*Article* **12-Year Trends in Active School Transport across Four European Countries—Findings from the Health Behaviour in School-Aged Children (HBSC) Study**

**Ellen Haug 1,2,\*, Otto Robert Frans Smith 3, Jens Bucksch 4, Catherina Brindley 4, Jan Pavelka 5, Zdenek Hamrik 5, Joanna Inchley 6, Chris Roberts 7, Frida Kathrine Sofie Mathisen <sup>1</sup> and Dagmar Sigmundová <sup>8</sup>**


**Abstract:** Active school transport (AST) is a source of daily physical activity uptake. However, AST seems to have decreased worldwide over recent decades. We aimed to examine recent trends in AST and associations with gender, age, family affluence, and time to school, using data from the Health Behaviour in School-Aged Children (HBSC) study collected in 2006, 2010, 2014, and 2018 in the Czech Republic, Norway, Scotland, and Wales. Data from 88,212 students (11, 13 and 15 years old) revealed stable patterns of AST from 2006 to 2018, apart from a decrease in the Czech Republic between 2006 and 2010. For survey waves combined, walking to and from school was most common in the Czech Republic (55%) and least common in Wales (30%). Cycling was only common in Norway (22%). AST differed by gender (Scotland and Wales), by age (Norway), and by family affluence (everywhere but Norway). In the Czech Republic, family affluence was associated with change over time in AST, and the effect of travel time on AST was stronger. The findings indicate that the decrease in AST could be levelling off in the countries considered here. Differential associations with sociodemographic factors and travel time should be considered in the development of strategies for AST.

**Keywords:** active school transport; trends; cross-national; HBSC; gender; age; SES

#### **1. Introduction**

In recent years, active travel has become an integral part of international initiatives aimed at increasing levels of physical activity within the population [1–4]. Walking and cycling to school have gained considerable attention as sources of young peoples' daily physical activity. More recently, there has also been an increased focus on active commuting as a sustainable form of transport that can reduce problems caused by motorised vehicles, with potentially significant economic benefits and public health impacts [3,5–8].

Reviews of the growing body of literature in the field strongly support a positive relationship between active school transport (AST) and levels of physical activity [9,10].

**Citation:** Haug, E.; Smith, O.R.F.; Bucksch, J.; Brindley, C.; Pavelka, J.; Hamrik, Z.; Inchley, J.; Roberts, C.; Mathisen, F.K.S.; Sigmundová, D. 12-Year Trends in Active School Transport across Four European Countries—Findings from the Health Behaviour in School-Aged Children (HBSC) Study. *IJERPH* **2021**, *18*, 2118. https://doi.org/10.3390/ ijerph18042118

Academic Editor: Paul B. Tchounwou Received: 22 December 2020 Accepted: 16 February 2021 Published: 22 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Positive associations have also been found between cycling to and from school and cardiovascular fitness [9–12]. A relationship between AST and body composition indicators is less clear [9,10]. Other potential co-benefits of AST relate to improved navigation and road safety abilities [13,14], better processing of the physical surroundings [15], higher activation (i.e., alertness and activity) during school hours [16], and the development of long-term physical activity and active transportation habits [17,18].

Despite the potential benefits, the number of children and youths that walk or cycle to school seems to have decreased worldwide in recent decades [19–28]. However, large cross-country variations are observed in the prevalence of AST and the magnitude of decline [21]. In the United States, the prevalence of AST among children dropped from 49 to 13% between 1969 and 2009 [25], whereas it declined from 44 to 21% among 10–14 year old children in Australia between 1971 and 2003 [24]. Findings from Europe display a more mixed picture, with generally higher proportions of AST being reported, but decreasing trends have also been observed in European studies [22,26,29,30]. Many of the existing studies on trends in AST span over several decades, covering a period when use of motorised vehicles increased dramatically. Furthermore, the studies are limited to data from individual countries preventing reliable cross-national comparisons. More recently, policy initiatives and national programs to promote active school commuting have been initiated in many countries [3,6,31–33], which may have affected schoolchildren's travel behaviours.

Contemporary studies suggest that AST is associated with a wide range of factors, such as demographic (gender, ethnicity, age), family (parental education, household income, car ownership), social (individual and parental attitudes and concerns, social and cultural norms), environmental (school distance, safety, walkability, traffic calming, infrastructure, recreational facilities, centralization), and policy-related factors [6,25,29,34–36]. It is to be expected that some of the underlying drivers of AST will vary between and within countries [23]. However, a large body of research has identified the distance between home and school and time taken to travel to school as the strongest predictors of AST [6,14,29,36,37], and findings from several countries point towards an increase in the distance to school over time [20,27,29,38].

Gender, age, and socioeconomic status (SES) have been identified as potential moderators of AST [39]. Findings from North America, Australia, New Zealand, and the Czech Republic indicate higher levels of AST for boys more often than girls [14,32,36,40,41], while gender differences have not been observed in studies from Switzerland [26] and Norway [11]. Regarding gender difference trends in recent times, there is no clear pattern, with no differences [41,42], a decline only for girls [43] and a decline in boys and an increase in girls [27] having been observed. The relationship between age and AST is expected to be curvilinear, with an initial age-related increase due to more independent mobility and parental allowances, followed by a decrease because of generally longer distances to secondary schools compared with localised primary schools [14]. Socioeconomic differences in AST have been less well studied. The literature from North America and New Zealand generally shows that low-income households and lower parental education correlate with more AST [14,36]. Nevertheless, the lack of standardised measures and comparable control variables across different studies makes it difficult to compare, aggregate and interpret findings [35,36,44,45].

Cross-national studies of current time trends with a comparable methodological approach are of interest because they can provide unique insights into how recent developments, as well as national and local level policies, may have had an impact on AST. It has been suggested that future research should also consider changes in key AST correlates over time to support the development of new policies, regulations, designs, and programme interventions [36]. To improve the understanding of young peoples' transport to school in different regions in Europe between 2006 and 2018, the current study aimed to examine secular trends in AST and their associations with gender, age, SES, and time

to school across Northern Europe (Norway), western Europe (Scotland and Wales) and central Europe (the Czech Republic).

#### **2. Materials and Methods**

#### *2.1. Study Population and Procedures*

The Health Behaviour in School-Aged Children study (HBSC) is unique in collecting comparable cross-national data on representative samples of young peoples' health behaviours every fourth year. A standardised international protocol ensures the consistency of measures, sampling, and implementation procedures prepared by the HBSC International Coordinating Centre [46].

