RESEARCH ARTICLE

Neighborhood-based physical activity

differences: Evaluation of the effect of health
promotion program
Amanda Cristina de Souza Andrade1,2*, Sueli Aparecida Mingoti2,3, Amanda
Paula Fernandes1,2, Roseli Gomes de Andrade1,2, Amélia Augusta de Lima Friche1,2,
César Coelho Xavier2,4, Fernando Augusto Proietti2,4, Ana V. Diez-Roux5, Waleska
Teixeira Caiaffa1,2
1 Faculty of Medicine, Federal University of Minas Gerais, Belo Horizonte, Brazil, 2 Belo Horizonte
a1111111111
Observatory for Urban Health, Belo Horizonte, Brazil, 3 Institute of Exact Sciences, Federal University of
a1111111111 Minas Gerais, Belo Horizonte, Brazil, 4 Faculty of Health and Human Ecology, Vespasiano, Brazil, 5 School
a1111111111 of Public Health, Drexel University, Philadelphia, Pennsylvania, United States of America
a1111111111
a1111111111 * amandasouza_est@yahoo.com.br

Abstract
OPEN ACCESS
Introduction
Citation: Andrade ACdS, Mingoti SA, Fernandes
AP, Andrade RGd, Friche AAdL, Xavier CC, et al.
The practice of physical activity is an important factor in the prevention of health problems.
(2018) Neighborhood-based physical activity However, a small portion of the population is physically active. Recent reviews show that
differences: Evaluation of the effect of health physical activity classes in community settings have the potential to increase population lev-
promotion program. PLoS ONE 13(2): e0192115.
els of physical activity and reduce health inequalities.
https://doi.org/10.1371/journal.pone.0192115

Editor: Jaroslaw Harezlak, Indiana University,

Objective
UNITED STATES
To evaluate the effect of the Academias da Cidade Program in Belo Horizonte on the prac-
Received: March 22, 2017
tice of physical activity in leisure time (PALT) by non-users living near the program centers.
Accepted: January 18, 2018

Published: February 5, 2018 Methods

Copyright: © 2018 Andrade et al. This is an open We conducted a home-based health survey in Belo Horizonte (2008–2009) with 1,581
access article distributed under the terms of the
adults who were non-users of the program and who lived within a 1,500-meter radius of one
Creative Commons Attribution License, which
permits unrestricted use, distribution, and active program center (exposed group) and two nonoperational centers with sites reserved
reproduction in any medium, provided the original for their construction (unexposed group). We collected data on PALT levels (150 minutes/
author and source are credited. week), which was measured with the Physical Activity International Questionnaire and ana-
Data Availability Statement: The data underlying lyzed with binary logistic regression using the Generalized Estimating Equations method.
this study are third party data and are available to The propensity score was used as an adjustment variable to control the potential confusion
all interested researchers. To submit a proposal the
OSUBH Coordinating Center at http://site.medicina.
in the measures of effect of exposure studied.
ufmg.br/osubh/contato/. The authors do not have
any special access privileges to these data and Results
confirm that interested researchers may apply for
access to these data in the manner described. The overall prevalence of the PALT was 26.5% in the exposed group and 22.7% in the unex-
posed group. The exposed group was more likely to be active in leisure time (OR = 1.05; CI
Funding: This work was supported by the National
Health Fund of the Ministry of Health, Fapemig, 95%: 1.01–1.10). When considering the interaction between exposed group and distance,
CNPq, NIH / Fogarty International Center. The individuals in the exposed group who lived less than 500 meters from the program center

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 1 / 13

 Evaluation of the effect of health promotion program

CAPES for a research scholarship to ACSA and were more likely to be active in leisure time (OR = 1.18, CI 95%: 1.03–1.35) compared to
APF. CNPq for a research productivity scholarship their counterparts.
to the researcher WTC. The funders had no role in
study design, data collection and analysis, decision
to publish, or preparation of the manuscript. Conclusion
Competing interests: The authors have declared Promoting physical activity in the community can favorably affect PALT levels among resi-
that no competing interests exist. dents, especially those living closest to intervention centers. We believe the Academias da
Cidade Program is a promising strategy to facilitate the access to appropriate spaces for the
practice of physical activity and contribute to increase the levels physical activity within
populations.

