A Systematic Review of Empirical and Simulation Studies Evaluating The Health Impact of Transportation Interventions

Environmental Research 186 (2020) 109519
Contents lists available at ScienceDirect
Environmental Research
journal homepage: www.elsevier.com/locate/envres
A systematic review of empirical and simulation studies evaluating the T

health impact of transportation interventions
Ivana Stankova,∗, Leandro M.T. Garciab, Maria Antonietta Mascollic, Felipe Montesd,
José D. Meiself, Nelson Gouveiac, Olga L. Sarmientoe, Daniel A. Rodriguezg, Ross A. Hammondh,i,
Waleska Teixeira Caiaffaj, Ana V. Diez Rouxa
a
Urban Health Collaborative, Dornsife School of Public Health, Drexel University, 3600 Market St, 7th Floor, Philadelphia, PA, 19104, USA
b
UKCRC Centre for Diet and Activity Research (CEDAR), MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine, Cambridge, UK
c
Department of Preventive Medicine, University of São Paulo Medical School, São Paulo, Brazil
d
Department of Industrial Engineering, Social and Health Complexity Center, Universidad de Los Andes, Bogotá, Colombia
e
School of Medicine, Universidad de Los Andes, Cra 1 # 18a-10, Bogotá, Colombia
f
Facultad de Ingeniería, Universidad de Ibagué, Carrera 22 Calle 67, Ibagué, 730001, Colombia
g
University of California, Berkeley, USA; Department of City and Regional Planning and Institute for Transportation Studies, University of California, Berkeley, USA
h
Center on Social Dynamics and Policy, The Brookings Institution, 1775 Massachusetts Ave NW, Washington, DC, 20036, USA
i
Brown School at Washington University in St. Louis, One Brookings Drive, St Louis, MO, 36130, USA
j
Observatory for Urban Health in Belo Horizonte, School of Medicine, Federal University of Minas Gerais, Belo Horizonte, Brazil
ARTICLE INFO ABSTRACT
Keywords: Urban transportation is an important determinant of health and environmental outcomes, and therefore essential
Transportation to achieving the United Nation's Sustainable Development Goals. To better understand the health impacts of
Health transportation initiatives, we conducted a systematic review of longitudinal health evaluations involving: a) bus
Systematic review rapid transit (BRT); b) bicycle lanes; c) Open Streets programs; and d) aerial trams/cable cars. We also syn-
Natural experiment
thesized systems-based simulation studies of the health-related consequences of walking, bicycling, aerial tram,
Complex systems
bus and BRT use.
Two reviewers screened 3302 unique titles and abstracts identified through a systematic search of MEDLINE
(Ovid), Scopus, TRID and LILACS databases. We included 39 studies: 29 longitudinal evaluations and 10 si-
mulation studies. Five studies focused on low- and middle-income contexts. Of the 29 evaluation studies, 19
focused on single component bicycle lane interventions; the rest evaluated multi-component interventions in-
volving: bicycle lanes (n = 5), aerial trams (n = 1), and combined bicycle lane/BRT systems (n = 4). Bicycle
lanes and BRT systems appeared effective at increasing bicycle and BRT mode share, active transport duration,
and number of trips using these modes. Of the 10 simulation studies, there were 9 agent-based models and one
system dynamics model. Five studies focused on bus/BRT expansions and incentives, three on interventions for
active travel, and the rest investigated combinations of public transport and active travel policies. Synergistic
effects were observed when multiple policies were implemented, with several studies showing that sizable in-
terventions are required to significantly shift travel mode choices.
Our review indicates that bicycle lanes and BRT systems represent promising initiatives for promoting po-
pulation health. There is also evidence to suggest that synergistic effects might be achieved through the com-
bined implementation of multiple transportation policies. However, more rigorous evaluation and simulation
studies focusing on low- and middle-income countries, aerial trams and Open Streets programs, and a more
diverse set of health and health equity outcomes is required.
1. Introduction with more than half of all people living in urban areas (United Nations,
2014). According to the United Nations (United Nations, 2018), espe-
Over the course of the last two centuries, the proportion of the cially rapid rates of urbanization are projected to occur in low- and
world's population residing in cities has increased more than 10 fold, middle-income countries. While the dense intersection of social,
∗
Corresponding author. Urban Health Collaborative, Dornsife School of Public Health, Drexel University, 3600 Market St, 7th Floor, Philadelphia, PA, 19104, USA.
E-mail address: is379@drexel.edu (I. Stankov).
https://doi.org/10.1016/j.envres.2020.109519
Received 21 August 2019; Received in revised form 8 April 2020; Accepted 10 April 2020
Available online 15 April 2020
0013-9351/ © 2020 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/BY/4.0/).
I. Stankov, et al. Environmental Research 186 (2020) 109519
natural, and built environments in cities affords numerous health-re- In addition to health impact evaluations, research employing
lated benefits, it can also pose serious risks to human health, well-being, system-based simulation methods such as agent-based modelling (ABM)
and environmental sustainability (Vlahov et al., 2005; Rydin et al., and system dynamics (SD) can also be used to elucidate the potential
2012). health-related impacts of diverse policies. Large-scale transportation
Transportation is well recognised as an important feature of urban interventions can affect travel behavior, population health, and health
life and a significant determinant of health and well-being. Beyond the inequalities in multiple ways that are hard to anticipate due to the
direct associations between transport and traffic accident-related in- complex and dynamic relations between these policies and the system
juries (Litman, 2013), transportation may also influence health and on which they act. System-based simulation methods are able to cap-
well-being via indirect pathways. For instance, transportation serves an ture complex mechanisms not readily accommodated by other analy-
important function in facilitating social interaction and access to a wide tical approaches, including non-linearity, feedback loops, individual
range of health-related opportunities, including health care services, and collective adaptation to changes in environmental and social con-
employment and educational opportunities (Litman, 2013). The re- texts, self-organization, and emergence (Jayasinghe, 2011). Simulation
lationships between transportation and health may also transpire models are often used to assess the health-related impacts of large-scale,
through mediating factors such as physical inactivity, and air and noise complex population-level interventions that are often prohibitively
pollution arising from an overreliance on motorized forms of trans- expensive or impractical to trial in the real world (Hammond, 2015).
portation (Brunekreef and Holgate, 2002; Babisch, 2006; Lee et al., For example, elucidating the dynamic mechanisms through which
2012). The design of transportation systems, through their capacity to transportation policies impact population health can improve the de-
enable or constrain mobility, can also impact a city's economic growth sign of more effective initiatives and reduce the potential for undesir-
and urban structure (Becerra et al., 2013). As a determinant of health, able or unintended consequences. Another advantage of simulation
transport advantage or disadvantage, including inadequate access to modelling is its capacity to facilitate policy prioritization and planning
transportation infrastructure and services, can exacerbate social segre- by affording a platform through which the differential or combined
gation (Lucas, 2012) and impact health inequalities (Borrell et al., health impact of diverse policies may be compared (Hammond, 2015).
2013). While there have been a number of systematic reviews of simulation-
The importance of transportation is reflected within the based methods applied to the study of non-communicable diseases
Transformative Commitments of the New Urban Agenda (United (Nianogo and Arah, 2015; Li et al., 2016), there has been no attempt to
Nations, 2017) and the United Nation's Agenda for Sustainable Devel- review models simulating the impact of transportation mode choices
opment (UN General Assembly, 2015), which features transportation as and transport-focused initiatives on health outcomes, more broadly.
essential to achieving the 17 Sustainable Development Goals (United This systematic review has two broad aims. The first is to summarize
Nations, 2016). Recognising this call to action, international agencies existing evidence (longitudinal empirical and simulation-based) on the
and cities worldwide have expressed a strong interest in the design and influence of four innovative transportation policies, programs, and in-
implementation of transportation-based policies and initiatives capable vestments (i.e., BRT, bicycle lanes, Open Streets programs, and aerial
of addressing some of the unique challenges faced by rapidly urbanising trams) on health-related behavior and or health outcomes. And second,
cities. Four emerging and innovative policies in urban mobility, that based on the synthesized evidence, this review will outline key re-
have attracted attention and been implemented in both high and low- commendations for future research.
and middle-income countries (HIC and LMIC, respectively) include: 1)
bus rapid transit (or BRT) in over 160 cities (for example, the Mexico 2. Methods
City Metrobús, the Lagos BRT in Nigeria and the Spurbus in Germany)
(Becerra et al., 2013, BRT+ Centre of Excellence and EMBARQ, 2019); 2.1. Study design
2) bicycle paths (for example, Santiago's Mapocho Pedaleable in Chile
and Denmark's well-known network of paths and bicycle lanes) (Pucher A systematic review of the peer-reviewed literature was conducted
et al., 2010; Becerra et al., 2013); 3) Open Streets programs, which with a focus on identifying: 1) primary studies evaluating the influence
involve the temporary closure of main streets to motorized traffic in of BRT, bicycle lanes, Open Streets programs, and aerial trams on
order to encourage cycling and other modes of active transport health-related behavior and or health outcomes; and 2) system-based
(Kuhlberg et al., 2014; Sarmiento et al., 2017); and 4) aerial trams (i.e, simulation studies exploring the links between transportation mode
cable cars) designed to connect peripheral hillside neighborhoods or choice, BRT, bicycle lanes, Open Streets programs, and aerial tram
islands with downtown activity nodes (examples include: the Medellín policies, and health-related behavior and or health outcomes. The
Metrocable in Colombia, the Cable of Constantine in Algeria, and the PRISMA checklist was used to ensure methodological rigor (Moher
Roosevelt Island Tramway in New York City) (Alshalalfah et al., 2012). et al., 2009) and the systematic review was registered with the Inter-
Despite the growing interest in the health-related impacts of trans- national Prospective Register of Systematic Reviews (PROSPERO No.
portation policies, there have been few attempts to evaluate these in- CRD42018093172). A narrative synthesis was conducted to char-
itiatives. From a public health vantage, it is important to understand acterize studies and investigate the health impacts of the four types of
what, if any, impact these transportation policies have had on popu- applied and simulated policies and projects.
lation health and health inequities in both HIC and LMIC. By identifying
initiatives that have demonstrated positive public health effects, as well 2.2. Search strategy
as highlighting gaps in understanding, the synthesis of local and in-
ternational research can inform decision-making and policy design in A search strategy seeking to identify all relevant reviews and pri-
cities all over the world. mary studies was developed by IS and LMTG, and further refined
2
Table 1
Simplified search strategy.
Domain Transportation
MeSH “Bicycling” [Mesh: focus] OR

Keywords (Bike lane* OR bike way* OR bike path* OR bikeway* OR cicloruta* OR bicycl* OR cycling
Bus rapid transit OR BRT OR Aerial lift* OR aerial tram* OR cable car* OR metrocable OR gondola lift* OR gondola car* OR cable propelled transit OR (Ciclovia*OR
mass event* OR mega event* OR open street*).ti, ab.
Health
MeSH “Disease” [Mesh: focus] OR “Health” [Mesh: focus] OR “Urban Health” [Mesh: focus] OR “Public Health” [Mesh: focus] OR “Mortality” [Mesh: focus] OR “Wounds
and Injuries” [Mesh: focus] OR
Keywords (health OR disease* OR behavio* OR injur*OR fatal* OR mortalit*).ti, ab.
Study design
MeSH “Non-Randomized Controlled Trials as Topic” [Mesh: focus] OR “Follow-Up Studies” [Mesh: focus] OR “Controlled Before-After Studies” [Mesh: focus] OR
Keywords (Systematic review* OR quasi-experiment* OR social experiment OR natural experiment* OR difference in difference* OR pre-post OR evaluation OR impact
assessment* OR before and after OR Simulation OR systems model* OR agent-based model* OR multi-agent model* OR individual-based model* OR system
dynamics).ti, ab.
ti = title; ab = abstract.
through consultation with a project working group. Four electronic included, though they were used to help identify potentially relevant
databases formed the focus of the literature search: MEDLINE (Ovid), studies not identified directly through the search. While no limits were
Scopus, Transportation Research International Documentation (TRID) placed on participant demographics, where possible, the search was
and LILACS. These databases were selected because of their coverage of restricted to studies published in English, Spanish, and Portuguese. The
literature from both HIC and LMIC. Using a combination of keywords, Participants, Interventions, Comparators, Outcomes, and Study design
MeSH terms, and phrases, a relatively broad search strategy was em- (PICOS) approach was used to assess whether identified studies met the
ployed to ensure all relevant studies published from the year 2000 and review's inclusion criteria (Table 2). Disagreements related to eligibility
onwards were identified. The basic search strategy is outlined in were resolved through discussions within reviewer pairs.
Table 1 (please refer to Appendix 1 for details on the complete data- Third, full-text manuscripts of all potentially eligible studies were
base-specific search strategies enacted in English, Spanish, and Portu- retrieved and screened by IS and LMTG. An interlibrary loan request
guese). Following the database search, the reference lists of included was made for all manuscripts that could not directly be accessed.
studies were screened for other unidentified but potentially relevant Studies that appeared to meet all the inclusion criteria of the review
studies. were discussed by the two reviewers and any disagreements were re-
solved by reflecting on the inclusion criteria and through consultation
2.3. Study selection and inclusion criteria with the project working group. Systematic reviews and the reference
lists of included studies were searched to identify any other potentially
The study selection process included three key steps. First, two re- relevant papers.
viewers (IS and LMTG) searched all relevant electronic databases in
August and September 2017, imported search results into EndNote, 2.4. Data extraction and analysis
removed all reference types other than journal articles and conference
proceedings as well as papers in languages other than English, Spanish, Two data extraction tools (one for evaluations and another for
or Portuguese. Second, all duplicate records were removed from system-based simulation studies) were developed to capture idiosyn-
EndNote using exact match for author and title. The remaining citations cratic aspects of study design and implementation characteristics of
were then imported into Covidence (2017), a Cochrane review man- policy evaluations and simulation studies. Each tool was independently
agement platform. Additional duplicate citations identified within pilot tested on three studies and amended based on in-depth discussions
Covidence were also removed. with the project working group. The refined tool applied to evaluation
Second, the titles and abstracts of retrieved studies were screened by studies extracted information on: author, year, geography, study type,
two pairs of independent reviewers within Covidence and assessed for aims or scope, sample size, participants, policy characteristics, length of
inclusion in the review. Primary studies were selected for inclusion if follow-up, outcomes, the nature and significance of policy effects. The
they used empirical analyses or system-based simulation methods to extraction tool applied to simulation-based studies sought to capture
evaluate the health impact of Open Streets programs, BRT, bicycle information relating to key aims or scope, software, and model speci-
lanes, or aerial tram infrastructure. Health impact was defined broadly fication – including, conceptual models, agent properties, actions and
to include health outcomes or health-related behaviors such as bus, BRT rules, temporal structure and characteristics of the model environment
and aerial tram use, bicycling, walking and more generic physical ac- – modeled policy scenarios, outcomes, model parameters, calibration
tivity measures. Studies assessing mode share as an outcome were also and validation processes, and key findings.
included if they considered bus, BRT, bicycle and/or aerial tram use Five independent reviewers working in four pairs (FM & JDM,
within their definition of mode share. We included these studies be- LMTG & MAM, IS & MAM and LMTG & IS) conducted an in-depth re-
cause these modes represent active forms of transportation and there- view of all included studies and performed extractions in accordance
fore qualify as health-related behaviors. Systematic reviews were not with the two extraction tools described above. The extractions were
3
Table 2
PICOS criteria for study inclusion and exclusion.
INCLUDE EXCLUDE
Participants
Studies including participants of any age group Animal studies
Interventions
Studies evaluating the impact of new BRT, Open Streets programs, bicycle paths and/or Studies evaluating light rail transport systems, bicycle boxes, intersection crossings
aerial tram infrastructure on at least one health-related behavior and/or health and roundabouts.
outcome. Studies evaluating traffic-free bicycle infrastructure, such as multi-use trails, System-based simulation studies exploring transportation choices that do not
bridges and boardwalks that report on cycling outcomes separately. include bus, cycling or aerial trams, or do not consider health-related behavior or
Multicomponent interventions, including those implemented across different sites, health outcomes.
were included only if the impact of BRT, Open Streets programs, bicycle lanes and/or Simulation studies modelling route choice but not reporting on mode share
aerial trams was explicitly evaluated and reported on in at least one of these sites. including at least one of the aforementioned modes, or health-related behaviors or
Systems-based simulation studies exploring the impact of transportation mode choice health outcomes.
(including bus, bicycle and/or aerial trams) on at least one health-related behavior
and/or health outcome were included, even if no policy scenarios were simulated
Comparators
Evaluations and system-based simulation studies comparing the impact of BRT, bicycle
paths, Open Streets programs and aerial tram policies and/or transportation choices
including at least one of these modes, on health-related outcomes.
Outcomes
Studies that report on at least one health-related behavior or health outcome, including Studies reporting on intermediary outcomes, such as air pollution, intentionality for
injury, prevalence and counts of walking or cycling, time and distance walked or behavior change or car crashes, without making a link to health-related behavior or
cycled (surrogates of physical activity energy expenditure (PAEE)), cycling speed health outcomes.
(only if distance and/or time travelled are reported, thus enabling the assessment of
PAEE).
Studies reporting on mode share including bus, bicycle or aerial trams were also
included.
Study design
Studies published in peer-reviewed journals or as peer-reviewed conference proceedings. Studies that only collected follow-up data before the intervention was completed or
Quasi-experimental studies and natural experiments with longitudinal designs. that are cross-sectional in nature.
Systems-based simulation studies, such as agent-based models and system dynamics Qualitative studies providing no quantitative assessment of policy effects and
models simulation studies that are not systems-based (e.g., health impact models).
Commentaries and opinion pieces.
compared by each pair and discrepant information relating to any given or uncertainty analyses.
study was discussed until all conflicts were resolved. Given the het-
erogeneity in study designs, exposures and outcomes reported by pa-
pers included in this review, a meta-analysis was not possible. Instead, a 3. Results
narrative synthesis of information extracted for each of the two ex-
traction tool formats was conducted in accordance with Cochrane re- A total of 4197 records were identified by the search strategy im-
commendations (Higgins et al., 2016). plemented in each of the four databases; MEDLINE(Ovid), Scopus,
LILACS and TRID (see Fig. 1). Records written in a language other than
2.5. Quality assessment English, Spanish or Portuguese and those not referenced as journal
articles or conference proceedings were removed, as were 439 dupli-
The quality of included studies was assessed using a modified ver- cates. The remaining 3302 citations were screened for inclusion based
sion of the Newcastle-Ottawa Quality Assessment Scale for Cohort on title and abstract. Most records (n = 3186) did not meet our in-
Studies (Wells et al., 2013). Information relating to intervention/policy clusion criteria. We screened 116 full-text papers (including 15 sys-
characteristics, sampling strategy, representativeness of the target po- tematic reviews) and excluded the majority of these (n = 90) for the
pulation, comparability of study controls and target populations, nature reasons detailed in Fig. 1. The identified systematic reviews were all
of outcome assessment, timing and adequacy of follow-up, and rates of excluded after they were screened for potentially relevant primary
attrition was collected from all evaluation studies. Given the absence of studies. A total of 26 studies were identified using the employed search
tools or guidance relating to the quality assessment of system-based strategy. Thirteen additional studies were identified by: screening ex-
simulation studies, we extracted information that we believed would cluded systematic reviews (n = 2) and reference lists of included stu-
provide a good indication of the quality of a given model's simulated dies (n = 4), as well as by searching other sources (e.g., reviews and
output. This included information concerning sources used to inform papers identified by the authorship team (n = 7)). Ultimately, a total of
model parameters, transparency concerning the equations and as- 39 studies were included in the systematic review, including 29 em-
sumptions made, the presence of calibration, validation and sensitivity pirical studies and 10 simulation-oriented papers.
4
Fig. 1. PRISMA flowchart showing process of study selection.
3.1. Empirical studies 2016; Chang et al., 2017).

