Active living neighborhoods: is neighborhood walkability a key element for Belgian adolescents?

Abstract

In adult research, neighborhood walkability has been acknowledged as an important construct among the built environmental correlates of physical activity. Research into this association has only recently been extended to adolescents and the current empirical evidence is not consistent. This study investigated whether neighborhood walkability and neighborhood socioeconomic status (SES) are associated with physical activity among Belgian adolescents and whether the association between neighborhood walkability and physical activity is moderated by neighborhood SES and gender.
In Ghent (Belgium), 32 neighborhoods were selected based on GIS-based walkability and SES derived from census data. In total, 637 adolescents (aged 13-15 year, 49.6% male) participated in the study. Physical activity was assessed using accelerometers and the Flemish Physical Activity Questionnaire. To analyze the associations between neighborhood walkability, neighborhood SES and individual physical activity, multivariate multi-level regression analyses were conducted.
Only in low-SES neighborhoods, neighborhood walkability was positively associated with accelerometer-based moderate to vigorous physical activity and the average activity level expressed in counts/minute. For active transport to and from school, cycling for transport during leisure time and sport during leisure time no association with neighborhood walkability nor, with neighborhood SES was found. For walking for transport during leisure time a negative association with neighborhood SES was found. Gender did not moderate the associations of neighborhood walkability and SES with adolescent physical activity.
Neighborhood walkability was related to accelerometer-based physical activity only among adolescent boys and girls living in low-SES neighborhoods. The relation of built environment to adolescent physical activity may depend on the context.

Full text

Page 1

RESEARCH ARTICLE Open Access
Active living neighborhoods: is neighborhood
walkability a key element for Belgian
adolescents?
Femke De Meester1*, Delfien Van Dyck1, Ilse De Bourdeaudhuij1, Benedicte Deforche1,2, James F Sallis3and
Greet Cardon1
Abstract
Background: In adult research, neighborhood walkability has been acknowledged as an important construct
among the built environmental correlates of physical activity. Research into this association has only recently been
extended to adolescents and the current empirical evidence is not consistent. This study investigated whether
neighborhood walkability and neighborhood socioeconomic status (SES) are associated with physical activity
among Belgian adolescents and whether the association between neighborhood walkability and physical activity is
moderated by neighborhood SES and gender.
Methods: In Ghent (Belgium), 32 neighborhoods were selected based on GIS-based walkability and SES derived
from census data. In total, 637 adolescents (aged 13-15 year, 49.6% male) participated in the study. Physical activity
was assessed using accelerometers and the Flemish Physical Activity Questionnaire. To analyze the associations
between neighborhood walkability, neighborhood SES and individual physical activity, multivariate multi-level
regression analyses were conducted.
Results: Only in low-SES neighborhoods, neighborhood walkability was positively associated with accelerometer-
based moderate to vigorous physical activity and the average activity level expressed in counts/minute. For active
transport to and from school, cycling for transport during leisure time and sport during leisure time no association
with neighborhood walkability nor, with neighborhood SES was found. For walking for transport during leisure
time a negative association with neighborhood SES was found. Gender did not moderate the associations of
neighborhood walkability and SES with adolescent physical activity.
Conclusions: Neighborhood walkability was related to accelerometer-based physical activity only among
adolescent boys and girls living in low-SES neighborhoods. The relation of built environment to adolescent
physical activity may depend on the context.
Background
Increasing physical activity in youth is one of the key
public health strategies to conquer the alarming rise of
overweight, obesity and a cluster of risk factors asso-
ciated with cardiovascular disease and type 2 diabetes
[1,2]. To achieve substantial health benefits for school-
aged youth, participation in physical activity of at least
moderate to vigorous intensity for a minimum of 60
minutes per day is recommended [3,4]. A large propor-
tion of school-aged youth does not achieve the public
health recommendations [5-8]. In addition, adolescence
is marked by a decline in time spent in physical activity
which is more apparent in adolescent boys than in ado-
lescent girls [9-11].
Ecological models provide a framework for under-
standing the multiple facets that influence physical
activity. From the perspective of ecological models, an
interwoven relationship between individual and psycho-
social, sociocultural, policy and physical environmental
factors influences behaviors [12]. Past research has
* Correspondence: femke.demeester@ugent.be
1Department of Movement and Sport Sciences, Faculty of Medicine and
Health Sciences, Ghent University, Ghent, Belgium
Full list of author information is available at the end of the article
De Meester et al. BMC Public Health 2012, 12:7
http://www.biomedcentral.com/1471-2458/12/7
© 2011 Meester et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons
Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.

