streets 13 ± 11 11 ± 10 1.02 (0.98; 1.07) Amount of intersections 2 ± 2 3 ± 3 0.90 (0.75; 1.09) Crossing aids 372 ± 247 414 ± 310 0.95 (0.80; 1.12) Poor visibility when crossing a street 23 ± 54 4 ± 16 4.85 (0.80; 29.52) Well-maintained street segment 548 ± 267 531 ± 282 1.02 (0.86; 1.21) Streetlights 522 ± 265 536 ± 290 0.98 (0.83; 1.16) Parking facilities On street parking facilities 180 ± 173 122 ± 156 1.25 (0.93; 1.68) Parking facilities next to the street 210 ± 206 321 ± 297 0.84 (0.69; 1.02) Parking facilities on adjacent parking 18 ± 62 3 ± 15 2.98 (0.44; 20.35) No parking facilities 88 ± 193 33 ± 55 1.45 (0.86; 2.44) Slope Flat 546 ± 266 499 ± 297 1.06 (0.90; 1.25) Gentle to moderate slope 39 ± 62 41 ± 105 0.97 (0.56; 1.66) Swerving alternatives 407 ± 254 353 ± 266 1.08 (0.91; 1.30) Buildings No buildings with windows on street side 47 ± 182 30 ± 82 1.10 (0.78; 1.55) Few buildings with windows on street side 58 ± 78 22 ± 45 2.92 (1.12; 7.63)* Many buildings with windows on street side 391 ± 248 427 ± 293 0.95 (0.80; 1.13) Driveways No driveways 54 ± 136 65 ± 105 0.93 (0.63; 1.37) Approx. 25% of buildings have one driveway 181 ± 207 167 ± 238 1.03 (0.84; 1.27) Approx. 50% of buildings have one driveway 17 ± 29 35 ± 88 0.61 (0.25; 1.51) Most buildings have one driveway 244 ± 247 213 ± 259 1.05 (0.88; 1.26) Garages No garages 228 ± 240 187 ± 175 1.10 (0.88; 1.38) Approx. 25% of buildings have one garage 252 ± 228 270 ± 273 0.97 (0.81; 1.17) Approx. 50% of buildings or more have one garage 17 ± 32 22 ± 47 0.72 (0.22; 2.32) Reference = shortest cycling route. Results regarding ‘one road divided in two lanes each direction’ (road type) are not shown since this road type did not appear along the routes. For ease of interpretation of OR, distances were converted to hectometres (100 m/km) OR odds ratio, CI confidence interval *p ≤ 0.05; **p ≤ 0.01; p ≤ 0.1 Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 10 of 15 Table 4 Presence of items on cycling facilities along actual cycling routes compared to shortest cycling routes Item Actual cycling route Shortest cycling route OR (95% CI) (m/km; M ± SD) (m/km; M ± SD) Type of cycle lane Cycle lane separated from the road 74 ± 133 45 ± 75 1.30 (0.82; 2.08) Adjoining cycle lane (slightly increased) 76 ± 105 89 ± 141 0.91 (0.63; 1.33) Cycle lane is part of the road (white lines) 52 ± 84 180 ± 252 0.64 (0.44; 0.92)* Non-compulsory cycle lane or of a different colour 24 ± 101 7 ± 29 1.45 (0.65; 3.25) No cycle lane 271 ± 243 159 ± 209 1.25 (1.01; 1.56)* Width cycle lane Small 72 ± 106 166 ± 206 0.68 (0.48; 0.96)* Wide 242 ± 260 237 ± 276 1.01 (0.85; 1.20) Two-way cycle lane 224 ± 254 121 ± 203 1.23 (0.98; 1.54) Well-maintained cycle lane 294 ± 240 376 ± 286 0.89 (0.74; 1.06) Lighting covering cycle lane 174 ± 182 332 ± 284 0.75 (0.61; 0.94)** Surface cycle lane Bitumen 273 ± 177 260 ± 237 1.03 (0.83; 1.29) Continuous concrete 8 ± 37 5 ± 17 1.50 (0.27; 8.41) Paving bricks 181 ± 199 126 ± 131 1.22 (0.91; 1.63) Concrete slabs 80 ± 112 109 ± 181 0.73 (0.64; 1.20) Cobblestones 12 ± 30 16 ± 44 0.71 (0.20; 2.48) Gravel 32 ± 77 23 ± 92 1.13 (0.65; 1.98) Condition cycle lane Poor 28 ± 58 18 ± 78 1.25 (0.61; 2.58) Moderate 223 ± 182 257 ± 258 0.93 (0.76; 1.15) Good 335 ± 226 264 ± 229 1.15 (0.93; 1.41) Reference = shortest cycling route. For ease of interpretation of OR, distances were converted to hectometres (100 m/km) OR odds ratio, CI confidence interval *p ≤ 0.05; **p ≤ 0.01; p ≤ 0.1 Table 5 Presence of items on aesthetics along actual cycling routes compared to shortest cycling routes Item Actual cycling route (m/km; M ± SD) Shortest cycling route (m/km; M ± SD) OR (95% CI) Trees 459 ± 241 428 ± 294 1.04 (0.88; 1.24) Attractive buildings 60 ± 111 70 ± 134 0.94 (0.65; 1.37) Well-maintained buildings 501 ± 240 500 ± 287 1.00 (0.84; 1.19) Front yards 297 ± 257 328 ± 294 0.96 (0.81; 1.13) Well-maintained front yards 315 ± 247 398 ± 274 0.88 (0.73; 1.07) Attractive natural features 250 ± 310 163 ± 243 1.12 (0.95; 1.33) Graffiti and litter 120 ± 201 93 ± 205 1.07 (0.85; 1.35) Reference = shortest cycling route. For ease of interpretation of OR, distances were converted to hectometres (100 m/km) OR odds ratio, CI confidence interval *p ≤ 0.05; **p ≤ 0.01; p ≤ 0.1 Discussion arterial roads, cycle lanes separated from traffic by white The present study aimed to investigate differences in lines, small cycle lanes and cycle lanes covered by light- physical environmental characteristics between ado- ing were less frequently present along adolescents’ actual lescents’ actual cycling routes and the shortest possible cycling routes in comparison to the shortest possible cycling routes using a Google Street View-based audit. cycling routes. Besides, a speed limit of 30  km/h, roads A mixed land use, roads with commercial destinations, having few buildings with windows on street side and Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 11 of 15 roads without cycle lane were more frequently present routes included a larger part of the route where only few along actual cycling routes compared to the shortest pos- buildings with windows on the street were present, which sible cycling routes. is also an attribute of local roads. As already mentioned In line with previous studies [12–14], the present study above, the presence of buildings with windows on the showed that a short cycling distance is one of the most street refers to social control from people living in the important factors determining the route choice of ado- area [22]. Another study among 5-to-18-year-old youth lescent cyclists, as for 73.1% of the cycling trips partici- found similar results, participants in that study agreeing pants took the shortest possible route. In addition, for that ‘walkers and bikers on the streets in my neighbour- 35.1% of the cycling trips which were not the shortest hood can easily be seen by people in their homes’ were possible, adolescents still indicated that they chose that less likely to use active transport to school [39]. These route because they perceived it as the shortest/fastest findings could be explained by the fact that adolescents route. Thus, even if a route is not actually the shortest, perceive cycling along local roads with lower speed limits adolescents may choose this route because they perceive as more important than potential social control from res- it as the shortest route. For all cycling trips that were not idents. Among adults, Winters et al. [16] also found that the shortest possible, a mean difference of 516 m (15.6%) cyclists spent most of their travel distance along local between the actual and the shortest cycling route was roads. In the present study, arterial roads, such as road found. When only looking at those routes which were types which consist of two separate roads each divided not the shortest but adolescents perceived as the short- in two lanes each direction, were avoided. A number of est/fastest route, a mean difference of 431  m between previous studies also found that cyclists avoid busy, arte- the actual and the shortest cycling route was found, and rial roads [15, 16, 40] and roads with high traffic speed thus, the detour showed to be smaller. It is possible that [41]. In Flanders, if any type of cycle path is available adolescents do not notice the difference in cycling time along these busy roads it is typically a small cycle path between their actual cycling route they perceive as the separated from traffic by white lines, which explains why shortest and the shortest possible cycling route. the present study found that this type of cycle path is less Although some findings of the present study seem to be present along actual cycling routes. in contradiction with results of previous cross-sectional With regard to walking/cycling roads that are not studies [12], the findings generally have a clear explana - accessible for motorised traffic, no significant differ - tion. The present study showed that adolescents avoid ence in presence along actual and shortest cycling routes routes with a mixed land use where commercial destina- was found. Nevertheless, a previous study among adults tions are present. In contradiction, a US study showed found that cyclists spent more time on off-street paths that children and adolescents (5–18  years) were more [16]. In addition, among 10-to-15-year-old US girls, it likely to walk or cycle to school if their parents reported was found that the presence of walking/cycling trails in having stores in the neighbourhood environment [38]. the neighbourhood was associated with higher levels of However, in accordance with the present study, shops active transport to school [39]. Although, in the present and services were also less present along the actual study, these walking/cycling roads occurred relatively fre- cycling routes of adults in Austria [15]. Dessing et al. [17] quently along actual cycling routes (on average for 89 m/ found that children in The Netherlands mainly cycled to km), shortest cycling routes also included some amount school along residential areas to avoid busy streets. These of walking/cycling roads (62  m/km) since the city of findings are similar to results in our study since residen - Ghent already provides an extensive network of walking/ tial areas are, in general, characterised by a lower land cycling roads. These walking/cycling roads often serve as use mix and less commercial destinations. In the study shortcuts for pedestrians and cyclists. This could explain by Dessing et  al. [17], it was suggested that residential why no significant difference between actual and shortest streets may be perceived as safe, quiet streets to cycle for cycling routes was found for walking/cycling roads in this transport, even if separate cycle lanes are absent. This study. could be confirmed by the results of the present study which showed that adolescents mainly cycled along local Practical implications roads, such as roads for one-direction traffic and roads Based on the findings of the present study, some rec - which were not divided into lanes (trends towards sig- ommendations for policy and practice can be formu- nificance), where speed limits of 30 km/h apply. Further - lated. The present study showed that adolescents mainly more, our study also showed that actual cycling routes choose the (perceived) shortest route to cycle for trans- included more m/km road where no cycle lane was pre- port and that adolescents frequently use walking/cycling sent which is typical for residential streets in Flanders. roads that are not accessible for motorised traffic. It In addition, the present study found that actual cycling might thus be important for local authorities to provide Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 12 of 15 walking/cycling roads that are