Of the 29 studies, 19 were single-component interventions focused on
Empirical studies included in the review were published between the implementation of bicycle lanes. The remaining 10 studies evaluated
2005 and 2017 (Table 3). These studies were based in: North America, multicomponent interventions. Among these, one study evaluated newly
including the United States (n = 11) (Boarnet et al., 2005; Evenson built aerial tram infrastructure which was implemented in combination
et al., 2005; Burbidge and Goulias, 2009; Parker et al., 2011; Chen with other neighborhood improvements including, additional lighting in
et al., 2012; Parker et al., 2013; Dill et al., 2014; Brown et al., 2016; public spaces, pedestrian bridges and recreational centers (Cerdá et al.,
Cook et al., 2016; Ferenchak and Marshall, 2016; Brown et al., 2016b) 2012). Five studies evaluated the health effects of bicycle lanes which
and Canada (n = 1) (Bhatia et al., 2016); Europe, specifically, the were variously combined with a host of interventions such as light rail
United Kingdom (n = 9) (Goodman et al., 2013; Goodman et al. 2013b; and pedestrian infrastructure improvements, the creation of parking fa-
Goodman et al. 2014; Heinen et al., 2015; Panter and Ogilvie, 2015; cilities and additional lighting (Boarnet et al., 2005; Goodman et al.,
Heinen and Ogilvie, 2016; Panter et al., 2016; Panter and Ogilvie, 2017; 2013; Brown et al., 2016; Pazin et al., 2016; Brown et al., 2016b). The
Song et al., 2017) and Denmark (n = 1) (Jensen, 2008); and Australia remaining studies (n = 4) combined BRT and bicycle/pedestrian paths
(n = 4) (Greaves et al., 2015; Langdon, 2015; Rissel et al., 2015; with park-and-ride sites (Heinen et al., 2015; Heinen and Ogilvie, 2016;
Heesch et al., 2016). Only three studies were based in LMIC, specifi- Panter et al., 2016; Chang et al., 2017). Table 3 summarizes the char-
cally, Colombia, Mexico and Brazil (Cerdá et al., 2012; Pazin et al., acteristics of included studies.
5
Table 3
I. Stankov, et al.
Characteristics of included evaluation studies.

ID Author & year Intervention location & Intervention type Intervention description/definition of population Population/groups (n) Baseline age Female Outcome Type(s) Outcomes stratified;
period and groups (years) (%) analyzed groups
1 Bhatia et al. Toronto, CAN SC: cycling infrastructure; Seven lane segments in which a bicycle lane was Mainly Adults Unclear Unclear Injury (n = 3) No
(2016) 1993–2008 bike lanes (unprotected) painted between 1991 and 2010. The lane segments Collisions (329)
also had a higher collision cycle lane (i.e., Minimum
of 100 cycle-motor vehicle collisions between 1991
and 2010).
Population: Cycle-motor vehicle collisions
reported on the 7 lane segments with newly marked
bike lanes
2 Boarnet et al. Murrieta, USA unclear MC: cycling infrastructure; Bicycle facilities including on-street bike lanes were Combined adults & Unclear Unclear Active transit trips No
(2005) bike lanes (unprotected) installed at one site i.e., at Murrieta Elementary. children (n = 1)
“New sidewalks and sidewalk gap closures” were Cyclists/Pre (4)
also constructed around the school. [p.308] Post (14)
Population: Adult and child cyclists using the new
bike lanes
3 Brown et al. Salt Lake City, USA MC: cycling infrastructure; The complete streets intervention included the Adults (536) NR 51 Active transit mode No
(2016) 2013 bike lanes (unclear) completion of a sporadic bike lane and involved share (n = 3)
widening sections to make it a designated “high
comfort” bike lane (with speed limit reduced to
50 km/h) on the city bike map. The bike lanes were
painted on both sides of the street and the sidewalks
were improved and widened to create a shared bike
6
and pedestrian path. The intervention also included
a light rail line extension, narrowing of automotive
lanes and the creation of wider and better lit
sidewalks.
Intervention (near) group: Adults living
(≤800 m) from the complete street renovation
Control (far) group: Adults living 801–2000m
from complete street renovation
4 Brown et al. Salt Lake City, USA MC: cycling infrastructure; The complete streets intervention included the Adults Physiological (n = 2) No
(2016b) 2013 bike lanes (unclear) completion of a sporadic bike lane and involved Intervention Anthropometric
widening sections to make it a designated “high Cyclist group: (n = 1)
comfort” bike lane (with speed limit reduced to Never (434) 43 55 Active transit duration
50 km/h) on the city bike map. The bike lanes were Continuing (29) 40 17 (n = 1)
painted on both sides of the street and the sidewalks Former (33) 38 30
were improved and widened to create a shared bike New (40) 37 43
and pedestrian path. The intervention also included
a light rail line extension, narrowing of automotive
lanes and the creation of wider and better lit
sidewalks.
Population: Adults living within 2 km of the
complete streets intervention. Participants were
divided into four cycling groups based on their
cycling patterns pre and post intervention: never
cyclists (not cycled pre or post) which serve as the
reference group; continuing cyclists (cycled pre and
post); former cyclists (cycled pre but not post); new
cyclists (not cycled pre but cycled post).
(continued on next page)
Table 3 (continued)
I. Stankov, et al.
5 Burbidge and Salt Lake City, USA SC: cycling infrastructure; “A Class 1 trail (two-way multi-use trail separated Adults (98) 48 55 Active transit trips Yes; age
Goulias (2009) 2007 multi-use trail (separated) from existing roads and sidewalks) [was (n = 3)
constructed] on the existing canal right-of-way.” Active transit duration
“This trail creates a 4.025 km loop connecting two (n = 1)
existing sidewalks.” [p.79]
Population: Adult residents of West Valley City, a
suburb of Salt Lake City, who live within 1.6 km of
the trail
6 Cerdá et al. Medellin, COL 2004 MC; aerial trams A cable-propelled transit system (gondola) known Adolescents & Adults 36–61 yrs: Homicide (n = 1) No
(2012) as Metrocable was built using funding from the Intervention (225) 26% 67
municipal government of Medellin as part of “a Control (241) 27% 67
territorial plan to promote urban and rural
development”. It connects “an elevated train system
in the city centre to the impoverished Santo
Domingo neighborhood in the mountainous
periphery, with 4 stops covering a distance of
2072 m and reaching an elevation of 399 m”
[p.1046]
“The municipal government made other

