towards understanding the relation between transmilenio
TRANSCRIPT
TOWARDS UNDERSTANDING THE
RELATION BETWEEN TRANSMILENIO
AND WALKING FOR TRANSPORTATION
Pablo David Lemoine Arboleda
[Dirección de correo electrónico]
1
Agradecimientos
En esta etapa que termina quiero agradecer a todos los que hicieron el camino posible.
A Roberto, quien hace años creyó y siguió creyendo que un estudiante bueno para el rugby podría ser doctor.
A Olga Lucia, por siempre saber el momento apropiado para el rigor y la alegría
A Meisel, Felipe y Pablin, que me hicieron reír cuando todo parecía mal, y me hicieron reír cuando todo parecía bien.
A Juan Cordovez, por recibirme, asesorarme y confiar sin preguntar.
A Juan Alejandro, por hacer del problema más complejo una cosa menor.
A Camilo Olaya, por la ingeniería y el diseño.
A Nelson, Juan Camilo, Dumpa, Victor, Jorge, hermanos mayores que generaron admiración.
A Andrea, Juan Zambrano y Juan David Pinzón por ver en la relación entre TransMilenio y caminar un problema interesante
A papá por el consejo
A mamá por la alegría
A Julio por la amistad A Diana por el amor
A mis hermanos mayores Carlos Esteban, René y Mónica y sus familias porque somos familia
A Rafa, Max, Rodrigo, Felipe y los profesores Miguel Kiwi y José Rogan por hacer de Chile un país al que anhelo volver. Al Holding por la ilusión.
A todos los que en esta lista olvido, gracias.
2
CONTENTS
List of Figures ....................................................................................................................... 4
List of Tables ........................................................................................................................ 5
List of papers ........................................................................................................................ 6
Declaration of original research .................................................................................... 7
Abstract| ............................................................................................................................... 8
Introduction ........................................................................................................................... 9
1. Overweight, Obesity and Physical Inactivity .................................................... 14
2. Relationships between active school transport and body composition 1indicators in school age children from low-, middle- and high-income countries ............................................................................................................................ 16
2.1. Materials and Methods ........................................................................................ 18
2.1.1. Participants ......................................................................................................... 18
2.1.2. Measurements ................................................................................................... 19
2.1.2.1. Anthropometry ............................................................................................... 19
2.1.2.2. Active transport ............................................................................................. 20
2.1.3. Covariates ........................................................................................................... 20
2.1.4. Statistical Analysis ........................................................................................... 21
2.2. Results ..................................................................................................................... 22
2.2.1. Socio-demographic characteristics ............................................................. 22
2.2.2. Body composition ............................................................................................. 23
2.2.3. School transport ............................................................................................... 23
2.2.4. Associations between AST and body composition indicators ............ 25
2.3. Discussion .............................................................................................................. 28
3. Bus Rapid Transit Systems to promote Physicial activity ............................ 33
3.1. BRT a global trend ................................................................................................ 34
3.2. Setting ...................................................................................................................... 37
3.3. Study design .......................................................................................................... 37
3.4. Data collection ....................................................................................................... 38
3.5. Physical activity measurements ....................................................................... 38
3.5.1. Subjective physical activity measurements .............................................. 38
3.5.2. Objective physical activity measurements ................................................ 39
3.6. Independent variables ......................................................................................... 40
3
3.6.1. Individual variables .......................................................................................... 40
3.6.2. Geographic units and built-environment characteristics ...................... 40
3.7. Statistical analyses .............................................................................................. 42
3.7.1. Cross-sectional analysis using 2010-11 data ............................................ 42
3.7.2. Subsample of adults with objective physical activity measures ......... 43
3.7.3. Cohort analysis using both the 2005 and 2010-11 data ......................... 43
3.8. Results ..................................................................................................................... 44
3.8.1. Cross-sectional sample based on 2010-11 data ....................................... 44
3.8.2. Subsample of adults with objective physical activity measures ......... 45
3.8.3. Cohort analyses using 2005 and 2010-11 data ......................................... 46
3.9. Discussion .............................................................................................................. 50
4. The effect of increased public transport infrastructure on walking for transportation ................................................................................................................... 54
4.1. Agent Based Model .............................................................................................. 55
4.2. Model Development ............................................................................................. 56
4.2.1. Setting .................................................................................................................. 56
4.2.2. Agents .................................................................................................................. 57
4.2.3. Modes of transportation .................................................................................. 58
4.2.4. Mode decision .................................................................................................... 59
4.2.5. Calibration ........................................................................................................... 61
4.2.6. Model assessment results for the calibration scenario ......................... 62
4.2.7. Scenarios ............................................................................................................ 64
4.3. Results from the different scenarios ............................................................... 66
4.4. Discussion .............................................................................................................. 68
5. Final remarks and future work .............................................................................. 72
4
List of Figures
Figure 1. Bus Rapid Transit system growth throughout the world ...................................... 11 Figure 2.Estimated fractions of BMI categories by age group and SES. ............................. 24 Figure 3. : Typology distribution according to active school transport and income level. .. 25 Figure 4. Percentage of adults that meet physical activity recommendations in Bogotá and
Colombia. ............................................................................................................................. 34 Figure 5. Constructed and projected TM in Bogotá ............................................................. 36 Figure 6. Sampling design for cross-sectional and cohort studies ....................................... 37
Figure 7. Sampling design for cross-sectional and cohort studies ....................................... 41 Figure 8. Relationship between moderate and vigorous physical activity ........................... 46 Figure 9.Median of minutes walking for transportation ....................................................... 48
Figure 10. Bogota's inspired city of 100〖km〗^2 .............................................................. 57
Figure 11. Distance decay function for walking................................................................... 62
Figure 12 Prevalence of walking per travel time per SES in the calibration test case ......... 63 Figure 13. Comparison mode distribution per SES for Bogotá and the ABM model .......... 64
Figure 14. Walking minutes TM users vs non users. ........................................................... 66 Figure 15. Average minutes of walking for transport according to TM and bus densities .. 68
5
List of Tables
Table 1. Associations of body composition variables with active school transport in 6,797
9-11 year old children in the International Study of Childhood Obesity, Lifestyle and the
Environment (ISCOLE). ....................................................................................................... 27 Table 2. Percentage of agents, average daily expenditure on transport, percentage of daily
budget spent on transport, probability of owning a car, and α for each SES ....................... 58
Table 3. Costs, access time and speed of the different modes of transportation. ................. 59 Table 4. Average distance from home to the closest bus and BRT station plus average
distance from work to the closest BRT and bus station according to number of lanes ........ 65 Table 5. Average distance from home to the closest bus or BRT station plus average
distance from work to the closest BRT and bus station ....................................................... 65
6
List of papers
This thesis is based on the following papers:
1. Relationships between active school transport and adiposity indicators in school age
children from low-, middle- and high-income countries. Olga L. Sarmiento, Pablo D.
Lemoine, Silvia A. Gonzalez, Stephanie T. Broyles, Kara, D. Denstel, Richard
Larouche, Tiago V. Barreira, Jean-Philippe Chaput, Mikael Fogelholm, Gang Hu,
Rebecca Kuriyan, Anura Kurpad, Estelle V. Lambert, Carol Maher, Jose Maia, Victor
Matsudo, Timothy Olds, Martyn Standage, Mark S. Tremblay, Catrine Tudor-Locke,
Pei Zhao, Timothy S. Church, and Peter T. Katzmarzyk (Submitted)
2. Walking and Bus Rapid Transit in Bogotá. Pablo D. Lemoine, Olga L.
Sarmiento, Jose David Pinzón3, Jose David Meisel, Felipe Montes, Dario
Hidalgo, Michael Pratt, Juan Manuel Zambrano, Juan Manuel Cordovez,
Roberto Zarama. (Submitted)
3. The effect of increased public transport infrastructure on walking for
transportation. Pablo D. Lemoine, Juan Manuel Cordovez, Juan Manuel
Zambrano, Olga Lucia Sarmiento, José David Meisel, Juan Alejandro
Valdivia1, 2, Roberto Zarama. (Submitted)
7
Declaration of original research
In this work we produced the following original contributions to the Public Health
field and to Engineering.
By showing a negative association between active transport and body
composition indicators. This association does not vary between the
assessed countries or sex (Sarmiento et al. 2015).
By showing a great variance in terms of mode use by country and showing
the possibilities in each country for increasing active travel to school through
trips that are less than five minutes long (Sarmiento et al. 2015).
By showing there must be different approaches to increasing active travel
according to a choice based framework and a need based framework
(Sarmiento et al. 2015).
By showing a positive association between using TransMilenio and walking
for transport more than 150 minutes per week. We also showed a positive
association between averaging more than 22 minutes per day of moderate
to vigorous physical activity (MVPA) and using TransMilenio (Lemoine et al.
2015)
By showing that the moments of the day in which the difference in the
number of minutes of MVPA from users and non-users is largest during
peak hours of transportation. (Lemoine et al. 2015)
8
By showing that people in Bogotá have increased the number of minutes of
walking for transportation from 2005 to 2010-11 and showing that the
neighborhoods that most increased the number of minutes walking for
transport were those neighborhoods which acquired access to TM (Lemoine
et al. 2015)
By calculating approximate health savings resulting from the association
between using TransMilenio and walking for transport (Lemoine et al. 2015).
By creating a model that explains the difference in minutes walking for
transport from users and not users (Lemoine et al. 2015).
By simulating the trend walking for transport is likely to follow as
TransMilenio increases in Bogotá (Lemoine et al. 2015)
By establishing a possible casual link between the use of TransMilenio and
the increased number of minutes walking for transport (Lemoine et al.
2015).
I declare that this thesis is my own work as well as the graphics and tables
presented here, except where I explicitly state the contrary.
Abstract|
The leading causes of death and disability have changed from communicable
diseases in children to non-communicable diseases (NCDs) in adults. Today
9
Physical inactivity (PI) is associated with more than 5.3 million deaths per year. One
potential way to diminish this trend is through active transport. We show that children
that walk or travel by bicycle to school are less likely to be obese, have lower waist
circumference and lower percentage body fat. This relation does not differ by country
or sex. Bus Rapid Transit (BRT) systems are a growing trend around the globe.
There is a positive association between using TM and walking more than 150
minutes per week, there is also a positive association between access to TM and
walking more than 150 minutes per week, also subjects living in neighborhoods in
which TM was built increased more the number of minutes walking for transport and
the likelihood of walking for transport more than 150 minutes. Using an agent based
model (ABM) I show that the relation between use and access of TM and walking for
transport could be explained through agents that decide the mode of transportation
based on resources and time. According to the model TM increases the number of
minutes walking for transport only in environments with a high density of bus
stations. TM has a positive relation with walking for transport in Bogotá due to
Bogotá´s high bus density, and TM´s price and speed. Understanding the system in
terms of actual decision making processes driven by actors and their environment
may prove helpful in designing policies to increase walking and PA without harming
people´s quality of life.
Introduction
This thesis is the culmination of five years of work on different scientific articles
related to non-communicable chronic diseases and transport. The objective is to
10
understand the relation between Bogota's BRT called TransMilenio and walking for
transport in Bogota. As I state this objective two questions arise. The questions are
of course, why is this important? And what does it mean to understand this relation?
These questions will be briefly answered in this introduction and in more depth
throughout the thesis.
Global health has experienced a rapid transformation. People are living longer than
ever before. The leading causes of death and disability have changed from
communicable diseases in children to non-communicable diseases (NCDs) in
adults1. Today Physical inactivity (PI) is associated with more than 5.3 million deaths
per year and it is estimated that if PI was decreased by 25%, over 1.3 million deaths
could be prevented every year 2, mostly in middle and low income countries 3.
This trend is also happening in Colombia4.
Overweight and obesity are product of an unbalanced consumption and expenditure
of energy. One possible way to shift the current trends is increasing energy
expenditure through physical activity (PA). Today the world is not only concerned
with obesity levels, it is also concerned with Physical inactivity (PI), climate change,
increased traffic, and declining oil supplies. Active transport is increasingly being
recognized as a possibility to mitigate these problems 5. Active transport refers to all
travel by non-motorized means including part of a trip chain for public transport
users. Active travel modalities such as walking are associated with reductions in
cardiovascular risk 6; a reduction in type-2 diabetes, cancer; improvement in overall
fitness, and obesity 7. Accumulating short periods of PA by walking for transportation
contributes to meeting PA recommendations 6 8 9. Therefore transportation
infrastructure and active transport could play an important role in decreasing PI and
11
obesity 10 11. Chapter three further establishes this relation in a study with children.
This Chapter is based on the paper “Relationships between active school
transport and adiposity indicators in school age children from low-, middle-
and high-income countries”. In this article we present a twelve-country study with
10 year-old children. Results show that there is an inverse association between
active transport and adiposity indicators. The article also presents the potential trips
per country that may occur through active transport.
The World health Organization (WHO) has recommended interventions for
increasing PA. Some of the strategies are designed to motivate and educate
individuals. However, in contrast to interventions targeting individuals, which tend to
have weak and short-term effects, change in environments and policies can affect
entire populations on a relatively permanent basis12.
A world trend related to the built environment, which can also be observed in
Colombia is the growth of the Bus Rapid Transit (BRT) system which in Bogota is
called TransMilenio (TM) (Figure 1).
Figure 1. Bus Rapid Transit system growth throughout the world
12
Source: Hidalgo, D., & Gutiérrez, L., BRT and BHLS around the world: Explosive
growth, large positive impacts and many issues outstanding, Research in
Transportation Economics (2012)
There is evidence that access to TM is associated with walking more than 150
minutes per week13. However the research used a cross sectional design and only
assessed access. Chapter four is based on the article “Walking and Bus Rapid
Transit in Bogotá”. In this article we show that there is a positive association
between meeting PA guidelines and using TM. This chapter also includes a cohort
study from 360 individuals evaluated in 2005 and 2010. This analysis shows that
people from our sample are walking more in 2010 than in 2005 and also that people
who live within a 1000 meter network to TM are more likely to meet PA guidelines.
So far I have established the importance of the objective of the thesis. Obesity´s
growth is a mayor health concern; Active transport has the potential to decrease the
current obesity trend; The number of BRT systems in the world is growing specifically
in Bogotá we have three phases and there is a plan to build up to eight phases and
people who use or have access to TM are more likely to meet PA guidelines.
The study of Human decisions cannot be separated from their environment. Since
our environment is changing with the growth of TM and there is a relation between
the BRT system in Bogota and the city´s levels of PA it is important to understand
this relation. However, even though we have established a positive association
between walking and use and access to TM we still do not understand why it occurs
nor do we have any solid ideas as to what will happen to Bogotá in terms of walking
for transport if new TM lanes are built.
13
Environmental effects on walking have received much attention as a strategy to
increase PA. Meta-Analysis of ecological models have found consistent associations
between environmental variables such as a high residential density, street network
design and land-use mix, variables which are associated with proximity to
destinations12,14. In terms of individual variables walking for transport has been
associated with not having a car and being from a less affluent social class15. These
results have been very useful, however this data-based approach excludes the
notion of decision-making actors16 which is essential in understanding the decision
of walking for transport.
Knowledge according to Valente17, is the capacity to explain, that is, the capacity to
explain how we can logically expect a state of the world to lead to a certain
consequence. In this sense the associations we have established so far are mere
descriptions. Since they cannot explain how TM, street network design, residential
density, land use mix, or SES affect the decision of walking for transport.