not accessible for motor- were performed in a (sub)urban area. u Th s, results can - ised traffic and could serve as shortcuts for cyclists [42]. not be generalized to rural areas where less alternative These shortcuts free from motorised traffic also meet the routes are available because of a less dense street net- preference of adolescents to avoid cycling along arterial work. Fifth, because of the limited time window of the roads. However, since it is not always possible for an indi- study (i.e. 2 days of data per participant), it is difficult to vidual to avoid to cycle along busy, arterial roads, these draw generalizable conclusions regarding the impact of roads should be made more bike-friendly by providing the physical environment on adolescents’ route choice adequate cycling infrastructure. while cycling. Sixth, the majority of adolescents was of a higher SES family, which could have influenced the Strengths and limitations results. Previous research showed that children and ado- A first strength of the present study was that the evalu - lescents living in lower SES neighbourhoods perceive the ated routes were actual cycling routes which were objec- neighbourhood environment as less attractive and safe, tively recorded using a GPS device. Using objective GPS and more often report heavy traffic in their neighbour - data limits recall bias related to route choice. Particularly hood compared to those living in higher SES neighbour- for young people such as adolescents it may be difficult hoods [43, 44]. It is thus possible that adolescents from to recall and indicate on a map which route they took at lower SES families attach importance to other factors on a particular moment. Second, information obtained by their cycling route compared to adolescents from higher the structured one-on-one interviews enabled to cor- SES families. Thus, caution is needed when generalizing rect trip mode when this was misclassified by PALMS results to the overall adolescent population. Seventh, data (e.g. when a car trip was classified as a bicycle trip due to were collected during autumn/winter which may have traffic congestion). Furthermore, this was the first study influenced the results. Bad weather and less hours of day - to collect subjective information regarding route choice light may influence adolescents’ choice of transport mode via one-on-one interviews, and combine this with audits. and route choice. Finally, results of the present study do This allowed participants to indicate their actual reason not enable to draw conclusions regarding non-cyclists. for choosing a particular route. Third, the presence of physical environmental characteristics along the routes Recommendations for future research was measured objectively using a tool (EGA-Cycling) Since data collection was very time-consuming and the that showed acceptable reliability and validity [22], which burden on participants was rather high, future studies limits the bias of results compared to self-reported ques- should consider to make use of dedicated smartphone tionnaires. Nevertheless, some limitations should be applications to identify adolescents’ actual cycling routes. acknowledged. First, virtual audit tools showed to be less Adolescents generally carry their smartphones with them accurate for measuring micro-environmental features. during the day, thus running dedicated mobile apps may However, Google Street View showed to be more accu- be less considered as a burden compared to wearing rate in measuring micro-environmental features com- portable GPS-devices. The use of dedicated smartphone pared to other virtual audit tools [28]. Nevertheless, a applications would enable to include a larger sample discrepancy in physical environmental factors may exist and would allow to track adolescents’ mobility patterns between the Google Street View images and the period in over a longer time period [45]. Thus, this method has which GPS data were collected. The Google Street View the advantage that much more actual cycling routes can images ranged from March 2009 till April 2015, whereas be identified in more diverse areas. The introduction of participants’ GPS data were collected between Septem- smartphones with a longer battery life and a higher stor- ber and December 2015. Second, the sample size was rel- age space and memory capacity should be able to facili- atively small as only 38 actual cycling routes (spread over tate this type of data collection. Nevertheless, this would 22 adolescents) that were not the shortest cycling route imply a huge burden on the researchers to audit such a could be identified and were evaluated. A small sample large set of cycling routes. Geographic Information Sys- size increases the likelihood of a type II error and, thus, tems (GIS) could be used instead, since GIS makes use the chance that an effect was not detected when there of existing data sources (e.g. governmental data sources) was one to be detected. Third, some characteristics (i.e. to measure physical environmental characteristics that poor visibility when crossing a street, parking facilities have some spatial reference [46]. However, since for some on adjacent parking) were only present along a few street locations (i.e. rural or suburban areas) GIS data may segments. This resulted in wide 95% confidence inter - not be available [46] and micro-environmental factors vals due to insufficient variability. Fourth, as data collec - are also not commonly available in GIS databases [46, tion took place among adolescents attending secondary 47], virtual or on-street audits may be used to comple- schools in the city of Ghent, the majority of cycling trips ment GIS data where needed. In addition, future studies Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 13 of 15 should consider to investigate potential moderators (e.g. environmental factors besides distance. Local authorities individual and social environmental factors) of the rela- should provide shortcuts free from motorised traffic in tionships between the presence of physical environmen- order to meet the preference of adolescents to cycle along tal factors and adolescents’ route choice while cycling for the shortest route and to avoid cycling along busy, arte- transport. It may be interesting to investigate whether rial roads. In addition, it may also be important to pro- the relationship between the presence of certain physi- vide adequate cycling infrastructure along busy, arterial cal environmental factors and adolescents’ preference roads since these roads cannot always be avoided. for a certain cycling route is moderated by, for example, Additional files psychosocial factors. It could be that adolescents with lower self-efficacy or less social models for active trans - Additional file 1. Structured one-on-one interview. port attach importance to other physical environmental Additional file 2. EGA-Cycling checklist. factors when choosing a cycle route compared to adoles- cents with higher  self-efficacy or more social models for active transport. Results of the present study only enable Authors’ contributions to draw conclusions for the general adolescent popula- HV, LVH, DVD, NVDW, BD and JVC designed the protocol for this study. HV tion, no specific conclusions for subgroups of adolescents and LVH collected the data and coordinated the data collection. HV, LVH and (e.g. those with a low psychosocial profile towards cycling TB contributed to the data processing. HV performed the statistical analyses and drafted the manuscript. HV, LVH, DVD, NVDW, PC, BD, JVC critically revised for transport or the least regular cyclists) can be drawn. and helped to draft the manuscript. All authors read and approved the final Since this study was one of the first exploring the factors manuscript. associated with route choice among adolescent cyclists, Author details results are valuable. However, future studies may consider Department of Public Health, Faculty of Medicine and Health Sciences, to conduct moderation analyses among larger samples in 2 Ghent University, Corneel Heymanslaan 10, 9000 Ghent, Belgium. Physical order to be able to formulate recommendations to target Activity, Nutrition and Health Research Unit, Faculty of Physical Education and Physical Therapy, Vrije Universiteit Brussel, Pleinlaan 2, 1050 Brussels, specific subgroups of adolescents. More research investi - 3 4 Belgium. Research Foundation - Flanders (FWO), Brussels, Belgium. Depart- gating adolescents’ route choice for cycling is needed. In ment of Movement and Sport Sciences, Faculty of Medicine and Health Sci- order to be able to formulate recommendations regard- ences, Ghent University, Watersportlaan 2, 9000 Ghent, Belgium. Department of Geography – CartoGIS, Faculty of Sciences, Ghent University, Krijgslaan 281, ing which factors may stimulate adolescents to cycle for 9000 Ghent, Belgium. transport, future studies should also investigate which factors along a route are important among non-cyclists. Acknowledgements We would like to thank the schools and the adolescents who participated in An experimental study which aimed to investigate ado- the study. Furthermore, we would like to thank master’s students for assisting lescents’ preferences towards cycling for transport using with the data collection and processing. We would also like to thank R. Col- manipulated photographs, showed that the least regular man for her help with the data analyses. cyclists in that study attached most importance to cycling Competing interests distance when indicating which route they preferred to The authors declare that they have no competing interests. cycle along [48]. The most regular cyclists in that study Availability of data and materials attached most importance to being able to cycle together The datasets used and/or analysed during the current study are available from with a friend. However, no associations with actual par- the corresponding author on reasonable request. ticipation in cycling for transport were investigated in Consent for publication that study. Not applicable. Ethics approval and consent to participate Conclusions Passive informed consent was obtained from adolescents’ parents. If parents For 73.1% of the cycling trips, participants took the did not agree to let their child participate in the study, they had to sign a form. Furthermore, researchers also obtained active informed consent from shortest route possible which confirmed the importance adolescents. The study protocol was approved by the Ethics Committee of the of cycling distance for adolescents. When not taking University Hospital of Ghent University (EC 2015/0317). the shortest cycling route, adolescents avoided to cycle Funding for transport along arterial roads with a small cycle lane HV and LVH are supported by the Research Foundation Flanders (FWO, http:// separated from traffic by white lines. Local roads with www.fwo.be/en, 3GOA8514). JVC is supported by a FWO postdoctoral fellow- a speed limit of 30  km/h in an area with a low land use ship (FWO - 12I1117N). DVD is supported by a FWO postdoctoral fellowship (FWO12/PDO/158). mix where few commercial destinations are located were more frequently used, even when no cycle lane was avail- Publisher’s Note able. In general, the ability to cycle along quiet, local Springer Nature remains neutral with regard to jurisdictional claims in pub- roads overruled the importance of all other physical lished maps and institutional affiliations. Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 14 of 15 Received: 8 January 2018 Accepted: 22 May 2018 20. Pikora TJ, Bull FC, Jamrozik K, Knuiman M, Giles-Corti B, Donovan RJ. Developing a reliable audit instrument to measure the physical envi- ronment for physical activity. Am J Prev Med. 2002;23:187–94. 21. Day K, Boarnet M, Alfonzo M, Forsyth A. The Irvine-Minnesota inven- tory to measure built environments: development. Am J Prev Med. 2006;30:144–52. References 22. Vanwolleghem G, Van Dyck D, Ducheyne F, De Bourdeaudhuij I, Cardon 1. Cohen AJ, Ross Anderson H, Ostro B, Pandey KD, Krzyzanowski M, Kunzli G. Assessing the environmental characteristics of cycling routes to N, Gutschmidt K, Pope A, Romieu I, Samet JM, Smith K. The global burden school: a study on the reliability and validity of a Google Street View- of disease due to outdoor air pollution. J Toxicol Environ Health A. based audit. Int J Health Geogr. 2014;13:19. 2005;68:1301–7. 23. Brownson RC, Hoehner CM, Brennan LK, Cook RA, Elliot MB, McMullen 2. Woodcock J, Edwards P, Tonne C, Armstrong BG, Ashiru O, Banister D, KM. Reliability of 2 Instruments for Auditing the Environment for Physi- Beevers S, Chalabi Z, Chowdhury Z, Cohen A, et al. Public health benefits cal Activity. J Phys Act Health. 2004;1:189–207. of strategies to reduce greenhouse-gas emissions: urban land transport. 24. Badland HM, Opit S, Witten K, Kearns RA, Mavoa S. Can virtual Lancet. 2009;374:1930–43. streetscape audits reliably replace physical streetscape audits? J Urban 3. Int Panis L, de Geus B, Vandenbulcke G, Willems H, Degraeuwe B, Bleux Health. 2010;87:1007–16. N, Mishra V, Thomas I, Meeusen R. Exposure to particulate matter in 25. Taylor BT, Fernando P, Bauman AE, Williamson A, Craig JC, Redman S. traffic: a comparison of cyclists and car passangers. Atmos Environ. Measuring the quality of public open space using Google Earth. Am J 2010;44:2263–70. Prev Med. 2011;40:105–12. 4. Oja P, Titze S, Bauman A, de Geus B, Krenn P, Reger-Nash B, Kohlberger T. 26. Rundle AG, Bader MD, Richards CA, Neckerman KM, Teitler JO. Using Health benefits of cycling: a systematic review. Scand J Med Sci Sports. Google Street View to audit neighborhood environments. Am J Prev 2011;21:496–509. Med. 2011;40:94–100. 5. Ortega FB, Konstabel K, Pasquali E, Ruiz JR, Hurtig-Wennlof A, Maestu J, 27. Kelly CM, Wilson JS, Baker EA, Miller DK, Schootman M. Using Google Lof M, Harro J, Bellocco R, Labayen I, et al. Objectively measured physical Street View to audit the built environment: inter-rater reliability results. activity and sedentary time during childhood, adolescence and young Ann Behav Med. 2013;45(Suppl 1):S108–12. adulthood: a cohort study. PloS ONE. 2013;8:e60871. 28. Ben-Joseph E, Lee JS, Cromley EK, Laden F, Troped PJ. Virtual and actual: 6. Chillon P, Ortega FB, Ruiz JR, De Bourdeaudhuij I, Martinez-Gomez D, relative accuracy of on-site and web-based instruments in auditing the Vicente-Rodriguez G, Widhalm K, Molnar D, Gottrand F, Gonzalez-Gross environment for physical activity. Health Place. 2013;19:138–50. M, et al. Active commuting and physical activity in adolescents from 29. Gullon P, Badland HM, Alfayate S, Bilal U, Escobar F, Cebrecos A, Diez J, Europe: results from the HELENA study. Pediatr Exerc Sci. 2011;23:207–17. Franco M. Assessing walking and cycling environments in the streets 7. Chillon P, Ortega FB, Ruiz JR, Veidebaum T, Oja L, Maestu J, Sjostrom M. of Madrid: comparing On-field and virtual audits. J Urban Health. Active commuting to school in children and adolescents: an oppor- 2015;92:923–39. tunity to increase physical activity and fitness. Scand J Public Health. 30. Statistics Belgium: Bodemgebruik in België 1834–2017. 2015. 2010;38:873–9. 31. Structuur van de bevolking volgens woonplaats: grootste gemeenten. 8. Bere E, Seiler S, Eikemo TA, Oenema A, Brug J. The association between http://statb el.fgov.be/nl/stati stiek en/cijfe rs/bevol king/struc tuur/ cycling to school and being overweight in Rotterdam ( The Netherlands) woonp laats /groot /. and Kristiansand (Norway). Scand J Med Sci Sports. 2011;21:48–53. 32. Lien N, Friestad C, Klepp KI. Adolescents’ proxy reports of parents’ 9. Nettleton S, Green J. Thinking about changing mobility practices: how a socioeconomic status: how valid are they? J Epidemiol Community social practice approach can help. Sociol Health Illn. 2014;36:239–51. Health. 2001;55:731–7. 10. Sallis JF, Cervero RB, Ascher W, Henderson KA, Kraft MK, Kerr J. An ecologi- 33. Kerr J, Normam G, Godbole S, Raab F, Demchak B, Patrick K. Validating cal approach to creating active living communities. Annu Rev Public GPS data with the PALMS system to detect different active transporta- Health. 2006;27:297–322. tion modes. Med Sci Sports Exerc. 2012;44:647. 11. Wong BY, Faulkner G, Buliung R. GIS measured environmental correlates 34. The Personal Activity and Location Measurement System (PALMS). of active school transport: a systematic review of 14 studies. Int J Behav https ://ucsd-palms -proje ct.wikis paces .com/. Nutr Phys Act. 2011;8:39. 35. Demchak B, Kerr J, Raab F, Patrick K, Krüger IH. PALMS: a modern coevo- 12. Panter JR, Jones AP, van Sluijs EM. Environmental determinants of active lution of community and computing using policy driven development. travel in youth: a review and framework for future research. Int J Behav In 45th Hawaii international conference on system sciences; 2012. Nutr Phys Act. 2008;5:34. 36. Carlson JA, Jankowska MM, Meseck K, Godbole S, Natarajan L, Raab F, 13. Nelson NM, Foley E, O’Gorman DJ, Moyna NM, Woods CB. Active com- Demchak B, Patrick K, Kerr J. Validity of PALMS GPS scoring of active muting to school: how far is too far? Int J Behav Nutr Phys Act. 2008;5:1. and passive travel compared with SenseCam. Med Sci Sports Exerc. 14. Verhoeven H, Ghekiere A, Van Cauwenberg J, Van Dyck D, De Bourdeaud- 2015;47:662–7. huij I, Clarys P, Deforche B. Which physical and social environmental 37. MAPS GLOBAL audit tool. http://salli s.ucsd.edu/measu re_maps.html. factors are most important for adolescents’ cycling for transport? An 38. Kerr J, Rosenberg D, Sallis JF, Saelens BE, Frank LD, Conway TL. Active experimental study using manipulated photographs. Int J Behav Nutr commuting to school: associations with environment and parental Phys Act. 2017;14:108. concerns. Med Sci Sports Exerc. 2006;38:787–94. 15. Krenn PJ, Oja P, Titze S. Route choices of transport bicyclists: a com- 39. Evenson KR, Birnbaum AS, Bedimo-Rung AL, Sallis JF, Voorhees CC, parison of actually used and shortest routes. Int J Behav Nutr Phys Act. Ring K, Elder JP. Girls’ perception of physical environmental factors and 2014;11:31. transportation: reliability and association with physical activity and 16. Winters M, Teschke K, Grant M, Setton E, Brauer M. How far out of the way active transport to school. Int J Behav Nutr Phys Act. 2006;3:28. will we travel? Built environment influences on route selection for bicycle 40. Timperio A, Ball K, Salmon J, Roberts R, Giles-Corti B, Simmons D, Baur and car travel. Transp Res Rec J Transp Res Board. 2010;2190:1–10. LA, Crawford D. Personal, family, social, and environmental correlates of 17. Dessing D, de Vries SI, Hegeman G, Verhagen E, van Mechelen W, Pierik active commuting to school. Am J Prev Med. 2006;30:45–51. FH. Children’s route choice during active transportation to school: dif- 41. McMillan TE. The relative influence of urban form on a child’s travel ference between shortest and actual route. Int J Behav Nutr Phys Act. mode to school. Transp Res Part A. 2007;41:69–79. 2016;13:48. 42. Pucher J, Buehler R. Making cycling irresistible: lesson from Europe. 18. Carpiano RM. Come take a walk with me: the “go-along” interview as a Transp Rev. 2008;28:495–528. novel method for studying the implications of place for health and well- 43. Giles-Corti B, Donovan RJ. Socioeconomic status differences in being. Health Place. 2009;15:263–72. recreational physical activity levels and real and perceived access to a 19. Clifton KJ, Livi Smith ADL, Rodriguez D. The development and test- supportive physical environment. Prev Med. 2002;35:601–11. ing of an audit for the pedestrian environment. Landsc Urban Plan. 2007;80:95–110. Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 15 of 15 44. Timperio A, Crawford D, Telford A, Salmon J. Perceptions about the local 47. Adams MA, Ryan S, Kerr J, Sallis JF, Patrick K, Frank LD, Norman GJ. Valida- neighborhood and walking and cycling among children. Prev Med. tion of the Neighborhood Environment Walkability Scale (NEWS) items 2004;38:39–47. using geographic information systems. J Phys Act Health. 2009;6(Suppl 45. Vlassenroot S, Gillis D, Bellens R, Gautama S. The use of smartphone 1):S113–23. applications in the collection of travel behaviour data. Int J Intell Transp 48. Verhoeven H, Ghekiere A, Van Cauwenberg J, Van Dyck D, De Bourdeaud- Syst Res. 2015;13:17–27. huij I, Clarys P, Deforche B. Subgroups of adolescents differing in physical 46. Brownson RC, Hoehner CM, Day K, Forsyth A, Sallis JF. Measuring the built and social environmental preferences towards cycling for transport: a environment for physical activity: state of the science. Am J Prev Med. latent class analysis. Prev Med. 2018;112:70–5. 2009;36(S99–123):e112. Ready to submit your research ? Choose BMC and benefit from: fast, convenient online submission thorough peer review by experienced researchers in your field rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png International Journal of Health Geographics Springer Journals

Differences in physical environmental characteristics between adolescents’ actual and shortest cycling routes: a study using a Google Street View-based audit

Free
15 pages

Loading next page...
 
/lp/springer_journal/differences-in-physical-environmental-characteristics-between-DU3tta0e3J
Publisher
BioMed Central
Copyright
Copyright © 2018 by The Author(s)
Subject
Medicine & Public Health; Public Health; Geographical Information Systems/Cartography; Human Geography; Epidemiology; Remote Sensing/Photogrammetry; Health Promotion and Disease Prevention
eISSN
1476-072X
D.O.I.