improvements to neighborhoods serviced by the
gondola, including additional lighting for public
7
spaces; new pedestrian bridges and street paths;
‘‘library parks’’; buildings for schools, recreational
centers, and centers to promote microenterprises;
more police patrols; and a family police station next
to a gondola station.” [p.1046]
Intervention group: Residents (12–16 years) of
one of 25 neighborhoods (city districts 1 and 2)
where the gondola system was installed.
Control group: Residents of one of 23
neighborhoods located in comparable city districts
(4 and 8) that were not serviced by the gondola
system.
7 Chang et al. Mexico City, MEX 2013 MC: BRT & cycling The Metrobus Line 5 corridor was added to an Adults Mean age: Active transit Yes; gender,
(2017) infrastructure; bike lanes existing BRT network with four existing lines in Intervention (1420) 47 52 frequency (n = 3) education,
(separated) Mexico City. The Line 5 corridor is 10 km long with Control (1067) employment type
18 stations, with an average of 625 m between
stations. Its service features include: “(i) articulated
and bi-articulated high floor buses, (ii). Exclusive
bus lanes, and (iii) off-board fare collection”
[p.339]. The BRT intervention was combined with a
Complete Street intervention which included a host
of streetscape interventions including protected
bike lanes and parking, widened sidewalks,
redesigned junctions, and the recovery of public
and green space throughout the corridor.
Population: Adults residing 500 m either side of
the Line 5 corridor.
I. Stankov, et al.
Table 3 (continued)
8 Chen et al. New York City, USA SC: cycling infrastructure; Intervention included the creation of 69 km of Unclear Unclear Unclear Injury (n = 2) No
(2012) 1996–2006 bike lanes (unprotected) bicycle lanes on 61 streets not protected by a Cyclists
parking lane, in the 5 boroughs of New York City Intervention
from 1996 through 2006. Pre (4360)
Intervention group: Road users in New York City Post (1349)
travelling on roadways where on-street bicycle Control
lanes (not protected by a parking lane) had been Pre (12,365)
installed from 1996 through 2006 (a total length of Post (3578)
about 69.2 km on 61 streets).
Control group: Road users travelling on roadways
without bicycle lanes but with segment- or
intersection-level characteristics comparable to
those of the treatment group.
9 Cook et al. Durham, USA 2014 SC: cycling infrastructure; A 3.2 km long bicycle and pedestrian bridge-link Unclear 26–54 yrs 45 Active transit mode Yes; household
(2016) multi-use trail (separated) was created to connect the northern segment Trail users share (n = 4) income
(11.3 km long) of the trail to the southern trail Pre (1301 survey; 9266 Active transit duration
segment (21.7 km long) to form a continuous 35 km counts) (n = 4)
shared use, separated path. Post (2245 survey;
Population: Users of the trail segments both to the 21,365 counts)
north and south of the bridge linkage
8
10 Dill et al. (2014) Portland, USA unclear SC: cycling infrastructure; A new bicycle boulevard was installed on 8 street Adults Active transit trips No
bike lanes (unprotected) segments (1.45 km–6.76 km long) in Portland, Intervention (182) 43 63 (n = 1)
Oregon. Eleven control street segments (1.6–9.2 km Control (168) 41 67 Active transit mode
long) were also monitored as part of the evaluation. share (n = 3)
Intervention group: Adults residing within 300 m Active transit duration
the 8 streets selected for bicycle boulevard (n = 2)
installation. Physical activity
Control group: Adults residing within 300 m of the (n = 1)
11 control street segments.
11 Evenson et al. Durham, USA 2002 SC: cycling infrastructure; This study evaluated a railway track segment that Adults (366) 43% 65 Physical activity No
(2005) multi-use trail (separated) was converted to a paved, 3-m-wide multi-use trail, ≥50 yrs (n = 5)
which extended an existing 5.1 km trail segment by Active transit duration
another 4.5 km, along with a 3.2 km spur. (n = 3)
Population: Adults living within 3.2 km of the new
trail segment.
12 Ferenchak and Chicago, USA 2008-10 SC: cycling infrastructure; 259 block groups that had only sharrows (shared Unclear Unclear Unclear Active transit trips No
Marshall (2016) bike lanes & sharrows (both lane markings) installed (overall 54 km of Cyclists (n = 2)
unprotected) sharrows), and 292 block groups that overall had Intervention Bike lane Active transit mode
168 km of bike lanes installed. (1621 ridership; 2046 share (n = 2)
Intervention group: safety outcome) Injury (n = 2)
Bike lane; census block groups that had bike lanes or Sharrow (259 ridership;
trails installed between 2000 and 2010. 89 safety outcome)
Sharrow group; census block groups that had only Control (292 ridership;
sharrows installed between 2000 and 2010. 39 safety outcome)
Control group: Block groups that had no bike
infrastructure installed between 2000 and 2010.
Table 3 (continued)
I. Stankov, et al.
13 Goodman et al. 18 townsa, GBR 2005- MC: cycling infrastructure; The town-level initiatives involved mixtures of Adults Unclear Unclear Active transit mode Yes; area-level
(2013) 11 bike lanes (unprotected) & capital investment (e.g. cycle lanes) and revenue Intervention group share (n = 3) deprivation
bike tracks (protected) investment (e.g. cycle training), tailored to each (1,266,337)
town. “Each town implemented a different mixture Comparison group
of infrastructure, tailored to its specific context. In Matched comparison
total, 98 km of on-road lanes and 264 km of off-road (969,605)
paths were created between 2008 and 2011. This Unfunded comparison
represented a 28% increase in the length of such (4,195,540)
routes previously available (based on 16 of 18 National comparison
towns reporting sufficient data on pre-intervention (10,356,452)
facilities)” [p.230]. The capital investment
component involved an increase in cycle lanes and
paths ranging from as low as 9% (Stoke-on-Trent)
to as high as 105% in Brighton & Hove.
Intervention group: 17 urban towns and one city
outside of London selected to be part of initiative.
Comparison groups: Matched comparison; “largest
urban regions within the English local authority
‘most similar’ similar’ to each intervention local
authority.” Similarity was based on demographic,
socioeconomic, employment and industry
characteristics. Unfunded comparison; “largest urban
region within the 67 local authorities which applied
9
unsuccessfully” for the initiative. National
comparison; “all non-intervention, urban areas
outside London with a population of > 30,000
(close to the size of the smallest intervention town)”
[p.231].
14 Goodman et al. Cardiff, Kenilworth & SC: cycling infrastructure; Three Connect2 projects were evaluated. These Adults ≥50 yrs Active transit mode Yes; education,
(2013b) Southampton, GBR bike lanes (separated) were based in Cardiff, where a new 140 m long, 4 m Intervention share (n = 4) income
2010-11 wide traffic-free bridge with integral lighting was 1-year follow-up sample 66% 54
built over Cardiff Bay; Kenilworth, where a traffic- (1849)
free bridge was built over a busy trunk road to link 2-year follow-up sample 70% 57
the town to a rural greenway; and Southampton, (1510)
where an informal riverside footpath (impassable at
high tide) was turned into a new 400 m boardwalk.
Population: Adults residing within 5 km by road
network from the core Connect2 projects in each of
the three towns
15 Goodman et al. Cardiff, Kenilworth & SC: cycling infrastructure; Three Connect2 projects were evaluated. These Adults ≥50 yrs Active transit duration No
(2014) Southampton, GBR bike lanes (separated) were based in Cardiff, where a new 140 m long, 4 m Intervention (n = 6)
2010-11 wide traffic-free bridge with integral lighting was 1-year follow-up sample 66% 56 Physical activity
built over Cardiff Bay; Kenilworth, where a traffic- (1796) (n = 1)
free bridge was built over a busy trunk road to link 2-year follow-up sample 70% 57
the town to a rural greenway; and Southampton, (1465)
where an informal riverside footpath (impassable at
high tide) was turned into a new 400 m boardwalk.
the three towns
Table 3 (continued)
I. Stankov, et al.
16 Greaves et al. Sydney, AUS 2014 SC: cycling infrastructure; “The intervention comprised a 2.4 km length of Adults 45–55 yrs Active transit trips No
(2015) bike tracks (separated) separated bi-directional cycleway linking the inner- Intervention (184) 46% 61 (n = 5)
city suburbs of Green Square in the south with the Control (251) unclear unclear Active transit duration
Central Business District (CBD) through Redfern (n = 1)
and Waterloo”. “The George Street cycleway adds Active transit time
to several pre-existing bi-directional cycleways share (n = 1)
within the City of Sydney Local Government Area
(LGA) totaling a distance of 11 km (as of October
2014)” [p.4].
Intervention group: Geographic area
encompassing new cycleway in inner Sydney
Control group: Neighboring area of similar
demographics with no new planned bicycle
infrastructure
17 Heesch et al. Brisbane, AUS 2013 SC: cycling infrastructure; “The V1 is a dedicated 17-km long, 3-m wide Unclear Active transit trips No
(2016) bike tracks (separated) exclusive off-road bikeway.” “The V1 has been Cyclists Unclear 15 (n = 1)
delivered in stages with Stage A (about 1.4 km) Intervention (169)
completed in July 2010, and Stage B (about 900 m) Reference
completed in May 2011. Completion of these stages Pre (132) Unclear 14
extended existing V1 bikeway infrastructure farther Post (99) Unclear 20
south. Stage C (about 2.3 km) is 7 km north of these
earlier improvements via the existing V1
infrastructure. It opened June 25, 2013 to extend
the V1 farther north towards the city centre”
[p.368].
10
Intervention group: Cyclists on the newly created
bikeway i.e., the Veloway 1 (or V1).
Reference group: Cyclists on the South East
Freeway Bikeway (or SEFB).
18 Heinen and Cambridge, GBR 2011 MC: cycling infrastructure; “The busway comprises a 25 km off-road guideway Adults (470) ≥51 yrs 67 Active transit mode No
Ogilvie (2016) multi-use trail (separated) & for buses, with a parallel path that can be used for 34% share (n = 10)
BRT walking and cycling, in two sections: one between
the market town of St Ives and the northern edge of
Cambridge, and the other between Cambridge
railway station and the southern fringe at
Trumpington” [p.2]. It also includes three park-
and-ride sites.
Intervention group: Adults aged 16 or over,
working in areas of Cambridge to be served by the
busway and living within approximately 30 km of
the city centre.
19 Heinen et al. Cambridge, GBR 2011 MC: cycling infrastructure; “The busway comprises a 25 km off-road guideway Adults (466) ≥51 yrs 67 Active transit mode No
(2015) multi-use trail (separated) & for buses, with a parallel path that can be used for 34% share (n = 6)
BRT walking and cycling, in two sections: one between
the market town of St Ives and the northern edge of
Trumpington. It also includes three park-and-ride
sites” [18, p.2].
the city centre.
Table 3 (continued)
I. Stankov, et al.
20 Jensen (2008) Copenhagen, DNK SC: cycling infrastructure; Construction of one-way bicycle track (2–2.5 m Unclear Unclear Unclear Injury (n = 30) No
1978–2003 bike lanes (unprotected) & wide) on both sides of a 20.6 km road and marking Collisions
bike tracks (separated) of one -way bicycle lanes (1.5–2 m wide) on both Intervention
sides of a 5.6 km road in Copenhagen, Denmark. On Bike lanes: Crashes
Intervention group: Collisions on roads with (700); Injuries (219).
newly constructed bike lanes and tracks. On Bike tracks: Crashes
General comparison group: Collisions on (5,898);
unchanged roads with known developments in Injuries (2,413).
traffic volume. This group consists of 110 km of Comparison
roads with 170 locations, where motor vehicle and Crashes (24,369)
bicycle/moped traffic is counted yearly or every Injuries (8,648)
forth to sixth year.
21 Langdon (2015) Brisbane, AUS 2001-10 SC: cycling infrastructure; The intervention included the installation of Unclear Unclear Unclear Active transit mode No
bike lanes (unclear) “bridges, missing links and end of trip facilities for Cyclists (8,600) share (n = 1)b
cyclists” [p.1]. The bridges evaluated include the: Active transit trips
Goodwill Bridge 2001; Go-between Bridge & (n = 1)b
Kurilpa Bridge; Eleanor Schonell Bridge; Sir Leo
Hielsher (Gateway) Bridge; and the Ted Smout
Memorial Bridge. The missing links included the:
Normanby Pedestrian Cycle Link; Western Freeway
Bikeway & Toowong Overpass; and the Veloway 1
(V1) Stage C.
11
Population: Bicycle commuters on the bridges and
along missing links installed.
22 Panter and Southampton, SC: cycling infrastructure; Three Connect2 projects were evaluated. These Adults (967) ≥50 yrs 52 Active transit duration No
Ogilvie (2015) Cardiff & Kenilworth, bike lanes (separated) were based in Cardiff, where “a new 140 m long, 65% (n = 1)
GBR 2010-11 4 m wide traffic-free bridge with integral lighting”
was built over Cardiff Bay; Kenilworth, where “a
traffic-free bridge was built over a busy trunk road
to link the town to a rural greenway”; and
Southampton, where an informal riverside footpath
(impassable at high tide) was turned into “a new
400 m boardwalk” [p.2].
the three towns
23 Panter et al. Cambridge, GBR 2011 MC: cycling infrastructure; “The busway comprises a 25 km off-road guideway Adults (469) Mean age: 44 67 Physical activity No
(2016) multi-use trail (separated) & for buses, with a parallel path that can be used for (n = 2)
BRT walking and cycling, in two sections: one between Active transit trips
the market town of St Ives and the northern edge of (n = 6)
Trumpington. It also includes three park-and-ride
sites” [18, p.2].
the city centre.
Table 3 (continued)
I. Stankov, et al.
24 Panter and Southampton, Cardiff & SC: cycling infrastructure; Three Connect2 projects were evaluated. These Adults (1258) ≥50 yrs 55 Active transit trips No
Ogilvie (2017) Kenilworth, GBR 2010- bike lanes (separated) were based in Cardiff, where “a new 140 m long, 72% (n = 6)
11 4 m wide traffic-free bridge with integral lighting”
400 m boardwalk” [22, p.2].
the three towns
25 Parker et al. New Orleans, USA SC: cycling infrastructure; “The 5.0 km dedicated bike lane on St. Claude Combined adults & NR Mean Active transit trips Yes; gender
(2011) 2008 bike lanes (unprotected) Avenue, also known as Louisiana Highway 46”, was children number (n = 1)
completed in the spring of 2008. “Bike lanes were Cycling trips
striped on both sides of the road and are 1.5 m Intervention
wide” [p.S99]. Pre (mean: 121) (13)
Population: Adults and children cycling along St. Post (mean: 188) (29)
Claude Avenue.
26 Parker et al. New Orleans, USA SC: cycling infrastructure; “The 1.6 km dedicated bike lane on S. Carrollton Combined adults & NR Mean Active transit trips Yes; race, gender
(2013) 2010 bike lanes (unprotected) Avenue”, New Orleans, completed in June 2010. children number (n = 1)
“Bike lanes were striped on both sides of the road Cycling trips
12
and are 1.5 m wide. There is one 3.4 m wide travel Intervention
lane on either side of the road, separated by a 18 m (mean: 257) (15)
wide median” [p.S102]. Comparison
Intervention group: Adults and children cycling (mean: 37) (33)
along S. Carrollton Avenue
Comparison group: Adults and children cycling on
two adjacent side streets; Short and Dublin St.
27 Pazin et al. Florianópolis, BRA MC: cycling infrastructure; A new walking and cycling route (2.3 km long) was Adults (519) 55–85 yrs: Physical activity No
(2016) 2010 bike lanes (unprotected) inaugurated in [an area known as Beira-Mara Intervention 41% 58 (n = 4)
Continental,] in the continental coast of 0-500m (192) 43% 55 Active transit mode
Florianópolis, SC, Brazil.” “The project included a 501-1000m (137) 46% 54 share (n = 1)
new avenue, parking lots, and an on-road walking 1001-1500m (190) 37% 65 Active transit
and cycling route, all along the seashore” [p.19]. frequency (n = 1)
Population: Adult residing within 0–500 m,
501–1000 m and 1001–1500 m of the new walking
and cycling route.
28 Rissel et al. Sydney, AUS 2014 SC: cycling infrastructure; “New 2.4 km bi-directional separated bicycle path Adults (512) ≥45 yrs: 63 Active transit trips No
(2015) bike track (separated) built” as part of its expanding bicycle network in Intervention (240) 37% (n = 2)
Sydney [p.1]. Comparison (272) Active transit
Intervention group: Adults living no more than frequency (n = 1)
2.5 km from the new bicycle path.
Comparison group: Adults living in
“neighborhoods a similar distance from the central
business district and with a similar demographic
profile, and where local council had no plans to
modify infrastructure during the study period”
[p.2].
SC: Single component intervention; MC: Multicomponent intervention; NR: not reported; CAN: Canada; USA: United States of America; DNK: Denmark; GBR: United Kingdom; AUS: Australia; BRA: Brazil; COL: Colombia;
18 towns: Darlington, Derby, Brighton & Hove, Aylesbury, Exeter, Lancaser with Morecambe, York, Cambridge, Colchester, Southend-on-Sea, Leighton Buzzard, Woking, Bristol, Shrewsbury, Stoke-on-Trent, Chester,
Southport & Ainsdale, Blackpool; CAN Canada, USA United States of America, DNK Denmark, GBR United Kingdom, AUS Australia, BRA Brazil, COL Colombia NOTE: For multicomponent interventions, this table reports
All 29 empirical evaluation studies employed a longitudinal study
NOTE: For multicomponent interventions, this table reports information relating to the interventions relevant to our review, and the population characteristics and outcomes assessed only for relevant interventions.
Outcomes stratified;
design and included pre- and post-intervention observations. Of these
29 studies, 15 included a control or reference group (Jensen, 2008;
Cerdá et al., 2012; Chen et al., 2012; Goodman et al., 2013; Parker
et al., 2013; Dill et al., 2014; Greaves et al., 2015; Panter and Ogilvie,
groups
2015; Rissel et al., 2015; Brown et al., 2016; Heesch et al., 2016; Pazin
No
Active transit duration

et al., 2016; Brown et al., 2016b; Panter and Ogilvie, 2017; Song et al.,
2017). The proportion of studies with a control or reference group did
Active transit distance
Active transit distance
not differ by intervention type; around half of all single component (10
Active transit time
Non-active transit
Non-active transit
Outcome Type(s)
duration (n = 2)
distance (n = 2)
of 19 studies) and multi-component intervention (5 of 10 studies) re-
share (n = 2)
share (n = 2)
ported either a control or reference group. Four studies employed a

analyzed
random sampling approach to recruit participants, including three

(n = 6)
(n = 6)
multi-component (Cerdá et al., 2012; Pazin et al., 2016; Chang et al.,

2017) and one single component intervention study (Evenson et al.,
2005). A further eleven studies attempted to sample the entire study
Female
area, including nine single component (Chen et al., 2012; Goodman

(%)
56
et al. 2013b, 2014; Dill et al., 2014; Panter and Ogilvie 2015, 2017;
Bhatia et al., 2016; Ferenchak and Marshall, 2016; Song et al., 2017)
Baseline age
and two multi-component intervention studies (Goodman et al., 2013;

Brown et al., 2016).
(years)
Exposure to the transportation intervention was operationalized in

57
several different ways among the 29 empirical studies. Some studies

information relating to the interventions relevant to our review, and the population characteristics and outcomes assessed only for relevant interventions.
compared geographic locations, specifically, street segments, before

Population/groups (n)
Intervention (1489)
and after an intervention (Boarnet et al., 2005; Jensen, 2008; Parker

et al. 2011, 2013; Chen et al., 2012; Bhatia et al., 2016; Cook et al.,
2016; Heesch et al., 2016). Most studies however, considered inter-
vention impacts among people living in a fixed area of influence. These
Adults
areas of influence were most commonly defined as buffers centred on

the focal point of an intervention (Evenson et al., 2005; Burbidge and
Goulias, 2009; Dill et al., 2014; Panter and Ogilvie, 2015; Rissel et al.,

4 m wide traffic-free bridge with integral lighting”

were based in Cardiff, where “a new 140 m long,

Intervention description/definition of population
2015; Brown et al., 2016b; Chang et al., 2017), although some studies
Three Connect2 projects were evaluated. These
used predefined geographies such as towns (Goodman et al., 2013),

block groups (Langdon, 2015; Ferenchak and Marshall, 2016) and

neighborhoods (Cerdá et al., 2012) in and/or surrounding intervention
sites. Only 9 studies used a distance-based metric to assess health im-
pacts among people living different distances from an intervention
400 m boardwalk” [22, p.2].
(Goodman et al. 2013b, 2014; Heinen et al., 2015; Brown et al., 2016;
Heinen and Ogilvie, 2016; Panter et al., 2016; Pazin et al., 2016; Panter
and Ogilvie, 2017; Song et al., 2017). We were unable to determine
the three towns
how exposure to the intervention was defined in one study (Greaves

et al., 2015).
and groups
A total of 183 outcomes relating to 19 unique interventions were

assessed by the 29 empirical studies included in our review. Of these,
21 outcomes were described but statistical tests were not conducted
(see Appendix 2) (Burbidge and Goulias, 2009; Dill et al., 2014; Greaves
SC: cycling infrastructure;
et al., 2015; Langdon, 2015; Song et al., 2017). The 162 outcomes
bike lanes (separated)
tested for statistical significance varied both with respect to outcome

Intervention type
type and the frequency with which they were evaluated (Fig. 2A;
Appendix 3, Table 1). The four most frequently assessed outcome types
included, active travel duration (25%), injury (23%), mode share (i.e.,
the share of walking, cycling, public transport and or car trips) (22%),
Outcomes not tested for statistical significance.
and number of active trips (including walking or cycling trips) (11%).