In this sense to understand something is to be able to establish a simple set of rules
that explains it18, that causes it. With this definition in mind chapter four based on
the article “The effect of increased public transport infrastructure on walking
for transportation” presents simple rules to explain the relation of walking for
transport and TM. It should be stated that it is not the purpose of the chapter to
explain every detail of the decision making process of transport nor the results of
such decisions. The object is to identify the basic mechanisms of the decision
process and to show the results in walking patterns behavior from this mechanism
in a hypothetical city. The Agent Based Model (ABM) based on resources and time
is able reproduce some results that have been observed in walking for transport as
14
TM lanes increase. We show that as TM lanes increase, walking for transport also
increases as established in empirical studies. However the increment of TM lanes
has a decreasing effect on walking for transport and for high densities of TM stations
it may even reduce the number of minutes walking for transport. Finally chapter five
presents a brief summary of the contributions and future research.
1. Growing public concern with Overweight, Obesity and Physical Inactivity
The object of this chapter is to state that overweight, obesity and PI are a growing
problem in the world.
Overweight and obesity are a major growing public health problem worldwide.19,20
The rising prevalence of overweight and obesity in developed and developing
countries has been called a global pandemic21. Increases in the prevalence of
overweight and obesity have been substantial, widespread in developed and
developing countries, and have arisen over a short time. Overweight and obesity
have been associated with noncommunicable diseases (NCDs) like diabetes,
cancer, and cardiovascular diseases.22–25. High body mass index (BMI), was the
sixth leading risk factor for premature death and disability in 201026. According to the
World Health Organization (WHO) it is projected that by 2030 there will be an
increase in total deaths due to NCDs especially cardiovascular disease27.
Latin American countries are experiencing a nutritional transition too. In the case of
Colombia, according to The Colombian Nutritional Survey in 20104 (ENSIN) 34.6%
of Colombians are overweight and 16.5% are obese for a total of 51,2% of the
population with high BMI index (BMI > 25.0). In 2005 the proportion of adults with a
15
high BMI was 45.9%. These represent a 5.3% increment in the proportion of adults
with a high BMI. The prevalence of overweight and obesity is higher in urban settings
compared to rural settings. These results from the Nutritional Survey give support to
the decision taken by the government to declare through the law 1355 from 2009
obesity and associated NCDs a national public health priority to adopt measures
such as guaranteeing the availability of fruits and vegetables in all cafeterias that
belong to an educational institution and the promotion of active transport in order to
control and prevent the obesity pandemic.
Attempts to explain the large increases in obesity have focused on several potential
contributors including increases in calorie intake, changes in the composition of diet,
decreasing levels of physical activity, and changes in the gut microbiome18. The
relative contribution of changes in energy intake versus energy expenditure has
been vigorously debated27,28. However in the ENSIN adults with overweight and
obesity were less likely to report meeting PA recommendations through spare time
PA.
PA itself is considered one of the main pillars of NCDs prevention. PI is the fourth
mortality risk factor according to the WHO. Strong evidence shows that meeting PA
recommendations is associated with diminishing the risk of premature death,
cardiovascular disease, and some types of cancer. Therefore the study of any type
of intervention that may help shift the current trends is of great value. The following
chapters of this thesis will take a closer look at the relation between transport, body
composition and PA.
16
2. Relationships between active school transport and body composition
1indicators in school age children from low-, middle- and high-income
countries
As mentioned before obesity is a major problem not only in adults but in all ages. In
less than one generation, childhood and adolescent obesity has increased
worldwide29. Many low and middle-income countries (LMIC) have shown a similar or
even more rapid increments of childhood obesity compared to high-income countries
(HIC)30,31. Although the increment of obesity in some HIC seems to be leveling off,
their prevalence remains very high29. Unfortunately, the data for time trends in
physical activity (PA) and sedentary behaviors among children and adolescents from
LMIC are extremely sparse32. Nonetheless, in some HIC, PA patterns among school-
age children are decreasing while sedentary behaviors are increasing32.
Within this context of nutrition33 and PA transition34, in which PA patterns are often
the result of environmental and societal changes, it is important to understand the
role of activities that can be incorporated into everyday life, including active school
transport (AST). The prevalence of AST, unfortunately, has declined in several HIC
including Canada35, USA36, Australia37, and Switzerland38. In LMIC the data is
limited, but a study in Brazil showed that the AST trend in this country mirrored HIC
trends39.
Active travel to school is one way in which children can increase their levels of PA
and prevent obesity40. Recent systematic reviews showed that there is conflicting,
17
and very low-quality evidence, regarding the association between body composition
indicators and AST41–43. Specifically, Larouche et al. found that in only 36% of the
studies, AST was associated with more favorable body composition41. Furthermore,
most of these studies assessed only body mass index (BMI), 27% measured body
fat and 12% measured waist circumference41.
In addition, 82% of the studies assessing the association between body composition
indicators and AST have been conducted in HIC in North America40,44–52,
Australia53,54, and Europe55–75 with few studies extending findings to LMIC such as
the Philippines76,77, Indonesia78, China 79,80 Brazil52,81,82, Colombia83 and Kenya84 .
The interpretation of different patterns of body composition indicators and AST
associations across different world regions requires common standardized methods
that have not been employed. Further, the limited variability in obesity, AST patterns
and nutrition and PA transition within each country may have underestimated the
strength of the associations. Thus, only international studies using comparable
methods can help to elucidate the extent to which associations between obesity and
AST are generalizable across countries or are country specific. Such findings could,
in turn, support international and country-specific interventions to prevent obesity
and could inform global efforts, such as the global strategy on diet, physical activity
and health of the World Health Organization85, the United Nations political
declaration on non-communicable diseases86 and the World Bank commitment to
sustainable transport87.
18
In this context, the International Study of Childhood Obesity, Lifestyle and the
Environment (ISCOLE) provides a unique opportunity to assess whether the
relationship between obesity and AST differs across different environmental and
socio-cultural settings. The objective of this study was to assess the associations
between body composition indicators (i.e., obesity, percentage body fat and waist
circumference) and AST in children from 12 different countries.-sites.
2.1. Materials and Methods
The International Study of Childhood Obesity, Lifestyle and the Environment
(ISCOLE) is a multi-country, cross-sectional study conducted in 9-11 year-old
children from 12 countries in 5 continents (Australia [Adelaide], Brazil [São Paulo],
Canada [Ottawa], China [Tianjin], Colombia [Bogota], Finland [Helsinki, Espoo and
Vantaa], India [Bangalore], Kenya [Nairobi], Portugal [Porto], South Africa [Cape
Town], the United Kingdom [Bath and North East Somerset], and the United States
[Baton Rouge]). Additional details on study design, participating countries, and
methodology have been published elsewhere88. The Institutional Review Board at
the Pennington Biomedical Research Center (coordinating center) approved the
overarching protocol, and the Institutional/Ethical Review Boards at each
participating institution also approved local protocols. Written informed consent was
obtained from parents or legal guardians, and child assent was also obtained as
required by local Institutional/Ethical Review Boards. The data were collected from
September 2011 through December 2013.
2.1.1. Participants
19
Of the 7,372 children enrolled in the ISCOLE study, 6,797 remained in the analytic
dataset after excluding participants who did not have valid/information data on BMI
(N=31), PBF (N=68), waist circumference (N=6), main mode of transportation to
school (N=61), travel time to school (N=2), parental education (N=368), and motor
vehicle ownership (N=39). The participants who were excluded in the present
analysis were more likely to have parents who achieved less than high school
(P=0.002), report walking to school (P<0.001) and report trips to school of less than
5 minutes (P=0.004).
2.1.2. Measurements
2.1.2.1. Anthropometry
Anthropometric data (i.e., height, weight, PBF, waist circumference) were directly
measured by trained ISCOLE researchers during an in-school visit according to
standardized procedures88. Weight (to the nearest 0.1 kg) and PBF (to the nearest
0.1%) were measured using a portable Tanita SC-240 Body Composition Analyzer
(Arlington Heights, IL, USA), after outer clothing and shoes were removed. The
Tanita SC-240 has shown acceptable accuracy for estimating PBF when compared
with dual-energy X-ray absorptiometry, supporting its use in field studies89. Height
was measured with a portable Seca 213 stadiometer (Hamburg, Germany) at the
end of a deep inhalation with participant’s head in the Frankfort Plane. Waist
circumference was measured with a non-elastic tape at the end of gentle expiration
held midway between the lower rib margin and the iliac crest90. Each measurement
was repeated and the average was used for the analysis. The BMI was calculated
and then categorized using the 2007 World Health Organization growth reference
20
tables91. The participants were classified as obese (BMI z-score [BMIz] > +2 SD) or
non-obese (BMIz ≤ +2SD).
2.1.2.2. Active transport
Active school transport was assessed through questions adapted for each country
from the Canadian component of the 2009-2010 Health Behaviour in School-aged
Children Study92. The children were asked about the main mode of transport they
used to go to school during the last week. The response options included active
modes such as walking, bicycle, roller blades, scooter, non-active modes such as
motorized vehicles (e.g., car, motorcycle, bus, train, tram, underground or boat), and
others according to country specific modes of transport. Other modes of
transportation included active modes such as running and jogging, and non-active
modes such as the school van, matatu, bus feeder, pedicab, wheelchair and riding
on the top tube of the bike’s frame. For this analysis, we classified children into two
categories, active transport vs non-active transport. To assess biking and other
wheeled modes of transport independent from walking we also classified a
subsample of children into two categories (non-active vs biking or other wheeled
modes of transport). In addition, a question regarding the time spent during the
journey from home to school was included. The response options were: less than 5
minutes, 5 to 15 minutes, 16 to 30 minutes, 31 minutes to 1 hour and more than 1
hour. Based on the AST and travel time variables, we created an AST-composite
variable with the following categories non active-AST, less than 5 minutes to 15
minutes, and at least 16 minutes. The common referent category of this composite
variable was non active-AST.
2.1.3. Covariates
21
The socio-demographic variables included age, sex, highest parental education and
motorized vehicle ownership. Age was computed from birth and anthropometry
dates. Sex and parental education was recorded on the demographic and family
history questionnaire. The highest parental education variable was created based
on the highest education level of the mother or the father. Motorized vehicle
ownership was reported as the number of motorized vehicles available for use in the
household. Motor vehicles included cars, motorcycles, moped and/or trucks.
In addition, the time spent in moderate-to-vigorous physical activity (MVPA) was
obtained from 24-h, waist-worn accelerometry. An Actigraph GT3X+ accelerometer
(ActiGraph, LLC, Pensacola, FL, USA) was worn at the waist on an elasticized belt
on the right mid-axillary line. The participants were encouraged to wear the
accelerometer 24 hours per day (removing only for water-related activities) for at
least 7 days (plus an initial familiarization day and the morning of the final day),
including weekends. The minimal amount of accelerometer data that was considered
acceptable for inclusion in the sample was 4 days with at least 10 hours of wear time
per day, including at least one day of the weekend. MVPA was defined as all activity
≥574 counts per 15 seconds. This protocol provided reliable estimates93.
2.1.4. Statistical Analysis
The descriptive characteristics included the means and standard deviations (SD) for
continuous variables and the frequencies of categorical variables by study site.
Associations between AST and obesity were estimated in terms of odds ratios using
generalized linear mixed models (SAS PROC GLIMMIX). Associations between AST
and continuous body composition variables (i.e., BMIz, PBF and waist
circumference) were estimated using a linear mixed model (SAS PROC MIXED).
22
The Models were adjusted for age, sex, highest level of parental education and
motorized vehicles. To assess effect modification by study site, an AST*study site
interaction term was included in the multivariate model. To assess dose-response
relationships between body composition and AST, we used the AST-composite
variable of travel time. In addition to the primary analyses, three sets of sensitivity
analyses were conducted. First, analyses were conducted with sub-samples that
included weekend MVPA as a covariate. We did not adjust for mean weekly MVPA
because it is an intermediate factor in the conceptual model. Second, use of the
public bus, which could include walking as part of the trip94, was reclassified within
the active mode category. Third, the category of biking and other wheeled modes of
transport was analyzed separately from walking in a sub-sample. Study sites and
schools nested within study sites were considered as having random effects. The
denominator degrees of freedom for statistical tests pertaining to fixed effects were
calculated using the Kenward and Roger approximation95. All statistical analyses
were conducted using SAS version 9.3 (SAS Institute, Cary, North Carolina, USA).
2.2. Results
2.2.1. Socio-demographic characteristics
Reflecting the variability in the ISCOLE sample, selected countries differed in several
socio-economic and transport indicators. According to the World Bank
classifications, ISCOLE countries differed in income level and income distribution
(Table 1). Likewise, ISCOLE sites also differed in number of motor vehicles with the
US having the highest value (809 per 1000 inhabitants) and India having the lowest
value (15.0 per 1000 inhabitants)96. According to the World Health Organization
indicator on road traffic death rates, sites showed large differences with South Africa
23
having the largest rate (31.9 per 100000 population ) and UK having the lowest rate
(3.7 per 100000 population)97.
Table 1 also shows descriptive individual characteristics of participants stratified by
study site. The children from the study were on average 10.4 (SD=0.6) years old and
46.3% were male. Overall parental highest education differed by site with India
having the highest percentage of parents with at least a college education (47.3%)
and Colombia having the lowest percentage (17.5%). Overall, 76.7% of parents
reported motorized vehicles in their households ranging from 24.4% in Colombia to
97.7% in Australia.
2.2.2. Body composition
The overall percentage of obese children was 12.5%, which ranged from 5.3% in
Finland to 23.7% in China. The mean PBF was 20.9% (SDS=7.7), and the mean
waist circumference was 64.3 cm (SD=9.0). The mean PBF ranged from 16.6% in
Kenya to 23.1% in Brazil and the mean waist circumference ranged from 62.2 cm in
Kenya to 66.9 cm in Brazil.
2.2.3. School transport
Sites also differed by main mode of transport to school (Figure 1) and travel time
(Table 1). Within the active mode category, the percentage of children reporting
walking to school ranged from 3.8% in India to 71.5% in Colombia. Less than five
percent of the children reported other active modes of transport like biking and
wheeled modes of transport, ranging from 0.7% in the US to 24.7% in Finland. Within
the non-active mode category, 22.7% reported some kind of public transportation
ranging from 3.2% in the UK to 61.8% in India. About a third of the children reported
car or motorcycle as their main modes of transport ranging from 7.4% in Colombia
24
and Finland to 63.7% in Australia. In the subsample of children from the study that
reported AST, 26.1% reported spending less than 5 minutes commuting to school,
53.2% spent 5 to 15 minutes commuting to school and 20.7%spent more than 15
minutes (Table 1 and Figure7A). The active time spent commuting varied largely by
site. In Australia 88.3% of the children reported less than 15 minutes of AST and in
Kenya 19.7% of the children reported more than 30 minutes of AST.
Figure 2.Estimated fractions of BMI categories by age group and SES. a) Not-overweight population of lower
SES a) Overall transport to school mode distribution per site b) Mode distribution per site among trips
shorter than five minutes
a)
b)
0
10
20
30
40
50
60
70
80
90
100
Walking Bicycle and other wheels Bus, train, metro, boat Car and other motorized vehicles Other
Active Non-active
25
Figure 3. : Typology distribution according to active school transport and income level. Size of the dots
represent proportion of obesity by country-site.
