Effects of a Community-Based, Professionally Supervised Intervention on Physical Activity Levels Among Residents of Recife, Brazil

Abstract

Objectives. We evaluated the effects of a community-based intervention, the Academia da Cidade program (ACP), on increasing leisure-time physical activity among residents of Recife, Brazil.

Methods. We used the International Physical Activity Questionnaire to assess leisure-time physical activity and transport physical activity (i.e., activities involved in traveling from place to place) levels in a random sample of 2047 Recife residents surveyed in 2007. We also examined factors related to exposure to ACP (participation in the intervention, residing near an intervention site, hearing about or seeing intervention activities). We estimated prevalence odds ratios (ORs) of moderate to high leisure-time and transport physical activity levels via intervention exposures adjusted for sociodemographic, health, and environmental variables.

Results. Prevalence ORs for moderate to high levels of leisure-time physical activity were higher among former (prevalence OR = 2.0; 95% confidence interval [CI] = 1.0, 3.9) and current (prevalence OR = 11.3; 95% CI = 3.5, 35.9) intervention participants and those who had heard about or seen an intervention activity (prevalence OR = 1.8; 95% CI = 1.3, 2.5). Transport physical activity levels were inversely associated with residing near an ACP site.

Conclusions. The ACP program appears to be an effective public health strategy to increase population-level physical activity in urban developing settings.

The global burden of chronic diseases is large^1–3 and growing.^4,5 Given that regular physical activity has a strong preventive effect against chronic disease,^6,7 the Ministry of Health of Brazil began a program in 2003 designed to promote healthful living among the country's population.^8,9 An important component of this program is the promotion of leisure-time physical activity and better nutrition together with the prevention of tobacco use and injuries. Still in its early phases, the physical activity component of this program is already funding 201 Brazilian cities in their efforts to promote physical activity. However, the prevalence of moderate to high levels of leisure-time physical activity in Brazil's population is low, with rates across the 26 state capitals and the Federal District of Brasília ranging from 10.5% (São Paulo) to 14.7% (Recife) and 21.5% (Brasília).¹⁰

The US Guide to Community Services provides recommendations for physical activity programs in the United States and Europe, but it is unclear whether these recommendations apply to or can be adapted for use in other regions of the world.^11–13 Thus, a transdisciplinary, multisector partnership was established in Brazil to review evidence supporting strategies to promote physical activity in Latin America.¹⁴ In its categorization of innovative physical activity interventions, this evidence-based review identified the Academia da Cidade program (ACP), an ongoing community-based program designed to promote leisure-time physical activity in the city of Recife, in northeastern Brazil.¹⁵ ACP fits the profile of providing “physical activity classes in community settings,” 1 of 3 new categories of community-based physical activity interventions identified in the Latin American evidence-based review.¹⁴

We used a cross-sectional design to evaluate the effects of ACP on levels of leisure-time physical activity in the adult population of Recife. Our primary interest was in examining whether past or current participation in ACP activities, seeing or hearing about ACP activities, and living in a neighborhood near an ACP site were directly associated with moderate to high levels of leisure-time physical activity and inversely associated with transport physical activity (i.e., activities involved in traveling from place to place, including to destinations such as workplaces, stores, and movies).

METHODS

Intervention

To our knowledge, few studies have been conducted on ACP-like health promotion strategies.¹⁴ ACP is a government-funded intervention that provides supervised leisure-time physical activity for community members in 21 public spaces of Recife (e.g., parks, beaches, and recreation centers).¹⁵ An ACP site is created after an environmental assessment of public spaces and in response to community demands. The city of Recife identifies public spaces feasible for ACP development and then implements engineering and beautification plans that may cost between US $325 000 and $650 000 (M. das Gracas Cavalcante, MD, oral communication, April 2007).