Data stem from the HBSC studies conducted in 2006, 2010, 2014, and 2018 in the Czech Republic, Norway, Scotland, and Wales. Data on AST were obtained from a total of 88,212 children across the four time points (2006: *n* = 18,317, 50.3 % girls; 2010: *n* = 18,902, 51.0% girls; 2014: *n* = 17,699, 51.3 % girls; 2018: *n* = 33,294, 50.6 % girls). The age and gender distributions were fairly stable across countries and survey years (Table S1). Students were surveyed to produce representative national estimates for 11, 13, and 15 year old children. Classes within schools were selected with variations in sampling criteria which allowed us to fit country-level circumstances (e.g., national regions, type of school, and size of schools). Ethical approval for the surveys was obtained at the national level. Participation was voluntary and the children were informed about confidentiality and anonymity. Classroom teachers or trained administrators conducted the survey and consent (explicit or implicit) was given from school administrators and/or parents before participation. More details on the HBSC study procedures can be found elsewhere [46].

#### *2.2. Survey Items*

#### 2.2.1. Active School Transport

Mode of travel to and from school was assessed with two questions: "On a typical day is the main part of your journey to school made by ... ?" and "On a typical day is the main part of your journey from school made by ... ?". Response options were "Walking", "Bicycle", "Bus, train, tram, underground or boat", "Car, motorcycle or moped" or "Other means". A slightly different version of the AST items used earlier in the 1985/86 HBSC study have been examined, with the reliability in terms of Cronbach's alpha found to be 0.83 and a correlation with the total weekday physical activity score—measured by accelerometers—of 0.20 (*p* < 0.01) [47]. In the present study, only 1–1.5% reported "Other means"; this category was included in non-active transport. For the prevalence and trend analyses, AST was both used as a categorical variable based on 4 categories (walking both ways, cycling both ways, active one-way only, non-active transport), and as a categorical variable based on 2 categories (active transport both ways vs. one-way only or non-active transport). The latter categories in both cases were used as a reference in the analysis that follows.

Time to school was assessed with one question "How long does it usually take you to travel to school from your home?" and was used as a proxy for distance to school. Response options were "Less than 5 min", "5–15 min", "15–30 min", 30 min to 1 h" and "More than 1 h". For the analyses, this variable was recoded into three categories (Less than 5 min, 5–15 min, >15 min). Again, the latter category was used as reference. Travel time to school increased between 2006 and 2018 in Czech Republic and Wales, whereas this remained fairly stable in Norway and Scotland (Supplementary Table S1).

#### 2.2.2. Sociodemographic Variables

Gender (boys vs. girls), age groups (11, 13, and 15 year olds), and individual family affluence (FAS—Family Affluence Scale (FAS-II)) were included in the analysis. The latter is a validated HBSC measure of SES [48]. Family affluence is a composite sum score, which resembles a valid measure of household material affluence derived from participants' responses to 4 items describing the material conditions of their household (respondents'

own household bedrooms, family holidays, family vehicle ownership, and PC ownership). FAS has changed through time but this version was used in 2006, and was therefore applied in the current study. Responses to the individual items are summed on a 9-point scale with set cut-points for low (0 to 3), medium (4 to 5), and high (6 to 9) affluence. The individual FAS responses were combined and standardised by using ridit transformation to give a linear SES-score (0–1). The regression coefficient of the FAS score can be directly interpreted as the predicted difference in AST between the least deprived individual and the most deprived individual. When using this procedure, ordered categorical variables are converted to cumulative probabilities, and the individuals are thus ranked on this continuum. Ridit transformation has previously been applied in inequality studies using SES scales with ordinal measurements [49–51] and is recommended for comparisons of the effects of FAS [52]. Family affluence increased between 2006 and 2018 in the Czech Republic, Scotland, and Wales, and remained relatively stable in Norway (Supplementary Table S1).

#### *2.3. Data Analysis*

All analyses were conducted using Stata version 15 (StataCorp LLC, College Station, Texas, USA). Stata's survey command (svyset) was used to adjust for sampling weight, clustering, and stratification in the sampling design. The alpha level was set to 0.001 given the large sample size and the number of tests. Joint significance of regression terms containing polytomous categorical variables was determined by means of adjusted F-tests. Secular trends were examined both for AST based on 4 categories and for AST based on 2 categories by means of multinomial and logistic regression, respectively. The initial model included age, gender, and country (Section 3.1). In the next step, the country-bysurvey year interaction was added (Section 3.2). Pending statistical significance of this interaction term, age- and gender-adjusted results were presented separately for each country. Results were adjusted for age and gender to make sure that changes in AST could not be attributed to changes in age or gender distributions across survey years. Separate analyses were conducted for each survey year, modelled as a categorical variable and as a continuous variable (linear trend). For categorical time, backward difference coding was used to allow for the comparison between consecutive survey years. To determine whether SES and time to school were related to trends in AST, these two variables were added to the model as main effects (Section 3.2). These factors were considered to be potentially explanatory when the OR associated with survey year was reduced by ≥10% [53]. For ease of interpretation, the remaining models were only conducted for AST based on 2 categories and with survey year modelled as a categorical variable (Section 3.3). Country differences were explored further by adding the two-way interactions of country by gender, age, SES, and time to school, respectively. Finally, potential country differences over time were explored by testing a model with three-way interactions of country by survey year by, respectively, gender, age, SES, and time taken to travel to school. For statistically significant interaction terms, country differences were examined across survey years.

#### **3. Results**

#### *3.1. AST by Country across Surveys*

For AST based on four categories, the adjusted F-test for country was F(9, 2258) = 204.5, *p* < 0.001, and for AST based on two categories F(3, 2264) = 136.1, *p* < 0.001. This indicated significant variation in AST by country for all survey waves combined (Table 1). Walking to school both ways was most common in the Czech Republic (55%) and least common in Wales (30%), whereas cycling to school was limited, with the exception of Norway (22%). One-way AST was relatively uncommon and there was also only modest variation in its prevalence between countries (6–10%). When considering active travel both ways (Table 1) the prevalence in the Czech Republic (57%) and Norway (59.4%) was significantly higher as compared to Scotland (46%) and Wales (31%).