Introduction
The promotion of physical activity is one of the top public health priority strategies as lack of
physical exercise is one of the main risk factors for the development of chronic non-transmit-
table diseases, and frequently observed among populations. The adoption of an active lifestyle
not only depends on individual behaviors and choices, but also on the interaction between
individuals, environment and public policies[1].
The impact of the physical and social environment in health and health-related events has
been increasingly studied[1]. A growing number of studies have shown that the environment
plays an important role in promoting global physical activity or even in specific behaviors[2].
Access to places where exercise can be carried out is one of the most important factors in pro-
moting higher levels of physical activity within the population[3–5].
In this context, community-based interventions have been designed as strategic actions for
the mitigation of inequities, through the promotion of physical activity and changes in the
environment that allow the access to appropriate spaces[6–8]. In Brazil, programs to promote
physical activity include the Academias da Cidade Program, which was initially implemented
in some Brazilian cities and, more recently, The Health Academy Program, implemented at
the national level. Both are models of intervention for health promotion, aimed at encouraging
and guiding the practice of physical activity in public spaces with appropriate infrastructures
and qualified professionals[9].
The evaluation of these interventions is essential to determine their effectiveness on the
health and access of the population, as well as the quality of the actions performed[9]. Two
recent systematic reviews have highlighted the lack of studies focused on the assessment of
physical activity interventions at the community level in Latin America[7, 8]. In Brazil, longi-
tudinal data on these types of programs are lacking, and experiments designed in places where
the programs have already been implemented and are ongoing for a long time are not feasible
[6, 10]. The use of more robust analysis methods is necessary, including natural experiments
and the use of the propensity score to evaluate the effectiveness of an intervention, even in the
absence of randomization between the comparison groups[6, 11, 12].
Thus, the objective of this study was to evaluate the effect of the Academias da Cidade Pro-
gram in Belo Horizonte, Brazil, on the practice of physical activity in leisure time of non-users
living near the program centers. Non-users of the program correspond to individuals residing
in the radius of up to 1,500 meters from the active program center (named exposed group)
and to the residents in the same radius definition in two other neighborhoods planned for
future construction (named the unexposed group). We hypothesized that the program, an
intervention model for health promotion in the community, was able to increase the physical
activity levels of the general population, and not only of its direct users.

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 2 / 13

 Evaluation of the effect of health promotion program

Materials and methods

Academias da Cidade Program in Belo Horizonte
The Academias da Cidade Program in Belo Horizonte was created in 2005 as a community
intervention by the Municipal Health Department of the city of Belo Horizonte and supported
by the Brazilian Ministry of Health. The first center was inaugurated in Leste district of the city
and, between 2007 and 2008, the program expanded, with the construction and planning of
new centers, to other regional cities. Currently, the municipality has 75 program centers inte-
grated into the Brazilian Unified Health System and composed of spaces with infrastructure,
equipment and qualified human resources to guide bodily practices, physical activity and
healthy lifestyles[13, 14].
The program aims to promote physical activity and improve the quality of life of the popu-
lation by providing access to free activities such as gymnastics, dance, games, sports, fights,
and walking, all of which are guided by a Physical Education professional. The activities are
offered preferably for adults over 18 years, three times a week and one hour a day in up to two
times per day (morning, afternoon or evening). Admission is by referral from the basic health
unit or by spontaneous desire. The centers are strategically installed in areas of social vulnera-
bility and in public places[13, 14].

BH Health Study
To evaluate the effect of the Academias da Cidade Program on the practice of physical activity
in the leisure time of the population assigned to the program units, a home-based survey
named BH Health Study was applied between 2008–2009 by the Belo Horizonte Observatory
for Urban Health of the Federal University of Minas Gerais. The survey collected data from
two of the nine districts of Belo Horizonte, Oeste and Barreiro, which were selected due to
their geographic proximity and internal heterogeneity regarding several demographic, socio-
economic and health indicators. The evaluation of the program was one of the main objectives
of the study as a whole[15].
At the time of the survey, one program center had been active for six months and three
were planned to be constructed. The center was inaugurated in October 2008 in a sports center
built to respond to a collective demand from the residents, which was possible due to the Par-
ticipatory Budget[14].
The sample of the study was obtained using a stratified sampling design and three-stage
cluster: (a) census tract, selected with different probabilities and with a sample size propor-
tional to the total of stratum census tracts; (b) domicile, selected with a simple random sample
of households registered in the database of the Belo Horizonte City Hall; (c) an adult resident
(aged 18 or more)[15]. The stratification factor, which allowed the proportional presence of all
socioeconomic levels in the sample[16], was the Health Vulnerability Index, a multidimen-
sional indicator resulting from the combination of social, demographic, economic and health
indicators geocoded by the census tract. The final sample of the original study was composed
of 4,048 adults (18 years or more) not enrolled in the program, referred to as “non-users” in
this study.
To ensure the representativeness of residents in the vicinity of the Academias da Cidade
Program, the probabilities of selecting each census tract were differentiated according to the
distance to the centers. The two census tracts closest to the active and planned centers were
included in the survey without randomization. Those located less than 500 meters and between
500 and 1,000 meters were, respectively, 8 and 4 times more likely to be selected compared to
census tracts located at more than 1,000 meters from any program center.