At ACP sites, physical education teachers contracted by the city offer free supervised leisure-time physical activity sessions, nutrition education, and health monitoring (i.e., blood pressure measurements, anthropometric and nutrition assessments). Typical leisure-time physical activity sessions include calisthenics, aerobics, walking groups, stretching, and dance classes. All activities are offered on weekdays, every hour from 5 am to 10 am and 5 pm to 10 pm, to approximately 20 participants per session. No registration is required except in the case of individuals who need health monitoring. Since 2002, the program is estimated to have enrolled more than 10 000 members per year and to have resulted in a total of 880 000 person-sessions (M. das Gracas Cavalcante, oral communication, April 2007).

ACP is integrated into the public health care system. ACP participants identified as being at high risk for hypertension and overweight are referred to primary health care clinics for further evaluation and management. ACP also offers free supervised leisure-time physical activity sessions to patients at primary health care clinics in selected areas.

Study Design, Sampling, and Data Collection

Our study had a cross-sectional design, with ACP exposures and study outcomes (i.e., leisure-time physical activity levels and transport physical activity levels) assessed at the same point in time. We derived data from the 2007 Recife Physical Activity Survey, a random-digit-dialing telephone survey that sampled noninstitutionalized residents of the city who were 16 years or older; the Vigilância de Fatores de Risco e Proteção para Doenças Crônicas por Inquerito Telefonico survey (VIGITEL; Chronic Disease Risk Factor Surveillance Telephone Survey) methodology was used in stratifying and clustering the multistage sample.^10,16 Recife neighborhoods were divided into strata with and without an ACP site. The survey randomly sampled 2400 households with at least 1 landline telephone from each stratum, generating 12 clusters of 200 telephone numbers.

We completed interviews with 2046 of the 3632 potentially eligible respondents (an eligible respondent was categorized as any individual 16 years or older who answered the telephone in a household), for a 56% crude response rate (i.e., number of completed interviews divided by total units) and an adjusted rate of 64.5% (i.e., number of completed interviews that would have been obtained under perfect conditions).¹⁷ As a result of missing or inappropriate values, 8 and 62 observations were excluded from the descriptive and regression modeling analyses, respectively.

Measures

Physical activity.

The survey questions on physical activity were derived from the long-form International Physical Activity Questionnaire (IPAQ); only 2 of the IPAQ's original 4 domains of activity, leisure-time physical activity and transport physical activity, were included.¹⁸ All questions were translated into Portuguese and back-translated into English. In the case of both leisure-time physical activity and transport physical activity, participants were instructed to answer questions related only to activities that were at least 10 minutes in duration.

We used the equations included in IPAQ's guidelines for data processing and analysis to create a metabolic-equivalent-minutes-per-week score for both leisure-time physical activity (e.g., moderate and vigorous walking) and transport physical activity (e.g., biking and walking activities).¹⁸ We used the IPAQ total physical activity algorithm to categorize leisure-time physical activity and transport physical activity levels as low, moderate, or high.^18,19 The moderate and high categories approximated levels of physical activity recommended by the American College of Sports Medicine and the Centers for Disease Control and Prevention.^7,20

We used other definitions of leisure-time physical activity and transport physical activity to evaluate their associations with exposure to ACP. We adapted the 3-level algorithm developed for the VIGITEL survey¹⁰ to classify leisure-time physical activity levels as low (no activity reported or some activity reported but not at a moderate or high level), moderate (5 or more days of moderate physical activity of at least 30 minutes per day), and high (3 or more days of vigorous physical activity of at least 20 minutes per day).

Based on the distribution of metabolic-equivalent minutes per week in the sample, we grouped leisure-time physical activity levels into 2 categories: low (less than 792 metabolic-equivalent minutes per week) and moderate to high (792 or more metabolic-equivalent minutes per week). In a similar manner, we classified transport physical activity levels as low (less than 485 metabolic-equivalent minutes per week) and moderate to high (485 or more metabolic-equivalent minutes per week).