**Table 1.** Age/gender-adjusted prevalence of active school transport (AST) by country across survey years.

Ref. active travel 4 cat. = no AST based on multinominal logistic regression, ref. active travel 2 cat. = no AST/one-way AST. Estimates in bold = *p* < 0.001, logistic regression.

#### *3.2. Secular Trends in AST by Country*

The country-by-categorical survey year interaction was statistically significant for both AST based on four categories (F(9, 2258) = 204.5, *p* < 0.001) and AST based on two categories (F(9, 2258) = 204.5, *p* < 0.001), indicating that the effect of survey year varied across countries. Similar results were obtained for treating time as a continuous variable. As shown in Table 2, AST changed significantly over time in Czech Republic, but remained stable in the other three countries. In the Czech Republic, there was a relatively sharp decrease in walking both ways between 2006 and 2010, followed by a stable pattern between 2010 and 2018. A similar, though less pronounced, pattern was found for cycling both ways. A small linear increase over time was observed for one-way AST in the Czech Republic. For AST based on two categories, this translated into a decrease in AST both ways between 2006 and 2010, followed by a stable pattern between 2010 and 2018. Despite the overall decrease in the Czech Republic between 2006 and 2018, the prevalence of AST both ways remained higher when compared to Scotland and Wales (Figure 1).

Adding family affluence and time to school to the basic country-specific models (age, gender, survey year) did not change the effect of survey year in Norway, Scotland, and Wales. In the Czech Republic, the odds ratio representing the change from 2006 to 2010 in walking both ways (based on four categories) changed from 0.57 to 0.64 after adding family affluence, 0.59 after adding time to school, and 0.65 after adding both variables. This equates to an OR-change of, respectively, 16, 5, and 19%. Family affluence and time to school did not change the significant effect of survey year on cycling both ways, neither did these variables change the linear effect of survey year on one-way AST. For AST based on two categories, the odds ratio representing a change from 2006 to 2010 in AST both ways changed from 0.54 to 0.60 after adding family affluence, 0.56 after adding time to school, and 0.62 after adding both variables. This equates to OR-change of 13, 4, and 17%, respectively. Overall, these results indicated that change over time in family affluence was associated with a change in walking both ways in the Czech Republic. It should be noted that all the mentioned effects of survey year in the Czech Republic remained statistically significant after adding family affluence and time to school to the model.

*IJERPH* **2021**, *18*, 2118


 Ref. AST 4 cat. = no AST, ref. AST 2 cat. = no AST/one-way AST. Estimates in bold = *p* < 0.001. The reported results of the adjusted F-tests and linear trends for the 4-category active travel variable arecountry and within the relevant outcome category. The joint significance of survey year across outcome categories was only statistically significant for Czech Republic (*<sup>p</sup>* < 0.001).

**Figure 1.** Age and gender adjusted prevalence of AST both ways by country and survey year. (CZ = Czech Republic, NO = Norway, SCT = Scotland, WLS = Wales).

#### *3.3. Country Differences in AST Both Ways by Gender, Age Group, Family Affluence and Time to School*

All four two-way interactions of gender, age group, family affluence, and time to school by country were statistically significant (*p* < 0.001), indicating that the effects of these variables varied by country. As shown in both Table 3 and Figure 2, there were significant gender differences in Scotland and Wales, with boys being more likely to exhibit AST both ways as compared to girls in these two countries. There was a particularly strong age group effect in Norway with 11 year olds being much more likely to exhibit AST both ways as compared to 13 and 15 year olds. Family affluence did not impact the probability of AST in Norway, whereas children in the other countries from more affluent families were less likely to have AST both ways as compared to their counterparts from less affluent families. The estimated probability difference of AST between children coming from the least affluent families (0) and children coming from the most affluent families (1) is displayed in Figure 2. Finally, the effect of time to school on AST was much stronger in the Czech Republic when compared with the other three countries. Children from the Czech Republic with a travel time of less than 5 min were more likely to engage in AST (86%), whereas this was much less likely in Scotland (62%) and Wales (50%). This was also true for travel times between 5–15 min, but the country differences were somewhat less pronounced for this category. Across surveys, the prevalence of travel time less than 5 min was 21% for the Czech Republic, 17% for Norway, 22% for Scotland, and 14% for Wales, whereas the prevalence of travel time between 5–15 min was 48% for the Czech Republic, 46% for Norway, 44% for Scotland, and 44% for Wales.



\* Estimates by country for model AST = Survey year + gender + Age group + Time to school + Family affluence. Ref. AST 2 cat. = no AST/one-way AST. CZ = Czech Republic, NO = Norway, SCT = Scotland, WLS = Wales. Ref. Age group = 11 year olds, Ref. Time to school = >15 min. Family affluence ridit transformed to a linear score (0–1). Estimates in bold = *p* < 0.001.

**Figure 2.** Country differences in AST both ways by gender, age, family affluence, and time to school.

None of the four three-way interactions of country by survey year by, respectively, gender, age, SES, and time to school were statistically significant, indicating that the observed country differences did not change over time.

#### **4. Discussion**

The current study provides recent trends in AST and their associations with gender, age, SES, and time to school across four European countries. The findings demonstrate that, apart from the Czech Republic, there were generally stable patterns of partly low levels of AST in the period from 2006 to 2018. However, the prevalence and association with moderating factors varied considerably between countries. For all survey waves combined, walking to school both ways were most common in the Czech Republic (55%), followed by Scotland (45%), Norway (37%), and Wales (30%). Cycling to school was only common in Norway, with the highest prevalence of total AST (59%). One-way AST was relatively uncommon with modest variation between countries. Overall, the results show that active travel both ways was considerably higher in the Czech Republic and Norway when compared to Scotland and Wales. Although, there was a general direction towards a decline across all four countries, this was non-significant. The stable trends in AST observed in the current study contrast with most studies [19–28] but are in line with one study from Australia that assessed trends between 2004–2010 [54] and the findings from 28 studies in Spain between 2010 to 2017 [41]. The findings suggest that a general decline in AST may have levelled out in various regions in Europe.