2.2.4. Associations between AST and body composition indicators
0
10
20
30
40
50
60
70
80
90
100
Walking Bicycle and other wheels Bus, train, metro, boat Car and other motorized vehicles Other
Active Non-active
26
Multi-level analyses of the associations between AST and body composition are
presented in Table 2. There were negative associations between AST and obesity
(0.72, CI [0.60-0.87]), BMIz (-0.09, [SE=0.04]), PBF (-0.56, [SE=0.22]) and waist
circumference (-0.90, [SE=0.26]) after adjusting for age, sex, parental education and
car ownership. Similarly, when we analyzed only AST by bike and excluded walking,
there were negative associations between AST and BMIz (-0.02, [SE=0.08]), and
waist circumference (-1.27, [SE=0.59]) after adjusting for age, sex, parental
education and car ownership. No differences between AST and sex and AST and
site were apparent. We did not find a significant trend in the dose-response analysis
(p for trend=0.213). The estimates did not change significantly when adjusting for
weekend-
MVPA. When public bus was included in the active mode category the point
estimates decreased in magnitude.
27
Table 1. Associations of body composition variables with active school transport in 6,797 9-11 year old children in the
International Study of Childhood Obesity, Lifestyle and the Environment (ISCOLE).
Unadjusted Adjusteda
OR 95% CI p-value OR 95% CI p-value AST*site
Obesityb
Boys N=3149
Active transportc 0.69 (0.55-0.87) 0.002 0.69 (0.55-0.88) 0.002
Girls N =3648
Active transport 0.76 (0.59-0.99) 0.038 0.74 (0.56-0.96) 0.025
Total Sample
Active transport 0.74 (0.62-0.88) 0.001 0.72 (0.60-0.87) <0.001 0.279
Bicycle or other wheels 0.76 (0.51-1.14) 0.185 0.72 (0.48-1.09) 0.124 Did not converge
β SE p-value Β SE p-value AST*site
BMIzd
Boys N=3149
Active transport -0.14 0.05 0.007 -0.12 0.05 0.026
Girls N =3648
Active transport -0.12 0.05 0.012 -0.08 0.05 0.082
Total Sample
Active transport -0.11 0.04 0.002 -0.09 0.04 0.013 0.132
Bicycle or other wheels 0.16 0.08 0.049 -0.17 0.08 0.036 0.313
Percentage body fat (%)
Boys N=3149
Active transport -1.01 0.30 0.001 -0.88 0.30 0.004
Girls N =3648
Active transport -0.60 0.30 0.043 -0.49 0.30 0.105
Total Sample
Active transport -0.81 0.22 <0.001 -0.66 0.22 0.002 0.315
Bicycle or other wheels -1.11 0.51 0.031 -0.88 0.49 0.077 0.603
Waist circumference (cm)
Boys N=3149
Active transport -1.17 0.38 0.002 -1.10 0.38 0.004
Girls N =3648
Active transport -0.86 0.34 0.012 -0.87 0.35 0.012
Total Sample
Active transport -0.90 0.26 0.001 -0.90 0.26 0.001 0.168
Bicycle or other wheels -1.22 0.6 0.045 -1.27 0.59 0.034 0.589
a Models were adjusted for age, parental education, and motorized vehicle ownership. The combined analyses of boys
and girls were also adjusted for sex.
b Obesity defined as BMI WHO z-score > 2
28
2.3. Discussion
This study is the first to examine associations between AST and body composition
indicators among a multi-national sample of children from low-to high-income
countries. Our findings show that children who used AST were less likely to be
obese, had lower BMIz, lower PBF and a lower waist circumference, compared to
those who used a non-active mode of transport. Likewise, children who reported
biking as their main mode of transport had a lower BMIz and waist circumference.
Overall associations of obesity and AST did not differ by country or sex. The low
evidence of heterogeneity in the associations between AST and body composition
indicators among countries with a wide range of income distribution, transport
indicators and stages of PA and nutrition transition provides evidence of the
importance of promoting AST as one of the strategies to prevent obesity.
Our results are consistent with the few previous smaller studies that found that active
travelers had lower BMI and were less likely to be overweight and obese41. While in
the same direction, our 12-country-site study showed larger negative estimates
which could be due to a larger sample size with higher variability.
The pathway in which AST is associated with lower measures of body composition
indicators may occur in part through small increments of everyday levels of PA98. PA
could potentially be increased if non-active trips of less than 5 minutes were replaced
by active commuting in a suitably built environment with safe conditions. Specifically,
our study shows that 10.3% of all trips to school are non-active and take less than
c Active transport was defined as walking or riding a bike, roller blade, skateboard or scooter in the main part of the journey
to school during the last week.
d Body mass index z-score according to WHO reference data.
29
five minutes. For example, in the US (Baton Rouge) 76.5% of trips that take less
than 5 minutes are motor vehicle-dependent. In contrast, in Finland (Helsinki, Espoo
and Vantaa), only 15.6% of trips that take less than 5 minutes are motor- vehicle-
dependent.
Nonetheless, walking extremely large distances could be associated with a low
quality of life99,100. In our study, among the subsample of children who used AST,
19.8% of children in Kenya reported walking to school for more than 30 min and
10.1% walk more than one hour for one-way trip. Therefore, these trips should
potentially be made using multimodal transportation combining active and non-active
modes. Multimodal strategies that take into account AST should be implemented
before unintended consequences of development negatively affect transport-related
activity in those countries undergoing early stages of PA transition.
Despite not finding significant differences between AST and body composition
among the countries our results should be understood within the “need-based
framework“ of LMIC and the “choice-based framework” of HIC101,102. Specifically, in
LMIC where car ownership remains relatively low in comparison with HIC, AST may
be more reflective of need rather than choice since a significant proportion of the
children walk to school because they have no other option for transportation.
Therefore, our results could be used to classify countries into four typologies that
could be useful for future AST interventions (Figure 2). The first typology includes
LMIC with higher proportions of AST, including Colombia, Brazil South Africa and
Kenya. The second typology includes LMIC with lower proportions of AST, including
India and China. The third typology includes HIC with lower proportions of AST,
30
including USA, Portugal and Canada. Finally, the fourth typology includes HIC with
higher proportions of AST including Finland and the UK.
Sites like Colombia and Finland, where a large proportion of AST was observed,
have school transportation programs and built environment characteristics that
promote AST. For both of these sites, proximity to the school is a key factor. In
Bogota 90% of the children who attend public schools live within 2km103, and in
Helsinki 70% of the primary school students go to their nearest school104. In Bogotá,
the District Education Department has a School Transportation Program targeted
mainly to public schools from low socio-economic levels with 2 main strategies. The
first strategy promotes walking to school among children who live within 1 km of the
school under the supervision of an adult. The second strategy “Al colegio en bici”
promotes the use of the bicycle to go to school among children located within 1-2
km of the school103. In Finland, most of the children attending public schools, use
an active mode of transport to go to school, and the municipalities provide free public
transportation tickets for those children living within distances over 2 km; however,
regardless of the mode of transport, Finnish children are very independent in their
mobility104. Nonetheless, Finland differs significantly from Colombia, in car
ownership, safety and traffic accidents. Both countries are non-car-dependent for
different reasons; in Finland, by choice and in Colombia, by need due to low motor-
vehicle ownership.
This study has several strengths, including a large international sample of children
from 12 sites in five continents with different environmental and socio-economic
31
settings, multiple direct measures of body composition and standardized instruments
and rigorous training protocols to ensure the comparability among sites.
Nonetheless, our findings should be interpreted cautiously considering the following
limitations. First, the design of the study is cross-sectional; therefore, we are unable
to determine the direction of causality. Second, despite the large internationally
diverse sample included, none of the countries had a nationally representative
sample; hence the results may not be generalizable to country-sites. Third, AST was
defined based only on the “main” mode of transport for the journey “to school”. Thus
we assumed that both journeys were the same. However, the mode of transportation
by journey could differ and could be multimodal. This, in part, explained why we did
not find a dose-response relationship. Fourth, biking was combined with other
wheeled modes of transportation, and its low prevalence provided very imprecise
estimates. Finally, we did not independently assess the short active trips of public
transportation.
ISCOLE is the first multi-country study that shows associations between body
composition and AST in a sample of 9-11-year old children. Such findings could
inform global efforts to prevent obesity among school-age children. The large
differences among countries in terms of AST patterns underscore the importance of
considering the need-based and choice-based frameworks when designing
interventions to prevent obesity by promoting active commuting. As shown in
Chapter one obesity is a growing problem and is a problem for all ages. Given the
wide range of obesity in children around the world, The International Study of
Childhood and Environment (ISCOLE) was created. The object of the study is to
32
determine the relationship between lifestyle characteristics and obesity in children in
twelve countries. In this chapter we specifically study the relationship between using
active transport to school and obesity in children. Reviews have shown that children
that use active transport to school accumulate more minutes of PA 105. However
there is little evidence of the relation between active transport and obesity in children
and none of the studies include all continents.
33
3. Bus Rapid Transit Systems to promote Physicial activity
So far I have established that non communicable diseases are a major world
problem. This problem is related to overweight, obesity and physical inactivity.
Colombia suffers from overweight and obesity and simulations show the problem is
expected to grow. I also showed that there is a negative association between active
transport and obesity among children from twelve countries and that this association
does not vary by country. The following chapters of this document will focus on the
influence of the built environment on walking for transport. Specifically the BRT in
Bogotá.
In general, PA interventions can be classified into community-based, health care and
other sector interventions. Community-based interventions are interventions
designed for a specific community, such as active pauses in organizations or PA
programs in a school. Health care interventions are an initiative of the health care
system, for example a clinical medication of PA. Finally, other sector interventions
are those interventions whose main goal is not PA but as an externality promote it.
These interventions specifically in transport infrastructure will be our main focus.
Studies on access and use of public transit and walking have shown inconsistent
results. On one hand, positive associations are seen between public transit access
and walking for transportation. Walking has been positively associated with distance
to the nearest transit stop 106. Specifically, walking for transport has been associated
with access to Bus Rapid Transit (BRT) stations in Curitiba 107 and Bogotá 108 109,
use of trains in New Jersey 110, access to transit stops in Atlanta 111, and in public
transport use in Australia 112. On the other hand, studies assessing perception of
distance to public transit in 11 countries showed positive, negative and null
34
associations 113. Nonetheless, most of the studies have assessed only access
instead of use, have been cross-sectional, and have been conducted in cities in the
developed world 114.
3.1. BRT a global trend
The Colombian Nutritional Survey (La encuesta Nacional de Situación Nutricional
[ENSIN, 2010]) states that 53.5% of adults in Colombia meet PA recommendations,
19.9% meet the PA recommendations through spare time PA, 33.8% meet the
recommendations walking for transport and 5.6% through biking for transport. This
means that 73% of people that meet PA recommendations in Colombia meet them
through active transport. Bogotá is very similar, 57.8% of adults meet the PA
recommendations and 77% of those that meet the recommendations meet them
through transport (See Figure 9).
Figure 4. Percentage of adults that meet physical activity recommendations in Bogotá and Colombia.
35
Source: ENSIN 2010
In the world more than 4,000 km of BRT and bus corridors have been
implemented in 160 cities across the globe, serving 29 million passengers per day
115. BRT success can be partly attributed to the mode’s low cost, rapid
implementation, and impact on mobility 116. Bogotá has been part of that trend. The
TransMilenio (TM) system in Bogotá is a BRT referent for planners and practitioners
worldwide 117. Before 2000, Bogotá had public transport services offered by small-
vehicle owners with little supervision by the city. Old public transportation buses do
not have stops: they drop and pick passengers everywhere, and thousands of routes
exist. The result was an ample supply with relatively low costs, but with high negative
impacts, mainly congestion, air pollution, and road traffic incidents. In response, the
city initiated the implementation of an organized system with BRT corridors and
feeder routes, called TransMilenio (TM), which was expected to contribute to mobility
in the city by significantly decreasing traffic congestion and travel times 118. Several
evaluations of the TM system impacts have confirmed these expectations 116. 119.
According to Hidalgo et al.120, the first two phases, comprising 84 km of the BRT
network and 663 km of feeder routes, and serving 1.7 million passengers per day (in
2011), resulted in the equivalent of USD $3.1 billion in societal benefits from 2000 to
2008 (using a 12% discount rate). Benefits came from travel-time savings for transit
users, savings on the operation of traditional buses removed from service, and
reductions in deaths and illnesses from reduction in air pollution, reductions in
particulate matter, and increased traffic safety 120. However, these estimates did not
include benefits from increased PA. In 2012, TM was expanded to 106 km, and in
2013 it was moving close to two million passengers per day 115.
36
Initial plans included the construction of 388 km of BRT corridors with 85% spatial
coverage of the city by 2016 119.
Figure 5. Constructed and projected TM in Bogotá
The plan is advancing at a slower pace than originally envisioned: only 27%
of the system planned in 2000 was in operation by 2013. Nevertheless, TM is
expected to continue expanding over the upcoming years, including adding 34 km
to the system over the next three years, with complementary investments in rail
transit and cable cars 121. This study contributes to the field of health and
transportation research by assessing the association between the use of and access
to a growing BRT and walking behaviors using repeated cross-sectional studies
conducted in 2005 and 2010-11 a longitudinal analysis of a cohort sampled during
both periods and an agent based model which has the capacity to explain patterns
observed in the real world, e.g., in Bogota's mobility survey 122
37
3.2. Setting
The study was conducted in Colombia´s capital city, Bogotá, which has an
estimated population of 7.4 million 123. In Bogotá, 41% of trips 15 minutes or longer
are made using public transport, 28% are made on foot, 14% in automobile, 5% by
taxi, and the remainder using other modes of transportation 122.
3.3. Study design
The study is composed from two cross sectional studies. The first cross-
sectional study was conducted in 2005 108 and the second in 2010-11 as part of the
Colombian component of IPEN (International Physical Activity Environment
Network), a collaborative network of researchers studying the associations between
PA and the built environment in 12 countries 124. To assess the association between
walking and TM access, we used combined data from these two cross-sectional
studies in a post-hoc cross-sectional and a cohort analyses. The latter takes
advantage of a natural experiment: the replacement of the traditional-bus transport
system by TM in selected geographic units, allowing us to study the influence of a
faster and more organized mode of transit on walking for transport (Figure 1).
Figure 6. Sampling design for cross-sectional and cohort studies using 2005 and 2010-11 data.Group 1 =
adults living in geographic units with at least one TransMilenio station in 2005 and 2010. Group 2 = adults
living in geographic units with no TransMilenio station in 2005 but at least one in 2010. Group 3 = adults
living in geographic units with no TransMilenio station in either 2005 or 2010–2011.
IPEN 2010
1000 adults 18 to
70 years
2005-2010
360 adults
Cohort study
2005-2010
Group 1
2005-2010
Group 2
2005-2010 Group 3
38
The studies used a multistage stratified sampling design. First, a
representative sample was taken of 30 neighborhoods stratified by socioeconomic
status (SES), average slope of terrain, proximity to TM stations, and public park
provision 108. Next, five blocks were randomly selected within each neighborhood,
and within each block 10 households were randomly selected, and in each
household one adult was systematically surveyed. If the selected adult did not
answer the survey, then a replacement adult was chosen from the next house. For
the 2010-11 survey, in accordance with IPEN protocol, households were randomly
selected after being stratified by a walkability index. Although this procedure
preserved all the geographic units from the 2005 cross-sectional study, it led to
differences in the households and adults surveyed in 2010-11. Nonetheless, we
were able to follow a subsample of 360 adults for the cohort analysis
3.4. Data collection
The survey data for both cross-sectional studies (2005 and the 2010–11
IPEN-Colombia) were collected through face-to face interviews with subjects who
met the following criteria: had lived in the household for at least a year, were between
18 and 74 years old, and had no physical or cognitive disabilities. All the protocols
and questionnaires were reviewed and approved by the Institutional Review Board
of the Universidad de los Andes in Bogotá.