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 3 / 13

 Evaluation of the effect of health promotion program

The present study included a subsample of 1,581 non-users adults living in the 1,500 meter
radius of the active program center and of two centers with locations reserved for their future
construction. Residents around the fourth center, not implemented, were removed from the
analysis because their original building site was changed. This geographic delimitation was
adopted based on the information that most users of the active center resided within a radius
of 1,500 meters[17].
Data from the survey were collected through face-to-face interviews conducted by trained
interviewers, with a questionnaire containing household information, habits and behaviors,
sociodemographic data, and social and health determinants.

Outcome variable
Physical activity in leisure time was measured using the Physical Activity International Ques-
tionnaire and obtained by multiplying the frequency (days/week) and the average duration
(minutes/day) of walking, and of mild, moderate and vigorous activities, the latter multiplied
by two. Individuals with a physical activity score 150 minutes/week were considered active
[18].

Exposure variables
To measure the effect of the Academias da Cidade Program, two comparison groups were
defined: (a) exposed group, corresponding to non-users residing in a radius of up to 1,500
meters from the active program center; (b) unexposed group, which included non-users resid-
ing in the same geographic area of two nonoperational centers whose sites were reserved for
future construction.
For both groups, we measured the Euclidean distance in meters (0-500m; 501–1,000m;
1,001–1,500m) of the participants’ homes to the place where the program center was opera-
tional or the site was reserved for construction.

Adjustment variables
The adjustment variables considered included gender (female, male), age (in years), years of
schooling (0 to 8, 9 to 11 and more than 12 years), family income (<2, 2 to 5, 5 minimum
wages), residence time (in years) and census tract per capita income categorized as low,
medium and high tertiles. Income was obtained by the ratio between the total monthly nomi-
nal income of permanent private households and the total population of each census tract
available in the 2010 census[19].

Other variables
Two other variables were included. The perception of physical environment, measured by the
following: “In your neighborhood, how would you rate the following public sports and recrea-
tional areas?”, measured on a five-item Likert scale (very good to very bad). Barrier to physical
activity in leisure time among inactive individuals in the three months prior to the interview
was measured by the following question: "Does not know the proper / safe place to exercise"
(yes and no).

Statistical analysis
To control for the lack of homogeneity between the comparison groups in terms of sociode-
mographic variables, we used the propensity score as an adjustment variable to reduce possible
bias in the estimation of the intervention effect on the practice of physical activity in leisure

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 4 / 13

 Evaluation of the effect of health promotion program

time. The propensity score is defined as the conditional probability of an individual receiving
treatment or, in the case of this study, being exposed to the intervention, given its observed
covariates. It is represented by a single variable that simultaneously considers all potential
covariates of confusion. Thus, individuals with the same score have the same distribution of
observed covariates, regardless of their exposure condition[11, 12].
The propensity score was estimated using a binary logistic model with the maximum likeli-
hood method. The dependent variable was the group (exposed and unexposed groups) and the
independent variables were characteristics such as gender, age, residence length, years of
schooling and family income.
The association between physical activity in leisure time and the variables “group” and “dis-
tance to the center” was estimated by the odds ratio and its respective 95% confidence interval
(CI 95%) with binary logistic regression using the Generalized Estimating Equations method
(GEE). GEE and an exchangeable correlation structure were used because the observations
were grouped in one specific structure (i.e. individuals were nested within census tracts)[20].
The final models were adjusted for census tract per capita income and propensity score.
We used a significance level of 5%. All analyses were performed using the Stata software,
version 12 (Stata Corp., College Station, USA). The manipulation of the geographic data was
carried out using the MapInfo software, version 8.5 (MapInfo Corp. LP, New York, United
States).
This study was approved by the Research Ethics Committee of the Faculty of Medicine of
the Federal University of Minas Gerais (ETIC 253/06). All interviewees were informed about
the aims of the research and signed the Informed Consent Form, agreeing in participating of
the study.

Results
The sample used in this study was composed of 1,581 individuals, of which 508 were in the
exposed group (32.1%) and 1,073 in the unexposed group (67.9%). Of these, 60.4% were
female, mean age 43.4 (± 16.5) and 15.7 (± 12.3) residence years. Most of the interviewees
reported 0 to 8 years of study (51.0%), family income of 2 to 5 minimum wages (50.2%) and
48.4% resided in census tracts with per capita income classified in the mean position.
(Table 1).
The proportion of active individuals in leisure time (150 minutes/week) was 26.5% (CI
95%: 22.7–30.4) in the exposed group and 22.7% (CI 95%: 20.2–25.2) in the unexposed group.
In the exposed group, we also detected a gradient between the proportion of active individuals
and distance to the program center (32.1% for the 0-500m distance, 25.4% between 501 and
1,000m and 16.3% between 1,001 and 1,500m). This was not observed in the unexposed group
(22.4% for the 0-500m distance, 22.3% between 501 and 1,000m and 24.2% between 1,001 and
1,500m) (Fig 1; S1 Table). The results observed in Fi 1 suggest that there is an interaction
between the variables group and distance in relation to the practice of physical activity in lei-
sure in this study.
We verified that the estimated probability of being in the exposed group was higher for
women, individuals with family income between 2 and 5 minimum wages, older age and
shorter residence time (Table 2). The variable schooling was not statistically significant. How-
ever, it was maintained in the model due to its importance as a marker of the socioeconomic
level. Fig 2 shows the distribution of the estimated propensity score for the exposed and unex-
posed groups, and that the distribution was similar between the groups (p0,05; mean-com-
parison test). The propensity score allowed to balance the groups exposed and not exposed in
relation as sociodemographic variables (S2 Table).