The prevalence of AST changed significantly over time only in the Czech Republic, with a relatively sharp decrease observed between 2006 and 2010, followed by a stable pattern between 2010 and 2018. This is in line with a previous study [42]. A small linear increase over time was also observed for one-way AST in the Czech Republic. The extent to which the recent initiation of services, such as bike shares, has contributed to this increase is unknown, but a topic worth further investigation. The market demand for micro-mobility is also expected to grow significantly. The availability of these personal vehicles may also influence AST in the future. The negative trend observed in the Czech Republic might be explained by increasing car ownership [55], insufficient cycling infrastructure at a municipal and school level [56], and barriers dealing with safety concerns [57]. The Czech Republic was also the only country where family affluence was associated with change over time in walking both ways. It may, therefore, be the case that the change in family affluence contributed to some of the change in AST, for example, through increased availability of family cars. In all countries, except Norway, children from more affluent families were less likely to engage in AST both ways when compared to their counterparts from less affluent families. This finding is in line with studies from North America and New-Zealand [14,36]. In addition to the possibility of reduced access to a private car in lower-income families, constraints on the time available for single-parents to drive has been identified as a potential reason for family affluence differences [36].

The travel time to school increased between 2006 and 2018 in the Czech Republic and Wales, whereas this remained relatively stable in Norway and Scotland. School distance and time to school have been identified as strong correlates of AST [6,14,29,36,58]. It has been suggested that the development of bigger school units and the closing of neighbourhood schools and school choice policies are contributors to the increased distance and time to school [29,36]. Nevertheless, the literature suggests that many children live within a reasonable walking distance from school [59–61]. In the current study, the prevalence of a travel time less than 15 min was relatively similar across countries. An interesting finding was that the effect of time to school on AST was strongest in the Czech Republic, as Czech children with a travel time of less than 5 min were much more likely to engage in AST, as compared to children in Scotland and Wales. This was also observed for travel times between 5–15 min, but the country differences were less pronounced. Thus, there seems to be potential, especially for countries such as Scotland and Wales, to increase their levels of AST by targeting families who live within a "threshold" distance. This is

reported to be less than 1.5–2.0 km for walking and 3.0 km for biking [62–64]. However, environmental factors (e.g., how many busy roads to cross, presence of staff to help crossing roads) could also be influential in short distances. The findings highlight the need for comprehensive local approaches, which should also include some form of risk assessment and environmental modifications.

The relatively high and stable levels of AST observed in Norway may partly be a result of a substantial proportion of schoolchildren cycling to school. Cycling to school allows students to move faster and they can cover greater distances [65]. Unlike studies from many other European countries, schoolchildren from Scotland, Wales and the Czech Republic did not use cycling as a mode of travel. However, the findings are in line with studies from Ireland [66]. Higher rates of cycling versus walking to school among adolescents have been found in other Nordic countries, such as Denmark and Finland [29]. In contrast to Denmark, known for its longstanding cycling traditions, cycling-friendly infrastructure, and flat landscape, the climate, topography, and physical environmental conditions in Norway are diverse. The relatively high prevalence of cycling could be related to social and cultural norms. Nordic countries have a culture where outdoor activities, in general, play an important role in the "way of life" [67] and Norway has a long tradition for outdoor education in primary schools [68]. This could have a positive impact on attitudes towards independent mobility, which has been associated with cycling to school [69]. Another, but perhaps related issue is expectations around school dress codes. In New Zealand, school uniform requirements have been found to influence adolescents' motivation for cycling to school [70]. This could also be the case in Scotland and Wales, where school uniforms can be quite formal, whereas in Norway students do not use school uniforms. In a study from Scotland, wearing helmets were also a barrier for cycling to school, especially among older children [71]. Other essential factors that could explain why school children do not cycle to school, include lack of cycling infrastructure (e.g., dedicated cycle paths), limited facilities at school and children's lack of competence in terms of cycling [66]. The reality is most likely a combination of all the factors noted.

In the current study, an age effect was observed, with lower odds for AST for the 13 and 15 year olds in Norway and the 13 year olds in Scotland. In Wales 11 year olds are already at secondary school which would explain the lack of age effect. In terms of the secular trends in AST, we did not observe any age interactions. The effect of age was considerable in Norway. This is in line with previous studies that have found substantially higher levels of AST among primary as compared to secondary school children [11,47]. Secondary schools are typically bigger school units with increased travel distances for most students [14], which might explain this finding.

Combining the data from all study waves, it appears that gender was a significant factor influencing AST in Wales and Scotland. Gender differences have been confirmed in several previous studies [14,32,36,39,40]. In terms of secular trends in AST, we did not observe any gender interactions, suggesting that the gender differences in Wales and Scotland remained constant throughout the years. Findings from other trend studies have shown comparable results [25,34,41,72]. In two studies from Spain and Brazil, no gender differences were observed [27,43]. However, over time there was a decline for Spanish girls [43], and a decline for boys and an increase for girls among Brazilian youth [27]. Suggested explanations for girls showing lower AST levels relate mainly to safety concerns (e.g., traffic and crime) and independent mobility, with boys more likely to be allowed to explore their neighbourhood environment to a greater extent without supervision [10]. It has also been suggested that the current generation of parents have more concerns about safety that might be responsible for some of the differences in AST levels [31]. Nevertheless, the differential gender differences observed across countries, suggests cultural variations that future studies should seek to understand more in depth.

#### *Strengths and Limitations*

The key strengths of this study include the large sample size and ability to compare across four European countries, with representative samples of adolescents, using a shared research protocol. Furthermore, the use of standard methodology for data collection and measurement of AST, with consistent wording throughout the waves, ensured internationally comparable data and robust trend analyses. The analyses were conducted with rigour, by adjusting for sampling weight, clustering and stratification in the sampling design, and the alpha level was set to 0.001, given the large sample size and the number of tests. The possibility of differentiating between cycling and walking in this study was also important, as it gave a more detailed picture of students' travel behaviours to school.

However, there are also limitations. The analysis is based on a repeated cross-sectional design conducted every fourth year. Interpretation of trends should be made carefully because there is no information available between the different survey waves. Data collection was conducted by self-report, and therefore, may be susceptible to recall bias. However, it is a challenge to objectively assess AST because of its versatile nature [73]. The AST measure does not detect day-to-day variations in transport behaviour and the students were categorised based on their main method of travel, which could misrepresent the amount of activity. For example, respondents who used passive transport may have accumulated some amount of physical activity for a segment of the journeys, that could add to their daily physical activity uptake. Furthermore, distance to school was approximated by the included time to school variable, which was a suboptimal solution as time to school partly dependents on transportation mode and could also be influenced by time spent stopping somewhere during the journey and other environmental factors. Nevertheless, a strength of HBSC lies in its breadth, which as a consequence, unfortunately means that not all issues can be explored in great depth.