We also adapted the IPAQ total physical activity algorithm to categorize levels of leisure-time walking as low (no leisure-time walking reported or some walking reported but not at moderate or high levels), moderate (5 or more days of leisure-time walking of at least 30 minutes or 5 or more days of leisure-time walking of at least 600 metabolic-equivalent minutes per week), or high (7 or more days of leisure-time walking, accumulating at least 3000 metabolic-equivalent minutes per week).^18,19 In the case of all categorical analyses, the combined moderate–high level was compared with the low level.

Exposure to Academia da Cidade.

Respondents were classified as having been exposed to ACP if they had participated in the program at the time of the interview or at any time in the past. Otherwise, they were classified as having never been exposed to the program. In our secondary analyses, respondents who had heard about or seen an ACP activity and those who lived in a neighborhood with an ACP site were classified as having been exposed to ACP.

Environmental and other factors.

We used respondents’ answers to questions focusing on neighborhood traffic, safety, aesthetics, and walkability to create environmental variables (see the table available as a supplement to the online version of this article at http://www.ajph.org). On the basis of the sample distribution, age was coded as 16 to 34 years, 35 to 54 years, or 55 years or older. Education was coded as less than high school, high school, or college or more. On the survey questionnaire, items focusing on respondents’ perception of their race or skin color (white, black, light or dark brown, red [among indigenous populations], or yellow [among Asian populations]) were used to assess race/ethnicity. In all of our analyses, we coded skin color as white and other.

In terms of alcohol use, respondents were classified as ever or never having consumed alcohol. With respect to smoking status, respondents were classified as never, former, or current smokers; the latter 2 categories were combined. Respondents reported their health status as very good or excellent, normal or good, or fair or poor. We used self-reported weight and height to calculate body mass index (BMI; weight in kilograms divided by height in meters squared); respondents with BMIs below 18.5 kg/m² were classified as underweight, those with BMIs between 18.5 and 24.9 kg/m² were classified as being of normal weight, those with BMIs between 25.0 and 29.9 kg/m² were classified as overweight, and those with BMIs of 30.0 kg/m² or above were classified as obese.²¹

Statistical Analyses

We weighted data to compensate for the unequal sampling selection probabilities resulting from factors such as the study design, number of adults in a given household, and the unique telephone numbers in the different study households. Poststratification weighting based on age, gender, and education groups was used to partially adjust for nonresponse and low telephone coverage rates.²²

We generated frequencies for sociodemographic variables (i.e., age, race, education), chronic disease risk and protective factors (i.e., leisure-time physical activity and transport physical activity, environmental variables, BMI, smoking, alcohol use, self-reported health status), and exposures to ACP. We used logistic regression to generate prevalence odds ratios (ORs) for moderate to high levels of leisure-time physical activity, transport physical activity, and leisure-time walking by levels of exposures to ACP after adjusting for potential confounders (sociodemographic variables, chronic disease risk factors, and environmental factors).

The Surveymeans and Surveylogistics procedures in SAS version 9.1 were used to ensure that all of our analyses accommodated the weighting and complexity of the sample survey data.²³ In addition, we used diagnostic methods to assess the adequacy of our regression models.^24,25

RESULTS

The sample population was predominantly female (56.3%), older than 34 years (52.3%), and unmarried (57.2%). In addition, most of the respondents had a high school diploma or college degree (53.7%) and reported their skin color as other than white (65.9%; Table 1). Although only 5.2% of the respondents had participated in ACP activities, 37.9% lived in neighborhoods with an ACP site and 61.7% had either heard about or seen an ACP activity.

TABLE 1.