3.5. Physical activity measurements
3.5.1. Subjective physical activity measurements
39
PA was measured using the long version of the International Physical Activity
Questionnaire (IPAQ) 125, a validated self-report measurement tool 126. The PA
dimension assessed was walking for transport, classified based on three cut-off
points: (1) Walking for transport ≥10 minutes/week vs. walking ≤ 10 minutes/week,
which is characteristic of short trips walking to transit. (2) Walking for transport ≥120
minutes/week vs. walking ≤ 120 minutes/week, the inflection point of the continuous
measures of walking minutes and TM access using smoothed LOESS curves (e.g.
robust locally weighted regression) 127. (3) Meeting the weekly PA recommendation
of ≥150 minutes/week vs. walking ≤150 minutes/week.
3.5.2. Objective physical activity measurements
Objective measures provide more accurate measurement of PA frequency
and intensity, particularly when walking. PA was also measured objectively in a
subsample of 250 participants from the 2010-11 study. The inclusion criteria for this
analysis were the same as those used for the questionnaire with one addition: having
no plans to travel in the next week. These participants, who were evenly distributed
by neighborhood, were asked to wear accelerometers (model GT3X ActiGraphTM,
Pensacola, CA) twelve hours a day for seven days except when sleeping, showering,
or swimming.
In accordance with the IPEN-study protocol, all accelerometry data were
stored in 60-second epochs, and were scored using MeterPlus 4.2 based on
Freedson's cut-off points for adults 128. The intensity of moderate-to-vigorous activity
corresponded to ≥1,952 counts per minute. The results were considered valid if
participants wore the accelerometer for at least ten hours a day for five days. If not,
they were asked to re-wear the device for the days needed to comply with the
40
measurement requirements. An hour was considered invalid any time zero PA was
recorded for 60 minutes. The cut-off point used with the accelerometer data was an
average moderate and vigorous physical activity (MVPA) per day of more than 22
minutes.
3.6. Independent variables
3.6.1. Individual variables
To account for characteristics that may influence walking behavior 129,
individual variables included sex, age, marital status, education, monthly household
income, neighborhood SES, occupation over the last 30 days, years of residency in
the neighborhood, and motorcycle and car ownership. The 2010–11 study, as part
of its transportation module, also evaluated modes of transportation and minutes
spent on each mode (public bus, TM, feeder bus, car, taxi, motorcycle, and others)
in the previous seven days.
3.6.2. Geographic units and built-environment characteristics
The geographic unit was a 1,000-meter network around the block centroid in
each neighborhood (Figure 2). These units were classified into three groups: those
with at least one TM station in both 2005 and 2010, those with no TM station in 2005
but at least one in 2010, and those with no TM stations in either 2005 or 2010.
41
Figure 7. Sampling design for cross-sectional and cohort studies using 2005 and 2010-11 data. (a) Group 1
= adults living in geographic units with at least one TransMilenio station in 2005 (Phase 1) and 2010-11
(Phase 2). (b) Group 2 = adults living in geographic units with no TransMilenio station in 2005 (Phase 1)
but at least one in 2010-11 (Phase 2). (c) Group 3 = adults living in geographic units with no TransMilenio
station in either 2005 (Phase 1) or 2010-11 (Phase 2).
a) b) c)
The built-environment (BE) characteristics were slope of terrain, presence of
public park(s), proximity to TM, and walkability index. In line with the IPEN protocol,
geographic units were classified according to the walkability score derived as a
function of the z-score of four variables: (i) net residential density (ratio of residential
units to the land area devoted to residential use); (ii) land-use mix (diversity of land-
use types per block), whose normalized scores ranged from 0 (single use) to 1 (an
even distribution of area across several use types, including residential, retail,
entertainment, office, institutional); (iii) intersection density (connectivity of street
network measured as the ratio of number of intersections with three or more legs to
land area of the administrative unit); and (iv) retail floor-area ratio (the ratio of retail
42
building floor area to retail land area). Walkability index, which was subsequently
divided into high and low walkability, was calculated using the following function: (2
x z-intersection density) + (z-net residential density) + (z-retail floor area ratio) + (z-
land use mix) 130. The data for BE characteristics were collected from Bogotá’s
Planning Department, its Cadester’s 2010 digital map, the Corporation of
Universities in the Center of Bogotá, and Universidad Jorge Tadeo Lozano. All GIS
variables were generated using ArcGIS 9.3 (ESRI, Inc., Redlands, CA, USA).
3.7. Statistical analyses
3.7.1. Cross-sectional analysis using 2010-11 data
For the cross-sectional analysis of the 2010-11 data (N = 1,000 adults), our
analytic strategy involved three steps. First, we conducted bivariate analyses to
assess the association between ≥150 minutes/week of walking for transportation and
sociodemographic characteristics, BE characteristics, and transportation patterns.
Second, we evaluated collinearity between independent variables using Pearson
correlation coefficients. Third, we employed a multilevel multivariate Poisson 131
model to measure the association between TM use and walking for transportation
for ≥150 minutes/week. The model used is as follows:
𝐿𝑜𝑔{𝑢𝑖𝑗} = 𝑓(𝑥𝑖𝑗) + 𝜖𝑖,
Where 𝑢𝑖𝑗 = 𝐸(𝑦𝑖𝑗) and 𝑦𝑖𝑗~𝑃𝑜𝑖𝑠𝑠𝑜𝑛.
𝑳𝒐𝒈(𝑼𝑖𝑗) = 𝛽0𝑗 + 𝛽1𝑥𝑖𝑗 + 𝜀𝑖𝑗, to distinguish the different levels we have a random
intercept for each different unit in level two. Therefore 𝛽0𝑗 = 𝛾00 + 𝐾0𝑗 where
𝑖 indicates level-one unit (individuals)
𝑗 indicates level-two unit (geographical units)
43
𝑥𝑖𝑗 indicates explanatory variables at the individual level
𝛾00 is the average intercept
𝐾0𝑗 is the geographical unit effect with normal distribution (0,τ)
3.7.2. Subsample of adults with objective physical activity measures
For the subsample of adults who wore accelerometers (N = 250), the analytic
strategy involved three steps: (i) identifying the differences in sociodemographic
characteristics between the overall sample and the accelerometer subsample, (ii)
assessing the differences in MVPA minutes among TM users and nonusers, and (iii)
comparing MVPA minutes throughout the day between users and nonusers. These
comparisons were conducted using the generalized additive model (GAM) function
of the MGCV package in R version 3.0.1 (R Foundation for Statistical Computing)
with a smooth covariate function (time of day) for each group. We fitted the GAM to
the data using a Poisson distribution with a log-link function and used thin-plate
regression splines to estimate a smooth function 132. The model structure used for
users and nonusers was:
𝐿𝑜𝑔{𝑢𝑖} = 𝑓(𝑥𝑖) + 𝜖𝑖,
Where 𝑢𝑖 = 𝐸(𝑦𝑖) and 𝑦𝑖~𝑃𝑜𝑖𝑠𝑠𝑜𝑛.
𝑦𝑖 is MVPA minutes throughout the day, 𝑥𝑖 is time of day, 𝑓 is a smooth function of
the covariate 𝑥𝑖 represented using a thin-plate regression spline basis, and 𝜖𝑖 are
i.i.d. 𝑁(0, 𝜎2) random variables.
3.7.3. Cohort analysis using both the 2005 and 2010-11 data
44
For the cohort analysis (N = 360 adults), our analytic strategy involved four
steps. The first estimated the differences in sociodemographic and transportation
characteristics based on TM accessibility. The second calculated the differences in
walking ≥150 minutes/week in both 2005 and 2010 between Group 2 (geographic
units with no TM station in 2005 but at least one in 2010) and Group 3 (geographic
units with no TM stations in either 2005 or 2010). The fourth identified the association
between walking for transport in geographic units with TM stations Group 1
(geographic units with TM station in 2005 and 2010) in 2005 and in 2010 Groups 1
and 2 and walking for transportation. All these analyses except the GAM were
conducted using the new SAS (version 9.2) GLIMMIX procedure with a Poisson
distribution and log-link functions with robust error variance 131.
3.8. Results
3.8.1. Cross-sectional sample based on 2010-11 data
Individual characteristics
The 2010-11 survey was applied to 1,000 participants, with an overall
response rate of 56.6%. Of these 1,000 participants, 63.7 percent were female. The
mean age was 41.1 (SD 14.52), 34.3% were single, 53.1% were married or
partnered, and 12.6% were divorced, separated, or widowed. In terms of education,
39.0% had a technical, college, or graduate degree. Nevertheless, 47.6% of the
participants reported a household income of less than USD $350 per month, and
49.0% reported living in a low-SES area. Forty-eight point three percent of the
sample reported being employed, and 48.1% named “working” as their principal
activity in the last 30 days. The mean length residence in the neighborhood was 15.5
years (SD12.6). The socio-demographic characteristics resemble Bogotá’s
45
characteristics in terms of neighborhood SES. However, participants from the study
are more likely to be female, married, more educated, and older.
Transportation characteristics
Although 31.9% of the participants reported family car ownership, with 15.4%
having two or more cars, the most frequently used mode of transportation over the
last seven days was public bus (68.1%), followed by car (33.2%) and TM (29.3%).
The median minutes and days per week for use of each mode of transportation were
240 minutes over three days for the bus, 180 minutes over two days for the car, and
120 minutes over two days for TM.
Association between walking for transportation and TM use
About half the participants reported walking ≥150 minutes/week for
transportation (Table 1) (i.e., moderate aerobic physical activity). Bivariate analysis
results indicate users of TM were more likely to walk for transport ≥150 minutes/week
[1.2] 95% CI [1.0-1.4]. Correlation coefficients did not show correlations above 0.3
for the variables used in the analysis. Users of TM were more likely to report walking
for transport ≥150 minutes/week compared to non-users (57.6% vs. 48.1 %; PR
unadjusted [1.2] 95% CI [1.1-1.4]; PR adjusted [1.2] 95% CI [1.1-1.4]) (Table 2).
3.8.2. Subsample of adults with objective physical activity measures
Individual characteristics
The subsample of adults who wore accelerometers was older than the
general population, less educated, and with a lower SES. Sixty-eight percent were
female, and the mean age was 48.1 years (SD 13.4) (data not shown)
46
Physical activity patterns
Among these adults, 65.6% overall averaged more than 22 minutes of MVPA
per day. Twenty-six percent were TM users, and 74% were non-users. TM users,
however, had more minutes than nonusers, a median of 38.4 versus 28.0 minutes a
day ( Figure 3). TM users were also more likely to average over 22 minutes of MVPA
daily (87.7% vs. 59.4%: unadjusted PR = [1.4] 95% CI [1.2-1.7]; adjusted PR = [1.3],
95% CI [1.1-1.6]). The average difference in fitted MVPA minutes per day between
TM users and nonusers was 12 minutes. Based on trip distribution across the city,
the largest differences are consistent with peak transportation hours (6:15-7:15,
11:45-12:45, 17:30-18:30) (Figure 13) 122.
Figure 8. Relationship between moderate and vigorous physical activity (MVPA) minutes per hour and time
of day for users and nonusers of the TransMilenio (TM) (N = 250).
3.8.3. Cohort analyses using 2005 and 2010-11 data
Individual characteristics
The subsample of adults (N = 360) in the cohort study differed from the overall
sample in the 2010-11 cross-sectional analysis on four sociodemographic
0
0.5
1
1.5
2
2.5
3
3.5
4
0 2 4 6 8 10 12 14 16 18 20 22
MV
PA
min
ute
s p
er
ho
ur
Hour
Predicted MVPAminutes TM users
Predicted MVPAminutes Non TMusers95% Confidenceintervals
47
characteristics: they reported lower SES and were older, less likely to be single, and
less likely to be studying. The surveyed geographic units were 12.2% from Group 1,
23.6% from Group 2, and 64.2% from Group 3.
Transport characteristics and access to TM
Although family car ownership increased from 15.6% in the 2005 data set to
25.8% in the 2010-11 data set (Table 3), the mode of transportation most frequently
reported in 2010-11 was bus (54.7%), followed by TM (17.5%) and car (15.3%)
(Table 3). With the introduction of the second TM phase the number of adults with
TM stations available increased from 10.8% in 2005 to 32.2% in 2010. A
corresponding increase occurred in the percentage of adults with access to more
than one TM station: from 3.8% in 2005 to 17.0% in 2010.
Association between walking for transportation and TM access
The number of minutes walking for transportation increased in all three
groups. However the smallest increment in walking for transportation greatest
increment occurred in group two, the group that did not have access to TM in 2005
and acquired access in 2010
48
Figure 9.Median of minutes walking for transportation in 2005 and 2010 according to group
When comparing the prevalence of walking ≥150 minutes/week based on the 2005
data, Group 1 (TM access in both 2005 and 2010-11) had the highest prevalence of
walking ≥150 minutes/week for transportation (48.7%), followed by Group 2
(geographic units with no TM station in 2005 but at least one in 2010) (32.5%) and
then Group 3 (geographic units with no TM stations in either 2005 or 2010) (25.8%).
According to the 2010-11 data, Group 2 had the highest prevalence of walking for
transportation ≥150 minutes/week (63.6%), followed by Group 3 (47.5%) and Group
1 (46.2%).
Based on data from 2005, Group 2 was more likely than Group 3 to walk ≥150
minutes/week in 2010-11 (adjusted PR2005 = [1.4], 95% CI [0.8-2.4]); based on data
from 2010 this likelihood increased (adjusted PR 2010-11 = [1.5], 95% CI [1.0-2.2]) (not
shown in table 2).
49
Based on the combined 2005 and 2010-11 data using different cut-off points, adults
were more likely to walk for transportation ≥150 minutes/week and ≥120
minutes/week in 2010 than in 2005, and adults living in geographic units with TM
stations were also more likely to walk for transport ≥150 minutes/week (PR = [1.4],
CI 95% [1.0-2.0]) (see Table 2).
Moreover, whereas the 2005 data suggest that Group 1 was more likely to walk ≥150
minutes/week than Groups 2 and 3 (adjusted PR2005 = [1.6], 95% CI [0.80-3.0]), the
2010-11 data indicates a greater likelihood of walking ≥150 minutes/week for both
Group 1 and Group 2 than Group 3 (adjusted PR2010 = [1.4], 95% CI [1.0-2.0]) than
those living in geographic units without TM access. However the interaction term
taking into account group and time did not show any influence.
Cost-benefit ratio.
Our estimates could be useful for a sensitivity analysis to estimate the TM/PA
promotion cost-benefit ratio over the 2000-2013 period. This value denotes the
return in medical costs savings attributed to walking for transport for TM users for
every dollar spent on the TM average annual costs. First, to calculate the average
annual TM costs, we discounted the sum of the average construction (71%) and
operational (29%) costs in 2013 U.S. dollars by assuming a 13-year period of TM
usage (2000-2013). Considering the calculated costs, the number of users meeting
PA recommendations, and the adjusted PR (Table 2), we estimated the health-care
savings from transportation-related PA to be between 3% and 22% of every dollar
spent on TM costs. Every TM user has a difference in the expected value of medical
costs savings attributed to meeting the PA recommendations by walking for
transportation of USD $10 per year compared to a non-user 133, which corresponds
50
to the portion of the direct health benefit for PA in Colombia attributed to PA for TM
usage 134. According to the number of physically active users attributable to TM, we
estimated an annual direct health benefit derived from PA and attributed to TM of
between USD $3,310,278 and USD $22,068,518, which corresponds to
approximately USD $2.63-USD $17.55 per user per year. This benefit represents
between 1% and 5% savings on the average annual medical costs per user.