10.1186/s12942-018-0136-x
Publisher site
See Article on Publisher Site

Abstract

Background: The objective evaluation of the physical environmental characteristics (e.g. speed limit, cycling infra- structure) along adolescents’ actual cycling routes remains understudied, although it may provide important insights into why adolescents prefer one cycling route over another. The present study aims to gain insight into the physical environmental characteristics determining the route choice of adolescent cyclists by comparing differences in physi- cal environmental characteristics between their actual cycling routes and the shortest possible cycling routes. Methods: Adolescents (n = 204; 46.5% boys; 14.4 ± 1.2 years) recruited at secondary schools in and around Ghent (city in Flanders, northern part of Belgium) were instructed to wear a Global Positioning System device in order to identify cycling trips. For all identified cycling trips, the shortest possible route that could have been taken was cal- culated. Actual cycling routes that were not the shortest possible cycling routes were divided into street segments. Segments were audited with a Google Street View-based tool to assess physical environmental characteristics along actual and shortest cycling routes. Results: Out of 160 actual cycling trips, 73.1% did not differ from the shortest possible cycling route. For actual cycling routes that were not the shortest cycling route, a speed limit of 30 km/h, roads having few buildings with windows on the street side and roads without cycle lane were more frequently present compared to the shortest pos- sible cycling routes. A mixed land use, roads with commercial destinations, arterial roads, cycle lanes separated from traffic by white lines, small cycle lanes and cycle lanes covered by lighting were less frequently present along actual cycling routes compared to the shortest possible cycling routes. Conclusions: Results showed that distance mainly determines the route along which adolescents cycle. In addi- tion, adolescents cycled more along residential streets (even if no cycle lane was present) and less along busy, arterial roads. Local authorities should provide shortcuts free from motorised traffic to meet adolescents’ preference to cycle along the shortest route and to avoid cycling along arterial roads. Keywords: Active transport, Cycling, Route choice, Physical environment, Audit, Youth *Correspondence: hannah.verhoeven@ugent.be Department of Public Health, Faculty of Medicine and Health Sciences, Ghent University, Corneel Heymanslaan 10, 9000 Ghent, Belgium Full list of author information is available at the end of the article © The Author(s) 2018. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/ publi cdoma in/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 2 of 15 Compared to the shortest routes, land use mix (i.e. the Background extent to which several types of land use, such as resi- Air pollution, which is partially caused by vehicle emis- dential and industrial areas, shops, services, are included sions, is consistently related to acute respiratory infec- in an area) was significantly higher along actual cycling tions among young children, cardiopulmonary disease routes. A study among children in the Netherlands and lung cancer [1]. By replacing private car use (passive (8–12  years) found that there were significantly fewer transport) by active modes of transport such as cycling, trees, zebra crossings and sidewalks along actual cycling carbon dioxide emissions can be reduced substan- routes compared to the shortest routes [17]. In addition, tially [2]. Although the risk of a higher intake of carbon actual cycling routes had significantly more traffic lights, dioxide can be considered as a negative aspect of active junctions and a higher chance of being on residential transport [3], a growing body of evidence emphasizes streets compared to the shortest routes. Safety showed the potential benefits of cycling for transport for public thus to be an important factor among children in this health [2, 4]. Since adolescence is characterised by a steep study. According to Dessing et al. [17], most of the zebra decrease in physical activity levels [5], increasing cycling crossings in the Netherlands are located on or near busy for transport is also a promising strategy to meet the rec- streets, that were avoided by the children. Furthermore, ommended 60 min of daily physical activity among ado- when main roads have to be crossed children preferred lescents [4, 6]. Cycling for transport has been associated signalized intersections. Because of some inconsist- with higher levels of cardiorespiratory fitness [7] and ent results across these previous studies, similar studies lower levels of overweight [8] among adolescents and it among adolescents may provide additional insights into can easily be incorporated into their daily lives once the which physical environmental factors are related to an skills for cycling have been acquired [9]. individuals’ route choice. The role of the physical environment for health behav - Methodologies to assess the physical environment iours such as cycling for transport has been acknowl- include both subjective and objective measurements. edged by socio-ecological models and previous research Subjective measurements, such as self-reported ques- [10–12]. However, the majority of previous studies inves- tionnaires, encounter limitations such as recall bias [18] tigating physical environmental correlates of cycling for and may not accurately assess the effect of the actual transport focused on the neighbourhood environment physical environmental factors on cycling for transport close to home, although cycling for transport does not [11]. Therefore, observational field audits are frequently necessarily take place in the immediate neighbourhood applied as an objective tool for measuring the physical environment. Nevertheless, the evaluation of physi- environment related to physical activity [19–21]. Vanwol- cal environmental characteristics along adolescents’ leghem et al. [22] developed EGA-Cycling (Environmen- actual cycling routes remains understudied, although tal Google Street View Based Audit-Cycling) to virtually it is important to find out why individuals chose a spe - assess physical micro- and macro-environmental charac- cific cycling route. In addition, although previous studies teristics along cycling routes using Google Street View. emphasized the importance of distance for adolescents’ EGA-Cycling was based on existing audit instruments cycling for transport [12–14], it is likely that adoles- (e.g. Pikora-SPACES instruments [20], Audit Tool Check- cents do not always take the shortest cycling route. By list version [21], Irvine-Minnesota Inventory [23]), but comparing adolescents’ actual cycling routes with the was adapted to the Flemish street infrastructure. In the shortest possible cycling routes, important information last decade, using virtual technologies, such as Google regarding which physical environmental characteristics Street View, to assess the physical environment is gain- determine the route choice of adolescent cyclists may be ing attention [24–29]. Auditors are able to virtually walk obtained. Among adults, two recent studies compared through a street which is time- and cost-saving [24, 28] physical environmental characteristics of actual and and they are not exposed to unsafe (traffic) situations shortest cycling routes [15, 16]. Winters et al. [16] found compared to field audits. Previous studies showed good that actual cycling routes of Canadian adults had signifi - agreement between virtual and field audit tools [24, cantly more traffic calming facilities (e.g. traffic circles or 26, 29]. However, virtual audit tools showed to be less median barriers to slow or block motorized traffic) and accurate when measuring micro-environmental charac- participants cycled less along arterial (busy) roads and teristics (e.g. litter, sidewalk condition) [24, 26, 28]. Nev- more along local roads, off-street paths and roads with ertheless, Ben-Joseph et  al. [28] concluded that Google cycling facilities. Krenn et  al. [15] also found that Aus- Street View was more accurate in measuring small fea- trian cyclists avoid busy roads and prefer roads with cycle tures compared to Google Maps and MS Visual Oblique. lanes. Actual cycling routes included more green and The aim of the present study is to gain insight into the aquatic areas and had fewer traffic lights, fewer cross - physical environmental characteristics determining the ings and less hilly roads compared to the shortest routes. Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 3 of 15 route choice of adolescent cyclists by comparing differ - adolescents (response rate = 84.1%) participating in the ences in physical environmental characteristics between study. their actual cycling routes and the shortest possible cycling routes using a Google Street View-based audit Study protocol (EGA-Cycling). The study protocol consisted of two parts (see Fig.  1 for a flow chart). In the first part of the study, each participat - ing school was visited three times by the research team Methods between September and December 2015. During a first Participants visit, the purpose of the study was explained to the ado- A convenience sample of 12 secondary schools in and lescents and informed consent was obtained. Each par- around Ghent was contacted to participate in the study. ticipant received a unique ID number in order to be able Ghent is a city in Flanders, northern part of Belgium, to link data of all measurements. Participants completed that has 253,266 inhabitants and comprises an area of a questionnaire assessing socio-demographics. Further- 2 2 156.2  km (population density: 1622  h/km ) [30, 31]. In more, participants received a Global Positioning System the six schools that agreed to participate, school prin- (GPS) device and a charger for the device together with cipals or staff members randomly selected at least two verbal and written instructions on how and when to wear classes from the first to fourth grade (12–16  years). A the device. All participants were instructed to wear the total of 18 classes was selected and 283 adolescents were GPS device, which was attached to their waist with an invited to participate in the study. Only participants who elastic belt, during waking hours until the research team were present at school when measurement materials returned to the school to collect the devices (4–5  days were handed out, could be included in the study. Pas- later). During activities that could damage the GPS sive informed consent was obtained from adolescents’ device or during which it could be uncomfortable to parents. If parents did not agree to let their child partici- wear it (e.g. showering, swimming or rugby), the adoles- pate in the study, they had to sign a form. Furthermore, cents were asked to temporarily remove the GPS device. researchers also obtained active informed consent from They were also instructed not to turn off the GPS device adolescents. This procedure resulted in a group of 238 during data collection. Participants were asked for their Fig. 1 Flow chart Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 4 of 15 Structured one‑on‑one interview with personal travel maps mobile phone number. Two text messages per day (in the GPS data were stored in a PostgreSQL database with morning and evening) were sent to the participants will- PostGIS in order to generate a personal travel map ing to give their number in order to remind them to wear per day in the web application. This web application the GPS devices and to charge it. During a second visit, showed the geographical position of participants for researchers returned to the schools to collect the devices. every 30-second-interval. Figure  2 shows an example Afterwards, the GPS data were downloaded and a web of a personal travel map. These personal travel maps application was created in order to visualize the data on were used as a guide to conduct a structured one-on- a personal travel map. During the last visit, which took one interview discussing routes on two selected days. place within the first week after collection of the GPS The first week- and weekend day with complete data devices, researchers conducted a structured one-on-one (excluding the day the devices were handed out) were interview (Additional file  1) during which a researcher