Sample Characteristics: Academia da Cidade Program (ACP), Recife, Brazil, 2007

	Male (n = 756), %	Female (n = 1282),%	Total (N = 2038), %
Sociodemographic characteristics
Age, y
16–34	51.0	45.1	47.7
35–54	32.9	34.9	34.0
≥55	16.1	20.0	18.3
Skin color
White	34.0	34.2	34.1
Other	66.0	65.8	65.9
Educational level
Less than high school	43.9	48.1	46.3
High school diploma	40.6	37.6	38.9
College degree or higher	15.6	14.2	14.8
Marital status
Married	48.2	38.6	42.8
Single	46.9	45.9	46.3
Other	4.9	15.5	10.9
Primary risk and behavioral factors
Body mass indexa
Underweight/normal	53.9	61.7	58.1
Overweight	30.5	24.0	27.0
Obese	15.6	14.3	14.9
Total physical activity level (LTPA + TPA)b
Low	44.2	57.4	51.7
Moderate	36.9	38.2	37.6
High	19.0	4.4	10.7
LTPA levelb
Low	73.9	85.7	80.6
Moderate	15.5	12.7	13.9
High	10.6	1.7	5.5
TPA levelb
Low	66.3	75.3	71.3
Moderate	30.3	24.0	26.8
High	3.5	0.7	1.9
Alcohol use status
Drinker	54.9	23.8	37.4
Nondrinker	45.1	76.2	62.6
Smoking status
Never smoker	54.6	69.8	63.2
Past or current smoker	45.4	30.2	36.8
Other factors
ACP participation
Current	0.7	1.9	1.3
Former	1.5	5.6	3.9
Never	97.8	92.5	94.8
Lives near an ACP site
Yes	39.6	36.5	37.9
No	60.4	63.5	62.1
Heard about or seen an ACP activity
Yes	59.4	63.5	61.7
No	40.6	36.5	38.3
Self-reported health status
Excellent or very good	23.8	18.2	20.6
Normal or good	49.7	35.2	41.5
Fair or poor	26.5	46.7	37.9
Neighborhood aesthetics
Neighborhood not pleasant	32.1	27.9	29.7
Neighborhood somewhat pleasant	42.0	45.7	44.1
Neighborhood pleasant	25.9	26.4	26.2
Neighborhood safety
Neighborhood not safe	39.7	40.6	40.2
Neighborhood somewhat safe	29.0	33.3	31.4
Neighborhood safe	31.3	26.1	28.3
Neighborhood traffic interference
No interference	56.5	56.2	56.3
Some interference	20.1	20.8	20.5
Great deal of interference	23.4	23.0	23.2
Neighborhood walkability
Neighborhood not walkable	15.9	21.3	19.0
Neighborhood somewhat walkable	41.5	33.2	36.8
Neighborhood walkable	42.6	45.4	44.2

Note. LTPA = leisure-time physical activity; TPA = transport physical activity. Percentage estimates are weighted. Numbers of observations with valid information were as follows: LTPA, n = 2038; TPA, n = 2025; total physical activity, n = 2023.

^aRespondents with BMIs below 18.5 kg/m² were classified as underweight, those with BMIs between 18.5 and 24.9 kg/m² were classified as being of normal weight, those with BMIs between 25.0 and 29.9 kg/m² were classified as overweight, and those with BMIs of 30.0 kg/m² or above were classified as obese.

^bDetermined by equations included in the International Physical Activity Questionnaire’s guidelines for data processing and analysis.

Rates of moderate- to high-level leisure-time physical activity were 19.4% overall, 26.1% among men, and 14.4% among women (Table 1). Rates of moderate- to high-level transport physical activity were 28.7% overall, 33.8% among men, and 24.7% among women. Moderate to high levels of combined leisure-time and transport physical activity were 48.3% overall, 55.4% among men, and 42.6% among women. When the VIGITEL survey algorithm¹⁰ was used to assess the prevalence of moderate to high levels of leisure-time physical activity, rates were 18.3% overall, 23.8% among men, and 13.9% for women (data not shown).