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 5 / 13

 Evaluation of the effect of health promotion program

Table 1. Sociodemographic characteristics of the exposed and unexposed groups, The BH Health Study, Belo Horizonte, 2008–2009.
Variable Total Group
(n = 1,581) Exposed Unexposed
(n = 508) (n = 1,073)
Gender (%) 
Female 955 (60.4) 324 (63.8) 631 (58.8)
Male 626 (39.6) 184 (36.2) 442 (41.2)
Age in years (mean ± SD) 43.4 (16.5) 43.8 (16.1) 43.4 (16.7)
Residence time in years (mean ± SD)  15.7 (12.3) 14.5 (11.9) 16.3 (12.5)
Schooling years (%)
0 to 8 806 (51.0) 251 (49.4) 555 (51.7)
9 to 11 589 (37.2) 189 (37.2) 400 (37.3)
More than 12 186 (11.8) 68 (13.4) 118 (11.0)
Family income (%) 
<2 mw 467 (29.5) 133 (26.2) 334 (31.1)
2 to 5 mw 794 (50.2) 272 (53.5) 522 (48.7)
> 5 mw 320 (20.3) 102 (20.3) 217 (20.2)
Census tract per capita income (%) 
Low 669 (43.2) 124 (24.4) 545 (52.4)
Average 749 (48.4) 368 (72.4) 381 (36.7)
High 129 (8.3) 16 (3.2) 113 (10.9)

mw–minimum wages: R$ 415,00; SD–standard deviation


p<0,05.

https://doi.org/10.1371/journal.pone.0192115.t001

We then estimated the odds ratios and the confidence intervals of the association between
physical activity in leisure time and the variables group and distance, adjusted by the income
per capita of the census tract and propensity score (Table 3). In model 1, the exposed group
revealed to be more likely to be active in leisure time (OR = 1.05, CI 95% 1.01–1.10). Including
the variable “distance to the program center” in model 2 led to a loss of statistical significance
of the variable “group”, which suggests an interaction between the two variables. In model 3,
when considering the interaction between group and distance, a higher probability of being
active in leisure time was observed for individuals in the exposed group and who lived less
than 500 meters from the program center (OR = 1.18, CI 95%: 1.03–1.35).

Discussion
Community-based interventions to promote physical activity, such as the Academias da
Cidade Program in Belo Horizonte, may favourably affect the practice of leisure physical activi-
ties among non-users living near the intervention centers. Individuals in the exposed group
and who lived less than 500 meters from a program center were more likely to be active in lei-
sure time than residents living at a higher distance.
These intervention models have been considered to be promising strategies to increase lev-
els of physical activity of the population, as they promote environmental changes and commu-
nity empowerment, and reduce inequalities in the access to physical activity[6–8]. Studies
carried out in two Brazilian cities have demonstrated the potential of the Academias da Cidade
Program in influencing the practice of physical activity among non-users, measured by the
indirect effect of having seen or heard about the Program. These studies reported that people
exposed to the Program, either visually or by participation, were more active than those who
were not exposed[21–23].

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 6 / 13

 Evaluation of the effect of health promotion program

40
30
%
20
10

0 − 500 501 − 1,000 1,001 − 1,500

Distance (m)
Exposed
Unexposed 95% confidence interval

Fig 1. Proportion of active individuals in leisure time (150 minutes/week) in each group (exposed or unexposed
groups) and distance to the program center. The BH Health Study, Belo Horizonte, 2008–2009.
https://doi.org/10.1371/journal.pone.0192115.g001

A previous study carried out in the city of Belo Horizonte, with the same database used in
the present study and with an exposed and unexposed group, was the first to show the effect of
the Program on the practice of physical activity in leisure time among individuals residing at

Table 2. Binary logistic model for the estimation of the propensity score. The BH Health Study, Belo Horizonte,
2008–2009.
Variable OR CI 95% p
Gender (%)
Female 1.26 1.02–1.58 0.039
Male 1.00
Age in years 1.09 1.01–1.17 0.033
Residence time in years 0.98 0.97–0.99 0.001
Schooling years (%)
0 to 8 1.00
9 to 11 1.06 0.82–1.38 0.630
More than 12 1.33 0.90–1.96 0.143
Family income (%)
<2 mw 1.00
2 to 5 mw 1.36 1.05–1.76 0.018
> 5 mw 1.22 0.86–1.73 0.256

mw–minimum wages: R$ 415,00; OR–odds ratio; CI 95%: 95% confidence interval.