#### **5. Conclusions**

The study found stable patterns of AST in the period from 2006 to 2018, except for a reduction in the Czech Republic from the first to the second wave. These findings could indicate that the previously observed decrease in AST has been flattening off in the countries studied here. Still, the findings suggest that there is a great potential to increase the level of active commuting to school, especially in Scotland and Wales where the levels were low despite government action. This indicates that tackling active travel alone (e.g., with a focus on infrastructure) is not enough and points to the fact that action really does need to be cross-cutting and comprehensive. The variation in the prevalence of AST and the observed associations with gender, age, family affluence and time to school, suggest there are most likely country-specific factors influencing students' choice of travel mode to school.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/1660-460 1/18/4/2118/s1, Table S1: Sample size and descriptive statistics of independent variables.

**Author Contributions:** Conceptualization, E.H., J.I., C.R., J.B.; methodology, E.H., O.R.F.S., D.S.; software, E.H., O.R.F.S., D.S.; validation, E.H., O.R.F.S., D.S.; formal analysis, O.R.F.S.; investigation, E.H., O.R.F.S., J.B., C.B., J.P., Z.H., D.S., J.B., C.B., J.P., Z.H., D.S.; data curation, D.S., O.R.F.S., writing original draft preparation, E.H.; writing—review and editing, J.B., C.B., J.P., Z.H., D.S., F.K.S.M., J.I., C.R.; visualization, O.R.F.S.; supervision, E.H., J.B.; project administration, J.I. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research is based on data from the HBSC study. In Scotland, the HBSC study is funded by NHS Health Scotland. J.I. is supported by the UK Medical Research Council (MC\_UU\_12017/14) and Chief Scientist Office (SPHSU14). For further details on HBSC, see http://www.hbsc.org (accessed on 19 February 2021). The HBSC study in Norwegian partly funded by the Norwegian Directorate of Health. The HBSC study in the Czech Republic is supported by the European Regional Development Fund-Project "Effective Use of Social Research Studies for Practice (No. C.02.1.01/0.0/16\_025/0007294), by the Technology Agency of the Czech Republic (ÉTA TL01000335)

and by the Ministry of Education, Youth and Sports, Inter-Excellence, LTT18020. The HBSC study in Wales is funded by Welsh Government.

**Institutional Review Board Statement:** All participating countries within the HBSC network adhere to ethical guidelines and principles as described in the HBSC study protocols. The HBSC study is conducted according to the guidelines of the Declaration of Helsinki. Ethical approval for the surveys was obtained at the national level, and adherence to protocol requirements was managed by the HBSC International Data Management Centre in Bergen, Norway (see http://www.hbsc.org/methods/ (accessed on 19 February 2021)).

**Informed Consent Statement:** Informed consent (explicit or implicit) was given from school administrators and or parents before participation. Participation was voluntary and the children were informed about confidentiality and anonymity. More details on the HBSC study procedures can be found elsewhere [46].

**Data Availability Statement:** The University of Bergen is the data-bank manager for the HBSC study. Please contact the corresponding author for data requests.

**Acknowledgments:** HBSC is an international study carried out in collaboration with WHO/EURO. The HBSC International Coordinator was Candace Currie (2006, 2010, 2014 surveys) and Jo Inchley (2018 survey). The Data Bank Manager was Oddrun Samdal. We are grateful to the Principal Investigators in the participating countries. We also wish to thank all the participating students, staff and schools that took part in the study.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Article* **Health Impacts of Urban Bicycling in Mexico**

**David Rojas-Rueda**

Department of Environmental and Radiological Health Sciences, Colorado State University, Fort Collins, CO 80523, USA; David.Rojas@colostate.edu; Tel.: +1-(970)-491-7038; Fax: +1-(970)-491-2940

**Abstract: Background:** Bicycling has been associated with health benefits. Local and national authorities have been promoting bicycling as a tool to improve public health and the environment. Mexico is one of the largest Latin American countries, with high levels of sedentarism and non-communicable diseases. No previous studies have estimated the health impacts of Mexico's national bicycling scenarios. **Aim:** Quantify the health impacts of Mexico urban bicycling scenarios. **Methodology:** Quantitative Health Impact Assessment, estimating health risks and benefits of bicycling scenarios in 51,718,756 adult urban inhabitants in Mexico (between 20 and 64 years old). Five bike scenarios were created based on current bike trends in Mexico. The number of premature deaths (increased or reduced) was estimated in relation to physical activity, road traffic fatalities, and air pollution. Input data were collected from national publicly available data sources from transport, environment, health and population reports, and surveys, in addition to scientific literature. **Results:** We estimated that nine premature deaths are prevented each year among urban populations in Mexico on the current car-bike substitution and trip levels (1% of bike trips), with an annual health economic benefit of US \$1,897,920. If Mexico achieves similar trip levels to those reported in The Netherlands (27% of bike trips), 217 premature deaths could be saved annually, with an economic impact of US \$45,760,960. In all bicycling scenarios assessed in Mexico, physical activity's health benefits outweighed the health risks related to traffic fatalities and air pollution exposure. **Conclusion:** The study found that bicycling promotion in Mexico would provide important health benefits. The benefits of physical activity outweigh the risk from traffic fatalities and air pollution exposure in bicyclists. At the national level, Mexico could consider using sustainable transport policies as a tool to promote public health. Specifically, the support of active transportation through bicycling and urban design improvements could encourage physical activity and its health co-benefits.

**Keywords:** bicycling; transport; Mexico; health impact assessment; environmental health

#### **1. Introduction**

The United Nations has reported that more than 50% of the global population lived in urban settings in 2018, and the urbanization trend is expected to increase in the coming years [1]. Urban and transport planning has been suggested as a critical health determinant, impacting physical activity, air and noise quality, traffic safety, blue and green spaces, among others [2,3]. Specifically, bicycling has been suggested as a tool to promote physical activity [4–6].

Sedentarism is one of the leading risk factors for mortality worldwide [7]. The global prevalence of insufficient physical activity in 2016 was 23%, and the Latin American region had the highest prevalence of insufficient physical activity (39%) [8]. Mexico is the second most populated country in the Latin American region, with 127 million inhabitants [9], with more than 80% of its population living in urban areas [1]. Mexico has reported 29% of the population has insufficient physical activity [8].