selected for the interview. When no weekdays with chronologically discussed the personal travel maps. Per complete data were available, two weekend days were trip travelled, participants were asked about their trans- selected and vice versa. When only 1 day with complete port mode and why they took a particular route. Partici- data was available, the structured interview was com- pants who completed all measurements and returned the pleted for 1  day. During these interviews, a researcher GPS device received an incentive (i.e. movie ticket). chronologically identified, together with the partici - In the second part of the study, adolescents’ cycling pants, the trips they made during a day. For each iden- routes were selected, and for each actual cycling route tified trip, the participant was asked which transport the shortest cycling route was calculated using Google mode was used. For active trips (walking or cycling/ Maps. For each cycling route which was not the short- skateboard/…) the participant was also asked why est cycling route, an adapted version of EGA-Cycling he/she chose that particular route to reach his/her was used to obtain information about physical environ- destination. mental characteristics along adolescents’ actual cycling routes and along the corresponding shortest routes using Google Street View. The study protocol was approved by the Ethics Com - GPS data processing mittee of the University Hospital of Ghent University (EC Data processing was executed using the Personal Activ- 2015/0317). ity and Location Measurement System (PALMS©) [34, 35]. PALMS filtered invalid GPS data when extreme speed (> 150  km/h) or extreme changes in distance Measurements and data processing (> 1000 m) or elevation (> 100 m) between two consecu- Questionnaire tive data points were identified. The programming soft - Participants completed a paper-and-pencil questionnaire ware Python was used to combine the PALMS dataset assessing following socio-demographics: home address, with information on school schedules of each participat- gender, date of birth, grade (first to fourth year), educa - ing class, school addresses and home addresses of par- tional type (general, technical or vocational) and high- ticipants. PALMS categorised data into location (home, est education of parents (primary education, secondary school, leisure) or transport. Data were categorised in the education, tertiary education-non university, tertiary domain ‘transport’ when a trip was detected. A trip was education-university, I don’t know). Education of parents defined as a period of at least 3  min of movement with was used to as a proxy for socio-economic status (SES). the same transport mode, allowing for stationary periods Adolescents were identified as being ‘of a higher SES fam - of maximum 3 min. PALMS classified all trips into walk - ily’ when at least one parent completed tertiary education ing, cycling or motorised transport based on speed. A [32]. trip was classified as walking when speed was between 1 and 9 km/h, cycling between 10 and 24 km/h and motor- GPS device ised transport starting from 25 km/h [33, 36]. The geographical position of participants was recorded Subsequently, all data from the structured one-on-one by the QStarz BT-Q1000X GPS device. In addition, the interviews were inserted into the database. For trips or GPS device recorded participants’ speed which was used locations that were misclassified by PALMS [e.g. when a to define their transport mode [33]. The GPS devices car trip was classified as a bicycle trip due to traffic con - were set to collect data every 30  s using Q-travel soft- gestion (speed < 25 km/h)], corrections were made based ware. Furthermore, the devices were set to stop logging on the data of the structured interviews. The number when the memory was full (this did not occur during of corrections due to misclassification by PALMS was data collection). Q-travel software was used to download rather limited. the collected GPS data. Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 5 of 15 Fig. 2 Example of a personal travel map. Every 30 s a dot was placed on the map (temporal resolution: 30 s). The green arrow represents the first data point of the day registered by the GPS and the ‘finish flag’ represents the last registered data point of the day by the GPS EGA‑cycling shortest possible cycling routes were selected to be used EGA-Cycling (Additional file  2) consists of five subscales in subsequent analyses. All routes for which the actual and includes 37 items: (1) land use (8 items; e.g. commer- cycling route was not the shortest possible cycling route cial destinations, heavy industry and public destinations), were included, even if only one segment differed between (2) general characteristics of the street segment (12 the actual and the shortest cycling route. Differences in items; e.g. road type and speed limit), (3) cycling facilities distance between actual cycling routes and the short- (7 items; e.g. type and width of cycle lane), (4) pedestrian est possible cycling routes were calculated absolutely in facilities (3 items; e.g. presence and maintenance of the meters as well as relatively in percentage of the shortest sidewalk) and (5) aesthetics (7 items; e.g. trees and front cycling route (reported as ‘detour’). Google My Maps (a yards). EGA-Cycling shows acceptable reliability and Google Maps application) was used to visualize actual validity [22]. However, since measures about (safety at) cycling routes and the corresponding shortest cycling intersections are very limited from this tool, three addi- routes. Each cycling route was manually divided into tional items regarding this topic were added (i.e. amount several street segments (average distance: 342 ± 468  m), of side streets, amount of intersections and visibility at a new street segment started when participants turned the corners). The item regarding visibility at the corners into another street or when the street name changed. For is part of the Microscale Audit of Pedestrian Streetscape each street segment, EGA-Cycling was filled out by one (MAPS) Global tool [37]. Furthermore, another item was out of three trained observers (the first author and two included that assessed whether or not the street segment independent observers). Google Street View was used to concerned a walking/cycling road (i.e. a separate road perform the audits, which took approximately 2 weeks per only accessible for non-motorised traffic). Data on dif - observer (6  weeks in total). Google Street View images ferences in pedestrian facilities between actual cycling ranged from March 2009 till April 2015. The majority routes and the shortest possible cycling routes are not (53.0%) of images were taken between August 2014 and shown since they are not relevant for cycling. October 2014. Prior to auditing the pre-defined routes, two independent observers were trained by the first author. The training included specific instructions; all items of the EGA-Cycling tool were explained and illus- Auditing of actual and shortest cycling routes trated with photographs if necessary. Subsequently, the For all cycling trips that could be identified in the previ - observers audited three random street segments which ous steps, the shortest cycling route was calculated using enabled them to raise questions. Thereafter, five test Google Maps. Only actual cycling routes that were not the Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 6 of 15 routes (i.e. no routes that were part of the study) were the dataset as were participants who were absent when rated by the first author and the two independent observ - the structured interviews were completed (n = 17). A ers. Prior to auditing the pre-defined routes, 95% agree - final sample of 204 adolescents (85.7%) was used for ment with the first author’s scores was required. For the data analyses (46.5% boys, 14.4 ± 1.2  years). Table  1 pre- actual audits, only street segments for which there was no sents descriptive characteristics of the sample. Within overlap between the actual cycling route and the short- this sample, a total of 1126 trips was identified. Passive est cycling route were audited (516 segments). Distances transport (car, as a passenger) was used most frequently of segments were measured in Google My Maps. Figure 3 (34.6% of trips), followed by public transport (33.9% of shows examples of actual versus shortest cycling routes. trips). Active transport such as walking and cycling was used for 17.2 and 14.2% of trips, respectively. The pur - Data analyses pose of a trip and the transport mode used showed to be Data were analysed using IBM SPSS Statistics 24. A related to each other (Chi = 257.1; p < 0.001). For trips paired samples t-test was used to calculate the difference to and/or from school, the majority (57.2%) was done by in distance between actual and shortest cycling routes. public transport, 18.4% was done by bicycle, 12.6% by Because EGA-Cycling was developed to assess physical foot and 11.8% by passive transport. For leisure-related environmental characteristics along entire cycling routes instead of individual segments [22], a total score per cycling route was calculated for each item. Per item, the Table 1 Descriptive characteristics of the sample (n = 204) score for a particular segment was multiplied by the dis- Socio-demographic characteristics tance of that segment. These weighted item scores of sev - eral segments of a route were summed to obtain one total Gender (% boys) 46.5 score per route for that item. Subsequently, item scores Age (years; mean ± SD) 14.4 ± 1.2 were expressed in m/km in order to be able to compare Socio-economic status (SES) parents (%) the actual cycling route with the shortest cycling route Lower SES (% no parent completed tertiary education) 28.4 (for which the route length differed). Univariate multi - Higher SES (% at least one parent completed tertiary educa- 71.6 level logistic regression analyses were used to investigate tion) differences in physical environmental characteristics Grade (%) between actual and shortest cycling routes (three levels: 1st year of secondary school 8.3 participant, route and street segment). Statistical signifi - 2nd year of secondary school 7.4 cance was set at p < 0.05. 3rd year of secondary school 46.1 4th year of secondary school 38.2 Results Educational type (%) Sample characteristics General education 65.2 From the 238 adolescents participating, adolescents older Technical education 10.3 than 17 years (n = 4) and participants who did not wear/ Vocational education 24.5 charge the material properly (n = 13) were removed from Fig. 3 Examples of actual versus shortest cycling routes Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 7 of 15 trips, nearly half (49.6%) was done by passive transport, and the shortest possible cycling routes (detour) was 20.1% was done by foot, 17.8% by public transport and 516 ± 369 m (med = 400 m; min = 100 m; max = 1600 m). 