https://doi.org/10.1371/journal.pone.0192115.t002

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 7 / 13

 Evaluation of the effect of health promotion program

Unexposed Exposed

10
%
5
0

.1 .2 .3 .4 .5 .6.1 .2 .3 .4 .5 .6
Propensity score

Fig 2. Distribution of the estimated propensity score for the exposed and unexposed groups. The BH Health Study, Belo
Horizonte, 2008–2009.
https://doi.org/10.1371/journal.pone.0192115.g002

different distances to the centers[7]. Our results, in addition to corroborating the previous
finding, also include the propensity score. This method allows controlling for the potential of
confusion between intervention and outcome in observational studies, i.e. allows the compari-
son between the exposed and unexposed groups, such that the observed effect is attributed to

Table 3. Estimates of the odds ratios and confidence intervals (95%) of the association between physical activity in leisure time and the variables “group” (exposed
and unexposed groups) and “distance to the center” using a binary logistic model. The BH Health Study, Belo Horizonte, 2008–2009.
Variable Model 1 Model 2 Model 3
OR (CI 95%) p OR (CI 95%) p OR (CI 95%) p
Group
Exposed 1.05 (1.01–1.10) 0.045 1.04 (0.99–1.09) 0.090 0.94 (0.85–1.04) 0.260
Unexposed 1.00 1.00 1.00
Distance
0–500m 1.03 (0.97–1.10) 0.330 0.97 (0.90–1.05) 0.522
501–1.000m 1.01 (0.95–1.06) 0.916 0.98 (0.92–1.04) 0.476
1.001–1.500m 1.00 1.00
Group x Distance a
Intervention x 0–500m 1.18 (1.03–1.35) 0.015
Intervention x 501–1.000m 1.11 (0.98–1.26) 0.093
Intervention x 1.001–1.500m 1.00

Models adjusted by the census tract per capita income and propensity score
ªinteraction between group and distance; OR–odds ratio; CI 95%: 95% confidence interval.

https://doi.org/10.1371/journal.pone.0192115.t003

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 8 / 13

 Evaluation of the effect of health promotion program

the intervention and not to any differences between groups. The use of the propensity score is
enough to remove the bias associated to the observed covariates used in its estimation[11, 12,
24].
The access to public spaces for the practice of physical activity is an important factor to
reduce environmental barriers related to physical inactivity[3]. In the present study, the effect
of the intervention was enhanced by a shorter distance to the program centers. People in the
exposed group had a 5% higher odds of being active at leisure time than people in the unex-
posed group. Also, when considering the interaction between distance and exposure, individu-
als in the exposed group who lived less than 500 meters from a center were 18% more likely to
be active in leisure time as compared to those in the unexposed group who lived less than 500
meters from a center. This result is in agreement with other studies reporting that a shorter
distance to physical activity is associated with higher levels of physical activity[4, 5, 17, 22].
When analysing how individuals evaluated the public venues for sports and leisure available
in their neighbourhood, we observed that 65.4% of those in the exposed group rated them as
“very good” and “good”, while only 25.2% individuals in the unexposed group did the same.
When considering the distance to the center, we observed a dose-response gradient evidenced
by the effect of proximity to the program for the exposed group (88.7% for the 0-500m dis-
tance, 58.8% between 501 and 1000m and 25.3% between 1.001 and 1.500 m), but this was not
observed in the unexposed group. In addition, those in the unexposed group who did not
engage in leisure physical activity in the three months prior to the interview more frequently
reported not knowing the place dedicated to sports and physical exercises (27.3% in the unex-
posed group versus 16.7% in the exposed group) (S3 Table). These results suggest that the Pro-
gram, to some extent, modifies the way the environment is configured in terms of the
availability of places for the practice of physical activity.
The adoption of an active lifestyle depends not only on the knowledge of the benefits associ-
ated with physical activity but also on the concrete conditions for its practice by different social
groups[25]. The availability of services and public spaces varies according to the socioeco-
nomic characteristics. Economically disadvantaged people are more likely to live in areas with
more precarious or irregularly available services[26–28].
Thus, the prioritisation of community-based programs to encourage physical activity, such
as the Academias da Cidade Program in Belo Horizonte, contributes to the reduction of inequi-
ties in the distribution of public equipment for physical activity for the general population,
especially in areas of social vulnerability. In this context, environmental interventions and the
formulation of public policies aimed at promoting physical activity, especially in urban areas,
have great potential to increase the levels of physical activity of the general population.
Aspects related to the accessibility, aesthetics, maintenance and safety can be an important
determinant for the use of public spaces for the practice of physical activity[5, 27]. In this
sense, intersectoral actions are necessary to expand this intervention model and propose modi-
fications to facilitate access to the equipment, such as more connected streets, the presence of
sidewalks, greater illumination and security.
Some limitations of the current study have to be highlighted. For instance, information on
the most variables analyzed in the current study was not available before and after the imple-
mentation of the intervention. However, the propensity score method was used as an adjust-
ment variable, since its application is indicated when it is not possible to obtain data from the
population before the intervention, or when the selection of the exposed and the unexposed
groups was not randomized[29].
In the present study, the propensity score was used as the analysis strategy to ensure that
the groups exposed and not unexposed to the intervention were comparable according to
sociodemographic variables. The propensity score attempts to simulate randomization of