Page 2

identified demographic (e.g. age, gender, pubertal sta-
tus), psychosocial (e.g. self-efficacy, perceived compe-
tence, parental and peer support), sociocultural (e.g.
ethnicity) and policy (e.g. extracurricular physical activ-
ity) correlates of physical activity among adolescents
[13-17]. More recently, the importance of the physical
environment as an opportunity to shape physical activity
has been established [18-21].
An important construct among the physical environ-
mental correlates is neighborhood “walkability”. Neigh-
borhoods considered walkable are characterized by
mixed land use, well-connected streets and high residen-
tial density [22,23]. These elements are synergistic and
can be objectively determined using Geographic Infor-
mation Systems (GIS) software [24]. The research into
the relationship between neighborhood walkability and
physical activity has only recently been extended to
young people and the current empirical evidence is not
consistent [25-28]. The review of Ding et al. [29] estab-
lished that in only 20% of the studies that investigated
the association between objectively determined neigh-
borhood walkability and objectively determined physical
activity among adolescents, a positive association was
found. Ding et al. stated that when investigating the
association between neighborhood environment and
youth physical activity, conclusions based on objectively
measured environmental attributes seem more credible
because of the lower measurement error associated with
objective measures. Furthermore, it was stated that self-
reported physical activity that captures specific domains
of activity allow for tests of association between concep-
tually matched environmental and physical activity
variables.
In research on adults, neighborhood “walkability” has
been supported as a key construct among the built
environmental determinants. Four studies with a similar
design investigated the relationship between objectively
determined neighborhood walkability and physical activ-
ity in adults: the Neighborhood Quality of Life Study
(NQLS) conducted in the US [30], the Physical activity
in Localities and Community Environments (PLACE)
study conducted in Australia [31], the Belgian Environ-
mental Physical Activity Study (BEPAS) [32] and the
Swedish Neighborhood and Physical activity (SNAP)
study [33]. The four studies examined neighborhood
socio-economic status (SES) as a possible moderator of
the association between neighborhood walkability and
physical activity. For each study, participants were
recruited from “quadrants” of neighborhoods; low-SES/
high-walkable, low-SES/low-walkable, high-SES/high-
walkable and high-SES/low-walkable neighborhoods
were defined to ensure diversity of environments. Living
in neighborhoods characterized by higher walkability
was found to be associated with more walking for
transport [30-33], more cycling for transport [32], more
walking for leisure [30,32,33] and more accelerometer-
based moderate to vigorous physical activity [30,32,33].
The results of the four studies concerning the moder-
ating effect of neighborhood SES on the association
between neighborhood walkability and physical activity
were not totally comparable. In NQLS, the association
between neighborhood “walkability” and walking for
transport was stronger in high-SES than in low-SES
neighborhoods [30]. In PLACE, BEPAS and SNAP the
benefits from neighborhood “walkability” were similar in
high- and low-SES neighborhoods [31-33].
The results of studies in youth, investigating the direct
association between neighborhood SES and physical
activity, showed a positive association between neighbor-
hood SES and physical activity [34-36]. However, to our
knowledge, the study of Kerr et al. [37] was the only
study investigating the interaction between neighbor-
hood walkability and neighborhood SES in youth. The
results of this US study revealed that physical activity
behavior of 5-18 year olds reported by the parents was
related to objectively determined neighborhood walk-
ability in high-income neighborhoods and not in low-
income neighborhoods.
Considering the discrepancies in needs and behaviors
between adults and youth, the relationships found in
adults may not be generalisable to youth. Youth are
dependent on adult rules governing travel and destina-
tion choices and are not licensed to use motor vehicles
under the age of 16. Consequently, they are more cap-
tive in their own neighborhood, and the influence of
local neighborhood environmental attributes may be
more pronounced in youth. Therefore, to design an
active living neighborhood built environment suitable
for both youth and adults, urban planners should
know whether and how neighborhood walkability is
related to physical activity among adolescent boys and
girls.
The aims of the Belgian Environmental Physical Activ-
ity study in Youth (BEPAS-Y) were (A) to investigate
the association between neighborhood walkability,
neighborhood SES and physical activity in youth (B) to
investigate whether the association between neighbor-
hood walkability and physical activity is moderated by
neighborhood SES and gender.
Methods
BEPAS-Y was a cross-sectional study conducted in
Ghent. Ghent, the capital of the Belgian province East-
Flanders occupies over 156.18 sq km (60.3 sq miles)
with 1,554.40 inhabitants per sq km (2009). The
research protocol of BEPAS in adults [32], which builds
on the protocols of NQLS [30] and PLACE [31], was
used as template for BEPAS-Y. BEPAS-Y received
De Meester et al. BMC Public Health 2012, 12:7
http://www.biomedcentral.com/1471-2458/12/7
Page 2 of 12