3.9. Discussion
Our analyses showed a positive association between walking for
transportation and TM use and access. Adults with TM access were more likely to
walk for transport, with nearly 6 out of 10 TM users walking at least 150
minutes/week and reporting a median MVPA per day of 38.4 minutes. Likewise, the
average difference in daily MVPA minutes between users and nonusers was 12
minutes, signaling that TM offers a potential health co-benefit. This positive
association may be relevant globally due to the high prevalence of NCDs 3 and the
worldwide growth in BRT as a mass transit mode 116.
The BRT in Bogotá has shown multiple positive externalities, including
decreased air pollution, carbon emissions, and automobile accidents and improved
urban environments for walking 120. Our study also underscores an additional
positive externality: walking for transportation. This behavior, as we hypothesized, is
positively associated with TM use and access, possibly because TM is the fastest
mode of transport, has fixed stations, and is linked to built-environment
transformations. TM is one of the fastest BRT systems in the world, as a result of the
combination of dedicated lanes with large stations with level access to the buses
through multiple doors, and overtaking lanes used by a combination of express and
51
local services. A commercial speed of 26 km/h allows for major time savings,
especially on long trips 120. TM construction is also associated with improved walking
infrastructure (pedestrian bridges and wider sidewalks) 135. These factors together
may attract users to walk longer distances to access BRT stations rather than other
modes of transportation.
Our results are consistent with a study conducted in Curitiba, which found that
a higher density of BRT stations was associated with walking at least 10 minutes per
week 136. The fact that this study used a lower cut-off point than our study
underscores the importance of assessing multiple thresholds in different settings. In
addition, consistent with Bogotá’s mobility survey from both 2005 and 2010 122 and
Colombia’s 2005-2010 National Nutrition Survey 4, our analysis shows an increased
prevalence of walking for transportation among participants with no TM access in
2005 and with access in 2010 (Group 2) and among those with no TM access in
either 2005 or 2010 (Group 3).
We hypothesize that this increment is related to three factors: (i) increased
private transportation costs, (ii) restrictive car-use policies, and (iii) increased travel
times due to a growth in car and motorcycle ownership in the face of a limited
transport infrastructure. Regarding the first, from 2005 to 2010, Bogota´s gasoline
prices increased more than 80%, from USD $0.61 to USD $1.11 per liter 137. This
cost is not only higher than in the United States, Canada, and Mexico; it ranks among
the highest in the Americas 138. Second, government policies to control car use
strengthened from a twice-weekly partial restriction (from 6:30 a.m. to 9:00 a.m. and
from 3:00 p.m. to 7:00 p.m.) to a twice-weekly stricter restriction (from 6:00 a.m. to
8:00 p.m.). Third, because the infrastructure for handling increased traffic is limited,
52
the growth in car ownership, from 15.5% in 2005 to 25.8% in 2010 in our cohort
(Table 3), has decreased speeds and increased travel times. More specifically,
average travel speed by car in Bogotá decreased from 32.2 km/h in 2005 to 23.6
km/h in 2010 139, making car and bus trips less attractive than walking short
distances. TM, with its exclusive corridors and average speeds of 28 km/h 120, is an
attractive and time-effective option.
Our estimates could be useful for a sensitivity analysis to estimate the TM/PA
promotion cost-benefit ratio over the 2000-2013 period. This value denotes the
return in medical costs savings attributed to walking for transport for TM users for
every dollar spent on the TM average annual costs. First, to calculate the average
annual TM costs, we discounted the sum of the average construction (71%) and
operational (29%) costs in 2013 U.S. dollars by assuming a 13-year period of TM
usage (2000-2013). Considering the calculated costs, the number of users meeting
PA recommendations, and the adjusted PR (Table 2), we estimated the health-care
savings from transportation-related PA to be between 3% and 22% of every dollar
spent on TM costs. Every TM user has a difference in the expected value of medical
costs savings attributed to meeting the PA recommendations by walking for
transportation of USD $10 per year compared to a non-user 133, which corresponds
to the portion of the direct health benefit for PA in Colombia attributed to PA for TM
usage 134. According to the number of physically active users attributable to TM, we
estimated an annual direct health benefit derived from PA and attributed to TM of
between USD $3,310,278 and USD $22,068,518, which corresponds to
approximately USD $2.63-USD $17.55 per user per year. This benefit represents
between 1% and 5% savings on the average annual medical costs per user.
53
Nevertheless, the association between BRT system growth and an increased
prevalence of walking for transportation in geographic units without TM access did
not hold for adults with TM access in 2005 and greater access in subsequent years
(i.e., an increased number of TM stations). We are currently assessing this trend
using simulations of projected TM lines in agent-based models used in public health
evaluations 140 141.
Our study included several limitations. First, the sample size used in the post
hoc cohort analysis and the fact that TM has been in service since 2000.
Nevertheless, the consistency between our cross-sectional and cohort results
provides evidence for the TM/walking for transportation association. To assess these
associations, projected BRTs like that in Rio de Janeiro should leverage the natural
experiment by measuring walking behaviors before and after BRT implementation.
Future studies might also assess multimodalities through which the interaction
between BRT and active transport could be enhanced. Second we do not have
constructs in our survey to better understand preferences towards walking for
transport.
Third biases in the sample include a higher response rate among women than
man But all models were adjusted by sex and the sample focused on low to middle
SES populations which represent 90% of the population. Finally, in terms of PA
measurement, because the IPAQ captures 10-minute bouts, and shorter transit trips
are not captured, future research might also consider incorporating objective
measures of PA, such as accelerometry, activity diaries, and global positioning
systems (GPS) to more accurately measure walking.
54
4. The effect of increased public transport infrastructure on walking for
transportation
So far we have shown in the cross sectional analysis that TM users were more
likely to report walking for transport ≥150 minutes/week compared to non-users, and
therefore more likely to meet physical activity recommendations.
In addition, the objective measurement of PA with accelerometers showed an
average difference of 12 MVPA minutes per day between TM users and non-users,
with the largest differences during peak transportation hours.
In the cohort analysis the relation between access to TM and walking for
transport was evaluated, showing a general increase in walking for transport from
2005 to 2011 in Bogotá.
In 2005 and 2010-11, neighborhoods with access to TM showed a higher
prevalence of walking for transport. And neighborhoods that gained access to TM in
2010 showed the largest increase in walking for transportation, compared to the
neighborhoods that already had TM and those that never had access to it.
This study, so far underscores the potential importance of BRT in increasing
walking for transport. It provides evidence of the benefit of incorporating
considerations of increasing walking into public transportation planning in order to
increase physical activity and prevent NCDs in the world’s rapidly urbanizing cities
11.
However, there is a need to examine if the stated hypothesis are sufficient to
explain the growth in walking for transportation and if substantial increases in BRT
access and use lead to an increase in PA 142. This is a complex problem, that
55
requires accounting for the distribution of people into neighborhoods, capture the
dynamical relationships (feedback loops) and interactions between the environment
and individuals 143, and an operational approach to understand how these dynamic
processes shape the distribution of health outcomes 16. Agent Based models (ABM)
represent an opportunity to analyze these complex relations since they help
understand the system in terms of the actual decision making processes driven by
actors 16. They naturally consider the relation between agents and the environment,
include feedback loops and emergent phenomena appears. ABMs have been used
to investigate ways in which physical environments contribute to health related
behaviors and to understand the impact of policy alternatives in the presence of
nonlinear relationships and feedback 143. Although ABMs have been used to study
pedestrian movement 143,144,145 applications in public health are scarce 143 and none
that we are aware of model the relation between walking for transportation and BRT
systems.
4.1. Agent Based Model
For the present case, we construct a hypothetical city inspired by Bogota, Colombia.
The model, which is applied at the fully-disaggregated level of persons can therefore
be considered an activity based travel demand model 146. The model considers
agents that must transverse different commuting paths according to some transport
matrix obtained from a statistical distribution. In this particular case, the model was
calibrated using Bogota´s mobility survey 122 to analyze the influence of the current
bus system and test the growth of the BRT system on the prevalence of walking for
transportation. In the case of Bogota, the BRT system is call “Transmilenio” (TM).
56
We will show that the model has the capacity to explain patterns observed in the real
world, e.g., in Bogota's mobility survey 122.
4.2. Model Development
The model was developed in Matlab® (Mathworks, Massachussets). It is a time
discrete model with a day time step. The model includes only adults on working days
that travel twice a day. Agents decide their mode of transportation according to: i)
time required to reach the destination and ii) cost among available modes of
transport 147. The model is based on two characteristics of commuting trips: they are
always to the same place 148 and they occur every day working 149.
4.2.1. Setting
The hypothetical city in the model is inspired by Bogotá, the capital city of Colombia
with an estimated population of 7.4 million. Trips 15 minutes or longer are distributed
as follows: 41% with public transport, 28% on foot, 14% in automobile, 5% by taxi,
and the remainder using other modes of transportation 122. The model represents a
city of 100 𝐾𝑚2 that is mapped to a 100 𝑥 100 grid space, so that each cell represents
a block of size 100 𝑥100 𝑚𝑡𝑠. The distribution of Socio Economic Stratum (SES) is
based on Bogota´s distribution 122: (a) lowest SES on the periphery, (b) SES two and
three in the south, and (c) SES four five and six in the north. The number of blocks
per SES and the number of agents per block are proportional to Bogotá. As in Bogotá
jobs were concentrated in the central east part of the city. The distance between two
points in the city is calculated using the Manhattan distance, which corresponds to
the simple sum of the absolute value of the difference of horizontal and vertical
components of each position in the grid. Routes and stops of public buses (not
associated with TM) and TM are established in the grid according to each scenario.
57
To calibrate and assess the model a scenario with 4 TM lanes and 29 bus lanes was
created (Figure 1). This is equivalent to the current number of TM lanes, however
the number of bus routes in Bogotá is much larger.
Figure 10. Bogota's inspired city of 100〖km〗^2 with four TM lanes, 29 bus lanes, distribution of SES, and
centralized work zone based on Bogotá's survey
4.2.2. Agents
The population that moves through the simulated city is 188.300 individuals. Each
individual is assigned a home and a place to work and is given the possibility of
choosing a car with a probability that depends on the SES 150 (see Table 1). The
neighborhood in which the individual is assigned determines the SES and therefore
the money the agent receives which is equal to the average portion of income
associated to transportation per SES 151 (see Table 1). If agents do not use all of
58
their available resources for a given day, they may use up to 50% of their savings
the next day for transportation.
Table 2. Percentage of agents, average daily expenditure on transport, percentage of daily budget spent on
transport, probability of owning a car, and α for each SES
SES
Percentage
of agents
Income per
person per day
for
transportation
only US $
Percentage
spent on
transport
Probability of
having a car
Relative
importance
of resources
α
SES 1 9.5% $1,85 12,4% 0.04 1
SES 2 40.3% $1,97 9,9% 0.05 0,8
SES 3 36.2% $3,08 9,0% 0.17 0,72
SES 4 9.2% $5,04 7,0% 0.49 0,56
SES 5 3.0% $6,16 5,0% 0.64 0,40
SES 6 1.5% $7,70 5,0% 1.00 0,40
4.2.3. Modes of transportation
For simplicity we take 5 modes of transportation, which are the most important
modes of transportation in Bogotá, namely, car, taxi, bus, TM, and walking. As
mentioned above, agents choose the mode of transportation according to the total
trajectory costs (c) and time (t) needed to cover the distance (see Table 2). Each
mode of transportation has its own characteristic speed. For the modes car, bus and
walking the time needed to get to the destination is the waiting time (𝑤𝑡) plus the
59
time required to cover the distance from home to work 𝑑ℎ𝑤 at the particular mode´s
speed. For the bus and the TM the distances 𝑑ℎ𝑠 (home – station) and 𝑑𝑤𝑠 (work –
station) needed to walk to and from the nearest station, respectively is taken into
account, this is the access time (see table 5).
To calculate the cost of a trip, we use the fixed cost of the bus or TM ticket and the
variable costs from the car and taxi (see Table 5).
Table 3. Costs, access time and speed of the different modes of transportation.
Mode Cost (US dollars) Speed
(km/h)
Access time (min) Waiting time
(min)
Car 0,50* 𝑑ℎ𝑤 26,88 0 2
Taxi 0,83+0,33*𝑑ℎ𝑤
Minimum $1,71
22,38 0 5
Bus 0,7 18,28 13,69(𝑑ℎ𝑠 + 𝑑𝑤𝑠) 10
TM 0,85 27,96 13,69(𝑑ℎ𝑠 + 𝑑𝑤𝑠) 10
Walking 0 4,38 0 0
4.2.4. Mode decision
For each agent of a given SES (Table 2) we calculate the probability of taking the
mode 𝑘, as a function of the time and resources needed to arrive at the destination.
60
We define the probability 𝑝(𝑘) of taking the mode 𝑘 is a weighted geometric average
of the influence of time 𝑢𝑘 and resources 𝑟𝑘 given by
2)( kk urkp (1)
We assume that α, reflects the relative weight agents assign to money when
choosing a transportation mode. Different SES have a different α (Table 2). We
conducted a sensitivity analysis on the α values, but we did not find a significant
effect on walking patterns.
We first compute the resource-based index (𝑟𝑘) of using mode 𝑘 according to the
daily resources and the savings accumulated:
(2)
Where 𝐺 is the set of available modes (restricted by the available resources), 𝑆 is
the amount of savings that an agent has from the past, 𝐼 is the daily’s income for the
agent and 𝐶𝑘 is the cost of traveling in mode 𝑘.
Second, since exponential forms yield a better fit to the distribution of walking trips
over distances152, the time-based index 𝑢𝑘 of taking mode 𝑘 is inversely proportional
to the exponent of the total time 𝑡 expected to get to the destination.
(3)
Gk k
k
k
CIS
CISr
)(
Gk
t
k
t
k
k
e
eu
61
From these three equations the probability 𝑝(𝑘) of taking the mode𝑘 is established,
therefore, at each time step we can use a random number between 0 and 1 to
choose a particular transportation mode for each agent.
4.2.5. Calibration
Model assessment had three parts. First, model output for walking behavior was
compared to Bogotá´s distance decay for walking calculated from Bogota´s mobility
survey 122. Distance decay functions have been used in geography to mathematically
describe how a given phenomenon varies as a function of distance 153. According to
Fortheringham 154 estimates of distance decay are functions of spatial structure as
well as interaction behavior.
The second step was to compare the prevalence of walking according to distance
and SES, empirical studies have found that lower SES tend to walk more for
transportation 155 156. Finally, the third step was to calibrate the use of the transport
modes by SES and to account for the fact that Bogotá is a centralized city where
jobs are found throughout but are concentrated in the central east part 157. To this
end, jobs were distributed 50% in the central area and 50% uniformly within 20
blocks of each agent´s home except for the SES 6 were 50% of the jobs were
distributed in the central area and 50% within 10 blocks of home given the ability of
these people to live near their jobs. Commuting by foot depends largely on closeness
between home and work 106. As we will see below, with this setup, a pattern similar
to Bogota´s mode distribution per SES 122 was achieved with the model.
Since walking levels became stable after 10 days each scenario and the calibration
was run four times for forty days and the last 30 days were used for the analysis.
62
4.2.6. Model assessment results for the calibration scenario
A specific distance-decay function fitted to real data is a precise description of the
distribution of walking trips over distances 152. The calibration model results were
similar to Bogota´s Distance-decay functions calculated from the survey (Figure 16).