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 9 / 13

 Evaluation of the effect of health promotion program

individuals as occurs in randomized experimental studies. However, unlike randomization,

the balancing achieved by propensity score is only on observed covariates rather than all possi-
ble confounders[11, 12].
The selection of only one active center and two with planned construction restricts the gen-
eralisation of the results obtained. On the other hand, it must be emphasised that the sample
design of the basic survey sought to guarantee the representativeness of the affiliated popula-
tion[15]. Also, the use of self-reported information, such as the practice of physical activity in
leisure time, may be under- or overestimated in terms of duration and intensity of the activities
performed by the individuals.
The Euclidean distance in meters between the homes of the interviewees and the program
centers, one in operation and two with sites reserved for construction, was determined
through the use of the Geographic Information System. The use of distance allowed the com-
parison of results between the groups with different levels of exposure to the intervention,
increasing the internal validity of the study. Although this measure does not represent the
actual displacement of individuals, because they do not consider the presence of natural or
built obstacles and segment networks (streets and sidewalks), distance represents an objective
measure of the environment and is not subjected to restrictions in self-reported data[30].
To better understand the effectiveness of physical activity interventions at the community
level, it is recommended that the centers and their surrounding environment are evaluated
using subjective methods and measurement objectives[30]. Future studies should include
information before and after the implementation of the program, the control of effects not
attributable to the intervention, and the use of natural experiments that allow making causal
inferences about the effect of the intervention studied[31].

Conclusions
Despite its limitations, the present study shows that the implementation of physical activity
sites, such as the Academias da Cidade Program in Belo Horizonte, can increase the levels of
physical activity of non-users of the program, and proximity to the centers of the program is a
relevant factor to increase the effect of the intervention. The results of this study can serve as
subsidies for actions of health promotion at the population level aimed at reducing the preva-
lence of inactive people, increasing levels of physical activity and coping with non-transmitta-
ble chronic diseases. This work also presents practical implications given the expansion of this
intervention model at the federal level and its insertion in health promotion public policies.

Supporting information
S1 Table. Proportion of active individuals in leisure time (150 minutes/week) in each
group (exposed or unexposed groups) and distance to the program center. The BH Health
Study, Belo Horizonte, 2008–2009. CI 95%: 95% confidence interval.
(DOCX)
S2 Table. Propensity score for the exposed and unexposed groups. The BH Health Study,
Belo Horizonte, 2008–2009.
(DOCX)
S3 Table. Perception of public area for sports and leisure available in neighbourhood of
the exposed and unexposed groups, The BH Health Study, Belo Horizonte, 2008–2009. 1
84 missing; 2 Did not engage in leisure physical activity in the three months prior to the inter-
view.
(DOCX)

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 10 / 13

 Evaluation of the effect of health promotion program

Acknowledgments
The research team of the Belo Horizonte Observatory for Urban Health of which participated
in The BH Health Study.

Author Contributions
Conceptualization: Amanda Cristina de Souza Andrade, Sueli Aparecida Mingoti, Amanda
Paula Fernandes, Amélia Augusta de Lima Friche, César Coelho Xavier, Fernando Augusto
Proietti, Ana V. Diez-Roux, Waleska Teixeira Caiaffa.
Formal analysis: Amanda Cristina de Souza Andrade, Sueli Aparecida Mingoti, Ana V. Diez-
Roux, Waleska Teixeira Caiaffa.
Funding acquisition: César Coelho Xavier, Fernando Augusto Proietti, Waleska Teixeira
Caiaffa.
Investigation: César Coelho Xavier, Fernando Augusto Proietti.
Methodology: Amanda Cristina de Souza Andrade, Sueli Aparecida Mingoti, Amanda Paula
Fernandes, Ana V. Diez-Roux, Waleska Teixeira Caiaffa.
Project administration: Waleska Teixeira Caiaffa.
Writing – original draft: Amanda Cristina de Souza Andrade, Sueli Aparecida Mingoti,
Amanda Paula Fernandes, Roseli Gomes de Andrade, Amélia Augusta de Lima Friche,
César Coelho Xavier, Fernando Augusto Proietti, Ana V. Diez-Roux, Waleska Teixeira
Caiaffa.
Writing – review & editing: Amanda Cristina de Souza Andrade, Sueli Aparecida Mingoti,
Amanda Paula Fernandes, Roseli Gomes de Andrade, Amélia Augusta de Lima Friche,
César Coelho Xavier, Fernando Augusto Proietti, Ana V. Diez-Roux, Waleska Teixeira
Caiaffa.