In comparison with those who had never participated in ACP, former participants were twice as likely to engage in moderate to high levels of leisure-time physical activity (prevalence OR = 2.0; 95% confidence interval [CI] = 1.0, 3.9), and current participants were approximately 11 times as likely to do so (OR = 11.3; 95% CI = 3.5, 35.9; Table 2). Prevalence odds ratios for moderate to high levels of leisure-time physical activity were significantly increased among respondents who had ever heard about or seen an ACP activity (OR = 1.8; 95% CI = 1.3, 2.5), among those with a high school diploma (OR = 1.9; 95% CI = 1.3, 2.7), and among those who had completed at least some college (OR = 2.2; 95% CI = 1.4, 3.4). Prevalence odds ratios for moderate to high levels of leisure-time physical activity were significantly reduced among women (OR = 0.4; 95% CI = 0.3, 0.5), those reporting normal or good health (OR = 0.5; 95% CI = 0.3, 0.7), and those reporting fair or poor health (OR = 0.5; 95% CI = 0.3, 0.8).

TABLE 2.

Prevalence Odds Ratios (ORs) for Moderate to High Levels of Leisure-Time Physical Activity, by Sociodemographic Characteristics, Levels of Exposure to Intervention, and Self-Reported Health Status: Academia da Cidade Program (ACP), Recife, Brazil, 2007

	Moderate to High Leisure-Time Physical Activity, %	Moderate to High Leisure-Time Physical Activity, Crude OR (95% CI)	Moderate to High Leisure-Time Physical Activity, Adjusted ORa (95% CI)
Sociodemographics
Age, y
16–34 (Ref)	22.1	1.0	1.0
35–54	17.3	0.7 (0.6, 0.9)	1.2 (0.8, 1.9)
≥55	16.0	0.7 (0.5, 0.9)	1.4 (0.9, 2.2)
Gender
Male (Ref)	26.1	1.0	1.0
Female	14.3	0.5 (0.4, 0.6)	0.4 (0.3, 0.5)
Skin color
White (Ref)	20.1	1.0	1.0
Other	18.9	0.9 (0.7, 1.2)	1.0 (0.8, 1.4)
Educational level
Less than high school (Ref)	12.5	1.0	1.0
High school	23.7	2.1 (1.6, 2.7)	1.9 (1.3, 2.7)
College degree or higher	28.4	2.7 (2.0, 3.7)	2.2 (1.4, 3.4)
Marital status
Married (Ref)	16.4	1.0	1.0
Single	21.7	1.4 (1.1, 1.8)	1.2 (0.7, 2.2)
Other	21.4	1.4 (1.0, 2.0)	1.3 (0.8, 2.2)
Exposure to ACP
Participation in ACP
Current	67.6	9.3 (4.1, 21.0)	11.3 (3.5, 35.9)
Former	29.3	1.8 (1.1, 3.0)	2.0 (1.0, 3.9)
Never (Ref)	18.3	1.0	1.0
Lives near an ACP site
Yes	22.8	1.4 (1.1, 1.8)	1.2 (0.9, 1.6)
No (Ref)	17.2	1.0	1.0
Heard about or seen an ACP activity
Yes	22.9	1.8 (1.4, 2.3)	1.8 (1.3, 2.5)
No (Ref)	13.9	1.0	1.0
Health
Self-reported health status
Excellent or very good (Ref)	32.4	1.0	1.0
Normal or good	17.3	0.5 (0.3, 0.6)	0.5 (0.3, 0.7)
Fair or poor	14.5	0.4 (0.3, 0.5)	0.5 (0.3, 0.8)

Note. CI = confidence interval. Leisure-time physical activity were classified based on the International Physical Activity Questionnaire’s guidelines.

^aAdjusted for health status, sociodemographic characteristics (age, gender, education, marital status), risk factors (smoking, alcohol use, overweight or obesity), and neighborhood environment variables (traffic interference, safety, aesthetics, walkability).