Few studies assessed physiological (Brown et al., 2016b), anthropo-
Intervention location &
Cardiff, Kenilworth &
metric (Brown et al., 2016b), and car travel-related outcomes (Greaves

Southampton, GBR
et al., 2015; Song et al., 2017). The three studies that evaluated the
health effects of interventions in LMIC focused on homicide in Co-
lombia (n = 1) (Cerdá et al., 2012), active transit frequency in Mexico
2010-11
and Brazil (n = 3 and n = 1, respectively) (Pazin et al., 2016; Chang

period
et al., 2017), as well as physical activity (n = 4) and active mode share

(n = 1) in Brazil (Pazin et al., 2016). Appendix 2 provides further in-
Table 3 (continued)
formation on assessed outcomes and the nature of the associations re-

Author & year
ported by included studies.

Song et al.
The 19 studies reporting on single-component interventions (all

(2017)
focusing on bicycle lanes; Fig. 2B) predominantly reported on injury

outcomes (n = 37) (Jensen, 2008; Chen et al., 2012; Bhatia et al., 2016;
29
ID
b
a
Ferenchak and Marshall, 2016), active travel duration outcomes
13
Fig. 2. Frequency of outcomes, by type, for A all included empirical studies (29 studies, 162 distinct outcomes), B single component interventions (19 studies, 117
outcomes) focusing on the creation of bicycle lanes, and C multicomponent interventions (10 studies, 45 outcomes). The bars in blue represent the total number of
outcomes, by type, while the green bars represent the total number of statistically significant outcomes. Of the statistically signficant outcomes, the bar in orange
represents the number of outcomes that were in the unexpected direction for each outcome type. (For interpretation of the references to colour in this figure legend,
the reader is referred to the Web version of this article.) NB: Figures do not include 21 outcomes reported by 5 studies because these outcomes were not assessed using
statistical models.
14
(n = 30) (Evenson et al., 2005; Burbidge and Goulias, 2009; Dill et al., Yang and Diez-Roux, 2013; Macmillan et al., 2014; Okushima 2015,
2014; Goodman et al., 2014; Greaves et al., 2015; Panter and Ogilvie 2016; Yang et al., 2015; Lemoine et al., 2016a; Zellner et al., 2016; Zou
2015, 2017; Cook et al., 2016; Song et al., 2017) and active trip out- et al., 2016) were published between 2011 and 2016.
comes (n = 16) (Burbidge and Goulias, 2009; Parker et al. 2011, 2013; Five studies tested the impacts of expanding, improving (e.g.,
Dill et al., 2014; Greaves et al., 2015; Rissel et al., 2015; Ferenchak and creating exclusive bus lanes or changing the BRT system), or in-
Marshall, 2016; Heesch et al., 2016). Only 6 out of 37 injury outcomes centivizing public transportation formed the focus of five studies
showed statistically significant associations with new bicycle lanes (McDonnell and Zellner, 2011; Okushima and Akiyama, 2011;
(Jensen, 2008; Bhatia et al., 2016; Ferenchak and Marshall, 2016), and Okushima, 2015; Lemoine et al., 2016a; Zou et al., 2016), three of
of these, the majority (n = 4, all from the same study) were in the which were dedicated to BRTs (McDonnell and Zellner, 2011;
unexpected direction (i.e., the installation of bicycle lanes was posi- Okushima, 2015; Lemoine et al., 2016a). Three studies investigated
tively associated with injuries) (Jensen, 2008). The highest number of interventions for active travel promotion, such as the creation of cycling
statistically significant outcomes were observed for active travel dura- infrastructure or implementation of traffic safety measures to promote
tion (i.e., 15 of 30 outcomes) (Evenson et al., 2005; Dill et al., 2014; walking trips to school (Yang and Diez-Roux, 2013; Macmillan et al.,
Goodman et al., 2014; Cook et al., 2016; Panter and Ogilvie, 2017; Song 2014; Okushima, 2016). The remaining two studies investigated com-
et al., 2017). Of these 15 outcomes, 10 showed statistically significant binations of public transport and active travel policies, such as the
associations in the expected direction, that is, active travel duration implementation of transportation cost policies (e.g., public transit fares
increased following the bicycle lane intervention, whereas the inverse and parking fees), interventions aimed at changing attitudes towards
was observed in five outcomes. The second highest number of statisti- driving and cycling, and streetscape improvements (Yang et al., 2015;
cally significant associations were observed for active trips (i.e., 8 of 16) Zellner et al., 2016).
(Burbidge and Goulias, 2009; Parker et al. 2011, 2013; Greaves et al., Of the 10 simulation studies, four included active travel as the only
2015; Rissel et al., 2015; Ferenchak and Marshall, 2016; Heesch et al., health-related outcome (Yang and Diez-Roux, 2013; Yang et al., 2015;
2016) and active mode share (i.e., 8 of 14) (Goodman et al., 2013b; Lemoine et al., 2016a; Okushima, 2016). One study also analyzed road
Cook et al., 2016; Ferenchak and Marshall, 2016). Of these, seven ac- traffic injuries and all-cause mortality (Macmillan et al., 2014). The
tive mode share and six active trip outcomes were in the expected di- other five studies (McDonnell and Zellner, 2011; Okushima and
rection (i.e., the share of travel made using active modes and the Akiyama, 2011; Okushima, 2015; Zellner et al., 2016; Zou et al., 2016)
number of active trips increased following the creation of new bicycle reported only on transport-related outcomes, such as mode share,
lanes). mode-specific travel time or mode shift, featuring either bus and/or
Four single component intervention studies assessed the health BRTs as one of the investigated modes.
impacts of living different distances from the site of the intervention. Nine studies (McDonnell and Zellner, 2011; Okushima and
Two of these studies, found positive associations between proximity to Akiyama, 2011; Yang and Diez-Roux, 2013; Okushima 2015, 2016;
cycling infrastructure and the duration of active travel (for 7 of 12 Yang et al., 2015; Lemoine et al., 2016a; Zellner et al., 2016; Zou et al.,
outcomes) (Goodman et al., 2014; Panter and Ogilvie, 2017) as well as 2016) used ABM, and one (Macmillan et al., 2014) used the system
the share of trips made using active modes (n = 4 of 4 outcomes) dynamics framework. Only two studies modeled cities in LMIC, speci-
(Goodman et al., 2013b). fically, Bogota, Colombia (Lemoine et al., 2016a), and Beijing, China
The ten studies reporting on multicomponent interventions (Fig. 2C) (Zou et al., 2016), whereas three were based on highly stylized settings
included a study evaluating the impact of aerial tram infrastructure on (i.e., not grounded in any real location) (McDonnell and Zellner, 2011;
homicide (Cerdá et al., 2012), five studies evaluating bicycle lanes, and Yang and Diez-Roux, 2013; Yang et al., 2015). All studies explored both
four studies focused on combined BRT and bicycle lane interventions the mechanisms driving the behavior of the investigated systems and
(Heinen et al., 2015; Heinen and Ogilvie, 2016; Panter et al., 2016; tested the potential effects of policy interventions.
Chang et al., 2017). Among the multicomponent interventions, one Of the 10 simulation studies, five included both agent-agent (or,
study (Brown et al., 2016b) uniquely reported on physiological and equivalently, persons-persons in the system dynamics model) and
anthropometric outcomes such kilocalorie expenditure and body mass agent-environment (persons-environment) interactions in their models
index. The most frequently reported outcome among multicomponent (McDonnell and Zellner, 2011; Okushima and Akiyama, 2011; Yang
interventions was active travel mode share with just over half of these and Diez-Roux, 2013; Macmillan et al., 2014; Yang et al., 2015). Three
outcomes (i.e., 12 of 23) demonstrating a statistically significant and modeled agent-environment interactions only (Lemoine et al., 2016a;
positive association with the installation of bicycle lanes and/or BRT Zellner et al., 2016; Zou et al., 2016) and the remaining two included
infrastructure (Heinen et al., 2015; Brown et al., 2016; Heinen and just agent-agent interactions (Okushima 2015, 2016). In all studies,
Ogilvie, 2016; Pazin et al., 2016). policy alternatives were considered as new scenarios through the ma-
Of the 10 multicomponent interventions, five investigated whether nipulation of exogenous variables (i.e., there were no agents in the
living different distances from the intervention differentially impacted model that enacted policy decisions based on changes in the model).
health outcomes. These five studies found that people living closer to Of the nine ABM studies, eight (McDonnell and Zellner, 2011;
bicycle lanes, with or without an adjoining BRT line, had higher levels Okushima and Akiyama, 2011; Yang and Diez-Roux, 2013; Okushima
of physical activity (for 3 of 6 outcomes) (Pazin et al., 2016), engaged 2015, 2016; Yang et al., 2015; Lemoine et al., 2016a; Zellner et al., 2016)
in longer periods of active travel (for 3 of 7 duration outcomes) (Panter applied utility functions to represent the agents' decision-making process
et al., 2016; Pazin et al., 2016), and used a higher share of active modes and one used production (‘if-then’) rules (Zou et al., 2016). Decisions
(for 9 of 20 outcomes) (Heinen et al., 2015; Brown et al., 2016; Heinen between modes of transportation were informed by a variety of factors,
and Ogilvie, 2016; Pazin et al., 2016). but most frequently by distance, financial costs and travel time, mode
chosen by other persons (such as members of one's friendship network, or
3.2. Simulation studies those of the community at large), as well as traffic congestion and safety.
Mode chosen by other persons, traffic congestion (and, consequently,
The 10 system-based simulation studies (Table 4) included in this travel time), and traffic safety were the main mechanisms of interaction
review (McDonnell and Zellner, 2011; Okushima and Akiyama, 2011; between people and with their environment.
15
Table 4
Characteristics of system-based simulation studies.
ID Author & year Modelling Intervention type Setting modeled Model description Scenarios modeled Relevant outcomes Key findings
I. Stankov, et al.
method
30 Lemoine et al. ABM BRT Stylized Bogota, Colombia Agents: Adults Type 1 (6 scenarios): Increasing number of Walking for Increasing the number of BRT
(2016a) Interactions: Agent-environment only BRT lanes by substituting existing bus transportation lanes increases minutes walked for
Main processes: Agents choose the lanes; Type 2 (7 scenarios): Changing BRT transportation, observable also
mode of transportation based on total and bus density among nonusers of the BRT;
trajectory costs and time needed to walking time saturates as BRT
reach their workplace; the resource lanes are added to the system
cost of each trip is deducted from a
daily resource budget
Time step: 1 day
Period modeled: 40 days (only last 30
used for analysis)
31 Macmillan System Bikeways Auckland, New Zealand Main variables: Cyclists, normality of Type1 (1 scenario): Business as usual; Type Injury, physical Benefits of all the intervention
et al. (2014) dynamics cycling, cyclist injuries, safety in 2 (6 scenarios): Regional Cycle Network activity, all-cause policies outweighed harms
numbers, real and perceived risk of (development of a partial network of mixed mortality between 6 and 24 times; most
injury, investment in cycle-friendly cycling infrastructure); Type 3 (6 scenario): effective approach would involve
infrastructure, car commuters, Arterial segregated bicycle lanes (ASBL - physical segregation on arterial
average car speed, vehicles on the gradual implementation of a one-way roads and low speed, bicycle-
road at commute time physically segregated lane on each side of friendly local streets; greatest
Interactions: Person-person and every arterial road); Type 4 (6 scenarios): benefits accrued from reduced all-
person-environment Self-explaining roads (SER - gradual cause mortality due to physical
Main processes: Balancing Causal transformation of all local roads into low- activity
Loop 1: Very little change in bicycling speed streets using nontraditional, endemic
infrastructure combined with more road features, such as street narrowing,
people bicycling to work results in trees, and art; Type 5 (12 scenarios):
16
more bicycle crashes, influencing Combination of ASBL and SER
perceived risk of injury and reducing interventions
bicycling. Balancing Causal Loop 2:
Significant shift from cars to bicycles
at peak times results in faster vehicles
and increased real and perceived risk
of bicycling injury, deterring cyclists.
Reinforcing Causal Loop 1: As the
number of cyclists increases, their
political influence rises, thereby
increasing investment in safer
bicycling facilities and encouraging
bicycling. Reinforcing Causal Loop 2:
More cycling to work can influence
the behavior of other road users
toward cycling. Reinforcing Causal
Loop 3: Significant mode shift from
cars to bicycle commutes results in
reduced vehicle numbers and,
therefore, lower likelihood of collision
and higher perceptions of bicycle
safety
Time step: 1 year
Period modeled: 1991 to 2051
(historical simulation from 1991 to
2012)
Table 4 (continued)
method
I. Stankov, et al.
32 McDonnell ABM BRT Stylized setting (generic Agents: Households (car-using Scenario 1: Business as usual (no exclusive Bus share, travel Addition of exclusive bus lane
and Zellner single-direction 5-km road household or bus-using household), bus lane or ancillary measures); Scenario 2: time, bus travel time, increased bus mode share and
(2011) running to a destination bus riders, cars, buses, car entrances, Exclusive bus lane introduced; Scenario 3: car travel time, reduced travel time for bus and car
representing a central bus stops Exclusive bus lane + off-board ticketing length of rush hour users; addition of pre-boarding
business district) Interactions: Agent-agent and agent- introduced; Scenario 4: Exclusive bus ticket machines is the most
environment lane + express bus stops introduced; effective means of increasing bus
Main processes: Each household Scenario 5: Exclusive bus lane + more modal share and reducing journey
produces either bus riders or car users frequent buses introduced; Scenario 6: All times, followed by introduction of
according to their modal choice; mode measures introduced higher bus frequencies
choice is represented as a decision to
switch from bus to car or vice-versa
when the travel time exceeds their
tolerance for lengthening commutes;
journey time of each agent depends on
the presence of other agents on the
road; car users on the road check the
road patch directly in front of them
and wait until a space becomes free;
bus riders need to wait for the bus;
buses start departing as per the
frequency parameter; buses check for
the presence of bus riders and their
capacity; if the bus is already at full
capacity, it does not stop at any more
bus stops
17
Time step: 3 s (20 iterations = 1 min)
Period modeled: 20 days
33 Okushima and ABM Public transport policies Stylized Tokushima, Japan Agents: Individuals Type 1 (1 scenario): Business as usual; Type Share of public Improvement of public transport
Akiyama Interactions: Agent-agent and agent- 2 (5 scenarios): improvement of public transport service combined with discount of
(2011) environment transport service + discount of bus fare; bus fare achieved the highest
Main processes: Agents choose the Type 3 (6 scenarios): Promotion of impacts in modal shift to bus,
commuting mode (car vs. public individual eco-consciousness by followed by strengthening the
transport) based on the level of population-wide environmental education; interaction between individuals
service of each mode and the agents' Type 4 (5 scenarios): Strengthening the
eco-consciousness; level of service of interaction (cooperation) between
each mode is defined by travel time individuals
(calculated based on traffic flow and
capacity) and travel cost; eco-
consciousness consists of evaluating
the traffic environment in terms of
satisfaction with commuting mode
and importance of reduction of CO2
emissions for the agent; eco-
consciousness is also influenced by a
strongly-linked agent (friend) such
that it is improved if the friend has
higher eco-consciousness
Time step: 1 week
Period modeled: 10 years
I. Stankov, et al.
Table 4 (continued)
method
34 Okushima ABM BRT Stylized Tokushima, Japan Agents: Individuals Type 1 (1 scenario): Business as usual; Type Share of sustainable Economic incentive policies
(2015) Interactions: Agent-agent only 2 (1 scenario): Implementation of BRT transport modes maximize the share of sustainable
Main processes: Agents decision system; Type 3 (3 scenarios): economic transport modes; BRT
process consists of vehicle purchase incentive policies (cordon pricing, distance implementation resulted in
and modal shift (car vs. sustainable tolls, and green tax) marginal benefits
transport modes); these processes are
based on environmental concern (fuel
efficiency), mode share in its social
network (social conformity), and
commuting travel time and cost;
vehicle type purchased, and
commuting mode chosen feedback to
decisions made by other agents in the
social network
Time step: 1 week
35 Okushima ABM Active travel promotion Stylized Tokushima, Japan Agents: Individuals Scenario 1: Business as usual; Scenario 2: Active travel Interventions did not result in
(2016) Interactions: Agent-agent only population-wide health education (health increased active travel, with
Main processes: Agents choose the concerns assumed as factor for modal similar results across scenarios
commuting mode (car vs. active travel shift); Scenario 3: promotion of high-
– walking or cycling) based on health- performance bicycles; Scenario 4:
18
consciousness, difficulty in modal population-wide health
shift, and trip distance; health- education + promotion high-performance
consciousness is influenced by social bicycles
network such that it is improved if
friends have higher health-
consciousness
Time step: 1 week
36 Yang and ABM Walking promotion Stylized setting (generic Agents: Households, children, schools Type 1 (1 scenario): School location evenly Walking trips to School locations should be evenly
Diez-Roux city of 64 km2) Interactions: Agent-agent and agent- distributed and children attend nearest school distributed over space and
(2013) environment school; Type 2 (3 scenarios): Changing children should be assigned to the
Main processes: Children choose schools' location and catchment area; Type closest school to maximize the
travel to school by walking or are 3 (16 scenarios): Changing schools' size and number of children who walk to
driven by their parents based on population density; Type 4 (9 scenarios): school; beneficial impacts of
distance to school, traffic safety and Changing traffic safety levels across the city smaller catchment areas and
children's attitude towards walking; higher population density
each day, traffic safety for each area is observed; to improve traffic safety,
updated as a function of the total targeting a smaller area around the
number of people who walk by that school with greater intensity
area, and attitude towards walking is seemed to be more effective
updated as a function of the total
number of children who walk to
school
Time step: 1 day
Period modeled: 5–10 days
I. Stankov, et al.
Table 4 (continued)
method
37 Yang et al. ABM Public transport Stylized setting (generic Agents: Individuals (nested in Type 1 (5 scenarios): Different types of Percentage of Segregation of land use and
(2015) policies + walking city of 64 km2) households) and non-residential segregation (by income, safety level and walking trips relative concentration of mixed
promotion locations land use); Type 3 (4 scenarios): land uses are important
Interactions: Agent-agent and agent- Implementation of transportation cost determinants of income
environment policies (public transit fares, fuel price, and differences in walking; safety and
Main processes: Every day, agents parking fee); Type 4 (3 scenarios): Policies income segregation on their own
choose the travel mode (private aimed at changing attitudes towards do not have large influences on
automobile, public transportation, or driving and cycling + transportation cost income differences in walking;
by walking) to non-residential policies very strong disincentives placed on
locations based on the attitude driving and a few incentives
towards each mode modified by the offered for taking the bus had the
cost of each mode; attitude is potential for greatly increasing
influenced by traffic congestion and walking trips, particularly at lower
traffic safety: more people driving income levels
along the route decreases a person's
attitude towards driving and higher
level of safety increases a person's
attitude towards walking; attitude is
also influenced by travel mode of the
19
agents' friends and family members
Time step: 1 day
38 Zellner et al. ABM Public transport Stylized four Agents: Individuals Type 1 (1 scenario): Business as usual; Type Mode share Shuttles have a significant impact
(2016) policies + active travel neighborhoods in the Interactions: Agent-environment only 2 (1 scenario): Streetscape improvements; and may be robust policies in low-
promotion Chicago Metropolitan Main processes: There are two types of Type 3 (1 scenario): Transportation cost density neighborhoods that have
Region, USA commuters: those going to downtown policies (public transit fares, fuel price, few transportation alternatives to
(Loop commuters) and those going to parking fees); Type 4 (4 scenarios): downtown areas or poor coverage
other places; the latter are randomly Provision of shuttles to the neighborhood of bus service to train station; in
assigned a mode of travel based on transit station; Type 5 (1 scenario): Ideal areas already offering a good
current mode share and do not change streetscape improvements + ideal coverage and reliable provision of
their travel mode during a simulation; transportation cost policy + Ideal bus service to train stations; shifts
Loop commuters choose a mode (walk provision of shuttles to the neighborhood were reinforced by streetscape
to train, bike to destination, bike to transit station; Type 6 (21 scenarios): Same improvements targeted to areas
train, bus to destination, drive to as type 5 but modifying distances from close to shuttle stops; policies that
destination, and shuttle to train) transit stations increase the cost of driving can
based on the probabilities computed reinforce the benefits of improving
from a utility function comprising the provision of public
monetary cost, time, and perceived transportation
safety; perception of safety is updated
based on the observed presence of
other pedestrians and cyclists in the
previous day
Time step: 1 h
Time steps varied significantly across studies, ranging from just
congestion charge can increase the