Page 3

approval from the Ethics Committee of Ghent Univer-
sity Hospital.
Selection neighborhoods
Ghent consists of 201 statistical sectors, the smallest
administrative entities for which statistical data pro-
duced by the Belgian National Institute of Statistics
(NIS) are available. Belgian census data derived from the
NIS were used to define SES, and geographical informa-
tion derived from the available GIS databases was used
to define walkability. To obtain neighborhoods with a
sufficient number of inhabitants as a recruitment pool
(approximately 1,000) [30,32], adjacent statistical sectors
characterized by comparable walkability (within the
same quartile based on the walkability index) and SES
(within the same decile based on the median annual
household income) were used to define a neighborhood.
Consequently, the geographical area of low-walkable
neighborhoods was larger than of high-walkable neigh-
borhoods (1.8 km2vs. 0.4 km2), and the average popula-
tion density was lower (1535.6 inhabitants/km2vs.
6201.2 inhabitants/km2). The selection process involved
two steps: 1) the statistical sectors were stratified on
neighborhood walkability (GIS-based) and SES and 2)
neighborhoods (one or more adjacent statistical sectors)
meeting criteria for high/low walkability and high/low
SES were identified. The selection resulted in 32 neigh-
borhoods: 8 high-walkable/low-SES, 8 high-walkable/
high-SES, 8 low-walkable/low-SES, and 8 low-walkable/
high-SES (Figure 1). Each neighborhood comprised 1-5
contiguous statistical sectors.
Neighborhood walkability
For each statistical sector a walkability index was calcu-
lated using three objective GIS-based measures: residen-
tial density, intersection density, and land use mix,
which have been consistently related to physical activity
[38,39]. Geographical cadastral data (residential land
use, street centerline data, zoning data) and census data
provided by the Service for Environmental Planning in
Ghent were integrated in a GIS database and used to
determine the walkability components.
Net residential density represents the ratio of residen-
tial units to the land area devoted to residential use per
statistical sector. Connectivity is the ratio between the
number of true intersections (three or more legs) to the
land area of each statistical sector. Land use mix is an
indication of the degree to which a diversity of land use
types were present in each statistical sector. Five land
uses were considered: residential, retail (supermarkets,
bakeries, butchers, banks, and clothing shops), office,
institutional, and recreational (sport and non-sport).
The corresponding values were normalized and z-scores
were calculated. The three-component walkability-index
was created by weighing the z-scores of the environ-
mental features, using the following expression: walk-
ability = (2*z-connectivity) + (z-residential density) + (z-
land use mix). The formula used is an adapted version
of the formula of Frank and colleagues [40]. Because no
GIS data were available, “retail floor area ratio” was
omitted from the formula. Based on their walkability
index, the statistical sectors were ranked and quartiles
were constructed. The highest quartile constituted the
Legend
low-SES/low-walk
high-SES/high-walk
high-SES/low-walk
low-SES/high-walk
Figure 1 Distribution of neighborhoods in Ghent, Belgium. The distribution of the 32 selected neighborhoods in Ghent, Belgium: 8 high-
walkable/low-SES, 8 high-walkable/high-SES, 8 low-walkable/low-SES, and 8 low-walkable/high-SES.
De Meester et al. BMC Public Health 2012, 12:7
http://www.biomedcentral.com/1471-2458/12/7
Page 3 of 12