When we combined former (n = 138) and current (n = 38) ACP participants, we found that model-adjusted odds of moderate to high levels of leisure-time physical activity in this group were approximately 3 times those of individuals who had never participated in ACP (OR = 3.3; 95% CI = 1.8, 5.9; data not shown). Sensitivity analyses in which other definitions of leisure-time physical activity were used did not substantially change the strength or direction of associations between leisure-time physical activity and exposure to ACP (Table 3). The OR for moderate to high levels of leisure-time walking was significantly higher among respondents who had participated in ACP (OR = 2.7; 95% CI = 1.6, 4.9) than among those who had never participated.

TABLE 3.

Prevalence Odds Ratios (ORs) for Moderate to High Levels of Leisure-Time Physical Activity and Leisure-Time Walking, by Levels of Exposure to Intervention: Academia da Cidade Program (ACP), Recife, Brazil, 2007

	Leisure-Time Physical Activity: METs Method,a OR (95% CI)	Leisure-Time Physical Activity: VIGITEL Method,b OR (95% CI)	Leisure-Time Walking,c OR (95% CI)
Participation in ACP
Ever	2.8 (1.8, 4.4)	3.1 (1.8, 5.5)	2.7 (1.6, 4.9)
Never (Ref)	1.0	1.0	1.0
Lives near an ACP site
Yes	0.9 (0.6, 1.2)	1.1 (0.8, 1.5)	1.1 (0.8, 1.6)
No (Ref)	1.0	1.0	1.0
Heard about or seen an ACP activity
Yes	1.5 (1.0, 2.1)	1.9 (1.3, 2.6)	2.0 (1.1, 3.5)
No (Ref)	1.0	1.0	1.0

Note. METs = metabolic equivalents; CI = confidence interval. Odds ratios were fully adjusted for health status, sociodemographic characteristics (age, gender, education, marital status), risk factors (smoking, alcohol use, overweight or obesity), and neighborhood environment variables (traffic interference, safety, aesthetics, walkability).

^aBased on sample distribution of metabolic-equivalent minutes per week.

^bBased on VIGITEL survey algorithm.

^cBased on International Physical Activity Questionnaire algorithm.

ORs for moderate to high levels of transport physical activity were significantly increased among respondents who reported their skin color as other than white and respondents who had a high school diploma, whereas the OR was significantly reduced among women (Table 4). Respondents who lived in a neighborhood with an ACP site were less likely to engage in moderate to high levels of transport physical activity than were those who did not live in such a neighborhood (OR = 0.7; 95% CI = 0.5, 0.9). Sensitivity analyses in which other definitions of transport physical activity were used did not substantially change the strength or direction of associations with ACP (data not shown).

TABLE 4.

Prevalence Odds Ratios (ORs) for Moderate to High Levels of Transport Physical Activity, by Sociodemographic Characteristics, Levels of Exposure to the Intervention, and Self-Reported Health Status: Academia da Cidade Program (ACP), Recife, Brazil, 2007

	Moderate to High Transport Physical Activity, %	Moderate to High Transport Physical Activity, Crude OR (95% CI)	Moderate to High Transport Physical Activity, Adjusted ORa (95% CI)
Sociodemographics
Age, y
16–34 (Ref)	33.0	1.0	1.0
35–54	28.2	0.8 (0.6, 1.0)	1.2 (0.8, 1.8)
≥55	17.7	0.4 (0.3, 0.6)	0.8 (0.3, 1.8)
Gender
Male (Ref)	34.3	1.0	1.0
Female	24.1	0.6 (0.5, 0.7)	0.6 (0.4, 0.8)
Skin color
White (Ref)	22.3	1.0	1.0
Other	31.8	1.6 (1.3, 2.0)	1.5 (1.1, 2.0)
Educational level
Less than high school (Ref)	25.0	1.0	1.0
High school	37.2	1.8 (1.4, 2.2)	1.7 (1.2, 2.4)
College degree or higher	17.4	0.6 (0.5, 0.9)	0.8 (0.6, 1.1)
Marital status
Married (Ref)	27.7	1.0	1.0
Single	32.4	1.2 (1.0, 1.5)	1.0 (0.7, 1.4)
Other	15.4	0.5 (0.3, 0.7)	0.7 (0.4, 1.2)
Exposure to ACP
Participation in ACP
Current	25.0	0.8 (0.5, 1.4)	1.1 (0.3, 4.1)
Former	24.9	0.8 (0.3, 2.0)	1.0 (0.5, 2.0)
Never (Ref)	28.7	1.0	1.0
Lives near an ACP site
Yes	23.1	0.6 (0.5, 0.8)	0.7 (0.5, 0.9)
No (Ref)	31.9	1.0	1.0
Heard about or seen an ACP activity
Yes	26.4	0.8 (0.6, 0.9)	0.9 (0.7, 1.3)
No (Ref)	31.8	1.0	1.0
Health
Self-reported health status
Excellent or very good (Ref)	33.0	1.0	1.0
Normal or good	30.4	0.9 (0.7, 1.1)	1.0 (0.7, 1.6)
Fair or poor	24.0	0.6 (0.5, 0.8)	0.8 (0.5, 1.3)