12.5% by bicycle. The median distance for car trips was The average detour was 15.6% (med = 12.1%; min = 2.0%; 6312  m and for public transport a median distance of max = 45.7%) in comparison to the shortest possible 4934 m was found. Walking trips had a median distance cycling route. of 710 m, whereas for cycling trips a median distance of Table 2 presents results on differences in physical envi - 2633 m was found. ronmental characteristics concerning land use between Out of 160 actual cycling trips, 73.1% did not differ actual cycling routes and the shortest possible cycling from the shortest possible cycling routes. Thirty-eight routes. An increase in 100 m/km of mixed land use along unique cycling routes for which the actual route dif- the actual cycling route, resulted in 16% lower odds fered from the shortest possible cycling route could that the actual cycling route was chosen over the short- be identified (see Fig.  4). The 38 routes were spread est cycling route. In addition, an increase in 100  m/km over 22 adolescents, with a range of 1 to maximum 4 where commercial destinations are present along the routes per person. A significant difference in distance actual cycling route, resulted in 17% lower odds that the between actual and shortest possible cycling routes actual cycling route was chosen over the shortest cycling was found (t = 8.606; p < 0.001). Actual cycling routes route. had a mean distance of 4505 ± 2201  m (med = 4100  m; Table  3 presents results on differences in general char - min = 1000  m; max = 8800  m), whereas for the shortest acteristics between actual cycling routes and the short- possible cycling routes a mean distance of 3989 ± 2048 m est possible cycling routes. An increase in 100  m/km of (med = 3600  m; min = 700  m; max = 8100 m) was found. a road type which consists of two roads divided in two The mean difference between actual cycling routes lanes each direction (i.e. arterial road) along the actual Fig. 4 Overview of the 38 actual cycling routes that differed from the shortest possible cycling route Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 8 of 15 Table 2 Presence of items on land use along actual cycling routes compared to shortest cycling routes Item Actual cycling route Shortest cycling route OR (95% CI) (m/km; M ± SD) (m/km; M ± SD) Mixed land use 256 ± 226 386 ± 317 0.84 (0.71; 1.00)* Types of buildings Single buildings 155 ± 247 153 ± 256 1.00 (0.83; 1.21) Closed/semi-detached buildings 225 ± 190 139 ± 182 1.30 (0.99; 1.70) Apartment buildings 111 ± 200 161 ± 244 0.90 (0.73; (1.12) Commercial destinations 233 ± 229 367 ± 304 0.83 (0.69; 0.99)* Heavy industry 9 ± 39 17 ± 63 0.74 (0.29; 1.91) Public destinations 248 ± 196 355 ± 300 0.84 (0.70; 1.02) Recreational destinations 85 ± 105 121 ± 212 0.87 (0.65; 1.17) Natural features 315 ± 314 257 ± 286 1.07 (0.91; 1.25) Openness view Open view 65 ± 174 24 ± 81 1.29 (0.85; 1.96) Not open/closed view 354 ± 229 370 ± 271 0.98 (0.81; 1.17) Closed view 166 ± 168 146 ± 185 1.07 (0.82; 1.39) Reference = shortest cycling route. For ease of interpretation of OR, distances were converted to hectometres (100 m/km) OR odds ratio, CI confidence interval *p ≤ 0.05; **p ≤ 0.01; p ≤ 0.1 cycling route, resulted in 47% lower odds that the actual cycling route was chosen over the shortest cycling cycling route was chosen over the shortest cycling route. route. Finally, for an increase in 100  m/km of a cycle An increase in 100 m/km with a speed limit of 30  km/h lane that is covered by lighting along the actual cycling along the actual cycling route, resulted in 50% higher route, 25% lower odds that the actual cycling route was odds that the actual cycling route was chosen over the chosen over the shortest cycling route was found. shortest cycling route. Furthermore, an increase in Table  5 presents results on differences in aesthetics 100  m/km for roads where few buildings with windows between actual cycling routes and the shortest possible on the street are present along the actual cycling route, cycling routes. For none of the included variables, a sig- resulted in 192% higher odds that the actual cycling route nificant result was found. was chosen over the shortest cycling route. This last item Subjective results of the structured one-on-one inter- refers to crime safety/social control (i.e. if few buildings views showed that, for the 38 actual cycling routes that with windows on the street (or few buildings in general) differed from the shortest possible cycling route, ado - are present, few people have a clear view on the street lescents still indicated for 35.1% (n = 13) of the trips and there is thus less social control) [22]. that they chose that route because it was the shortest/ Table  4 presents results on differences in cycling fastest route. For 16.2% (n = 6) of these cycling trips, facilities between actual cycling routes and the short- participants indicated they chose that particular route est possible cycling routes. For an increase in 100  m/ to cycle together with friends/siblings/…. For another km of a cycle lane which is part of the road (cycle lane 16.2% (n = 6) of the trips, they chose that particular separated from traffic by white lines) along the actual route because of lower traffic density. Furthermore, for cycling route, 36% lower odds that the actual cycling 13.5% (n = 5) of the trips participants indicated that route was chosen over the shortest cycling route was route choice was determined by their parents and for found. For an increase in 100 m/km road with no cycle 5.4% (n = 2) of the trips they indicated to choose that lane along the actual cycling route, 25% higher odds particular route because of the presence of few/safe that the actual cycling route was chosen over the short- crossings. For another 5.4% (n = 2) of the trips, par- est cycling route was found. In addition, an increase in ticipants indicated they chose that particular route 100 m/km of a small cycle lane along the actual cycling because of a commercial destination they wanted to route, resulted in 32% lower odds that the actual visit. Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 9 of 15 Table 3 Presence of items on general characteristics along actual cycling routes compared to shortest cycling routes Item Actual cycling route Shortest cycling route OR (95% CI) (m/km; M ± SD) (m/km; M ± SD) Road type Walking/cycling road 89 ± 99 62 ± 111 1.29 (0.82; 2.03) One road for one-direction traffic 120 ± 168 56 ± 107 1.43 (0.97; 2.11) One road not divided into lanes 259 ± 232 160 ± 195 1.25 (0.99; 1.57) One road divided in one lane each direction 82 ± 125 123 ± 199 0.86 (0.64; 1.15) Two roads divided in one lane each direction 21 ± 58 7 ± 24 2.44 (0.63; 9.41) Two roads divided in two lanes each direction 14 ± 46 130 ± 245 0.53 (0.28; 0.99)* Speed limit 30 km/h 145 ± 185 58 ± 131 1.50 (1.02; 2.21)* 50 km/h 309 ± 261 316 ± 255 0.99 (0.83; 1.18) 70 km/h or more 43 ± 87 106 ± 225 0.78 (0.57; 1.08) Traffic calming measures 248 ± 222 344 ± 304 0.87 (0.73; 1.04) Amount of side streets 13 ± 11 11 ± 10 1.02 (0.98; 1.07) Amount of intersections 2 ± 2 3 ± 3 0.90 (0.75; 1.09) Crossing aids 372 ± 247 414 ± 310 0.95 (0.80; 1.12) Poor visibility when crossing a street 23 ± 54 4 ± 16 4.85 (0.80; 29.52) Well-maintained street segment 548 ± 267 531 ± 282 1.02 (0.86; 1.21) Streetlights 522 ± 265 536 ± 290 0.98 (0.83; 1.16) Parking facilities On street parking facilities 180 ± 173 122 ± 156 1.25 (0.93; 1.68) Parking facilities next to the street 210 ± 206 321 ± 297 0.84 (0.69; 1.02) Parking facilities on adjacent parking 18 ± 62 3 ± 15 2.98 (0.44; 20.35) No parking facilities 88 ± 193 33 ± 55 1.45 (0.86; 2.44) Slope Flat 546 ± 266 499 ± 297 1.06 (0.90; 1.25) Gentle to moderate slope 39 ± 62 41 ± 105 0.97 (0.56; 1.66) Swerving alternatives 407 ± 254 353 ± 266 1.08 (0.91; 1.30) Buildings No buildings with windows on street side 47 ± 182 30 ± 82 1.10 (0.78; 1.55) Few buildings with windows on street side 58 ± 78 22 ± 45 2.92 (1.12; 7.63)* Many buildings with windows on street side 391 ± 248 427 ± 293 0.95 (0.80; 1.13) Driveways No driveways 54 ± 136 65 ± 105 0.93 (0.63; 1.37) Approx. 25% of buildings have one driveway 181 ± 207 167 ± 238 1.03 (0.84; 1.27) Approx. 50% of buildings have one driveway 17 ± 29 35 ± 88 0.61 (0.25; 1.51) Most buildings have one driveway 244 ± 247 213 ± 259 1.05 (0.88; 1.26) Garages No garages 228 ± 240 187 ± 175 1.10 (0.88; 1.38) Approx. 25% of buildings have one garage 252 ± 228 270 ± 273 0.97 (0.81; 1.17) Approx. 50% of buildings or more have one garage 17 ± 32 22 ± 47 0.72 (0.22; 2.32) Reference = shortest cycling route. Results regarding ‘one road divided in two lanes each direction’ (road type) are not shown since this road type did not appear along the routes. For ease of interpretation of OR, distances were converted to hectometres (100 m/km) OR odds ratio, CI confidence interval *p ≤ 0.05; **p ≤ 0.01; p ≤ 0.1 Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 10 of 15 Table 4 Presence of items on cycling facilities along actual cycling routes compared to shortest cycling routes Item Actual cycling route Shortest cycling route OR (95% CI) (m/km; M ± SD) (m/km; M ± SD) Type of cycle lane Cycle lane separated from the road 74 ± 133 45 ± 75 1.30 (0.82; 2.08) Adjoining cycle lane (slightly increased) 76 ± 105 89 ± 141 0.91 (0.63; 1.33) Cycle lane is part of the road (white lines) 52 ± 84 180 ± 252 0.64 (0.44; 0.92)* Non-compulsory cycle lane or of a different colour 24 ± 101 7 ± 29 1.45 (0.65; 3.25) No cycle lane 271 ± 243 159 ± 209 1.25 (1.01; 1.56)* Width cycle lane Small 72 ± 106 166 ± 206 0.68 (0.48; 0.96)* Wide 242 ± 260 237 ± 276 1.01 (0.85; 1.20) Two-way cycle lane 224 ± 254 121 ± 203 1.23 (0.98; 1.54) Well-maintained cycle lane 294 ± 240 376 ± 286 0.89 (0.74; 1.06) Lighting covering cycle lane 174 ± 182 332 ± 284 0.75 (0.61; 0.94)** Surface cycle lane Bitumen 273 ± 177 260 ± 237 1.03 (0.83; 1.29) Continuous concrete 8 ± 37 5 ± 17 1.50 (0.27; 8.41) Paving bricks 181 ± 199 126 ± 131 1.22 (0.91; 1.63) Concrete slabs 80 ± 112 109 ± 181 0.73 (0.64; 1.20) Cobblestones 12 ± 30 16 ± 44 0.71 (0.20; 2.48) Gravel 32 ± 77 23 ± 92 1.13 (0.65; 1.98) Condition cycle lane Poor 28 ± 58 18 ± 78 1.25 (0.61; 2.58) Moderate 223 ± 182 257 ± 258 0.93 (0.76; 1.15) Good 335 ± 226 264 ± 229 1.15 (0.93; 1.41) Reference = shortest cycling route. For ease of interpretation of OR, distances were converted to hectometres (100 m/km) OR odds ratio, CI confidence interval *p ≤ 0.05; **p ≤ 0.01; p ≤ 0.1 Table 5 Presence of items on aesthetics along actual cycling routes compared to shortest cycling routes Item Actual cycling route (m/km; M ± SD) Shortest cycling route (m/km; M ± SD) OR (95% CI) Trees 459 ± 241 428 ± 294 1.04 (0.88; 1.24) Attractive buildings 60 ± 111 70 ± 134 0.94 (0.65; 1.37) Well-maintained buildings 501 ± 240 500 ± 287 1.00 (0.84; 1.19) Front yards 297 ± 257 328 ± 294 0.96 (0.81; 1.13) Well-maintained front yards 315 ± 247 398 ± 274 0.88 (0.73; 1.07) Attractive natural features 250 ± 310 163 ± 243 1.12 (0.95; 1.33) Graffiti and litter 120 ± 201 93 ± 205 1.07 (0.85; 1.35) Reference = shortest cycling route. For ease of interpretation of OR, distances were converted to hectometres (100 m/km) OR odds ratio, CI confidence interval *p ≤ 0.05; **p ≤ 0.01; p ≤ 0.1 Discussion arterial roads, cycle lanes separated from traffic by white The present study aimed to investigate differences in lines, small cycle lanes and cycle lanes covered by light- physical environmental characteristics between ado- ing were less frequently present along adolescents’ actual lescents’ actual cycling routes and the shortest possible cycling routes in comparison to the shortest possible cycling routes using a Google Street View-based audit. cycling routes. Besides, a speed limit of 30  km/h, roads A mixed land use, roads with commercial destinations, having few buildings with windows on street side and Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 11 of 15 roads without cycle lane were more frequently present routes included a larger part of the route where only few along actual cycling routes compared to the shortest pos- buildings with windows on the street were present, which sible cycling routes. is also an attribute of local roads. As already mentioned In line with previous studies [12–14], the present study above, the presence of buildings with windows on the showed that a short cycling distance is one of the most street refers to social control from people living in the important factors determining the route choice of ado- area [22]. Another study among 5-to-18-year-old youth lescent cyclists, as for 73.1% of the cycling trips partici- found similar results, participants in that study agreeing pants took the shortest possible route. In addition, for that ‘walkers and bikers on the streets in my neighbour- 35.1% of the cycling trips which were not the shortest hood can easily be seen by people in their homes’ were possible, adolescents still indicated that they chose that less likely to use active transport to school [39]. These route because they perceived it as the shortest/fastest findings could be explained by the fact that adolescents route. Thus, even if a route is not actually the shortest, perceive cycling along local roads with lower speed limits adolescents may choose this route because they perceive as more important than potential social control from res- it as the shortest route. For all cycling trips that were not idents. Among adults, Winters et al. [16] also found that the shortest possible, a mean difference of 516 m (15.6%) cyclists spent most of their travel distance along local between the actual and the shortest cycling route was roads. In the present study, arterial roads, such as road found. When only looking at those routes which were types which consist of two separate roads each divided not the shortest but adolescents perceived as the short- in two lanes each direction, were avoided. A number of est/fastest