Page 4

high-walkable sectors, the lowest quartile the low-walk-
able sectors.
Neighborhood SES
The socioeconomic environment of each statistical sec-
tor was defined in terms of the median annual house-
hold income, using data from the NIS (Belgium, 2008).
Corresponding to their income data, the statistical sec-
tors were ranked and divided into deciles. Outliers were
avoided by excluding sectors with annual household
income values less than €11,600 and greater than €
116,000. The second, third, and fourth deciles of the
ranking contained the low-SES sectors; the seventh,
eighth, and ninth deciles the high-SES sectors [32].
Procedure
After neighborhood selection, the addresses from all, 13-
15 year old adolescents (n = 1553) living in the selected
neighborhoods were acquired by the Public Service of
Ghent. Between October 2008 and May 2009 an infor-
mative letter about the study with an invitation to parti-
cipate was posted to the potential participants. One
week later, home visits took place. Participants were
recruited simultaneously in the four groups of neighbor-
hoods throughout the recruitment period to avoid sea-
sonal bias.
In May 2009, 1,399 adolescents were invited, 1,078
adolescents were found at home and 59.1% (n = 637;
49.4% boys) consented to participate. Because at that
moment, recruitment goals were achieved with approxi-
mately equal numbers of participants within the four
groups of neighborhoods, it was decided not to contact
the remaining 154 adolescents as they were mostly from
high SES neighborhoods. Written consent was obtained
from all participants and the adolescents’ parents or
legal guardian. During the home visit, a questionnaire
was delivered and an interview was conducted. The pro-
tocol of the accelerometer and non-wear activity diary
was explained and an appointment for a second home
visit was made to collect the accelerometer, diary and
questionnaire.
Measures
Physical activity
Objectively assessed physical activity Physical activity
was objectively assessed using accelerometers, model
GT1M (Actigraph MTI, Manufacturing Technology Inc.,
Pensacola, FL, USA) and model 7,164 (Computer
Science Application, Inc., Shalimar, FL, USA). Two
recent studies confirmed that the output of acceler-
ometers model GT1M and model 7,164 was similar and
that therefore the two models can be used in the same
study [41,42]. The adolescents were asked to wear an
accelerometer during waking hours, for 7 consecutive
days including 2 weekend days. Secured by an elastic
belt, the accelerometers were worn on the right hip,
above the iliac crest.
Non-wear time activity diaries were provided to regis-
ter activities for which the accelerometer was removed
(aquatic activities or activities that prohibit an acceler-
ometer). Adolescents recorded on a pre-printed form
when the accelerometer was removed, when they put it
back on, and the kind of the activities they were
involved in [43,44].
Data-reduction software, MeterPlus 4.2. [45], was used
to screen, clean and score the accelerometer data. In the
data reduction process, time periods of at least one hour
of consecutive zeros were removed, assuming the accel-
erometer was unworn [46,47]. Whenever applicable,
these consecutive number of zeros were, after the accel-
erometer data scoring process, replaced by the corrected
number of minutes moderate physical activity and vigor-
ous physical activity registered in the diaries [43,44]. To
score the accelerometer data, the thresholds of Puyau
(moderate physical activity: 3,200-8,199 counts/min and
vigorous physical activity: ≥ 8,200 counts/min; respec-
tively corresponding to activities 3-6 MET and activities
> 6 MET) [48] were used. For inclusion in the data ana-
lysis, the required total accumulated number of minutes
registered time by the accelerometer and diaries was
600 minutes for weekdays and 480 minutes for weekend
days. Furthermore, 3 valid weekdays and 1 valid week-
end day of monitoring were needed to obtain reliable
estimates [49-51]. The accelerometer data were used to
provide an indication of adolescents physical activity by
means of the average activity level expressed in counts/
minute (CPM) and mean minutes of moderate to vigor-
ous physical activity per day (MVPA). The average activ-
ity level gives an indication of the total level of physical
activity and is not dependent of the chosen cut-points.
The number of minutes of moderate to vigorous physi-
cal activity is an outcome that is relevant for the assess-
ment of activity relative to meeting public health
guidelines.
Self-reported physical activity The Flemish Physical
Activity Questionnaire (FPAQ) [52] (interview version)
was used to determine the duration (hours and minutes
per day) of specific physical activity behaviors underta-
ken in specific contexts: school related active transporta-
tion (walking and cycling to and from school), walking
and cycling for transport during leisure time, and sport
during leisure time. The FPAQ was found to be a reli-
able and reasonably valid questionnaire for the assess-
ment of different dimensions of physical activity in 12-
18 year old adolescents [52].
Demographic variables
Self-reported data included adolescents’ gender, age,
nationality and SES. Educational attainment and
employment of the adolescent’ parents were used as a
De Meester et al. BMC Public Health 2012, 12:7
http://www.biomedcentral.com/1471-2458/12/7
Page 4 of 12