three seconds in one study to one year in another. Comparatively, there
Travelers shift travel mode and
changing rate to a large extent

situation better when demand
significant because the mode

was less variation in the periods modeled, with four studies (McDonnell
changing rate is not so high;

departure time to make the
increases, but effect is not

and Zellner, 2011; Yang and Diez-Roux, 2013; Lemoine et al., 2016a;
Zellner et al., 2016) modelling one month or less, and four other studies
modelling 10 years or more (Okushima and Akiyama, 2011; Macmillan
et al., 2014; Okushima 2015, 2016).
Key findings
Five of the 10 studies explored economic incentives to encourage

the uptake of mass transit and/or discourage the use of cars (Okushima
and Akiyama, 2011; Okushima, 2015; Yang et al., 2015; Zellner et al.,
2016; Zou et al., 2016). Evaluated strategies included implementation
of congestion fees and green taxes, changes in fuel price and parking
Relevant outcomes
fees, and bus fare discounts. The effects of such policies were found to
provide significant improvements in health-related outcomes. Im-
Modal shift
provements of public transport services, including the implementation

of BRT systems, were also investigated by five studies (McDonnell and
Zellner, 2011; Okushima and Akiyama, 2011; Okushima, 2015;
Lemoine et al., 2016a; Zellner et al., 2016). Besides the implementation
2 (1 scenario): Increasing demand (number
Type 1 (1 scenario): Business as usual; Type
of BRT lanes, other interventions included increasing bus frequency, the

Increasing demand + congestion fees
introduction of express bus stops, and changes in the system density

of travelers; Type 3 (7 scenarios):
(coverage). Overall, these studies showed improvements in active travel

time and bus share.
Models demonstrated that sizeable modal shift and changes in
health outcomes can be achieved from the implementation of multiple,
Scenarios modeled
synergistic policies. For instance, Yang et al., (2015) observed that

combinations of decreasing attitudes towards driving, increasing atti-
tudes toward walking, and economic interventions that encourage
walking were more effective than any single intervention , in increasing
walking for all income levels. Similarly, McDonnell and Zellner, (2011)
observed a 14 percentage point increase (21%–35%) in bus share with
departure time selection; both choices
the road network then decide whether

to stick to the current travel mode and
departure time or switch to new ones;
if the agent finds a better travel mode
and/or departure time, it chooses the
Interactions: Agent-environment only
consequently, mode speed and travel

congestion on the road network and,
experience about the performance of
Main processes: Individuals face two
are influenced by costs, travel time,
the implementation of a BRT system, with the potential of reaching a 21

new options; the distribution of the
and level of congestion on the road
selection (car vs. bus/metro) and
network; individuals accumulate
percentage point increase (i.e., 42% of the mode share) with the ad-
travel modes affect the level of
types of pre-trip choices: mode
ditional introduction of off-board ticketing.

Period modeled: unclear
3.3. Quality appraisal

Agents: Individuals
Time step: unclear

Model description
3.3.1. Empirical studies

The studies included in the review employed a range of sampling
strategies including, attempts to sample the entire population within a
time
given study area (n = 11) (Chen et al., 2012; Goodman et al., 2013;
Goodman et al., 2013b; Dill et al., 2014; Goodman et al., 2014; Panter
and Ogilvie, 2015; Bhatia et al., 2016; Brown et al., 2016; Ferenchak
and Marshall, 2016; Panter and Ogilvie, 2017; Song et al., 2017);
Second ring road of
random sampling (n = 4) (Evenson et al., 2005; Cerdá et al., 2012;

Setting modeled
Beijing, China
Pazin et al., 2016; Chang et al., 2017); non-random or stratified sam-

pling (n = 6) (Burbidge and Goulias, 2009; Parker et al. 2011, 2013;
Langdon, 2015; Rissel et al., 2015; Panter et al., 2016); purposive
sampling (n = 4) (Greaves et al., 2015; Heinen et al., 2015; Heesch
et al., 2016; Heinen and Ogilvie, 2016); traffic counts (n = 2) (Boarnet
Public transport policies
et al., 2005; Cook et al., 2016). The remaining studies (Jensen, 2008;
ABM: agent-based modelling; BRT: bus rapid transit.
Brown et al., 2016b) provided no description of the sampling strategy

Intervention type
used. The representativeness of the recruited participants, as de-

termined by the participant response rate and the sampling strategy
used, was unclear for most studies (69%) (Evenson et al., 2005; Jensen,
2008; Parker et al., 2011; Goodman et al., 2013b; Dill et al., 2014;
Goodman et al., 2014; Greaves et al., 2015; Heinen et al., 2015;
Langdon, 2015; Panter and Ogilvie, 2015; Rissel et al., 2015; Brown
Modelling
et al., 2016; Cook et al., 2016; Heesch et al., 2016; Heinen and Ogilvie,
method
ABM
2016; Panter et al., 2016; Brown et al., 2016b; Chang et al., 2017;
Panter and Ogilvie, 2017; Song et al., 2017), only one study was
Table 4 (continued)
deemed to have a truly representative sample (Goodman et al., 2013),