route, a mean difference of 431  m between previous studies also found that cyclists avoid busy, arte- the actual and the shortest cycling route was found, and rial roads [15, 16, 40] and roads with high traffic speed thus, the detour showed to be smaller. It is possible that [41]. In Flanders, if any type of cycle path is available adolescents do not notice the difference in cycling time along these busy roads it is typically a small cycle path between their actual cycling route they perceive as the separated from traffic by white lines, which explains why shortest and the shortest possible cycling route. the present study found that this type of cycle path is less Although some findings of the present study seem to be present along actual cycling routes. in contradiction with results of previous cross-sectional With regard to walking/cycling roads that are not studies [12], the findings generally have a clear explana - accessible for motorised traffic, no significant differ - tion. The present study showed that adolescents avoid ence in presence along actual and shortest cycling routes routes with a mixed land use where commercial destina- was found. Nevertheless, a previous study among adults tions are present. In contradiction, a US study showed found that cyclists spent more time on off-street paths that children and adolescents (5–18  years) were more [16]. In addition, among 10-to-15-year-old US girls, it likely to walk or cycle to school if their parents reported was found that the presence of walking/cycling trails in having stores in the neighbourhood environment [38]. the neighbourhood was associated with higher levels of However, in accordance with the present study, shops active transport to school [39]. Although, in the present and services were also less present along the actual study, these walking/cycling roads occurred relatively fre- cycling routes of adults in Austria [15]. Dessing et al. [17] quently along actual cycling routes (on average for 89 m/ found that children in The Netherlands mainly cycled to km), shortest cycling routes also included some amount school along residential areas to avoid busy streets. These of walking/cycling roads (62  m/km) since the city of findings are similar to results in our study since residen - Ghent already provides an extensive network of walking/ tial areas are, in general, characterised by a lower land cycling roads. These walking/cycling roads often serve as use mix and less commercial destinations. In the study shortcuts for pedestrians and cyclists. This could explain by Dessing et  al. [17], it was suggested that residential why no significant difference between actual and shortest streets may be perceived as safe, quiet streets to cycle for cycling routes was found for walking/cycling roads in this transport, even if separate cycle lanes are absent. This study. could be confirmed by the results of the present study which showed that adolescents mainly cycled along local Practical implications roads, such as roads for one-direction traffic and roads Based on the findings of the present study, some rec - which were not divided into lanes (trends towards sig- ommendations for policy and practice can be formu- nificance), where speed limits of 30 km/h apply. Further - lated. The present study showed that adolescents mainly more, our study also showed that actual cycling routes choose the (perceived) shortest route to cycle for trans- included more m/km road where no cycle lane was pre- port and that adolescents frequently use walking/cycling sent which is typical for residential streets in Flanders. roads that are not accessible for motorised traffic. It In addition, the present study found that actual cycling might thus be important for local authorities to provide Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 12 of 15 walking/cycling roads that are not accessible for motor- were performed in a (sub)urban area. u Th s, results can - ised traffic and could serve as shortcuts for cyclists [42]. not be generalized to rural areas where less alternative These shortcuts free from motorised traffic also meet the routes are available because of a less dense street net- preference of adolescents to avoid cycling along arterial work. Fifth, because of the limited time window of the roads. However, since it is not always possible for an indi- study (i.e. 2 days of data per participant), it is difficult to vidual to avoid to cycle along busy, arterial roads, these draw generalizable conclusions regarding the impact of roads should be made more bike-friendly by providing the physical environment on adolescents’ route choice adequate cycling infrastructure. while cycling. Sixth, the majority of adolescents was of a higher SES family, which could have influenced the Strengths and limitations results. Previous research showed that children and ado- A first strength of the present study was that the evalu - lescents living in lower SES neighbourhoods perceive the ated routes were actual cycling routes which were objec- neighbourhood environment as less attractive and safe, tively recorded using a GPS device. Using objective GPS and more often report heavy traffic in their neighbour - data limits recall bias related to route choice. Particularly hood compared to those living in higher SES neighbour- for young people such as adolescents it may be difficult hoods [43, 44]. It is thus possible that adolescents from to recall and indicate on a map which route they took at lower SES families attach importance to other factors on a particular moment. Second, information obtained by their cycling route compared to adolescents from higher the structured one-on-one interviews enabled to cor- SES families. Thus, caution is needed when generalizing rect trip mode when this was misclassified by PALMS results to the overall adolescent population. Seventh, data (e.g. when a car trip was classified as a bicycle trip due to were collected during autumn/winter which may have traffic congestion). Furthermore, this was the first study influenced the results. Bad weather and less hours of day - to collect subjective information regarding route choice light may influence adolescents’ choice of transport mode via one-on-one interviews, and combine this with audits. and route choice. Finally, results of the present study do This allowed participants to indicate their actual reason not enable to draw conclusions regarding non-cyclists. for choosing a particular route. Third, the presence of physical environmental characteristics along the routes Recommendations for future research was measured objectively using a tool (EGA-Cycling) Since data collection was very time-consuming and the that showed acceptable reliability and validity [22], which burden on participants was rather high, future studies limits the bias of results compared to self-reported ques- should consider to make use of dedicated smartphone tionnaires. Nevertheless, some limitations should be applications to identify adolescents’ actual cycling routes. acknowledged. First, virtual audit tools showed to be less Adolescents generally carry their smartphones with them accurate for measuring micro-environmental features. during the day, thus running dedicated mobile apps may However, Google Street View showed to be more accu- be less considered as a burden compared to wearing rate in measuring micro-environmental features com- portable GPS-devices. The use of dedicated smartphone pared to other virtual audit tools [28]. Nevertheless, a applications would enable to include a larger sample discrepancy in physical environmental factors may exist and would allow to track adolescents’ mobility patterns between the Google Street View images and the period in over a longer time period [45]. Thus, this method has which GPS data were collected. The Google Street View the advantage that much more actual cycling routes can images ranged from March 2009 till April 2015, whereas be identified in more diverse areas. The introduction of participants’ GPS data were collected between Septem- smartphones with a longer battery life and a higher stor- ber and December 2015. Second, the sample size was rel- age space and memory capacity should be able to facili- atively small as only 38 actual cycling routes (spread over tate this type of data collection. Nevertheless, this would 22 adolescents) that were not the shortest cycling route imply a huge burden on the researchers to audit such a could be identified and were evaluated. A small sample large set of cycling routes. Geographic Information Sys- size increases the likelihood of a type II error and, thus, tems (GIS) could be used instead, since GIS makes use the chance that an effect was not detected when there of existing data sources (e.g. governmental data sources) was one to be detected. Third, some characteristics (i.e. to measure physical environmental characteristics that poor visibility when crossing a street, parking facilities have some spatial reference [46]. However, since for some on adjacent parking) were only present along a few street locations (i.e. rural or suburban areas) GIS data may segments. This resulted in wide 95% confidence inter - not be available [46] and micro-environmental factors vals due to insufficient variability. Fourth, as data collec - are also not commonly available in GIS databases [46, tion took place among adolescents attending secondary 47], virtual or on-street audits may be used to comple- schools in the city of Ghent, the majority of cycling trips ment GIS data where needed. In addition, future studies Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 13 of 15 should consider to investigate potential moderators (e.g. environmental factors besides distance. Local authorities individual and social environmental factors) of the rela- should provide shortcuts free from motorised traffic in tionships between the presence of physical environmen- order to meet the preference of adolescents to cycle along tal factors and adolescents’ route choice while cycling for the shortest route and to avoid cycling along busy, arte- transport. It may be interesting to investigate whether rial roads. In addition, it may also be important to pro- the relationship between the presence of certain physi- vide adequate cycling infrastructure along busy, arterial cal environmental factors and adolescents’ preference roads since these roads cannot always be avoided. for a certain cycling route is moderated by, for example, Additional files psychosocial factors. It could be that adolescents with lower self-efficacy or less social models for active trans - Additional file 1. Structured one-on-one interview. port attach importance to other physical environmental Additional file 2. EGA-Cycling checklist. factors when choosing a cycle route compared to adoles- cents with higher  self-efficacy or more social models for active transport. Results of the present study only enable Authors’ contributions to draw conclusions for the general adolescent popula- HV, LVH, DVD, NVDW, BD and JVC designed the protocol for this study. HV tion, no specific conclusions for subgroups of adolescents and LVH collected the data and coordinated the data collection. HV, LVH and (e.g. those with a low psychosocial profile towards cycling TB contributed to the data processing. HV performed the statistical analyses and drafted the manuscript. HV, LVH, DVD, NVDW, PC, BD, JVC critically revised for transport or the least regular cyclists) can be drawn. and helped to draft the manuscript. All authors read and approved the final Since this study was one of the first exploring the factors manuscript. associated with route choice among adolescent cyclists, Author details results are valuable. However, future studies may consider Department of Public Health, Faculty of Medicine and Health Sciences, to conduct moderation analyses among larger samples in 2 Ghent University, Corneel Heymanslaan 10, 9000 Ghent, Belgium. Physical order to be able to formulate recommendations to target Activity, Nutrition and Health Research Unit, Faculty of Physical Education and Physical Therapy, Vrije Universiteit Brussel, Pleinlaan 2, 1050 Brussels, specific subgroups of adolescents. More research investi - 3 4 Belgium. Research Foundation - Flanders (FWO), Brussels, Belgium. Depart- gating adolescents’ route choice for cycling is needed. In ment of Movement and Sport Sciences, Faculty of Medicine and Health Sci- order to be able to formulate