Active transport policies have been promoted extensively in Latin America, being the open street programs (where main streets in cities are closed for walking and cycling), one of the most known active transport policy originated in Latin America [10–12]. Although bicycling has played an essential role in personal mobility around the world, current trends

**Citation:** Rojas-Rueda, D. Health Impacts of Urban Bicycling in Mexico. *IJERPH* **2021**, *18*, 2300. https://doi.org/10.3390/ ijerph18052300

Academic Editors: Adilson Marques and Paul B. Tchounwou Received: 1 January 2021 Accepted: 23 February 2021 Published: 26 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

show that motorized traffic is gaining more relevance [13]. Compared to other modes of transportation, bicycles offer a convenient and affordable transport option that could capture a higher proportion of urban transport passengers than is currently the case [13].

Previous studies have estimated the health impacts of local bicycling transport scenarios, but most of them have been focused on developed countries [14–16]. To our knowledge, no study has assessed the health impact of bicycling scenarios in Mexico. This study aims to estimate the health impacts, risks, and benefits of Mexico bicycling scenarios at the national level.

#### **2. Methodology**

#### *2.1. Study Design and Data Collection*

This study follows a quantitative health impact assessment (HIA) approach, assessing bicycling scenarios in the urban population in Mexico. Transport data were collected from the "Global High Shift Cycling" study [13]. The "Global High Shift Cycling" study provides bicycling data at a national level, describing transport patterns such as trips per person per day, trip length, kilometers traveled by a person, and mode of transport (Table 1). Methods and descriptions of the "Global High Shift Cycling" study have been reported elsewhere [13]. National population data were obtained from the United Nations population forecast [1]. Mortality rates by age and country were collected from the year 2017, which was reported by the Global Burden of Disease (GBD) project [17]. Air pollution data of particulate matter less than 2.5 micrometers of diameter (PM2.5) annual average national concentration was collected from the World Health Organization (WHO) Global Ambient Air Quality Database [18]. National annual traffic fatalities by mode of transport were collected from the Road Safety Annual Reports [19] and the global observatory data from the World Health Organization from years 2009 to 2018 [20]. National physical activity data in metabolic equivalent of task (MET) were collected from scientific publications [21,22]. Dose–response functions used in this quantitative Health Impact Assessment (HIA) for physical activity and air pollution on all-cause mortality were collected from the published meta-analysis [23,24].



MET: Metabolic Equivalent of Task; PM2.5: Particulate Matter with a dimeter <2.5 μm.

#### *2.2. Scenarios*

Five scenarios were included in this study (Figure 1): (a) current bike levels in Mexico (based on the bike trips reported at the national level for adults in urban population, 1.07%) [13]; (b) double the national bike-share (assuming as transport goal doubling the current levels of bike trips, 2.13%); (c) arriving at bike levels reported in Brazil (Brazil was the Latin America country with the largest bike mode share reported, 3%) [13]; (d) achieving the Danish bike levels (Denmark is reference country for bicycling, 16%) [25]; and (e) achieving the Dutch bike levels (The Netherlands is the country with the largest bike

share in the world, 27%) [26]. All the scenarios assumed an 8% car-bike substitution based on the average reported substitution among 26 cities worldwide [14,27–30]. All scenarios assumed a conservative average bike trip distance in Mexico of 2 km.

**Figure 1.** Conceptual framework of the study.

#### *2.3. Quantitative Model*

A quantitative health impact assessment approach was followed to estimate the number of annual premature deaths related to each scenario and health determinant (Figure 1). All-cause mortality was estimated considering three different health determinants (physical activity, road traffic fatalities, and air pollution (PM2.5)). The "TAPAS (transportation, air pollution, and physical activities) tool" developed and used in previous quantitative HIA was used to estimate the health impacts in this study [6,14]. A detailed description of the TAPAS tool methods has been reported in the supplemental material and elsewhere [6,14,31,32]. TAPAS tool is a quantitative HIA run on Microsoft Excel for Office 365, version 2008 (Microsoft, Redmond, USA, 2020). The dose–response functions used in the TAPAS tool, between physical activity, PM2.5, and all-cause mortality, were selected from meta-analyses of cohort studies from adult populations. The risk estimated from traffic fatalities by kilometer traveled was obtained from national transport and health data. Levels of each determinant were estimated for each country and scenario. An all-cause mortality relative risk (RR) was estimated for each health determinant and scenario and transformed into a population attributable fraction (PAF). Using the Mexico mortality rate for adults (20–64 years old) and the national urban adult population (20–64 years old) in each scenario, the number of expected premature deaths was estimated for each scenario. Finally, the PAF from each scenario was multiplied with the corresponding expected number of premature deaths in the population to obtain the number of attributable premature deaths. For the economic assessment, the value of statistical life was used to estimate the economic impacts of preventing deaths in each scenario, using the value of statistical life reported for Mexico (US \$210,880) [33].

#### 2.3.1. Physical Activity

The physical activity level was estimated based on the trip duration, trip frequency, and physical activity intensity, using the metabolic equivalent of task (MET) (Table 1). The physical activity was defined as 6.8 METs for bikes and 2 METs for car travelers. The relative risk of all-cause mortality was based on the dose–response function (DRF) provided by a meta-analysis of cohort studies (RR = 0.81 (0.76–0.84) for each increment of 8.6 METs, with a power transformation of 0.25)) [24], assuming a non-linear DRF. The physical activity assessment considers the basal levels of physical activity in the Mexican population [21,22] to estimate the relative risk for each scenario before being translated into a population attributable fraction and then to the estimated attributable premature deaths (see Supplemental Material Figures S1 and S2).

#### 2.3.2. Air Pollution

The air pollution assessment focused only on the exposure to particulate matter with a diameter < 2.5 μm (PM2.5), which has shown a strong association with all-cause mortality [34–36]. We obtained the annual average PM2.5 concentrations in Mexico, using the World Health Organization database of air quality [18] (Table 1). We estimated the concentration of PM2.5 in each microenvironment (bike and car), using background/car or bike ratios provided by a previous meta-analysis [37], following a similar approach as reported in previous studies [14,16,31] (see Supplemental Material Figure S3 and Tables S1–S3). The inhaled dose was estimated using the minute ventilation according to the intensity of physical activity (in METs) in each mode of transport (bike and car), PM2.5 concentration in the mode of transport, and trip duration [14,16,31] (see Supplemental Material Tables S2 and S3). The DRF for PM2.5 and all-cause mortality from a meta-analysis were used (RR = 1.06 (1.04, 1.08)) for each increment of 10 μg/m3 of PM2.5) [23]. Finally, using the comparative risk assessment approach, we estimated the relative risk, population attributable fraction, and the expected number of premature deaths for each scenario, as reported before (see Supplemental Material Figure S3).