Page 5

proxy measure of adolescents’ SES. The educational
level of the adolescents’ mother and father was deter-
mined based on four options: less than high school,
completed high school, completed college or completed
university. The employment status of mother and father
(employed, unemployed) was coded into both parents
employed, one of the parents employed or both parents
unemployed.
Statistical analyses
To provide information about characteristics of the sam-
ple, descriptive statistical analyses were conducted using
SPSS 17.0. Tests for normal distribution revealed some
skewed physical activity variables. To obtain distribu-
tions that more closely approximated symmetry, loga-
rithmic transformations were conducted, and the
transformed variables were used in the analyses. For
ease of interpretation, summary data of untransformed
physical activity variables are reported in minutes/day.
To analyze the associations between neighborhood
walkability (dichotomous variable: low/high neighbor-
hood walkability), neighborhood SES (dichotomous vari-
able: low/high neighborhood SES) and physical activity,
multivariate regression analyses were conducted using
MLwin version 2.22. To examine if the association
between neighborhood walkability and physical activity
behavior was moderated by neighborhood SES, the
cross-product term “neighborhood walkability × neigh-
borhood SES” was entered in the regression model.
To examine if the association between neighborhood
walkability, neighborhood SES and physical activity
behavior was moderated by gender, the cross-product
terms “neighborhood walkability × gender”, “neighbor-
hood SES × gender” and “neighborhood walkability ×
neighborhood SES × gender” were separately included
in the regression model.
All analyses were controlled for three proxy measures
of individual SES (parental employment and educational
attainment of mother and father) [53]. Clustering of
individuals in neighborhoods was taken into account by
using multi-level modelling with adolescents at the first
level and neighborhoods at the second level. Neighbor-
hood-level attributes (neighborhood walkability and
SES) were handled as level-2 variables, individual attri-
butes (physical activity, parental employment and paren-
tal education) were handled as level-1 variables.
Results
From the 637 adolescents who consented for the study,
615 (96.5%) returned a complete physical activity ques-
tionnaire, and 513 (80.5%) had complete accelerometer
data. Table 1 represents descriptive statistics for demo-
graphic characteristics by type of neighborhood. Mean
age of the total sample was 14.6 ± 0.9 years, and 50.4%
Table 1 Descriptive statistics for demographic characteristics by type of neighborhood
total
(n = 637)
low-SES/low-walk
(n = 139)
low-SES/high-walk
(n = 158)
high-SES/low-walk
(n = 169)
high-SES/high-walk
(n = 171)
Age: mean (SD)
14.6 (0.9) 14.5 (0.9)14.6 (1.0) 14.5 (0.9)14.6 (0.9)
Gender: %
Male49.6 50.450.6 47.3 50.3
Female 50.449.649.4 52.749.7%
Educational level: %
Mother:
Less than high school9.9 19.816.72.0 3.9
Completed high school25.139.7 17.4 28.917.5
Completed college40.4 33.631.9 45.6 48.1
Completed University24.66.9 34.123.5 30.5
Father:
Less than high school7.518.6 7.92.1 4.0
Completed high school36.0 54.924.6 36.4 31.1
Completed college26.6 19.535.430.8 29.1
Completed University29.87.142.1 30.835.8
Employment status: %
Both employed 68.561.656.578.8 75.5
One parent unemployed26.729.7 36.4 20.6 20.9
Both unemployed4.2 8.7 7.10.63.7
De Meester et al. BMC Public Health 2012, 12:7
http://www.biomedcentral.com/1471-2458/12/7
Page 5 of 12