Author & year
while the rest were somewhat representative (n = 7) (Burbidge and

Zou et al.
Goulias, 2009; Cerdá et al., 2012; Chen et al., 2012; Parker et al., 2013;
(2016)
Bhatia et al., 2016; Ferenchak and Marshall, 2016; Pazin et al., 2016) or
not very representative (n = 1) (Boarnet et al., 2005).
39
ID
Most studies in the review did not include a control or a comparison
20
group (n = 14) (Boarnet et al., 2005; Evenson et al., 2005; Burbidge and other two (Yang and Diez-Roux, 2013; Zou et al., 2016) used categorical
Goulias, 2009; Parker et al., 2011; Goodman et al. 2013b, 2014; Heinen calibration only (i.e., searched for parameter values that produce model
et al., 2015; Langdon, 2015; Bhatia et al., 2016; Cook et al., 2016; results within a range acceptably close to data). Six studies (Okushima
Ferenchak and Marshall, 2016; Heinen and Ogilvie, 2016; Panter et al., and Akiyama, 2011; Yang and Diez-Roux, 2013; Okushima 2015, 2016;
2016; Chang et al., 2017), of those that did, the extent to which the Zellner et al., 2016; Zou et al., 2016) did not validate the results of
control and intervention groups were similar at baseline was unclear for baseline scenarios against expected (in purely stylized models) or ob-
eight studies (Jensen, 2008; Parker et al., 2013; Greaves et al., 2015; served (in models grounded on real locations) outcomes, and only two
Panter and Ogilvie, 2015; Brown et al., 2016; Brown et al., 2016b; Panter studies (McDonnell and Zellner, 2011; Macmillan et al., 2014) compared
and Ogilvie, 2017; Song et al., 2017), while the rest featured control and outcomes using either qualitative or quantitative means. Five studies
intervention groups that were comparable at baseline, providing either a (Yang and Diez-Roux, 2013; Okushima 2015, 2016; Yang et al., 2015;
quantitative characterization of the two groups (n = 5) (Cerdá et al., Zou et al., 2016) did not conduct either sensitivity or uncertainty ana-
2012; Chen et al., 2012; Dill et al., 2014; Heesch et al., 2016; Pazin et al., lyses, only one conducted both (Macmillan et al., 2014), and the re-
2016) or general statement to that effect (n = 2) (Goodman et al., 2013; maining four (McDonnell and Zellner, 2011; Okushima and Akiyama,
Rissel et al., 2015). Included studies featured more self-reported (n = 13) 2011; Lemoine et al., 2016a; Zellner et al., 2016) ran sensitivity analyses
(Evenson et al., 2005; Burbidge and Goulias, 2009; Goodman et al., only. Appendix 5 provides more information about the quality appraisal
2013; Goodman et al. 2013b; Goodman et al. 2014; Heinen et al., 2015; of included simulation studies.
Panter and Ogilvie, 2015; Heinen and Ogilvie, 2016; Panter et al., 2016;
Pazin et al., 2016; Chang et al., 2017; Panter and Ogilvie, 2017; Song 4. Discussion
et al., 2017) than objectively measured outcomes (n = 7) (Boarnet et al.,
2005; Jensen, 2008; Parker et al. 2011, 2013; Cerdá et al., 2012; Dill Of the 39 empirical and system-based simulation studies identified,
et al., 2014; Brown et al., 2016). There were also studies that reported on the majority focused on bicycle lanes and BRT systems. Notably, only
both self-reported and objective outcomes (n = 9) (Chen et al., 2012; one study evaluated aerial trams, and none investigated Open Streets
Greaves et al., 2015; Langdon, 2015; Rissel et al., 2015; Bhatia et al., programs. Moreover, most studies (n = 24) focused on HIC, while only
2016; Cook et al., 2016; Ferenchak and Marshall, 2016; Heesch et al., five studies explored cities in LMIC. In empirical studies, bicycle lane
2016; Brown et al., 2016b). interventions were associated with increases in physical activity and
The timing and duration of pre-versus post-intervention assessments active transport. Similarly, BRT systems with an adjacent bicycle lane
varied substantively across studies. Pre-intervention assessments were were found to promote active travel and walking for transport and re-
commonly collected over a period spanning an average of 22 months creation. The sole aerial tram study reported a significant decrease in
and ranging from three months to five years. On average, these as- homicides following the aerial tram installation. There was also some
sessments were conducted 14 months before the intervention and evidence to suggest that multiple component interventions may be
ranged from immediately before to eight years pre-intervention. Post- more effective than single component interventions in increasing phy-
intervention assessments were collected on average over a period sical activity. System-based simulation studies showed that economic in-
spanning 17 months and ranging from five months to two and a half centives designed to disincentivise car use, and policies designed to
years. Intervention effects on outcomes were assessed on average eight improve the public transportation system, can have positive impacts on
and a half months after the intervention though some studies conducted active travel time and bus share. Consistent with empirical studies
post-intervention assessments immediately after, while others con- evaluating multicomponent interventions, systems-based simulations
ducted their first assessments four years after. often reported synergistic effects of multiple interventions and policies.
Of the 29 empirical studies, 17 studies experienced loss to follow-up The quality of included studies was mixed. Most empirical studies did
to differing degrees: < 30% attrition was reported by two studies not include a control or comparison group and for those that did, it was
(Cerdá et al., 2012; Dill et al., 2014), 30–59% by 13 studies (Evenson largely unclear to what extent the control and intervention groups were
et al., 2005; Goodman et al., 2013b; Goodman et al., 2014; Greaves similar at baseline. Empirical studies also reported a range of sampling
et al., 2015; Heinen et al., 2015; Panter and Ogilvie, 2015; Rissel et al., approaches. There existed substantive uncertainty about the general-
2015; Brown et al., 2016; Heinen and Ogilvie, 2016; Panter et al., 2016; izability of study findings because 21 of the 29 included studies fea-
Pazin et al., 2016; Brown et al., 2016b; Song et al., 2017), and 60–89% tured populations that were either not representative of the study re-
by two studies (Burbidge and Goulias, 2009; Panter and Ogilvie, 2017). gion or whose representativeness was unclear. System-based simulation
Please refer to Appendix 4 for more information about the quality ap- studies commonly provided justifications for the assumptions made and
praisal of included studies. the equations used, although several were highly abstract models. Most
simulation studies were also informed by empirical data. However,
3.3.2. Simulation studies calibration and validation procedures and uncertainty analysis were
Overall, the system-based simulation studies made explicit the infrequently conducted, and the periods modeled in a handful of studies
models’ assumptions and structure. Nine studies provided justifications appeared relatively short (less than 1 month), which may have im-
for all or most of their assumptions (McDonnell and Zellner, 2011; pacted study findings.
Okushima and Akiyama, 2011; Yang and Diez-Roux, 2013; Macmillan The most common policy evaluated by empirical studies pertained
et al., 2014; Okushima 2015, 2016; Yang et al., 2015; Zellner et al., to bicycle infrastructure. These studies generally reported beneficial ef-
2016; Zou et al., 2016). Only one study did not provide justifications for fects on bicycle mode share and active transport duration and number
the equations used (McDonnell and Zellner, 2011) while the four other of trips. Some adverse effects on injuries were documented although
studies (Macmillan et al., 2014; Yang et al., 2015; Lemoine et al., these were reported by a single study. However, only one of these
2016a; Zellner et al., 2016) provided justifications only for some of the longitudinal studies was based in or explored cities in LMIC. Prior re-
equations. All 10 system-based simulation studies used empirical views have also identified research on bicycle infrastructure in LMIC as
sources to inform their parameters. important gaps (Fraser and Lock, 2010; Yang et al., 2010). Cross-sec-
Calibration and validation procedures and sensitivity or uncertainty tional studies focused on LMIC report consistent effects of bicycle lanes
analysis were infrequently assessed. Among the four studies with some on active travel. For example, Florindo et al. found that people living
parameters that could not be informed by empirical data (McDonnell and within 500 meters of bicycle paths in Sao Paolo, Brazil were more than
Zellner, 2011; Okushima and Akiyama, 2011; Yang and Diez-Roux, 2013; twice more likely to engage in cycling for transportation than those who
Zou et al., 2016), two (McDonnell and Zellner, 2011; Okushima and lived further away (Florindo et al., 2018). Another study based in
Akiyama, 2011) did not calibrate the unknown parameter values and the Taiwan, found an association between residents’ perceptions of bicycle
21
lanes in their neighborhood and past week cycling for transportation important initiatives are necessary.
among adults (Liao et al., 2015). Additional longitudinal evaluation We observed several points of alignment between the system-based
studies are needed to determine the impact of the rapid growth in bi- simulation papers and the longitudinal evaluation studies included in our
cycle infrastructures that is occurring in many LMIC, in contexts that review. For example, both empirical and simulation studies focused
are very different from those of HIC. predominantly on estimating the impact of policies encompassing ex-
The second most explored policy focused on BRT systems. These pansions and or improvements in public transport infrastructure (e.g.,
studies, mostly focused on HIC, found that BRT systems increase active express bus lanes, creation of BRT system) and service delivery (e.g.,
travel and walking for transport and recreation. There are however over increased frequency of buses), or incentives for public transportation
160 cities around the world with operating BRT systems (BRT+ Centre (e.g., fare changes) on health. Evidence across both bodies of literature
of Excellence and EMBARQ, 2019). Among them are cities from a range suggests that multi-pronged interventions may be more effective than
of LMIC in Latin America (e.g., Brazil, Colombia, Argentina, Mexico), single-component interventions in shaping some health outcomes. This
Asia (e.g., China, India, Taiwan, Vietnam) and the African continent observation is in keeping with published research (van Sluijs et al.,
(e.g., Morocco, South Africa, Nigeria, Uganda). While few longitudinal 2007) and Social Ecological Theory which posits that ecological ap-
studies have investigated these transport systems in LMIC, cross-sec- proaches, which seek to enact change at multiple levels of a system, are
tional analyses suggest that BRT use may positively impact health more effective than those targeted toward just one level (Green et al.,
outcomes. For example, using cross-sectional survey data, Lemoine 1996; Sallis et al., 2008). However, it was unclear from our systematic
et al., 2016b found that BRT use in Bogota, Colombia, was associated review which of the intervention components in a given multi-compo-
with around 12 min of moderate-to-vigorous physical activity each day nent study were the drivers of the observed health impacts.
(Lemoine et al., 2016b). Similarly, Bartels et al. used cross-sectional Another important observation was that all simulation studies
intercept surveys of BRT passengers in Cape Town, South Africa, and captured by our review used empirical sources to inform the selection
found that BRT-users engaged in significantly longer periods of physical and characterization of model parameters, highlighting the importance
activity per week and were over twice more likely to achieve re- of robust empirical studies focused on exploring the influence of
commended physical activity guidelines than non-users (Bartels et al., transportation policies on health outcomes across a range of contexts,
2016). These findings align closely with the general findings of our not just HIC. This reliance on empirical data and the fact that most
review which suggest that BRT interventions represent effective means empirical studies we identified were conducted in HIC could explain
for increasing physical activity and active travel. why most of the simulation-based papers included in our review also
Aerial trams form part of the transit infrastructure, both in LMIC and explored high-income contexts.
HIC around the world (Alshalalfah et al., 2012). However, our review of We observed heterogeneity in how studies operationalized people's
the literature indicates that aerial trams remain relatively under-stu- exposure to a given transportation intervention. Some studies compared
died, particularly with respect to their health-related impacts. We found geographic locations while others considered intervention impacts
only one longitudinal evaluation of aerial trams and no system-based among people living in an area of intervention influence. These areas
simulation studies. The sole study included in our review, which capi- were either defined using existing geographic units (e.g., census blocks)
talized on a natural experiment, found that the construction of the or buffers centred on the focal point of an intervention. Only a subset of
Metrocable (i.e., aerial tram) in Medellin, Colombia, resulted in sig- empirical studies compared the health effects of a given intervention for
nificant reductions in homicide rates in the neighborhoods surrounding those living different distances from the intervention site, despite
the new aerial tram (Cerdá et al., 2012). Other studies which did not contemporary debates advocating for a pluralistic measurement ap-
meet the inclusion criteria of our review investigated the impacts of proach (Laatikainen et al., 2018). Most of these studies reported more
aerial trams on relatively distal outcomes indirectly related to health significant and positive effects on health outcomes for those living
and health-related behavior, for example, employment access and closer to the site of an intervention than those living further away; a
travel time. One of these studies found that the Metrocable significantly finding consistent with existing research (McCormack and Shiell, 2011;
improved access to the central business district and thereby more than Djurhuus et al., 2014).
doubled employment opportunities for aerial tram users, including low- The studies in our review assessed a wide range of health outcomes,
income groups (Bocarejo et al., 2014). Moreover, in their study of the however, anthropometric/physiological measures, and mortality out-
Mi Teleférico aerial tram in La Paz, Bolivia, Garsous et al., (2019) found comes were infrequently reported. Moreover, none of the studies in our
that users of the aerial tram reduced their travel time by around 20%. review considered disease outcomes such as diabetes, or respiratory and
Travel time reductions such as these have variously been linked to mental health outcomes. And, strikingly, few studies addressed issues of
improved access to health care, employment and education as well as equity by exploring intervention effects using stratified analyses which
opportunities for leisure-time activities including physical activity would have enabled critical insights into health inequalities by socio-
(Garsous et al., 2019). economic status and demographic factors such as race and gender. This
Despite the increasing prominence of Open Streets programs in both relative underrepresentation of equity-grounded research has been ob-
LMIC and HIC, our review did not identify any longitudinal evaluations served in other reviews of transportation systems in LMIC (Yanez-Pagans
of the health impacts of these initiatives. Sarmiento et al. identified 38 et al., 2018) and those investigating built environment influences on
Open Streets programs implemented across 11 different countries in physical activity and active transport more broadly (Smith et al., 2017).
2010, most of which were concentrated in Latin America (Sarmiento The accessibility of transportation is an important predictor of
et al., 2010). Another review identified Open Streets initiatives hosted health care access (Syed et al., 2013), and employment, which has been
in 47 different US cities (Kuhlberg et al., 2014). Furthermore, cross- linked to a range of health-related behaviors through its influence on
sectional evidence suggests that Open Streets initiatives can confer both income and time scarcity (Venn and Strazdins, 2017). Importantly,
meaningful public health benefits. For example, using information research based in Latin America has shown that patterns in access to
synthesized from the Open Streets programs, Sarmiento et al. estimated BRT, by income, can vary from one city to another. For example, the
important potential contributions of Ciclovias to physical activity BRT system in Lima, Peru has been shown to predominantly benefit
(Sarmiento et al., 2010). Positive associations have also been observed middle- and higher-SES groups due to the systems limited coverage of
among school children. For example, Triana et al. found that frequent areas with high concentrations of poor residents (Oviedo et al., 2019).
Ciclovia use among Colombian school children was associated with On the other hand, in Cali, Colombia, the highest levels of access to a
significantly lower levels of sedentary time and higher moderate-to- new BRT system were observed among middle income groups, while
vigorous physical activity on Sundays, but interestingly, not weekdays residents of predominantly low and high income neighborhoods had far
(Triana et al., 2019). However, longitudinal evaluations of these more limited access (Delmelle and Casas, 2012). Such inequalities
22
however are not only observed in LMIC, they have also been reported in represents an important area for future research. Participatory pro-
HIC such as Australia (Ricciardi et al., 2015), with observed inequalities cesses such as group model building (Hovmand, 2014) which seek to
in access spanning both age and the socioeconomic spectrum. Thus, elicit the perspectives of diverse stakeholders, can play an important
additional evaluations on the impact of these types of interventions on role in informing the design of simulation models. Moreover, they have
equity outcomes is sorely needed. the potential to elucidate novel evaluation targets and foster inter-
sectoral and community partnerships which can play an important role
4.1. Limitations in the sustainability and longevity of interventions.
Third, prospective studies, both empirical and simulation-based,
This review should be considered with a few limitations in mind. Our focusing on BRT, bicycle lanes, aerial trams and Open Streets programs
review was necessarily limited through the exclusion of papers that as- ought to explore the potential environmental and health co-benefits of
sessed the impacts of BRT, aerial trams, bicycle lanes and Open Streets these policies. The high expansion of new programs in both HIC and
programs on outcomes that have implications for health, such as acci- LMIC provide a unique opportunity for natural experiments. For ex-
dents, crashes and traffic-related air pollution, but that are not them- ample, studies investigating bicycle lane interventions would be well
selves health outcomes. We also excluded studies employing non-long- placed to investigate changes in transport-related air pollution and the
itudinal study designs and other simulation methods such as health respiratory health of city residents alongside and in interaction with
impact assessment models, social network analysis, microsimulation or changes in mode share and physical activity during transport and lei-
more conceptual/qualitative models arising from participatory ap- sure time. More studies evaluating the influence of bicycle lanes on
proaches such as group model building. To focus the scope of our review, injury outcomes are also required. Fourth, studies employing stratified
we also excluded studies evaluating light rail transit systems, as well as analysis, by for example, gender, income, age and race are required to
those focused specifically on bicycle boxes, intersection crossings or explore the impact of these transport policies on health disparities
roundabouts as opposed to continuous street segments. Given time and across a range of outcomes including anthropometric and physiological
resource constraints we did not contact study authors for clarification measures as well as respiratory and other disease outcomes. Relatedly,
where information was missing or unclear. Most of the studies we in- to ensure study quality, the design of future system-based simulation
cluded in the review were based in HIC which may limit the general- studies ought to reflect an alignment between the outcomes of interest
izability of our findings to LMIC. Our assessment of the quality of system- and the timeframes being modeled.
based simulation studies may also be limited given the lack of guidance Finally, researchers seeking to advance research in this area, parti-
on how to assess the quality of these types of studies. Finally, several cularly in LMIC may benefit from the use of innovative and relatively
papers included in the review evaluated the same intervention. For ex- cost-effective data collection methods such as street imagery (e.g.,
ample, five different papers, all conducted as part of the iConnect Study, Google Street View and Bing StreetSide) to capture granular informa-
evaluated the same set of bicycle lane interventions implemented in tion about the physical environment and behavioral data. Street ima-
three cities in the United Kingdom (UK) (Goodman et al. 2013b, 2014; gery has widely been tested as a built environment audit tool, including
Panter and Ogilvie 2015, 2017; Song et al., 2017), while four papers its predictive capacity in documenting relatively small-scale historic
evaluated the same combined BRT and bicycle lane intervention in changes to the built environment (Candido et al., 2018). There is also
Cambridge, UK (Heinen et al., 2015; Heinen and Ogilvie, 2016; Panter evidence to suggest that it can be used to estimate pedestrian counts
et al., 2016; Chang et al., 2017). Given this overlap, our findings re- (Yin et al., 2015), and city-level travel patterns including census-re-
present outcomes for just 19 unique interventions variously evaluated by ported mode share (i.e., walking and public transit use, cycling, mo-
the 29 empirical studies included in the review. torcycle and car use) as well as survey-reported past-month participa-
tion in cycling (Goel et al., 2018). Given these features, and with years
4.2. Recommendations for future research of historical data available, street imagery may provide an avenue for
the conduct of retrospective longitudinal policy evaluations of transport
The findings of our review highlight several important areas for policies, and a promising means of complementing traditional data
future research. First, more evaluation and system-based simulation collection methods, particularly in LMIC.
studies are needed to assess the influence of bicycle lanes, BRT systems, Researchers can also benefit from using a citizen science approach
aerial trams and Open Streets programs on health outcomes, particu- which “empowers residents to collect diagnostic information about
larly in LMIC. This is particularly important as differences between HIC their community environment, prioritize areas of concern, and engage
and LMIC have been observed in studies investigating associations be- in cross-sector collaboration to generate practical and impactful solu-
tween built environment characteristics and physical activity, for ex- tions” (King et al., 2016, p.31). These approaches have successfully
ample (Cleland et al., 2019). In the case of empirical studies, rigorous been used in Latin American countries to collect neighborhood-level
designs including representative population samples, valid comparison information as well as qualitative data relating to a range of initiatives,
groups, and before and after assessments are needed. On the other including Open Streets programs (King et al., 2016). Other emerging
hand, the use of evidence to justify model rules and parameters is cri- approaches that might be leveraged to advance future studies seeking to
tical for simulation studies. Second, there also exists a need for studies monitor the impact of transport policies include drone technology,
replicating policy evaluations in different cities to determine the extent which has been used to collect data on pedestrian counts (Park and