Figure 11. Distance decay function for walking obtained in the calibration model compared to Bogotas
distance decay for walking to
Consistent with other studies 155 156 lower SES have a higher prevalence of walking
long distances. Twenty percent of agents from SES six with a 5 minute walk to work
decided to walk. Similarly, in the case of the SES one 75% of agents with a 10 minute
walk made the same decision. (Figure 17).
63
Figure 12 Prevalence of walking per travel time per SES in the calibration test case
Figure 18 compares mode distribution per SES between Bogota and the model. In
both cases a higher tendency of walking for transport and a higher likelihood to use
public transport can be seen for low SES and a higher likelihood to use the car and
taxi from the higher SES. The biggest difference between the model and Bogotá is
in the use of the taxi.
64
Figure 13. Comparison mode distribution per SES for Bogotá and the ABM model for the calibration case.
Bogotá´s mode of transportation
distribution for trips from home to
work per SES taking into account
only walking bus BRT taxi and car 122.
Model´s mode of transportation
distribution per SES in the calibration
test
As can be seen from figures 2, 3 and 4, the model produces patterns that have been
observed in empirical studies: a similar distance decay function, the prevalence of
walking according to SES and mode distribution per SES.
4.2.7. Scenarios
In order to understand TM and bus influence on walking for transportation, we
created two types of scenarios. In the first type, we simulated the growth of the BRT
system through increase number of lanes which was varied from 0 to 10 lanes. BRT
lanes replaced bus lanes therefore the number of bus lanes was decreased from 33
to 23 lanes
0%
20%
40%
60%
80%
100%
SES 1 SES 2 SES 3 SES 4 SES 5 SES 6
Walking Bus BRT Taxi Car
0%
20%
40%
60%
80%
100%
SES 1 SES 2 SES 3 SES 4 SES 5 SES 6
Walking Bus BRT Taxi Car
65
Table 4. Average distance from home to the closest bus and BRT station plus average distance from work
to the closest BRT and bus station according to number of lanes
Number of Bus
Lanes
33 31 29 27 25 23
Number of BRT
Lanes
0 2 4 6 8 10
Bus (𝑑ℎ𝑠 + 𝑑𝑤𝑠 ) 1.92 2.26 2.44 2.62 2.91 3.17
BRT (𝑑ℎ𝑠 + 𝑑𝑤𝑠 ) - 29.98 17.54 12.32 10.10 9.00
For the second type of scenarios we randomly distributed throughout the city Bus
and BRT stations with different densities. The densities tested were 0%, 3%, 7%,
10%, 13%, 17%, and 20%. For example in the scenario with 3% bus density and 7%
BRT each block in the city has a 3% probability of having a bus station and a 7%
probability of having a BRT station. The densities are related to the number of blocks
an agent must walk to get to a bus or BRT station (Table 3). In all cases the same
distribution of agents and jobs throughout the grid, and the parameters of velocity,
price and income stay the same.
Table 5. Average distance from home to the closest bus or BRT station plus average distance from work to
the closest BRT and bus station
Density 0% 3% 7% 10% 13% 17% 20%
BRT/Bus (𝑑ℎ𝑠 + 𝑑𝑤𝑠 ) - 6.8 4.6 3.6 3.1 2.7 2.1
66
4.3. Results from the different scenarios
Scenarios with the added BRT lanes
It is interesting to note that increasing the number of TM lanes produces an
increment in the number of minutes walking for transportation.
Figure 14. Walking minutes TM users vs non users depending on the number of TM lanes.
Figure 19 shows an increased number of minutes of walking for transportation as
the number of TM lanes increases. This increment can also be seen among Non-
users of TM. These results show that walking time saturates as TM lanes are added
to the system. Average walking minutes from TM users increases one minute
between two TM lanes and four TM lanes, while the increment is only 0.5 minutes
between 8 TM lanes and 10 TM lanes. A similar behavior is observed among non-
11.412.4
13.4 13.7 14.1
5
6.5 6.7 6.8 7 7
5
78
99.9 10.4
0
2
4
6
8
10
12
14
16
0 2 4 6 8 10
Walking minutes TM users
Walking minutes TM Non-users
City´s average minutes of walking for transport
Number of TM lanes
67
users of TM. The introduction of the first two TM lanes had an increment of 1.5
minutes, while between 8 BRT lanes and 10 BRT lanes there was practically no
increment. The city´s average minutes of walking for transport goes from 5.0 to 10.4.
Again the biggest increment happens with the introduction of the first two lanes and
slowly diminishes.
Scenarios with random distribution of BRT and bus stations
Figure 6a is the three dimensional space average minutes walking for transport
according to the TM and Bus densities evaluated, Figure 6b has the same
information but only in two dimensions. These figures show that without bus and TM
agents must walk great distances. The average number of minutes walking in a
scenario without TM and bus goes up to 93.4 minutes. Prevalence of walking in SES
1 and 2 is above 90% in SES 3 is above 50% in higher SES (4,5 and 6) the
prevalence of walking is similar to the calibration scenario. As public transport
becomes available the number of walking minutes decreases. However, it is
important to note that that the increase of public transport does not necessarily
decreases walking for transport. The introduction of BRT at a 3% density in
scenarios that already have bus densities above 3% increments the number of
minutes of walking per transport, with a 3% density and below the introduction of TM
lanes reduces the number of minutes walking for transport. Higher densities of BRT
decrease walking for transport in all scenarios.
68
Figure 15. Average minutes of walking for transport according to TM and bus densities
a)
b)
4.4. Discussion
This simple model is able to reproduce behaviors that have been observed in other
studies such as a high probability of walking for short trips and a low probability of
walking long trips155 156. We also reproduced the higher willingness from the lower
69
SES (24), and the difference in walking between users and non-users of TM, as has
been reported by other empirical studies (7) (10) (11) (12) .
Model results suggests that walking for transport emerges from the relation between
people with their built transportation environment and the existing modes of
transportation. Hence, this emergent behavior could be used to design transport
systems in cities that increase the time users spent walking in their way to work, in
an attempt to reach reasonable levels of PA. For example in our hypothetical city
moving from 2 TM lanes to 4 lanes increased the average walking time for the city
in 1 minute per day (See figure 5). As such, combining large surveys and agent
based models, we have included some of the robust elements that may shape the
distribution of time spend walking by the users and the shifting of transport system
in a city such as Bogota. The results show that certain intuitive perturbations of the
system, may produce unexpected results, like a threshold number of TM above
which walking time does not increase. Also in figure 6 we found that above 3% TM
density, regardless of the number bus stops, walking time diminishes.
Results are consistent with the increment in walking for transport that has been
observed due to the construction and use of TM or similar systems (12) (13)(27).
This increment is also present in non-users of TM due to the fact that TM lanes
replace regular bus lanes. The increase in PA is expected to continue as more TM
lanes are added but with decreasing benefits, unless changes to transportation
environment are implemented such as a strong reduction of the non TM bus system.
70
The model also shows that the PA brought by BRT is related to other existing modes.
According to the model’s results, Figure 6b shows that BRT increases PA but only
in situations where bus densities are high (e.g. Bogotá). At Bus stops densities below
3% increased TM produces lower walking times. At bus densities above 3%, the
increased PA activity is due to BRT’s velocity, which is higher than all other modes
of transport available in the city. Its high speed makes it an attractive mode of
transportation worth walking a few extra blocks to save time. This result explains
why in cities with low bus density BRT may decrease PA. The model also stresses
the importance of high accessibility to public transport, especially for low SES. In
Figure 5 when TM and bus densities were low; walking for transport increased
substantially, especially among the lowest SES that live furthest from the city center
which is where most jobs are located. Of course, walking such distances has been
related to a low quality of life (8).
According to the model it is expected that the construction of a BRT network will
continue leading to more minutes of PA as long as pedestrians walk more than 700
meters per trip. This result is within international standard for accessibility which are
usually between 400 and 800 meters160. Inaccessibility or bad accessibility of public
transport means that the distance or time to walk to access public transport terminal
is longer.
The model, however, has some limitations. Bogotá is much larger than our
hypothetical city. Therefore trips are shorter and thus taxi and car trips are less
expensive, which may partly explain the large participation of the taxi in our results.
71
Mode choice is solely based upon resources and time but other variables such as
comfort, weather and safety might play an important role (9). These factors actually
could explain the small participation of motorcycle and bicycles in Bogotá despite of
being inexpensive and fast 151 162. Another limitation is that we assumed that all
agents from the same SES get the same amount of money for transport equal to the
average income (only the portion associated to transportation) per SES in Bogotá.
This amount of money is enough to have access to the bus or TM in any day, but
this is not necessarily true for a city like Bogotá where 10,2% of Bogotá´s population
lives under the poverty line (income below 100 US dollars per month)163, as a result
they do not have enough resources to always access public transportation. This is
most frequent among the lowest SES who live in the periphery and are therefore
trapped or bound to walk great distances. This use of averages partly explains why
in figure 2 the model´s distance decay function is steeper than Bogotá´s distance
decay for large walking times. This difference in distance decay functions implies
that results from the model underestimate people’s willingness to walk specially in
low SES.
Finally, Agent Based Models (ABM) give the opportunity to have a deeper
understanding of how physical environment affects attitudes toward walking and the
actual probability of walking since they are based on understanding the system in
terms of actual decision making processes driven by actors and their environment.
This understanding may prove helpful in designing policies to increase walking and
PA without harming people´s quality of life.
72
5. Final remarks and future work
Chapters 2, 3 and 4 are based on articles. Therefore, each chapter makes a
contribution and stands for itself. Chapter one offers results to observe changes on
body composition. These results show that obesity is growing. In terms of this
document the object of the chapter is to show that overweight and obesity are a
growing concern.
The second chapter presents a large international study with children with strict
protocols so that comparisons among countries may be made. The large sample
allows great variability and therefore to establish better estimators of the effect of
active transport on body composition indicators. Results show that there is great
variance in terms of the modes of transport used to school. However despite this
variation in transport mode use, the effect of active transport does not vary between
countries or sex. This stable relation between active transport and body composition
had not been presented before. This chapter also offers a connection between
overweight and obesity and transport. The pathway of this association may occur in
part through small increments of everyday levels of PA
Chapter three presents associations between access and use of TM with walking
more than 150 minutes per week. This chapter establishes through a cohort study
that walking for transport has increased from 2005 to 2010. The natural experiment
presented in this chapter strengthens the relation between BRT and walking for
transport. The contribution of this chapter is in presenting this relation through a
natural experiment, stepping one step further on establishing causality between TM
and walking for transport.
73
Chapter four presents an ABM based on the actual decision making process. This
chapter shows that the entrance of TM with a high access to bus in an environment
in which transport decisions are based on resources and time are sufficient
conditions to increment the number of minutes walking. The model also shows
through the randomized environment that it is not the distance between the stations
but the speed of the TM that makes agents walk more. In this sense the chapter
offers an explanation of the observed data, which constitutes knowledge and offers
understanding since we offer a simple set of rules that explain the observed relation
in chapter three.
The document as a whole presents a small contribution on the old debate of non-
operational and an operational modes of reasoning. On one hand non-operational
approaches to gain knowledge such as the associations presented on chapter 3
have been criticized. These associations are made through a very sophisticated
model designed to establish the variables that influence walking for transport in
Bogota. The model contains a raft of variables woven together in a complex
equation. But nowhere in the model did trips appear. If one asks how walking for
transport is actually generated, one discovers that trips are essential. This could
have perfectly been the example for Olaya´s article. In Olaya´s words this
associations “are not explained beyond some obscure force of mechanism (scientist
like the term law) that the researcher, passively instructed by data, attempts to
uncover […] In a changing world this obscurity entails a larger issue: induction is
perhaps the most contradictory way to source knowledge in a non-uniform world.
[…] It is easy to appreciate the unreliable way to meet a changing world through
analysis of data”
74
On the other hand Operational approaches like the one presented in chapter five
have also been criticized. First they involve simulation models. Therefore, they
cannot be classified as formally inductive, and as such, researchers cannot be
confident of the range over which hypothesized relationships between model
parameters and outcomes hold. Second, operational approaches generally
incorporate complex dynamic relationships, implying that many possible outcomes
are possible, and that small changes in parameter values and initial conditions may
result in large changes in model outcomes. Finally decisions rules are established
by the modeler, different decision rules may produce similar results never knowing
how accurate decision rules are under different circumstances.
This document uses both approaches to understand a system. I propose that both
approaches are essential to gaining knowledge. The associations between use and
access of BRT presented in chapter four without an explanation of why they occur
would be mere descriptions of the system and not knowledge and might lead
someone to believe that more TM means more walking for transport. Furthermore
The ABM without TM or the established association with walking would not have
anything to explain and would only become of value when the data became
available. This happened to Einstein´s and Darwin´s theories. They became useful
and of value when there was data to explain. Which are essential conditions in
engineering.
Future research should focus on operationalizing the first two associations
presented in this document. Chapter two presents a trend in the world of growing
overweight and obesity, and others have established a set of associations. However,
we do not know in terms of the actual decision making processes why this is
75
happening. Chapter three also presents an association between active transport and
body composition variables. The mechanism for this association has not been
completely explained. Finally there are two paths of improvement for the model. First
thanks to the available data we know the model is not perfect. This could be due to
the initial parameters and future work might run the model with more accurate
parameters (more realistic city, better income distribution). The second path is to
prove the rules under different environments. For example cities in the U.S. have
high incomes, large distances between jobs and homes and a high access to car.
According to the model´s predictions these cities have a low prevalence of walking
for transport. It would be interesting to test the model with data from other cities.
76
References
1. The Global Burden of Disease: Generating Evidence, Guiding Policy. at
<http://www.healthdata.org/policy-report/global-burden-disease-generating-evidence-
guiding-policy>
2. Lee, I.-M. et al. Effect of physical inactivity on major non-communicable diseases
worldwide: an analysis of burden of disease and life expectancy. Lancet 380, 219–229
(2012).
3. WHO, W. H. Global Health Risks: Mortality and Burden of Disease Attributable to
Selected Major Risks. (World Health Organization, 2009).
4. Instituto Colombiano de Bienestar Familiar (ICBF). Encuesta Nacional De La
Situación Nutricional En Colombia 2010. (Instituto Colombiano de Bienestar
Familiar;, 2011). at
<http://www.icbf.gov.co/portal/page/portal/PortalICBF/NormatividadC/ENSIN1>
5. Giles-Corti, B., Foster, S., Shilton, T. & Falconer, R. The co-benefits for health of
investing in active transportation. New South Wales Public Health Bull. 21, 122–127
(2010).
6. Hamer, M. & Chida, Y. Active commuting and cardiovascular risk: a meta-analytic
review. Prev. Med. 46, 9–13 (2008).
7. Gordon-Larsen, P. et al. Active commuting and cardiovascular disease risk: the
CARDIA study. Arch. Intern. Med. 169, 1216–1223 (2009).
8. Besser, L. M. & Dannenberg, A. L. Walking to public transit: steps to help meet
physical activity recommendations. Am. J. Prev. Med. 29, 273–280 (2005).
77
9. Freeland, A. L., Banerjee, S. N., Dannenberg, A. L. & Wendel, A. M. Walking
associated with public transit: moving toward increased physical activity in the United
States. Am. J. Public Health 103, 536–542 (2013).
10. Sallis, J. F. Measuring physical activity environments: a brief history. Am. J. Prev.
Med. 36, S86–92 (2009).