#### 2.3.3. Road Traffic Fatalities

The road traffic fatalities in Mexico were obtained from the annual traffic fatalities reported at the national level through transport mode from years 2009 to 2018 (Table 1). For each scenario, we estimated the number of kilometers traveled by car and bike. The expected traffic fatalities by mode of transport were estimated using the traffic fatalities per billion kilometers traveled and the distance traveled in each mode of transport [31,34]. Then a relative risk of traffic fatalities for cyclists compared with car drivers was estimated. The relative risk was translated to an attributable fraction and a final number of prevented premature deaths in each scenario (see Supplemental Material Table S1 and Figure S4).

#### **3. Results**

The national bike share in Mexico was 1.07% of all trips. We estimated an average of 2,068,750 daily bike trips among adults in urban settings in Mexico (Table 1). The number of bike trips per day (<2 km) was estimated to substitute car trips in Mexico where 165,500. In all the scenarios, the health benefits (in preventable deaths) of physical activity related to bicycling outweighed the health risks associated with traffic fatalities and air pollution inhalation (Table 2).


**Table 2.** Results of current and hypothetical bicycling scenarios in Mexico.

#### *3.1. Impacts of Current Bicycling Levels in Mexico*

It was estimated that the current levels of bike trips in Mexico (that are expected to substitute car trips, 165,500 trips per day) resulted in 9 (95% UI: 6–25) premature deaths avoided each year among the urban adult population. In terms of economic values, it was estimated that the current bike trips could result in US \$1,897,920 annual health economic benefits related to mortality (Table 2). In terms of risks and benefits, traffic fatalities were estimated to increase 2 annual deaths and air pollution exposure 1 annual death. Physical activity resulted in the prevention of 12 annual deaths (Figure 2 and Supplemental Material Table S4)

**Figure 2.** Risks and benefits of bicycling scenarios in Mexico, in annual premature deaths by scenario and risk factor.

#### *3.2. Impacts of Future Bicycling Scenarios in Mexico*

If Mexico doubles the current levels of bike trips to 2.13% (assuming similar trips substitution from cart to bike, that current levels), the annual premature deaths prevented could arrive at 17 (95% UI: 11–49), with an economical translation of US \$3,584,960. If Mexico achieves the bike trip levels reported in Brazil (3%), the annual benefits could arrive at 24 (95% UI: 16–69), with an economic impact of US \$5,061,120. If Mexico arrives at bike trip levels reported in Denmark (16% of bike trips), the health impacts will be translated into 129 (95% UI: 84–370) annual prevented deaths and US \$27,203,520. Finally, suppose Mexico achieves bike trip levels similar to those reported by The Netherlands. In that case, the health impacts will be an annual reduction of 217 (95% UI: 142–625) prevented deaths, with an annual health economic benefit of US \$45,760,960.

#### **4. Discussion**

This study found that bikes in Mexico have the potential to prevent up to 217 annual premature deaths if bike trip levels, similar to those reported in the Netherlands, are achieved with an annual economic benefit of more than 45 million US dollars (Table 1). In the current situation, bike trip levels in Mexico are expected to prevent 9 premature deaths each year (Table 1). In the five scenarios assessed, the health benefits (due to physical activity) outweighed the health risks (air pollution inhalation and traffic incidents) (Figure 2).

This is the first study assessing the health impacts of national bicycling scenarios in Mexico. This study included the 51,718,756 adult urban inhabitants in Mexico. This study includes five different bicycling scenarios comparing the current bike levels in Mexico with reference counties in Latin America (Brazil) and worldwide (Denmark and the Netherlands, the global reference countries for bicycling trends). This study provides a conservative estimation of the bicycle health impacts in Mexico. The analysis only includes a small portion of bike trips (those assumed to came from cars (8% of all bike trips)), assuming short trip distances 2 km, including only adult population (20–64 years old) and urban settings. The overall health impacts of bicycling in Mexico are expected to be larger if all bicycle trips and populations are counted.

These results are in accordance with previous quantitative health impact assessment studies on bicycling scenarios using similar exposures (physical activity, air pollution, and traffic fatalities) [14,31,38–40]. A previous study in seven European cities found that achieving 10% of bike trips will prevent between 0 to 31 deaths in Antwerp and Vienna [38]. In this study was assumed that 28% of bicycling increments came from cars [38]. Another study on the health impacts of bike-sharing systems in Barcelona, Spain, was found that if 90% of the bike-sharing trips came from cars (around 38,000 trips per day), 12 deaths could be avoided each year [31]. Another study with more ambitious scenarios from six European cities estimated the health impacts of bicycling scenarios in Paris, Prague, Warsaw, Basel, and Barcelona [2]. In this study, the aim was to assess "what if" those cities achieve the bike share from Copenhagen (35%) [2]. In this case, the study estimated among 5 to 113 premature deaths prevented each year between the six European cities [2]. Unlike previous studies that have been focused on single cities [2,6,14,31], our study focused on the national urban populations, providing a broader perspective of policy scenarios in Lat America. Like previous studies, this analysis focused on car trip substitution, considering that sifting car trips to active transportation will have larger health benefits and important climate co-benefits [2].

This study found that the current bicycling levels in Mexico will benefit public health at a national scale, preventing nine premature deaths annually among adult urban populations (that are expected to shift from car to bike). Those results were also translated into economic impacts related to mortality, using the value of statistical life, a standard metric used by transport planners and engineers to measure traffic safety impacts. We estimated that Mexico's current car-bike substitution levels have an economic benefit of up to 1.8 million US dollars annually. This study also included different hypothetical policy

scenarios related to bike share. We selected four extra bike scenarios to compare "what if" Mexico increases their bike levels to double the current bike share (2.13%); or to those reported by Brazil (3%) that was the Latin American country with the highest bicycling levels; or to those reported in Denmark (16%) or the Netherlands (27%), the reference countries for bicycling around the globe. In those scenarios, the health benefits ranged between 17 to 217 annual premature deaths that could be prevented each year, with an economic benefit between 3.5 to 45 million US dollars annually (Table 1). These results highlight the importance of active transportation policies in Mexico and the potential of transport policies to support public health.