to which city-level factors impact the effectiveness of interventions Ewing) and deep learning image analysis, which may afford an espe-
overall and for different population subgroups. These studies may in cially promising and cost-effective means of estimating local environ-
turn inform simulation-based studies which have the capacity to iden- mental exposures and spatial inequalities in income, education, em-
tify under what conditions a given policy or combinations of policies ployment and health, particularly in LMIC (Suel et al., 2019;
may be most effective in promoting health outcomes across the socio- Weichenthal et al., 2019).
economic and demographic spectrum. Systems-based simulation
methods can be especially useful as policy decision-tools, particularly in 5. Conclusion
LMIC where the assessment of large-scale population-level interven-
tions may be prohibitively expensive or impractical to test in the real The literature base encompassing longitudinal evaluations, and
world (Hammond, 2015). These methods can also raise new questions system-based simulation studies exploring the health impacts of BRT sys-
and in turn inform the focus of empirical research and evaluation stu- tems, bicycle lane and aerial tram infrastructure and Open Streets pro-
dies (Diez Roux, 2019). To support the use of system-based simulation grams varies widely by transportation policy and geographical context.
methods, tools guiding the assessment of quality for these studies This review contributes to the literature by highlighting several important
23
gaps in knowledge. Specifically, it highlights an underrepresentation of Babisch, W., 2006. Transportation noise and cardiovascular risk: updated review and
certain types of transportation policies (i.e., aerial trams and Open Streets synthesis of epidemiological studies indicate that the evidence has increased. Noise
Health 8 (30), 1–29.
programs), outcomes (e.g., physiological, anthropometric and health Bartels, C., Kolbe-Alexander, T., Behrens, R., Hendricks, S., Lambert, E.V., 2016. Can the
equity measures), and countries (i.e, LMIC) within the literature. By syn- use of Bus Rapid Transit lead to a healthier lifestyle in urban South Africa? The SUN
thesizing the available research, this review also identifies bike lanes and Study.". J. Trans. Health 3 (2), 200–210.
Becerra, J., Reis, R., Frank, L., Ramirez-Marrero, F., Welle, B., Arriaga Cordero, E.,
BRT systems as promising transportation initiatives for promoting physical Mendez Paz, F., Crespo, C., Dujon, V., Jacoby, E., Dill, J., Weigand, L., Padin, C.,
activity and active travel at the population-level. Finally, it provides a 2013. Transport and health: a look at three Latin American cities. Cad. Saúde Pública
series of recommendations for future research designed to bridge critical 29 (4), 654–666.
Bhatia, D., Richmond, S.A., Loo, C.K.J., Rothman, L., Macarthur, C., Howard, A., 2016.
gaps in understanding, and to support the advancement of the public "Examining the impact of cycle lanes on cyclist-motor vehicle collisions in the city of
health agenda through transportation policy. Toronto.". J. Trans. Health 3 (4), 523–528.
Boarnet, M., Day, K., Anderson, C., McMillan, T., Alfonzo, M., 2005. California's safe
routes to school program: impacts on walking, bicycling, and pedestrian safety. J.
Author contributions
Am. Plann. Assoc. 71 (3), 301–317.
Bocarejo, J.P., Portilla, I.J., Velásquez, J.M., Cruz, M.N., Peña, A., Oviedo, D.R., 2014. "An
IS initiated and conceptualised the review, IS and LMTG developed innovative transit system and its impact on low income users: the case of the
the search strategy, the extraction and reporting tools and screened Metrocable in Medellín.". J. Transport Geogr. 39, 49–61.
Borrell, C., Pons-Vigués, M., Morrison, J., Díez, È., 2013. Factors and processes influen-
studies for inclusion with the help of NG and BH. IS, LMTG, MAM, FM cing health inequalities in urban areas. Journal of Epidemiology and Community
and JDM performed duplicate extractions and assessment of quality. IS Health.
and LMTG synthesized the studies in the review. IS wrote the first draft Brown, B.B., Smith, K.R., Tharp, D., Werner, C.M., Tribby, C.P., Miller, H.J., Jensen, W.,
2016. A complete street intervention for walking to transit, nontransit walking, and
of the manuscript and critically revised subsequent versions with bicycling: a quasi-experimental demonstration of increased use.". J. Phys. Activ.
writing contributions from LMTG and AVDR. AVDR supervised the Health 13 (11), 1210–1219.
project and provided critical guidance throughout. All authors dis- Brown, B.B., Tharp, D., Tribby, C.P., Smith, K.R., Miller, H.J., Werner, C.M., 2016b.
Changes in bicycling over time associated with a new bike lane: relations with
cussed the search strategy, extraction and reporting tools and results, kilocalories energy expenditure and body mass index. J. Trans. Health 3 (3),
suggested revisions and contributed to the final manuscript. 357–365.
Global BRTData. Version 3.43 BRT+ Centre of Excellence and EMBARQ, 2019. EMBARQ
and centre of excellence for BRT. http://www.brtdata.org.
Funding sources Brunekreef, B., Holgate, S.T., 2002. "Air pollution and health.". Lancet 360 (9341),
1233–1242.
The Salud Urbana en América Latina (SALURBAL)/Urban Health in Burbidge, S.K., Goulias, K.G., 2009. "Evaluating the impact of neighborhood trail devel-
opment on active travel behavior and overall physical activity of suburban re-
Latin America project is funded by the Wellcome Trust, UK [grant
sidents.". Transport. Res. Rec. 2135 (1), 78–86.
205177/Z/16/Z]. More information about the project can be found at Candido, R.L., Steinmetz-Wood, M., Morency, P., Kestens, Y., 2018. Reassessing urban
www.lacurbanhealth.org. We acknowledge the support of SALURBAL health interventions: back to the future with Google street View time machine. Am. J.
investigators. For more information on SALURBAL and to see a full list Prev. Med. 55 (5), 662–669.
Cerdá, M., Morenoff, J.D., Hansen, B.B., Tessari Hicks, K.J., Duque, L.F., Restrepo, A.,
of investigators see https://drexel.edu/lac/salurbal/team/. Diez-Roux, A.V., 2012. Reducing violence by transforming neighborhoods: a natural
LMTG worked under the auspices of the Centre for Diet and Activity experiment in Medellín, Colombia.". Am. J. Epidemiol. 175 (10), 1045–1053.
Research (CEDAR), a UKCRC Public Health Research Centre of Chang, A., Miranda-Moreno, L., Cao, J., Welle, B., 2017. The effect of BRT implementa-
tion and streetscape redesign on physical activity: a case study of Mexico City.".
Excellence which is funded by the British Heart Foundation, Cancer Transport. Res. Pol. Pract. 100, 337–347.
Research UK, Economic and Social Research Council, Medical Research Chen, L., Chen, C., Srinivasan, R., McKnight, C.E., Ewing, R., Roe, M., 2012. "Evaluating
Council, the National Institute for Health Research, UK and the the safety effects of bicycle lanes in New York City.". Am. J. Pub. Health 102 (6),
1120–1127.
Wellcome Trust [grant MR/K023187/1]. JDM was funded by the Cleland, C., Reis, R.S., Ferreira Hino, A.A., Hunter, R., Fermino, R.C., Koller de Paiva, H.,
Universidad de Ibagué, Colombia [Project 17468 INT]. FM was funded Czestschuk, B., Ellis, G., 2019. Built environment correlates of physical activity and
by the National Institutes of Health Fogarty International Center, USA sedentary behaviour in older adults: a comparative review between high and low-
middle income countries.". Health Place 57, 277–304.
[grant D43TW010540], and by the FAPA grant from the Universidad de
Cook, T.J., O'Brien, S.W., Jackson, K.N., Findley, D.J., Searcy, S.E., 2016. "Behavioral
los Andes, Colombia. effects of completing a critical link in the american tobacco trail.". Transport. Res.
Rec. 2598 (1), 19–26.
Covidence, 2017. Covidence: A Cochrane Technology Platform. Covidence, Melbourne.
Declaration of competing interests
Delmelle, E.C., Casas, I., 2012. Evaluating the spatial equity of bus rapid transit-based
accessibility patterns in a developing country: the case of Cali, Colombia. Transport
The authors declare that they have no known competing financial Pol. 20, 36–46.
Diez Roux, A.V., 2019. The unique space of epidemiology: drawing on the past to project
interests or personal relationships that could have appeared to influ- into the future. Am. J. Epidemiol 188 (5), 886–889.
ence the work reported in this paper. Dill, J., McNeil, N., Broach, J., Ma, L., 2014. Bicycle boulevards and changes in physical
activity and active transportation: findings from a natural experiment. Prev. Med. 69,
S74–S78.
Acknowledgements Djurhuus, S., Hansen, H., Aadahl, M., Glümer, C., 2014. The association between access to
public transportation and self-reported active commuting. Int. J. Environ. Res. Publ.
The authors thank Brittany Hodges (BH) for her work as a second Health 11 (12), 12632–12651.
Evenson, K.R., Herring, A.H., Huston, S.L., 2005. "Evaluating change in physical activity
reviewer in screening abstracts for inclusion in the review. with the building of a multi-use trail.". Am. J. Prev. Med. 28 (2, Suppl. 2), 177–185.
Ferenchak, N.N., Marshall, W.E., 2016. The relative (in)effectiveness of bicycle sharrows
Appendix A. Supplementary data on ridership and safety outcomes. In: Transportation Research Board 95th Annual
Meeting. Washington DC, United States, Transportation Research Board: 17, .
https://www.sciencedirect.com/science/article/pii/S026427511731329X.
Supplementary data to this article can be found online at https:// Florindo, A.A., Barrozo, L.V., Turrell, G., Barbosa, J., Cabral-Miranda, W., Cesar, C.L.G.,
doi.org/10.1016/j.envres.2020.109519. Goldbaum, M., 2018. Cycling for transportation in Sao Paulo City: associations with
bike paths, train and subway stations.". Int. J. Environ. Res. Publ. Health 15 (4).
Fraser, S.D.S., Lock, K., 2010. "Cycling for transport and public health: a systematic re-
References view of the effect of the environment on cycling.". Eur. J. Publ. Health 21 (6),
738–743.
Alshalalfah, B., Shalaby, A., Dale, S., Othman, F.M.Y., 2012. Aerial ropeway transporta- Garsous, G., Suárez-Alemán, A., Serebrisky, T., 2019. Cable cars in urban transport: travel
tion systems in the urban environment: state of the art. J. Transport. Eng. 138 (3), time savings from La Paz-El Alto (Bolivia). Transport Pol. 75, 171–182.
253–262. Goel, R., Garcia, L.M., Goodman, A., Johnson, R., Aldred, R., Murugesan, M., Brage, S.,
24
Bhalla, K., Woodcock, J., 2018. Estimating city-level travel patterns using street Moher, D., Liberati, A., Tetzlaff, J., Altman, D.G., The PRISMA Group, 2009. Preferred
imagery: a case study of using Google Street View in Britain.". PloS One 13 (5), reporting items for systematic reviews and meta-analyses: the PRISMA statement.".
e0196521. PLoS Med. 6 (7), e1000097.
Goodman, A., Panter, J., Sharp, S.J., Ogilvie, D., 2013. Effectiveness and equity impacts of Nianogo, R.A., Arah, O.A., 2015. Agent-based modeling of noncommunicable diseases: a
town-wide cycling initiatives in England: a longitudinal, controlled natural experi- systematic review.". Am. J. Pub. Health 105 (3), e20–e31.
mental study.". Soc. Sci. Med. 97, 228–237. Okushima, M., 2015. "Simulating social influences on sustainable mobility shifts for
Goodman, A., Sahlqvist, S., Ogilvie, D., 2013b. Who uses new walking and cycling in- heterogeneous agents.". Transportation 42 (5), 827–855.
frastructure and how? Longitudinal results from the UK iConnect study. Prev. Med. Okushima, M., 2016. Multi-agent Simulation of Commuting Modal Shift Considering with
57 (5), 518–524. Health Conscious. 2016 Joint 8th International Conference on Soft Computing and
Goodman, A., Sahlqvist, S., Ogilvie, D., 2014. New walking and cycling routes and in- Intelligent Systems and 17th International Symposium on Advanced Intelligent
creased physical activity: one- and 2-year findings from the UK iConnect Study. Am. Systems. IEEE.
J. Pub. Health 104 (9), e38–e46. Okushima, M., Akiyama, T., 2011. "Multi-agent transport simulation model for eco-
Greaves, S., Ellison, R., Ellison, A., Crane, M., Rissel, C., Standen, C., 2015. Changes in commuting promotion planning.". J. Adv. Comput. Intell. Intell. Inf. 15 (7), 911–918.
Cycling Following an Infrastructure Intervention. 37th Australasian Transportation Oviedo, D., Scholl, L., Innao, M., Pedraza, L., 2019. Do bus rapid transit systems improve
Research Forum. Transportation Research Board: 13, Sydney, Australia. accessibility to job opportunities for the poor? The case of Lima, Peru.". Sustainability
Green, L.W., Richard, L., Potvin, L., 1996. "Ecological foundations of health promotion.". 11 (10), 1–24.
Am. J. Health Promot. 10 (4), 270–281. Panter, J., Heinen, E., Mackett, R., Ogilvie, D., 2016. "Impact of new transport infra-
Hammond, R.A., 2015. Appendix A Considerations and Best Practices in Agent-Based structure on walking, cycling, and physical activity.". Am. J. Prev. Med. 50 (2),
Modeling to Inform Policy. Assessing the Use of Agent-Based Models for Tobacco e45–e53.
Regulation. R. Wallace, A. Geller and V. Ogawa. National Academies Press (US), Panter, J., Ogilvie, D., 2015. "Theorising and testing environmental pathways to beha-
Washington, DC. viour change: natural experimental study of the perception and use of new infra-
Heesch, K.C., James, B., Washington, T.L., Zuniga, K., Burke, M., 2016. Evaluation of the structure to promote walking and cycling in local communities.". British Med. J. Open
Veloway 1: a natural experiment of new bicycle infrastructure in Brisbane, 5 (9), e007593.
Australia.". J. Trans. Health 3 (3), 366–376. Panter, J., Ogilvie, D., 2017. Can environmental improvement change the population
Heinen, E., Ogilvie, D., 2016. Variability in baseline travel behaviour as a predictor of distribution of walking? J. Epidemiol. Community 71 (6), 528.
changes in commuting by active travel, car and public transport: a natural experi- Park, K. and R. Ewing "The usability of Unmanned Aerial Vehicles (UAVs) for pedestrian
mental study. J. Trans. Health 3 (1), 77–85. observation." J. Plann. Educ. Res. 0(0): 0739456X18805154.
Heinen, E., Panter, J., Mackett, R., Ogilvie, D., 2015. "Changes in mode of travel to work: Parker, K.M., Gustat, J., Rice, J.C., 2011. Installation of bicycle lanes and increased ri-
a natural experimental study of new transport infrastructure.". Int. J. Behav. Nutr. dership in an urban, mixed-income setting in New Orleans, Louisiana. J. Phys. Activ.
Phys. Activ. 12 (1), 81. Health 8 (Suppl. 1), S98–s102.
Higgins, J., Lasserson, T., Chandler, J., Tovey, D., Churchill, R., 2016. Standards for the Parker, K.M., Rice, J., Gustat, J., Ruley, J., Spriggs, A., Johnson, C., 2013. "Effect of bike
Conduct of New Cochrane Intervention Reviews. Methodological Expectations of lane infrastructure improvements on ridership in one New Orleans neighborhood.".
Cochrane Intervention Reviews. J. Higgins, T. Lasserson, J. Chandler, D. Tovey and R. Ann. Behav. Med. 45 (Suppl. 1), S101–S107.
Churchill. Cochrane, London. Pazin, J., Garcia, L.M.T., Florindo, A.A., Peres, M.A., Guimarães, A.C.d.A., Borgatto, A.F.,
Hovmand, P., 2014. Group Model Building and Community-Based System Dynamics Duarte, M.d.F.d.S., 2016. Effects of a new walking and cycling route on leisure-time
Process. Community Based System Dynamics. Springer-Verlag, New York, pp. 17–30. physical activity of Brazilian adults: a longitudinal quasi-experiment.". Health Place
Jayasinghe, S., 2011. Conceptualising population health: from mechanistic thinking to 39, 18–25.
complexity science.". Emerg. Themes Epidemiol. 8 (2). Pucher, J., Dill, J., Handy, S., 2010. Infrastructure, programs, and policies to increase
Jensen, S.U., 2008. Bicycle tracks and lanes: a before-after study. In: Transportation bicycling: an international review. Prev. Med. 50, S106–S125.
Research Board 87th Annual Meeting, vol. 2095. Transportation Research Board, Ricciardi, A.M., Xia, J., Currie, G., 2015. Exploring public transport equity between se-
Washington DC, United States, pp. 15. parate disadvantaged cohorts: a case study in Perth, Australia. J. Transport Geogr. 43,
King, A.C., Winter, S.J., Sheats, J.L., Rosas, L.G., Buman, M.P., Salvo, D., Rodriguez, N.M., 111–122.
Seguin, R.A., Moran, M., Garber, R., Broderick, B., Zieff, S.G., Sarmiento, O.L., Rissel, C., Greaves, S., Wen, L.M., Crane, M., Standen, C., 2015. Use of and short-term
Gonzalez, S.A., Banchoff, A., Dommarco, J.R., 2016. "Leveraging citizen science and impacts of new cycling infrastructure in inner-Sydney, Australia: a quasi-experi-
information technology for population physical activity promotion.". Trans. J. Am. mental design. Int. J. Behav. Nutr. Phys. Activ. 12 (1), 129.
College Sports Med. 1 (4), 30–44. Rydin, Y., Bleahu, A., Davies, M., Dávila, J.D., Friel, S., De Grandis, G., Groce, N., Hallal,
Kuhlberg, J.A., Hipp, J.A., Eyler, A., Chang, G., 2014. Open streets initiatives in the P.C., Hamilton, I., Howden-Chapman, P., Lai, K.-M., Lim, C.J., Martins, J., Osrin, D.,
United States: closed to traffic, open to physical activity. J. Phys. Activ. Health 11 (8), Ridley, I., Scott, I., Taylor, M., Wilkinson, P., Wilson, J., 2012. "Shaping cities for
1468–1474. health: complexity and the planning of urban environments in the 21st century.".
Laatikainen, T.E., Hasanzadeh, K., Kyttä, M., 2018. "Capturing exposure in environmental Lancet 379 (9831), 2079–2108.
health research: challenges and opportunities of different activity space models.". Int. Sallis, J.F., Owen, N., Fisher, E.B., 2008. Ecological Models of Health Behaviour. In:
J. Health Geogr. 17 (1), 29. Glanz, K., Rimer, B.K., Viswanath, K. (Eds.), Health Behavior and Health Education:
Langdon, M., 2015. An Evidence-Based Assessment of the Impact of Cycling Infrastructure Theory, Research and Practice. Jossey-Bass, San Francisco, USA, pp. 465–485.
in South East Queensland. Australian Institute of Traffic Planning and Management Sarmiento, O., Torres, A., Jacoby, E., Pratt, M., Schmid, T.L., Stierling, G., 2010. The
National Conference. Queensland, Australia, Transportation Research Board: 18, Ciclovia-Recreativa: a mass-recreational program with public health potential.". J.
Brisbane. Phys. Activ. Health 7 (Suppl. 2), S163–S180.
Lee, I.M., Shiroma, E.J., Lobelo, F., Puska, P., Blair, S.N., Katzmarzyk, P.T., 2012. Effect of Sarmiento, O.L., Díaz del Castillo, A., Triana, C.A., Acevedo, M.J., Gonzalez, S.A., Pratt,
physical inactivity on major non-communicable diseases worldwide: an analysis of M., 2017. Reclaiming the streets for people: insights from ciclovías recreativas in
burden of disease and life expectancy. Lancet 380 (9838), 219–229. Latin America. Prev. Med. 103, S34–S40.
Lemoine, P.D., Cordovez, J.M., Zambrano, J.M., Sarmiento, O.L., Meisel, J.D., Valdivia, Smith, M., Hosking, J., Woodward, A., Witten, K., MacMillan, A., Field, A., Baas, P.,
J.A., Zarama, R., 2016a. "Using agent based modeling to assess the effect of increased Mackie, H., 2017. "Systematic literature review of built environment effects on
Bus Rapid Transit system infrastructure on walking for transportation.". Prev. Med. physical activity and active transport – an update and new findings on health
88, 39–45. equity.". Int. J. Behav. Nutr. Phys. Activ. 14 (1), 158.
Lemoine, P.D., Sarmiento, O.L., Pinzón, J.D., Meisel, J.D., Montes, F., Hidalgo, D., Pratt, Song, Y., Preston, J., Ogilvie, D., 2017. New walking and cycling infrastructure and modal
M., Zambrano, J.M., Cordovez, J.M., Zarama, R., 2016. “TransMilenio, a scalable bus shift in the UK: a quasi-experimental panel study." Transportation Research Part A:
rapid transit system for promoting physical activity.”. J. Urban Health 93 (2), Policy and Practice 95, 320–333.
256–270. Suel, E., Polak, J.W., Bennett, J.E., Ezzati, M., 2019. "Measuring social, environmental
Li, Y., Lawley, M.A., Siscovick, D.S., Zhang, D., Pagan, J.A., 2016. Agent-based modeling and health inequalities using deep learning and street imagery.". Sci. Rep. 9 (1),
of chronic diseases: a narrative review and future research directions.". Prev. Chronic 6229.
Dis. 13, E69. Syed, S.T., Gerber, B.S., Sharp, L.K., 2013. Traveling towards disease: transportation
Liao, Y., Wang, I.T., Hsu, H.-H., Chang, S.-H., 2015. "Perceived environmental and per- barriers to health care access. J. Community Health 38 (5), 976–993.
sonal factors associated with walking and cycling for transportation in Taiwanese Triana, C.A., Sarmiento, O.L., Bravo-Balado, A., González, S.A., Bolívar, M.A., Lemoine,
adults.". Int. J. Environ. Res. Publ. Health 12 (2), 2105–2119. P., Meisel, J.D., Grijalba, C., Katzmarzyk, P.T., 2019. Active streets for children: the
Litman, T., 2013. "Transportation and public health.". Annu. Rev. Publ. Health 34 (1), case of the Bogotá Ciclovía. PloS One 14 (5), e0207791.
217–233. UN General Assembly, 2015. Transforming our world : the 2030 agenda for sustainable
Lucas, K., 2012. Transport and social exclusion: where are we now? Transport Pol. 20, development, UN general assembly.
105–113. United Nations, 2014. World Urbanization Prospects: the 2014 Revision, Highlights,
Macmillan, A., Connor, J., Witten, K., Kearns, R., Rees, D., Woodward, A., 2014. The Department of Economic and Social Affairs; Population Division. United Nations.
societal costs and benefits of commuter bicycling: simulating the effects of specific United Nations, 2016. Mobilizing Sustainable Transport for Development: Analysis and
policies using system dynamics modeling.". Environ. Health Perspect. 122 (4), Policy Recommendations from the United Nations Secretary-General's High-Level
335–344. Advisory Group on Sustainable Transport. Department of Economic and Social
McCormack, G.R., Shiell, A., 2011. "In search of causality: a systematic review of the Affairs, United Nations, New York.
relationship between the built environment and physical activity among adults.". Int. United Nations, 2017. The new urban agenda. In: HABITAT III: United Nations
J. Behav. Nutr. Phys. Activ. 8 (125), 1479–5868. Conference on Housing and Sustainable Urban Development. Equador, Quito.
McDonnell, S., Zellner, M., 2011. "Exploring the effectiveness of bus rapid transit a pro- United Nations, 2018. Revision of world urbanization prospects, population division of
totype agent-based model of commuting behavior.". Transport Pol. 18 (6), 825–835. the united Nations department of economic and social affairs. pp. 2018.
25
van Sluijs, E.M.F., McMinn, A.M., Griffin, S.J., 2007. Effectiveness of interventions to towards policy intervention." Computers. Environ. Urban Syst. 51, 59–69.
promote physical activity in children and adolescents: systematic review of controlled Yang, Y., Diez-Roux, A., 2013. "Using an agent-based model to simulate children's active
trials. BMJ 335 (7622), 703. travel to school.". Int. J. Behav. Nutr. Phys. Activ. 10 (1), 67.
Venn, D., Strazdins, L., 2017. Your money or your time? How both types of scarcity Yin, L., Cheng, Q., Wang, Z., Shao, Z., 2015. ‘Big data’ for pedestrian volume: exploring
matter to physical activity and healthy eating. Soc. Sci. Med. 172, 98–106. the use of Google Street View images for pedestrian counts. Appl. Geogr. 63,
Vlahov, D., Galea, S., Freudenberg, N., 2005. The urban health “advantage”. J. Urban 337–345.
Health : Bull. N. Y. Acad. Med. 82 (1), 1–4. Zellner, M., Massey, D., Shiftan, Y., Levine, J., Arquero, M., 2016. Overcoming the last-
Weichenthal, S., Hatzopoulou, M., Brauer, M., 2019. A picture tells a thousand…ex- mile problem with transportation and land-use improvements: an agent-based ap-
posures: opportunities and challenges of deep learning image analyses in exposure proach. Int. J. Trans. 4 (1), 1–26.
science and environmental epidemiology.". Environ. Int. 122, 3–10. Zou, M., Li, M., Lin, X., Xiong, C., Mao, C., Wan, C., Zhang, K., Yu, J., 2016. An agent-
Yanez-Pagans, P., Martinez, D., Mitnik, O.A., Scholl, L., Vazquez, A., 2018. Urban based choice model for travel mode and departure time and its case study in Beijing.
Transport Systems in Latin America and the Caribbean: Challenges and Lessons Transport. Res. C Emerg. Technol. 64, 133–147.
Learned. Institute of Labor Economics (IZA) IZA Discussion Papers. Wells, G.A., Shea, B., O'Connell, D., Peterson, J., Welch, V., Losos, M., Tugwell, P., 2013.
Yang, L., Sahlqvist, S., McMinn, A., Griffin, S.J., Ogilvie, D., 2010. Interventions to pro- The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised stu-
mote cycling: systematic review. Br. Med. J. 341, c5293. dies in meta-analyses. http://www.ohri.ca/programs/clinical_epidemiology/nosgen.
Yang, Y., Auchincloss, A.H., Rodriguez, D.A., Brown, D.G., Riolo, R., Diez-Roux, A.V., pdf (accessed 15 April 2019).
2015. Modeling spatial segregation and travel cost influences on utilitarian walking:
26