11. Kahlmeier, S., Racioppi, F., Cavill, N., Rutter, H. & Oja, P. ‘Health in all policies’ in
practice: guidance and tools to quantifying the health effects of cycling and walking.
J. Phys. Act. Health 7 Suppl 1, S120–125 (2010).
12. Sallis, J. F. Environmental and Policy Research on Physical Activity is Going Global.
Res. Exerc. Epidemiol. 13, 111 – 117 (2011).
13. Cervero, R., Sarmiento, O. L., Jacoby, E., Gomez, L. F. & Neiman, A. Influences of
Built Environments on Walking and Cycling: Lessons from Bogotá. Int. J. Sustain.
Transp. 3, 203–226 (2009).
14. Ewing, R. & Cervero, R. Travel and the Built Environment. J. Am. Plann. Assoc. 76,
265–294 (2010).
15. Adams, J. Prevalence and socio-demographic correlates of ‘active transport’ in the
UK: analysis of the UK time use survey 2005. Prev. Med. 50, 199–203 (2010).
16. Olaya, C. Models that Include Cows: The Significance of Operational Thinking. in
(2012). at
<http://www.systemdynamics.org/conferences/2012/proceed/papers/P1375.pdf>
17. Valente, M. Knowledge as Explanations Methodological implications for simulation
modeling. (2011). at
<http://www.ie.ufrj.br/oldroot/datacenterie/pdfs/seminarios/pesquisa/texto0908.pdf>
78
18. Chaitin, G. The limits of reason. (2006). at
<http://www.academia.edu/5793776/The_limits_of_reason>
19. Ng, M. et al. Global, regional, and national prevalence of overweight and obesity in
children and adults during 1980–2013: a systematic analysis for the Global Burden of
Disease Study 2013. The Lancet (2014). doi:10.1016/S0140-6736(14)60460-8
20. Finucane, M. M. et al. National, regional, and global trends in body-mass index since
1980: systematic analysis of health examination surveys and epidemiological studies
with 960 country-years and 9·1 million participants. Lancet 377, 557–567 (2011).
21. Roth, J., Qiang, X., Marbán, S. L., Redelt, H. & Lowell, B. C. The obesity pandemic:
where have we been and where are we going? Obes. Res. 12 Suppl 2, 88S–101S
(2004).
22. Calle, E. E., Thun, M. J., Petrelli, J. M., Rodriguez, C. & Heath, C. W. J. Body-mass
index and mortality in a prospective cohort of U.S. adults. N. Engl. J. Med. 341,
1097–1105 (1999).
23. Wilson, P. W. F., D’Agostino, R. B., Sullivan, L., Parise, H. & Kannel, W. B.
Overweight and obesity as determinants of cardiovascular risk: the Framingham
experience. Arch. Intern. Med. 162, 1867–1872 (2002).
24. Fontaine, K. R., Redden, D. T., Wang, C., Westfall, A. O. & Allison, D. B. Years of
life lost due to obesity. JAMA J. Am. Med. Assoc. 289, 187–193 (2003).
25. Renehan, A. G., Tyson, M., Egger, M., Heller, R. F. & Zwahlen, M. Body-mass index
and incidence of cancer: a systematic review and meta-analysis of prospective
observational studies. Lancet 371, 569–578 (2008).
26. Institute for Health Metrics and Evaluation. The Global Burden of Disease:
Generating Evidence, Guiding Policy. (IHME, 2013). at
79
<http://www.healthdata.org/policy-report/global-burden-disease-generating-evidence-
guiding-policy>
27. Luke, A. & Cooper, R. S. Physical activity does not influence obesity risk: time to
clarify the public health message. Int. J. Epidemiol. 42, 1831–1836 (2013).
28. Blair, S. N., Archer, E. & Hand, G. A. Commentary: Luke and Cooper are wrong:
physical activity has a crucial role in weight management and determinants of obesity.
Int. J. Epidemiol. 42, 1836–1838 (2013).
29. Lobstein, T. et al. Child and adolescent obesity: part of a bigger picture. The Lancet
(2015). doi:10.1016/S0140-6736(14)61746-3
30. Rivera, J. Á. et al. Childhood and adolescent overweight and obesity in Latin
America: a systematic review. Lancet Diabetes Endocrinol. 2, 321–32 (2014).
31. Muthuri, S. K. et al. Evidence of an overweight/obesity transition among school-aged
children and youth in Sub-Saharan Africa: a systematic review. PloS One 9, e92846
(2014).
32. Hallal, P. C. et al. Global physical activity levels: surveillance progress, pitfalls, and
prospects. Lancet 380, 247–57 (2012).
33. Popkin, B. M. The nutrition transition: an overview of world patterns of change. Nutr.
Rev. 62, S140–3 (2004).
34. Katzmarzyk, P. T. & Mason, C. The physical activity transition. J. Phys. Act. Health
6, 269–80 (2009).
35. Buliung, R. N., Mitra, R. & Faulkner, G. Active school transportation in the Greater
Toronto Area, Canada: an exploration of trends in space and time (1986-2006). Prev.
Med. 48, 507–12 (2009).
80
36. McDonald, N. C. Active transportation to school: trends among U.S. schoolchildren,
1969-2001. Am. J. Prev. Med. 32, 509–16 (2007).
37. van der Ploeg, H. P., Merom, D., Corpuz, G. & Bauman, A. E. Trends in Australian
children traveling to school 1971-2003: burning petrol or carbohydrates? Prev. Med.
46, 60–2 (2008).
38. Grize, L., Bringolf-Isler, B., Martin, E. & Braun-Fahrländer, C. Trend in active
transportation to school among Swiss school children and its associated factors: three
cross-sectional surveys 1994, 2000 and 2005. Int. J. Behav. Nutr. Phys. Act. 7, 28
(2010).
39. Costa, F. F., Silva, K. S., Schmoelz, C. P., Campos, V. C. & de Assis, M. A. a.
Longitudinal and cross-sectional changes in active commuting to school among
Brazilian schoolchildren. Prev. Med. 55, 212–214 (2012).
40. Larouche, R., Lloyd, M., Knight, E. & Tremblay, M. S. Relationship between active
school transport and body mass index in grades 4-to-6 children. Pediatr. Exerc. Sci.
23, 322–30 (2011).
41. Larouche, R., Saunders, T. J., Faulkner, G. E. J., Colley, R. & Tremblay, M.
Associations between active school transport and physical activity, body composition,
and cardiovascular fitness: a systematic review of 68 studies. J. Phys. Act. Health 11,
206–227 (2014).
42. Faulkner, G. E. J., Buliung, R. N., Flora, P. K. & Fusco, C. Active school transport,
physical activity levels and body weight of children and youth: a systematic review.
Prev. Med. 48, 3–8 (2009).
81
43. Lee, M. C., Orenstein, M. R. & Richardson, M. J. Systematic review of active
commuting to school and childrens physical activity and weight. J. Phys. Act. Health
5, 930–49 (2008).
44. Heelan, K. A. et al. Active commuting to and from school and BMI in elementary
school children-preliminary data. Child Care Health Dev. 31, 341–9 (2005).
45. Madsen, K. A., Gosliner, W., Woodward-Lopez, G. & Crawford, P. B. Physical
activity opportunities associated with fitness and weight status among adolescents in
low-income communities. Arch. Pediatr. Adolesc. Med. 163, 1014–21 (2009).
46. Kong, A. S. et al. A pilot walking school bus program to prevent obesity in Hispanic
elementary school children: role of physician involvement with the school
community. Clin. Pediatr. (Phila.) 49, 989–91 (2010).
47. Saksvig, B. I. et al. Travel by walking before and after school and physical activity
among adolescent girls. Arch. Pediatr. Adolesc. Med. 161, 153–8 (2007).
48. Sirard, J. R., Riner, W. F., McIver, K. L. & Pate, R. R. Physical activity and active
commuting to elementary school. Med. Sci. Sports Exerc. 37, 2062–9 (2005).
49. Heelan, K. A., Abbey, B. M., Donnelly, J. E., Mayo, M. S. & Welk, G. J. Evaluation
of a walking school bus for promoting physical activity in youth. J. Phys. Act. Health
6, 560–7 (2009).
50. Rosenberg, D. E., Sallis, J. F., Conway, T. L., Cain, K. L. & McKenzie, T. L. Active
transportation to school over 2 years in relation to weight status and physical activity.
Obes. Silver Spring Md 14, 1771–6 (2006).
51. Mendoza, J. A. et al. Active commuting to school and association with physical
activity and adiposity among US youth. J. Phys. Act. Health 8, 488–95 (2011).
82
52. Pabayo, R., Gauvin, L., Barnett, T. A., Nikiéma, B. & Séguin, L. Sustained active
transportation is associated with a favorable body mass index trajectory across the
early school years: findings from the Quebec Longitudinal Study of Child
Development birth cohort. Prev. Med. 50 Suppl 1, S59–64 (2010).
53. Wen, L. M., Merom, D., Rissel, C. & Simpson, J. M. Weight status, modes of travel to
school and screen time: a cross-sectional survey of children aged 10-13 years in
Sydney. Health Promot. J. Aust. Off. J. Aust. Assoc. Health Promot. Prof. 21, 57–63
(2010).
54. Meron, D., Rissel, C., Reinten-Reynolds, T. & Hardy, L. L. Changes in active travel
of school children from 2004 to 2010 in New South Wales, Australia. Prev. Med. 53,
408–10 (2011).
55. Voss, C. & Sandercock, G. Aerobic fitness and mode of travel to school in English
schoolchildren. Med. Sci. Sports Exerc. 42, 281–7 (2010).
56. Baig, F. et al. Association between active commuting to school, weight and physical
activity status in ethnically diverse adolescents predominately living in deprived
communities. Public Health 123, 39–41 (2009).
57. Roth, M. A., Millett, C. J. & Mindell, J. S. The contribution of active travel (walking
and cycling) in children to overall physical activity levels: a national cross sectional
study. Prev. Med. 54, 134–9 (2012).
58. Ford, P., Bailey, R., Coleman, D., Woolf-May, K. & Swaine, I. Activity levels, dietary
energy intake, and body composition in children who walk to school. Pediatr. Exerc.
Sci. 19, 393–407 (2007).
59. Cooper, A. R., Page, A. S., Foster, L. J. & Qahwaji, D. Commuting to school: are
children who walk more physically active? Am. J. Prev. Med. 25, 273–6 (2003).
83
60. Landsberg, B. et al. Associations between active commuting to school, fat mass and
lifestyle factors in adolescents: the Kiel Obesity Prevention Study (KOPS). Eur. J.
Clin. Nutr. 62, 739–47 (2008).
61. Andersen, L. B., Lawlor, D. A., Cooper, A. R., Froberg, K. & Anderssen, S. A.
Physical fitness in relation to transport to school in adolescents: the Danish youth and
sports study. Scand. J. Med. Sci. Sports 19, 406–11 (2009).
62. Østergaard, L. et al. Cycling to school is associated with lower BMI and lower odds of
being overweight or obese in a large population-based study of Danish adolescents. J.
Phys. Act. Health 9, 617–25 (2012).
63. Cooper, A. R. et al. Longitudinal associations of cycling to school with adolescent
fitness. Prev. Med. 47, 324–8 (2008).
64. Cooper, A. R., Andersen, L. B., Wedderkopp, N., Page, A. S. & Froberg, K. Physical
activity levels of children who walk, cycle, or are driven to school. Am. J. Prev. Med.
29, 179–84 (2005).
65. Andersen, L. B. et al. Cycling to school and cardiovascular risk factors: a longitudinal
study. J. Phys. Act. Health 8, 1025–33 (2011).
66. Cooper, A. R. et al. Active travel to school and cardiovascular fitness in Danish
children and adolescents. Med. Sci. Sports Exerc. 38, 1724–31 (2006).
67. Klein-Platat, C. et al. Physical activity is inversely related to waist circumference in
12-y-old French adolescents. Int. J. Obes. 2005 29, 9–14 (2005).
68. Jansen, W., Mackenbach, J. P., Joosten-van Zwanenburg, E. & Brug, J. Weight status,
energy-balance behaviours and intentions in 9-12-year-old inner-city children. J. Hum.
Nutr. Diet. Off. J. Br. Diet. Assoc. 23, 85–96 (2010).
84
69. Bere, E., Seiler, S., Eikemo, T. A., Oenema, A. & Brug, J. The association between
cycling to school and being overweight in Rotterdam (The Netherlands) and
Kristiansand (Norway). Scand. J. Med. Sci. Sports 21, 48–53 (2011).
70. Wijga, A. H. et al. Diet, screen time, physical activity, and childhood overweight in
the general population and in high risk subgroups: Prospective analyses in the PIAMA
birth cohort. J. Obes. 2010, (2010).
71. Bere, E., Oenema, A., Prins, R. G., Seiler, S. & Brug, J. Longitudinal associations
between cycling to school and weight status. Int. J. Pediatr. Obes. IJPO Off. J. Int.
Assoc. Study Obes. 6, 182–7 (2011).
72. Aires, L. et al. A 3-year longitudinal analysis of changes in Body Mass Index. Int. J.
Sports Med. 31, 133–7 (2010).
73. Löfgren, B., Stenevi-Lundgren, S., Dencker, M. & Karlsson, M. K. The mode of
school transportation in pre-pubertal children does not influence the accrual of bone
mineral or the gain in bone size--two year prospective data from the paediatric
osteoporosis preventive (POP) study. BMC Musculoskelet. Disord. 11, 25 (2010).
74. Alwis, G. et al. Bone mineral accrual and gain in skeletal width in pre-pubertal school
children is independent of the mode of school transportation--one-year data from the
prospective observational pediatric osteoporosis prevention (POP) study. BMC
Musculoskelet. Disord. 8, 66 (2007).
75. Loucaides, C. A. & Jago, R. Differences in physical activity by gender, weight status
and travel mode to school in Cypriot children. Prev. Med. 47, 107–11 (2008).
76. Tudor-Locke, C., Ainsworth, B. E., Adair, L. S. & Popkin, B. M. Objective physical
activity of filipino youth stratified for commuting mode to school. Med. Sci. Sports
Exerc. 35, 465–71 (2003).
85
77. Tudor-Locke, C., Ainsworth, B. E., Adair, L. S. & Popkin, B. M. Physical activity in
Filipino youth: the Cebu Longitudinal Health and Nutrition Survey. Int. J. Obes.
Relat. Metab. Disord. J. Int. Assoc. Study Obes. 27, 181–90 (2003).
78. Collins, A. E., Pakiz, B. & Rock, C. L. Factors associated with obesity in Indonesian
adolescents. Int. J. Pediatr. Obes. IJPO Off. J. Int. Assoc. Study Obes. 3, 58–64
(2008).
79. Li, Y. et al. Determinants of childhood overweight and obesity in China. Br. J. Nutr.
97, 210–5 (2007).
80. Andegiorgish, A. K., Wang, J., Zhang, X., Liu, X. & Zhu, H. Prevalence of
overweight, obesity, and associated risk factors among school children and
adolescents in Tianjin, China. Eur. J. Pediatr. 171, 697–703 (2012).
81. Silva, K. S., Vasques, D. G., Martins, C. de O., Williams, L. A. & Lopes, A. S. Active
commuting: prevalence, barriers, and associated variables. J. Phys. Act. Health 8,
750–7 (2011).
82. Silva, K. S. & Lopes, A. S. Excess weight, arterial pressure and physical activity in
commuting to school: correlations. Arq. Bras. Cardiol. 91, 84–91 (2008).
83. Arango, C. M. et al. Walking or bicycling to school and weight status among
adolescents from Montería, Colombia. J. Phys. Act. Health 8 Suppl 2, S171–7 (2011).