recommendations regard- ences, Ghent University, Watersportlaan 2, 9000 Ghent, Belgium. Department of Geography – CartoGIS, Faculty of Sciences, Ghent University, Krijgslaan 281, ing which factors may stimulate adolescents to cycle for 9000 Ghent, Belgium. transport, future studies should also investigate which factors along a route are important among non-cyclists. Acknowledgements We would like to thank the schools and the adolescents who participated in An experimental study which aimed to investigate ado- the study. Furthermore, we would like to thank master’s students for assisting lescents’ preferences towards cycling for transport using with the data collection and processing. We would also like to thank R. Col- manipulated photographs, showed that the least regular man for her help with the data analyses. cyclists in that study attached most importance to cycling Competing interests distance when indicating which route they preferred to The authors declare that they have no competing interests. cycle along [48]. The most regular cyclists in that study Availability of data and materials attached most importance to being able to cycle together The datasets used and/or analysed during the current study are available from with a friend. However, no associations with actual par- the corresponding author on reasonable request. ticipation in cycling for transport were investigated in Consent for publication that study. Not applicable. Ethics approval and consent to participate Conclusions Passive informed consent was obtained from adolescents’ parents. If parents For 73.1% of the cycling trips, participants took the did not agree to let their child participate in the study, they had to sign a form. Furthermore, researchers also obtained active informed consent from shortest route possible which confirmed the importance adolescents. The study protocol was approved by the Ethics Committee of the of cycling distance for adolescents. When not taking University Hospital of Ghent University (EC 2015/0317). the shortest cycling route, adolescents avoided to cycle Funding for transport along arterial roads with a small cycle lane HV and LVH are supported by the Research Foundation Flanders (FWO, http:// separated from traffic by white lines. Local roads with www.fwo.be/en, 3GOA8514). JVC is supported by a FWO postdoctoral fellow- a speed limit of 30  km/h in an area with a low land use ship (FWO - 12I1117N). DVD is supported by a FWO postdoctoral fellowship (FWO12/PDO/158). mix where few commercial destinations are located were more frequently used, even when no cycle lane was avail- Publisher’s Note able. In general, the ability to cycle along quiet, local Springer Nature remains neutral with regard to jurisdictional claims in pub- roads overruled the importance of all other physical lished maps and institutional affiliations. Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 14 of 15 Received: 8 January 2018 Accepted: 22 May 2018 20. Pikora TJ, Bull FC, Jamrozik K, Knuiman M, Giles-Corti B, Donovan RJ. Developing a reliable audit instrument to measure the physical envi- ronment for physical activity. Am J Prev Med. 2002;23:187–94. 21. Day K, Boarnet M, Alfonzo M, Forsyth A. The Irvine-Minnesota inven- tory to measure built environments: development. Am J Prev Med. 2006;30:144–52. References 22. Vanwolleghem G, Van Dyck D, Ducheyne F, De Bourdeaudhuij I, Cardon 1. Cohen AJ, Ross Anderson H, Ostro B, Pandey KD, Krzyzanowski M, Kunzli G. Assessing the environmental characteristics of cycling routes to N, Gutschmidt K, Pope A, Romieu I, Samet JM, Smith K. The global burden school: a study on the reliability and validity of a Google Street View- of disease due to outdoor air pollution. J Toxicol Environ Health A. based audit. Int J Health Geogr. 2014;13:19. 2005;68:1301–7. 23. Brownson RC, Hoehner CM, Brennan LK, Cook RA, Elliot MB, McMullen 2. Woodcock J, Edwards P, Tonne C, Armstrong BG, Ashiru O, Banister D, KM. Reliability of 2 Instruments for Auditing the Environment for Physi- Beevers S, Chalabi Z, Chowdhury Z, Cohen A, et al. Public health benefits cal Activity. J Phys Act Health. 2004;1:189–207. of strategies to reduce greenhouse-gas emissions: urban land transport. 24. Badland HM, Opit S, Witten K, Kearns RA, Mavoa S. Can virtual Lancet. 2009;374:1930–43. streetscape audits reliably replace physical streetscape audits? J Urban 3. Int Panis L, de Geus B, Vandenbulcke G, Willems H, Degraeuwe B, Bleux Health. 2010;87:1007–16. N, Mishra V, Thomas I, Meeusen R. Exposure to particulate matter in 25. Taylor BT, Fernando P, Bauman AE, Williamson A, Craig JC, Redman S. traffic: a comparison of cyclists and car passangers. Atmos Environ. Measuring the quality of public open space using Google Earth. Am J 2010;44:2263–70. Prev Med. 2011;40:105–12. 4. Oja P, Titze S, Bauman A, de Geus B, Krenn P, Reger-Nash B, Kohlberger T. 26. Rundle AG, Bader MD, Richards CA, Neckerman KM, Teitler JO. Using Health benefits of cycling: a systematic review. Scand J Med Sci Sports. Google Street View to audit neighborhood environments. Am J Prev 2011;21:496–509. Med. 2011;40:94–100. 5. Ortega FB, Konstabel K, Pasquali E, Ruiz JR, Hurtig-Wennlof A, Maestu J, 27. Kelly CM, Wilson JS, Baker EA, Miller DK, Schootman M. Using Google Lof M, Harro J, Bellocco R, Labayen I, et al. Objectively measured physical Street View to audit the built environment: inter-rater reliability results. activity and sedentary time during childhood, adolescence and young Ann Behav Med. 2013;45(Suppl 1):S108–12. adulthood: a cohort study. PloS ONE. 2013;8:e60871. 28. Ben-Joseph E, Lee JS, Cromley EK, Laden F, Troped PJ. Virtual and actual: 6. Chillon P, Ortega FB, Ruiz JR, De Bourdeaudhuij I, Martinez-Gomez D, relative accuracy of on-site and web-based instruments in auditing the Vicente-Rodriguez G, Widhalm K, Molnar D, Gottrand F, Gonzalez-Gross environment for physical activity. Health Place. 2013;19:138–50. M, et al. Active commuting and physical activity in adolescents from 29. Gullon P, Badland HM, Alfayate S, Bilal U, Escobar F, Cebrecos A, Diez J, Europe: results from the HELENA study. Pediatr Exerc Sci. 2011;23:207–17. Franco M. Assessing walking and cycling environments in the streets 7. Chillon P, Ortega FB, Ruiz JR, Veidebaum T, Oja L, Maestu J, Sjostrom M. of Madrid: comparing On-field and virtual audits. J Urban Health. Active commuting to school in children and adolescents: an oppor- 2015;92:923–39. tunity to increase physical activity and fitness. Scand J Public Health. 30. Statistics Belgium: Bodemgebruik in België 1834–2017. 2015. 2010;38:873–9. 31. Structuur van de bevolking volgens woonplaats: grootste gemeenten. 8. Bere E, Seiler S, Eikemo TA, Oenema A, Brug J. The association between http://statb el.fgov.be/nl/stati stiek en/cijfe rs/bevol king/struc tuur/ cycling to school and being overweight in Rotterdam ( The Netherlands) woonp laats /groot /. and Kristiansand (Norway). Scand J Med Sci Sports. 2011;21:48–53. 32. Lien N, Friestad C, Klepp KI. Adolescents’ proxy reports of parents’ 9. Nettleton S, Green J. Thinking about changing mobility practices: how a socioeconomic status: how valid are they? J Epidemiol Community social practice approach can help. Sociol Health Illn. 2014;36:239–51. Health. 2001;55:731–7. 10. Sallis JF, Cervero RB, Ascher W, Henderson KA, Kraft MK, Kerr J. An ecologi- 33. Kerr J, Normam G, Godbole S, Raab F, Demchak B, Patrick K. Validating cal approach to creating active living communities. Annu Rev Public GPS data with the PALMS system to detect different active transporta- Health. 2006;27:297–322. tion modes. Med Sci Sports Exerc. 2012;44:647. 11. Wong BY, Faulkner G, Buliung R. GIS measured environmental correlates 34. The Personal Activity and Location Measurement System (PALMS). of active school transport: a systematic review of 14 studies. Int J Behav https ://ucsd-palms -proje ct.wikis paces .com/. Nutr Phys Act. 2011;8:39. 35. Demchak B, Kerr J, Raab F, Patrick K, Krüger IH. PALMS: a modern coevo- 12. Panter JR, Jones AP, van Sluijs EM. Environmental determinants of active lution of community and computing using policy driven development. travel in youth: a review and framework for future research. Int J Behav In 45th Hawaii international conference on system sciences; 2012. Nutr Phys Act. 2008;5:34. 36. Carlson JA, Jankowska MM, Meseck K, Godbole S, Natarajan L, Raab F, 13. Nelson NM, Foley E, O’Gorman DJ, Moyna NM, Woods CB. Active com- Demchak B, Patrick K, Kerr J. Validity of PALMS GPS scoring of active muting to school: how far is too far? Int J Behav Nutr Phys Act. 2008;5:1. and passive travel compared with SenseCam. Med Sci Sports Exerc. 14. Verhoeven H, Ghekiere A, Van Cauwenberg J, Van Dyck D, De Bourdeaud- 2015;47:662–7. huij I, Clarys P, Deforche B. Which physical and social environmental 37. MAPS GLOBAL audit tool. http://salli s.ucsd.edu/measu re_maps.html. factors are most important for adolescents’ cycling for transport? An 38. Kerr J, Rosenberg D, Sallis JF, Saelens BE, Frank LD, Conway TL. Active experimental study using manipulated photographs. Int J Behav Nutr commuting to school: associations with environment and parental Phys Act. 2017;14:108. concerns. Med Sci Sports Exerc. 2006;38:787–94. 15. Krenn PJ, Oja P, Titze S. Route choices of transport bicyclists: a com- 39. Evenson KR, Birnbaum AS, Bedimo-Rung AL, Sallis JF, Voorhees CC, parison of actually used and shortest routes. Int J Behav Nutr Phys Act. Ring K, Elder JP. Girls’ perception of physical environmental factors and 2014;11:31. transportation: reliability and association with physical activity and 16. Winters M, Teschke K, Grant M, Setton E, Brauer M. How far out of the way active transport to school. Int J Behav Nutr Phys Act. 2006;3:28. will we travel? Built environment influences on route selection for bicycle 40. Timperio A, Ball K, Salmon J, Roberts R, Giles-Corti B, Simmons D, Baur and car travel. Transp Res Rec J Transp Res Board. 2010;2190:1–10. LA, Crawford D. Personal, family, social, and environmental correlates of 17. Dessing D, de Vries SI, Hegeman G, Verhagen E, van Mechelen W, Pierik active commuting to school. Am J Prev Med. 2006;30:45–51. FH. Children’s route choice during active transportation to school: dif- 41. McMillan TE. The relative influence of urban form on a child’s travel ference between shortest and actual route. Int J Behav Nutr Phys Act. mode to school. Transp Res Part A. 2007;41:69–79. 2016;13:48. 42. Pucher J, Buehler R. Making cycling irresistible: lesson from Europe. 18. Carpiano RM. Come take a walk with me: the “go-along” interview as a Transp Rev. 2008;28:495–528. novel method for studying the implications of place for health and well- 43. Giles-Corti B, Donovan RJ. Socioeconomic status differences in being. Health Place. 2009;15:263–72. recreational physical activity levels and real and perceived access to a 19. Clifton KJ, Livi Smith ADL, Rodriguez D. The development and test- supportive physical environment. Prev Med. 2002;35:601–11. ing of an audit for the pedestrian environment. Landsc Urban Plan. 2007;80:95–110. Verhoeven et al. Int J Health Geogr (2018) 17:16 Page 15 of 15 44. Timperio A, Crawford D, Telford A, Salmon J. Perceptions about the local 47. Adams MA, Ryan S, Kerr J, Sallis JF, Patrick K, Frank LD, Norman GJ. Valida- neighborhood and walking and cycling among children. Prev Med. tion of the Neighborhood Environment Walkability Scale (NEWS) items 2004;38:39–47. using geographic information systems. J Phys Act Health. 2009;6(Suppl 45. Vlassenroot S, Gillis D, Bellens R, Gautama S. The use of smartphone 1):S113–23. applications in the collection of travel behaviour data. Int J Intell Transp 48. Verhoeven H, Ghekiere A, Van Cauwenberg J, Van Dyck D, De Bourdeaud- Syst Res. 2015;13:17–27. huij I, Clarys P, Deforche B. Subgroups of adolescents differing in physical 46. Brownson RC, Hoehner CM, Day K, Forsyth A, Sallis JF. Measuring the built and social environmental preferences towards cycling for transport: a environment for physical activity: state of the science. Am J Prev Med. latent class analysis. Prev Med. 2018;112:70–5. 2009;36(S99–123):e112. Ready to submit your research ? Choose BMC and benefit from: fast, convenient online submission thorough peer review by experienced researchers in your field rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions

Journal

International Journal of Health GeographicsSpringer Journals

Published: May 29, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off