References
1. Bauman AE, Reis RS, Sallis JF, Wells JC, Loos RJ, Martin BW, et al. Correlates of physical activity:
why are some people physically active and others not? Lancet. 2012; 380(9838):258–71. https://doi.
org/10.1016/S0140-6736(12)60735-1 PMID: 22818938.
2. Hoehner CM, Brennan Ramirez LK, Elliott MB, Handy SL, Brownson RC. Perceived and objective envi-
ronmental measures and physical activity among urban adults. Am J Prev Med. 2005; 28(2 Suppl
2):105–16. https://doi.org/10.1016/j.amepre.2004.10.023 PMID: 15694518.
3. Hallal PC, Reis RS, Parra DC, Hoehner C, Brownson RC, Simoes EJ. Association between perceived
environmental attributes and physical activity among adults in Recife, Brazil. J Phys Act Health. 2010; 7
Suppl 2:S213–22. PMID: 20702909.
4. Humpel N, Owen N, Leslie E. Environmental factors associated with adults’ participation in physical
activity: a review. Am J Prev Med. 2002; 22(3):188–99. PMID: 11897464.
5. Pazin J, Garcia LM, Florindo AA, Peres MA, Guimaraes AC, Borgatto AF, et al. Effects of a new walking
and cycling route on leisure-time physical activity of Brazilian adults: A longitudinal quasi-experiment.
Health Place. 2016; 39:18–25. https://doi.org/10.1016/j.healthplace.2016.02.005 PMID: 26922514.
6. Heath GW, Parra DC, Sarmiento OL, Andersen LB, Owen N, Goenka S, et al. Evidence-based in-
tervention in physical activity: lessons from around the world. Lancet. 2012; 380(9838):272–81.
https://doi.org/10.1016/S0140-6736(12)60816-2 PMID: 22818939; PubMed Central PMCID:
PMCPMC4978123.
7. Hoehner CM, Ribeiro IC, Parra DC, Reis RS, Azevedo MR, Hino AA, et al. Physical activity interven-
tions in Latin America: expanding and classifying the evidence. Am J Prev Med. 2013; 44(3):e31–40.
https://doi.org/10.1016/j.amepre.2012.10.026 PMID: 23415133; PubMed Central PMCID:
PMCPMC4217143.