84. Ojiambo, R. et al. Free-living physical activity and energy expenditure of rural
children and adolescents in the Nandi region of Kenya. Ann. Hum. Biol. 40, 318–23
(2013).
85. World Health Organization. Diet and physical activity: a public health priority.
(2004). at
86
<http://www.who.int/dietphysicalactivity/strategy/eb11344/strategy_english_web.pdf
>
86. United Nations. 2011 High Level Meeting on the Prevention and Control of Non-
communicable Diseases. in (United Nations, 2011).
87. African Development Bank et al. Commitment to Sustainable Transport. (2012).
88. Katzmarzyk, P. T. et al. The International Study of Childhood Obesity, Lifestyle and
the Environment (ISCOLE): design and methods. BMC Public Health 13, 900 (2013).
89. Barreira, T. V, Staiano, A. E. & Katzmarzyk, P. T. Validity assessment of a portable
bioimpedance scale to estimate body fat percentage in white and African-American
children and adolescents. Pediatr. Obes. 8, e29–32 (2013).
90. World Health Organization. Circumference and Waist-Hip Ratio: Report of a WHO
Expert Consultation. (2011).
91. de Onis, M. et al. Development of a WHO growth reference for school-aged children
and adolescents. Bull. World Health Organ. 85, 660–7 (2007).
92. Gropp, K., Janssen, I. & Pickett, W. Active transportation to school in Canadian
youth: should injury be a concern? Inj. Prev. J. Int. Soc. Child Adolesc. Inj. Prev. 19,
64–7 (2013).
93. Barreira, T. V. Reliability of accelerometer-determined physical activity and sedentary
behavior in children: A 12 country study. Int. J. Obes.
94. Giles-Corti, B., Foster, S., Shilton, T. & Falconer, R. The co-benefits for health of
investing in active transportation. New South Wales Public Health Bull. 21, 122–7
95. Kenward, M. G. & Roger, J. H. Small sample inference for fixed effects from
restricted maximum likelihood. Biometrics 53, 983–97 (1997).
96. The World Bank. World Development Indicators. (2011).
87
97. World Health Organization. Global status report on road safety. (2013).
98. Dentro, K. N. et al. Active Transportation and School Day Physical Activity in 9-11
year old Children from 12 Countries: The International Study of Childhood Obesity,
Lifestyle, and the Environment (ISCOLE). Int. J. Obes.
99. Daniels, R. & Mulley, C. Explaining walking distance to public transport: The
dominance of public transport supply. J. Transp. Land Use 6, 5 (2013).
100. Sarmiento, O. L. et al. Quality of life, physical activity, and built environment
characteristics among colombian adults. J. Phys. Act. Health 7 Suppl 2, S181–95
(2010).
101. Salvo, D., Reis, R. S., Sarmiento, O. L. & Pratt, M. Overcoming the challenges of
conducting physical activity and built environment research in Latin America: IPEN
Latin America. Prev. Med. 69, S86–S92 (2014).
102. Larouche, R. et al. Individual and neighborhood correlates of active transportation to
school among 10-year old children from twelve countries: The International Study of
Childhood Obesity, Lifestyle, and the Environment. Int. J. Obes.
103. Secretaria de Educación del Distrito. Movilidad Escolar. Múltiples soluciones para ir a
estudiar. (2015).
104. Broberg, A. & Sarjala, S. School travel mode choice and the characteristics of the
urban built environment : The case of Helsinki , Finland. Transp. Policy 37, 1–10
(2015).
105. Larouche, R., Saunders, T. J., Faulkner, G. E. J., Colley, R. & Tremblay, M.
Associations between active school transport and physical activity, body composition,
and cardiovascular fitness: a systematic review of 68 studies. J. Phys. Act. Health 11,
206–227 (2014).
88
106. Ewing, R. & Cervero, R. Travel and the Built Environment. J. Am. Plann. Assoc. 76,
265–294 (2010).
107. Hino, A. A. F., Reis, R. S., Sarmiento, O. L., Parra, D. C. & Brownson, R. C. The
built environment and recreational physical activity among adults in Curitiba, Brazil.
Prev. Med. 52, 419–422 (2011).
108. Cervero, R., Sarmiento, O., Jacoby, E., Gomez, L. & Neiman, A. Influences of Built
Environments on Walking and Cycling: Lessons from Bogota. World Transit Res.
(2009). at <http://www.worldtransitresearch.info/research/2474>
109. Gomez, L. F. et al. Characteristics of the built environment associated with leisure-
time physical activity among adults in Bogotá, Colombia: a multilevel study. J. Phys.
Act. Health 7 Suppl 2, S196–203 (2010).
110. Wener, R. E. & Evans, G. W. A Morning Stroll: Levels of Physical Activity in Car
and Mass Transit Commuting. Environ. Behav. 39, (2007).
111. Lachapelle, U. & Frank, L. D. Transit and health: mode of transport, employer-
sponsored public transit pass programs, and physical activity. J. Public Health Policy
30 Suppl 1, S73–94 (2009).
112. Villanueva, K., Giles-Corti, B. & McCormack, G. Achieving 10,000 steps: a
comparison of public transport users and drivers in a university setting. Prev. Med. 47,
338–341 (2008).
113. Ding, D. et al. Perceived neighborhood environment and physical activity in 11
countries: Do associations differ by country? Int. J. Behav. Nutr. Phys. Act. 10, 57
(2013).
89
114. Sugiyama, T., Neuhaus, M., Cole, R., Giles-Corti, B. & Owen, N. Destination and
route attributes associated with adults’ walking: a review. Med. Sci. Sports Exerc. 44,
1275–1286 (2012).
115. EMBARQ. Globar BRT Data. (2013). at <http://www.brtdata.org/#/location>
116. Hidalgo, D. & Gutiérrez, L. BRT and BHLS around the world: Explosive growth,
large positive impacts and many issues outstanding. Res. Transp. Econ. 39, 8–13
(2013).
117. Gwilliam, K. Cities on the Move. (The World Bank, 2002). at
<http://elibrary.worldbank.org/doi/book/10.1596/0-8213-5148-6>
118. Institute for Transportation and Development policy. The Life and Death of Urban
Highways. (ITDP, 2012). at <http://www.itdp.org/library/publications/the-life-and-
death-of-urban-highways/>
119. Consejo Nacional de Política Económica y Social. Documento CONPES 3093,
Sistema de Servicio Público Urbano de Transporte Masivo de Pasajeros de Bogotá –
Seguimiento –. 2000. (2000). at
<https://www.dnp.gov.co/LinkClick.aspx?fileticket=VHPsmMeAtDU%3D&tabid=10
63>
120. Hidalgo, D., Pereira, L., Estupiñán, N. & Jiménez, P. L. TransMilenio BRT system in
Bogota, high performance and positive impact – Main results of an ex-post evaluation.
Res. Transp. Econ. 39, 133–138 (2013).
121. Alcaldía Mayor de Bogotá. Aprobado cupo de endeudamiento para Bogotá | Portal
Bogota | Bogota.gov.co. (2013). at <http://www.bogota.gov.co/content/aprobado-
cupo-de-endeudamiento-para-bogot%C3%A1>
90
122. Unión Temporal Steer Davies - Centro Nacional de Consultoría. Informe de
indicadores, Encuesta de Movilidad de Bogotá 2011. (Unión Temporal Steer Davies -
Centro Nacional de Consultoría, 2012).
123. DANE. Proyecciones de población. (2013). at
<https://www.dane.gov.co/index.php/poblacion-y-demografia/proyecciones-de-
poblacion>
124. Kerr, J. et al. Advancing science and policy through a coordinated international study
of physical activity and built environments: IPEN adult methods. J. Phys. Act. Health
10, 581–601 (2013).
125. IPAQ. International Physical Activity Questionnaire (2002). at
<http://www.ipaq.ki.se/ipaq.htm>
126. Hallal, P. C. et al. Lessons learned after 10 years of IPAQ use in Brazil and Colombia.
J. Phys. Act. Health 7 Suppl 2, S259–264 (2010).
127. Selvin, S. Epidemiologic Analysis: A Case-oriented Approach. (Oxford University
Press, 2001).
128. Freedson, P. S., Melanson, E. & Sirard, J. Calibration of the Computer Science and
Applications, Inc. accelerometer. Med. Sci. Sports Exerc. 30, 777–781 (1998).
129. Hearst, M. O. et al. The relationship of area-level sociodemographic characteristics,
household composition and individual-level socioeconomic status on walking
behavior among adults. Transp. Res. Part Policy Pract. 50, 149–157 (2013).
130. Frank, L. D. et al. The development of a walkability index: application to the
Neighborhood Quality of Life Study. Br. J. Sports Med. 44, 924–933 (2010).
131. Deddens, J. A. & Petersen, M. R. Approaches for estimating prevalence ratios. Occup.
Environ. Med. 65, 481, 501–506 (2008).
91
132. Wood, S. Generalized Additive Models: An Introduction with R. (Chapman and
Hall/CRC, 2006).
133. Pratt, M., Macera, C. A. & Wang, G. Higher direct medical costs associated with
physical inactivity. Phys. Sportsmed. 28, 63–70 (2000).
134. Montes, F. et al. Do health benefits outweigh the costs of mass recreational programs?
An economic analysis of four Ciclovía programs. J. Urban Health Bull. N. Y. Acad.
Med. 89, 153–170 (2012).
135. Rodríguez, D. A., Brisson, E. M. & Estupiñán, N. The relationship between segment-
level built environment attributes and pedestrian activity around Bogota’s BRT
stations. Transp. Res. Part Transp. Environ. 14, 470–478 (2009).
136. Hino, A. A. F., Reis, R. S., Sarmiento, O. L., Parra, D. C. & Brownson, R. C. Built
Environment and Physical Activity for Transportation in Adults from Curitiba, Brazil.
J. Urban Health 91, 446–462 (2014).
137. MINISTERIO DE MINAS Y ENERGIA. MINISTERIO DE MINAS Y ENERGIA
Precios Combustible. (2013). at
<http://www.minminas.gov.co/minminas/hidrocarburos.jsp?cargaHome=3&id_catego
ria=159&id_subcategoria=878>
138. World Bank. Pump price for gasoline (US$ per liter) | Data | Table. (2013). at
<http://data.worldbank.org/indicator/EP.PMP.SGAS.CD>
139. El Tiempo. Bogotá se mueve a 23 km por hora, aunque hay vías más lentas - Archivo
- Archivo Digital de Noticias de Colombia y el Mundo desde 1.990. eltiempo.com
(2011). at <http://www.eltiempo.com/archivo/documento/CMS-10485084>
92
140. Yang, Y., Diez Roux, A. V., Auchincloss, A. H., Rodriguez, D. A. & Brown, D. G. A
spatial agent-based model for the simulation of adults’ daily walking within a city.
Am. J. Prev. Med. 40, 353–361 (2011).
141. Zhu, W. et al. Agent-based modeling of physical activity behavior and environmental
correlations: an introduction and illustration. J. Phys. Act. Health 10, 309–322 (2013).
142. Saelens, B. E., Vernez Moudon, A., Kang, B., Hurvitz, P. M. & Zhou, C. Relation
Between Higher Physical Activity and Public Transit Use. Am. J. Public Health 104,
854–859 (2014).
143. Yang, Y., Diez Roux, A. V., Auchincloss, A. H., Rodriguez, D. A. & Brown, D. G. A
spatial agent-based model for the simulation of adults’ daily walking within a city.
Am. J. Prev. Med. 40, 353–361 (2011).
144. Willis, A., Gjersoe, N., Havard, C., Kerridge, J. & Kukla, R. Human movement
behaviour in urban spaces: implications for the design and modelling of effective
pedestrian environments. Environ. Plan. B Plan. Des. 31, 805 – 828 (2004).
145. Varas, A. et al. Cellular automaton model for evacuation process with obstacles. Phys.
Stat. Mech. Its Appl. 382, 631–642 (2007).
146. Davidson, W. et al. Synthesis of first practices and operational research approaches in
activity-based travel demand modeling. Transp. Res. Part Policy Pract. 41, 464–488
(2007).
147. DeSalvo, J. S. & Huq, M. Mode Choice, Commuting Cost, and Urban Household
Behavior*. J. Reg. Sci. 45, 493–517 (2005).
148. Vega, A. & Reynolds-Feighan, A. A methodological framework for the study of
residential location and travel-to-work mode choice under central and suburban
employment destination patterns. Transp. Res. Part Policy Pract. 43, 401–419 (2009).
93
149. Akkerman, A. The Diurnal Cycle of Regional Commuter Systems: North Wales,
1991. Geogr. Anal. 32, 247–266 (2000).
150. Consejo Nacional de Política Económica y Social. Documento CONPES 3677, DE
MOVILIDAD INTEGRAL PARA LA REGIÓN CAPITAL BOGOTÁ-
CUNDINAMARCA . 2010. (2010). at
<https://colaboracion.dnp.gov.co/CDT/Conpes/3677.pdf>
151. Camara de comercio de Bogotá. Observatorios de Desarrollo Urbano y Movilidad. at
<http://camara.ccb.org.co/contenido/contenido.aspx?conID=3222&catID=721>
152. Yang, Y. & Diez-Roux, A. V. Walking Distance by Trip Purpose and Population
Subgroups. Am. J. Prev. Med. 43, 11–19 (2012).
153. Eldridge, J. D. & Jones III, J. P. Warped Space: A Geography of Distance Decay.
Prof. Geogr. 43, 500–511 (1991).
154. Fotheringham, A. S. Spatial Structure and Distance-Decay Parameters. Ann. Assoc.
Am. Geogr. 71, 425–436 (1981).
155. Kruger, J., Ham, S. A., Berrigan, D. & Ballard-Barbash, R. Prevalence of
transportation and leisure walking among U.S. adults. Prev. Med. 47, 329–334 (2008).
156. Besser, L. M. & Dannenberg, A. L. Walking to Public Transit: Steps to Help Meet
Physical Activity Recommendations. Am. J. Prev. Med. 29, 273–280 (2005).
157. Braake, S. ter. THE CONTRIBUTION OF CICLORUTA ON ACCESSIBILITY IN
BOGOTÁ A spatial analysis on Cicloruta’s contribution to job accessibility for
different social-economic strata. (University of Twente, 2013).
158. Saelens, B. E., Vernez Moudon, A., Kang, B., Hurvitz, P. M. & Zhou, C. Relation
Between Higher Physical Activity and Public Transit Use. Am. J. Public Health 104,
854–859 (2014).
94
159. Sarmiento, O. L. et al. Quality of life, physical activity, and built environment
characteristics among colombian adults. J. Phys. Act. Health 7 Suppl 2, S181–195
(2010).
160. Wibowo, S. S. & Olszewski, P. Modeling Walking Accessibility to Public Transport
Terminals: Case Study of Singapore Mass Rapid Transit. J. East. Asia Soc. Transp.
Stud. 6, 147–156 (2005).
161. Ortuzar, J. de D. & Willumsen, L. G. Modelling Transport. (2011).
162. Unión Temporal Steer Davies & Limited- Centro Nacional de Consultoría. Informe de
indicadores, Encuesta de Movilidad de Bogotá 2011. (Unión Temporal Steer Davies
& Limited- Centro Nacional de Consultoría, 2012).
163. DANE. Bogotá D.C.:Pobreza Monetaria 2013. (2014). at
<https://www.dane.gov.co/files/investigaciones/condiciones_vida/pobreza/Bogota_Po
breza_2013.pdf>