A Systematic Review of Empirical and Simulation Studies Evaluating The Health Impact of Transportation Interventions

Uploaded by

Copyright:

Available Formats

A Systematic Review of Empirical and Simulation Studies Evaluating The Health Impact of Transportation Interventions

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Systematic Review of Empirical and Simulation Studies Evaluating The Health Impact of Transportation Interventions

Uploaded by

Copyright:

Available Formats

Environmental Research 186 (2020) 109519

Contents lists available at ScienceDirect

A systematic review of empirical and simulation studies evaluating the T

ARTICLE INFO ABSTRACT

MeSH “Bicycling” [Mesh: focus] OR

Fig. 1. PRISMA flowchart showing process of study selection.

3.1. Empirical studies 2016; Chang et al., 2017).

Characteristics of included evaluation studies.

“The municipal government made other

Active transit duration

Active transit distance

ported either a control or reference group. Four studies employed a

random sampling approach to recruit participants, including three

multi-component (Cerdá et al., 2012; Pazin et al., 2016; Chang et al.,

area, including nine single component (Chen et al., 2012; Goodman

and two multi-component intervention studies (Goodman et al., 2013;

Exposure to the transportation intervention was operationalized in

several different ways among the 29 empirical studies. Some studies

compared geographic locations, specifically, street segments, before

and after an intervention (Boarnet et al., 2005; Jensen, 2008; Parker

areas of influence were most commonly defined as buffers centred on

network from the core Connect2 projects in each of

Population: Adults residing within 5 km by road

(impassable at high tide) was turned into “a new

used predefined geographies such as towns (Goodman et al., 2013),

block groups (Langdon, 2015; Ferenchak and Marshall, 2016) and

how exposure to the intervention was defined in one study (Greaves

A total of 183 outcomes relating to 19 unique interventions were

tested for statistical significance varied both with respect to outcome

and number of active trips (including walking or cycling trips) (11%).

Cardiff, Kenilworth &

metric (Brown et al., 2016b), and car travel-related outcomes (Greaves

and Brazil (n = 3 and n = 1, respectively) (Pazin et al., 2016; Chang

et al., 2017), as well as physical activity (n = 4) and active mode share

formation on assessed outcomes and the nature of the associations re-

ported by included studies.

The 19 studies reporting on single-component interventions (all

focusing on bicycle lanes; Fig. 2B) predominantly reported on injury

Ferenchak and Marshall, 2016), active travel duration outcomes

Time steps varied significantly across studies, ranging from just

congestion charge can increase the

changing rate to a large extent

significant because the mode

changing rate is not so high;

increases, but effect is not

Five of the 10 studies explored economic incentives to encourage

provements of public transport services, including the implementation

of BRT lanes, other interventions included increasing bus frequency, the

introduction of express bus stops, and changes in the system density

(coverage). Overall, these studies showed improvements in active travel

synergistic policies. For instance, Yang et al., (2015) observed that

the road network then decide whether

consequently, mode speed and travel

are influenced by costs, travel time,

the implementation of a BRT system, with the potential of reaching a 21

network; individuals accumulate

ditional introduction of off-board ticketing.

3.3. Quality appraisal

Time step: unclear