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 11 / 13

 Evaluation of the effect of health promotion program

8. Hoehner CM, Soares J, Parra Perez D, Ribeiro IC, Joshu CE, Pratt M, et al. Physical activity interven-
tions in Latin America: a systematic review. Am J Prev Med. 2008; 34(3):224–33. https://doi.org/10.
1016/j.amepre.2007.11.016 PMID: 18312811.
9. Malta DC, Silva MMAd, Albuquerque GM, Amorim RCdAd, Rodrigues GBA, Silva TSd, et al. Polı́tica
Nacional de Promoção da Saúde, descrição da implementação do eixo atividade fı́sica e práticas cor-
porais, 2006 a 2014. Revista Brasileira de Atividade Fı́sica & Saúde. 2014; 19(3):286–99. https://doi.
org/10.12820/rbafs.v.19n3p286
10. Reis RS, Yan Y, Parra DC, Brownson RC. Assessing participation in community-based physical activity
programs in Brazil. Med Sci Sports Exerc. 2014; 46(1):92–8. https://doi.org/10.1249/MSS.
0b013e3182a365ae PMID: 23846162; PubMed Central PMCID: PMCPMC4476502.
11. Austin PC. An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in
Observational Studies. Multivariate Behav Res. 2011; 46(3):399–424. https://doi.org/10.1080/
00273171.2011.568786 PMID: 21818162; PubMed Central PMCID: PMCPMC3144483.
12. D’Agostino RB Jr. Propensity score methods for bias reduction in the comparison of a treatment to a
non-randomized control group. Stat Med. 1998; 17(19):2265–81. PMID: 9802183.
13. Brasil. Ministério da Saúde. Secretaria de Vigilância em Saúde. Avaliação de Efetividade de Programas
de Atividade Fı́sica no Brasil. Brası́lia: Ministério da Saúde; 2011. p. 125–44.
14. Dias MAS, Bicalho K, Mourão M, Alves MN, Evangelista PA, Rodrigues RCLC, et al. Promoção à
saúde e articulação intersetorial. In: Magalhães Júnior HM, Org, editors. Desafios e inovações na
gestão do SUS em Belo Horizonte: a experiência de 2003 a 2008. 1 ed: Mazza; 2010. p. 63–99.
15. Saúde Urbana em Belo Horizonte. Friche AAL, Xavier CC, Proietti FA, Caiaffa WT, editors. Belo Hori-
zonte: Editora UFMG; 2015.
16. Caiaffa WT, Nabuco AL, Friche AAL, Proietti FA. Urban health and governance model in Belo Hori-
zonte, Brazil. In: Vlahov D, Boufford JI, Pearson C, Norris L, editors. Urban Health: Global Perspective.
1 ed. New York: The New York Academy of Medicine; 2010. p. 437–52.
17. Fernandes AP, Andrade ACdS, Ramos CGC, Friche AAdL, Dias MAdS, Xavier CC, et al. Leisure-time
physical activity in the vicinity of Academias da Cidade Program in Belo Horizonte, Minas Gerais State,
Brazil: the impact of a health promotion program on the community. Cadernos de Saúde Pública. 2015;
31:195–207. https://doi.org/10.1590/0102-311X00104514 PMID: 26648374
18. WHO. Global recommendations on physical activity for health. GENEVA: World Health Organization,
2010.
19. IBGE. Sinopse do Censo Demográfico 2010. Rio de Janeiro: Instituto Brasileiro de Geografia e Esta-
tı́stica; 2011.
20. Hanley JA, Negassa A, Edwardes MD, Forrester JE. Statistical analysis of correlated data using gener-
alized estimating equations: an orientation. Am J Epidemiol. 2003; 157(4):364–75. PMID: 12578807.
21. Mendonca BC, Oliveira AC, Toscano JJO, Knuth AG, Borges TT, Malta DC, et al. Exposure to a com-
munity-wide physical activity promotion program and leisure-time physical activity in Aracaju, Brazil. J
Phys Act Health. 2010; 7 Suppl 2:S223–8. PMID: 20702910.
22. Simoes EJ, Hallal P, Pratt M, Ramos L, Munk M, Damascena W, et al. Effects of a community-based,
professionally supervised intervention on physical activity levels among residents of Recife, Brazil. Am
J Public Health. 2009; 99(1):68–75. https://doi.org/10.2105/AJPH.2008.141978 PMID: 19008499;
PubMed Central PMCID: PMCPMC2636600.
23. Simoes EJ, Hallal PC, Siqueira FV, Schmaltz C, Menor D, Malta DC, et al. Effectiveness of a scaled up
physical activity intervention in Brazil: A natural experiment. Prev Med. 2017; 103S:S66–S72. https://
doi.org/10.1016/j.ypmed.2016.09.032 PMID: 27687538.
24. Basu S, Meghani A, Siddiqi A. Evaluating the Health Impact of Large-Scale Public Policy Changes:
Classical and Novel Approaches. Annu Rev Public Health. 2017; 38:351–70. https://doi.org/10.1146/
annurev-publhealth-031816-044208 PMID: 28384086.
25. Ribeiro MCSA, Barata RB. Saúde: vulnerabilidade social, vizinhança e atividade fı́sica. Cadernos
Metrópole. 2016; 18:401–20.
26. Rydin Y, Bleahu A, Davies M, Davila JD, Friel S, De Grandis G, et al. Shaping cities for health: complex-
ity and the planning of urban environments in the 21st century. Lancet. 2012; 379(9831):2079–108.
https://doi.org/10.1016/S0140-6736(12)60435-8 PMID: 22651973; PubMed Central PMCID:
PMC3428861.
27. Silva I, Mielke G, Nunes B, Böhm A, Blanke A, Nachtigall M, et al. Espaços públicos de lazer: distri-
buição, qualidade e adequação à prática de atividade fı́sica. Revista Brasileira de Atividade Fı́sica &
Saúde. 2015; 20(1):286–99. https://doi.org/10.12820/rbafs.v.20n1p82
28. WHO. Promoting physical activity and active living in urban environments. GENEVA: World Health
Organization, 2006.

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 12 / 13

 Evaluation of the effect of health promotion program

29. Moraes JR, Guimaraes PV, Paula Fde L, Ferreira ML, Guimaraes RM, Luiz RR. Relationship between
health care insurance and Papanicolaou exam: a propensity score application using a complex sample
inquiry. Rev Bras Epidemiol. 2011; 14(4):589–97. PMID: 22218658.
30. Hino AAF, Reis RS, Florindo AA. Ambiente construı́do e atividade fı́sica: uma breve revisão dos méto-
dos de avaliação. Revista Brasileira de Cineantropometria & Desempenho Humano. 2010; 12:387–94.
31. Craig P, Cooper C, Gunnell D, Haw S, Lawson K, Macintyre S, et al. Using natural experiments to evalu-
ate population health interventions: new Medical Research Council guidance. J Epidemiol Community
Health. 2012; 66(12):1182–6. https://doi.org/10.1136/jech-2011-200375 PMID: 22577181; PubMed
Central PMCID: PMCPMC3796763.

PLOS ONE | https://doi.org/10.1371/journal.pone.0192115 February 5, 2018 13 / 13