Among the exposures included in this quantitative health impact assessment, physical activity produced the most considerable health impacts (Figure 2). Physical activity is well known as a health-protective factor for multiple diseases and causes of death, such as cardiovascular, metabolic, and mental diseases, among others [16]. Our analyses focused on all-cause mortality as a health outcome because it has been proposed as the best indicator of health impacts on active transport assessments compared to morbidity [6,16]. This analysis utilized the "TAPAS tool," a quantitative health impact assessment tool for bicycling, walking, and public transport, reported in previous transport health impact assessments [2,6,14,31]. The "TAPAS tool" for bicycling estimated the health impacts of physical activity using a non-linear dose–response function (DRF) from a meta-analysis of cohort studies [24], and it was calibrated with the corresponded physical activity levels reported by the adult population in Mexico and applied to the exposure levels by each scenario. The non-linear function considers that those who already were physically active would gain fewer health benefits than those who are more sedentary. This non-linear approach results in a conservative result estimating fewer health benefits than using a linear DRF [2].

In this study, air pollution analysis only considers the exposure to PM2.5 inhalation during the trip. Although air quality improvements can be expected from changes in modal share, these health-related impacts were not in the scope of this study, and the study only focused on the PM2.5 exposure of bicyclists during the trip. PM2.5 was selected because it was expected to produce the largest health burden compared to other air pollutants such as NO2 or black carbon [6].

Traffic safety analysis was based on traffic fatalities. This study quantified fatal traffic incidents per billion kilometers traveled, using the reported national road safety estimates provided by the World Health Organization (WHO) [20]. This study only considered the traffic fatality risk by mode of transport (bike vs. car). It did not assess the impacts of other traffic risk factors such as the type of route or traveler demographics due to the lack of data available in these aspects.

Our study was limited by data availability and the necessity to make assumptions to model likely scenarios. In terms of the scenarios modeled, we select national bicycling goals similar to those that already exist in other nations in Latin America and globally. However, one limitation is the transferability of the policy scenarios to the Mexican context. In Denmark and the Netherlands, geographical and social characteristics differ from the Mexican context (i.e., population density, transport infrastructure, or land use and cartography). Another limitation was the lack of specific modal shift (car to bike) data from Mexico. Thus, the data available from 26 cities from China, Europe, and the US was summarized to estimate the average percentage of bike trips that can shift from car trips [14,27,29,30]. Our estimates' uncertainty was also assessed, providing uncertainty intervals composed of the input data's variability (maximum and minimum) and the confidence intervals from the DRF from air pollution and physical activity. Another limitation in this study was the need to assume an average trip distance in Mexico. In this study, we selected 2 km as a conservative scenario. But a sensitivity analysis was conducted to estimate the health impacts in the five scenarios if a similar bike trip length (5 km) was used as reported in a previous study in Europe [38]. In this sensitivity analysis, we found that the health benefits of urban bike trips in Mexico could be estimated between

15 to 384 preventable annual deaths among the five scenarios (see Supplemental Material Tables S5 and S6).

Furthermore, if national and local authorities improve traffic safety and air quality, in addition, to increase bike levels, more health benefits could be expected in Mexico. In all our scenarios, we assume only an 8% of car–bike trip substitution. This is of particular relevance because if authorities achieve attracting more car drivers and passengers to bicycles, the health benefits could increase largely in addition to the overall levels of bike trips. This study only considers a population between 20 and 64 years old. If policymakers and transport planners achieve the goal of attracting younger and older age groups to bicycles, the health benefits from bicycling in Mexico could be more extensive. As in many other countries, the aging process is also affecting the Mexican population [1]. Healthy aging starts with integrating a healthy lifestyle since the early stages of life, and bicycling could be used as a tool to promote healthy aging. Some general recommendations for policymakers and stakeholders to promote bicycling in Mexico are (a) the support of active transport policies, specifically on interventions to promote bicycling and reduce car driving; (b) support traffic safety and air quality improvements in urban settings in Mexico; and (c) improve data collection and quality improvement in terms of physical activity, traffic safety, air quality, and transport characteristics. For health practitioners, this study can help to dimension the relevance of transport policies to improve public health. Researchers should support local and national data collection on transport and health with a vision of harmonization and comparability. A summary of the policies needed to increase bicycling in Mexico is listed in Table 3.

**Table 3.** Policy recommendations to support bicycling in Mexico.

#### **Bicycling**


#### **Motorize transport**


#### **Urban planning**


#### **Environment**


#### **Public health**


#### **5. Conclusions**

The study found that bicycling promotion in Mexico would provide important health benefits. At the national level, Mexico could consider using sustainable transport policies as a tool to promote public health. Specifically, the support of active transportation through bicycling interventions could promote physical activity, reduce mortality and increase health economic benefits. The attraction of bike users could be supported by bike investments and interventions (e.g., bike lanes, bike parking, and bike-sharing systems), combined with interventions to reduce car use (e.g., parking pricing and reduction, and congestion pricing). To meet ambitious bicycling scenarios in Mexico, strong transport, urban planning, energy, environmental, and health policies should be adopted at national and local levels.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/1660-460 1/18/5/2300/s1, Figure S1. Physical activity model; Figure S2. Dose response functions (DRF) for physical activity and all cause mortality; Figure S3 Air pollution model; Figure S4. Traffic fatality model; Table S1. Relative risk formulas for each model; Table S2. General formulas; Table S3. Air pollution variables; Table S4. Results in annual premature deaths in each scenario by risk factor; Table S5. Sensitivity results in premature deaths prevented each year in each scenario, assuming a 5km bike trip length; Table S6. Sensitivity results in premature deaths prevented each year in each scenario, using the HEAT for walking and cycling V.3\* (5 km trip distance).

**Funding:** None.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Publicly available datasets were analyzed in this study. This data can be found in the references, supplemental material and here: [https://itdpdotorg.wpengine.com/ wp-content/uploads/2015/11/A-Global-High-Shift-Cycling-Scenario\_Nov-2015.pdf accessed on 1 March 2021] [https://www.who.int/airpollution/data/cities/en/ accessed on 1 March 2021] [https: //population.un.org/wup/Publications/Files/WUP2018-Report.pdf accessed on 1 March 2021] [https://www.itf-oecd.org/sites/default/files/docs/irtad-road-safety-annual-report-2019.pdf accessed on 1 March 2021]

**Conflicts of Interest:** The author declares they have no actual or potential competing financial interests.

#### **References**