Note. CI = confidence interval. Transport physical activity classifications were based on the International Physical Activity Questionnaire algorithm.

^aAdjusted for health status, sociodemographic characteristics (age, gender, education, marital status), risk factors (smoking, alcohol use, overweight or obesity), and neighborhood environment variables (traffic interference, safety, aesthetics, walkability).

DISCUSSION

Our results show that a community-based, professionally supervised physical activity intervention was independently associated with increased levels of leisure-time physical activity. We found an 18.3% prevalence of moderate to high levels of leisure-time physical activity in our sample, 25% higher than the rate observed in the VIGITEL survey conducted 2 years earlier (14.7%).¹⁰ Moreover, we found a positive association between leisure-time walking and participation in the ACP program; in fact, leisure-time walking was the most common ACP activity. Finally, we found an ACP awareness figure comparable to that observed for the independent, well-established Agita São Paulo physical activity promotion program.²⁶ To our knowledge, ACP represents the first instance in Latin America in which an intervention involving a professionally supervised physical activity class in a community setting has been evaluated with respect to its effectiveness at the population level.¹⁴

We did not find associations between engaging in transport physical activity and having participated in ACP or having heard about or seen an ACP activity. Living in a neighborhood without an ACP site was associated with higher transport physical activity levels. Residual confounding of poverty factors correlated with transport physical activity may explain this finding, in that the sociodemographic indicators available in our data set that were positively associated with transport physical activity (education and skin color) may have been inadequate to capture the full poverty effect.

Among Recife residents who are active in the workforce (i.e., those 10 years or older), approximately 44% live below the Brazilian poverty level (i.e., they earn less than $177 a month), and about 25% work in the informal economy (i.e., they are individuals with small unlicensed businesses, street vendors, and daily manual workers).^27,28 A large proportion of this low-income population is likely to walk or bike to work as well as to many other locations.

The confounding effects just described were compounded by the imbalanced socioeconomic distribution of survey respondents across the 2 study strata (i.e., neighborhoods with and without an ACP site). Approximately 60% of ACP sites are located in middle-class neighborhoods of Recife and out of reach of poor residents of the city. Previous Brazilian studies have shown that the primary determinant of commuting physical activity is socioeconomic status, and biking is a common form of transport.^29,30

Methodological Issues

We evaluated ACP only 5 years after it began, and the lack of baseline data precluded a longitudinal analysis. Therefore, although it is challenging to evaluate public health interventions via prevalence-based studies, we conducted a cross-sectional prevalence study.³¹ Prevalence-based studies have been criticized for not ensuring temporality between exposures and outcomes.^32,33 In our analysis, however, we found that former ACP participants still had significant and substantially higher levels of leisure-time physical activity than did individuals who had never participated in the program, providing some assurance that exposures preceded outcomes. For this reason, the direct association between ACP and leisure-time physical activity may indicate that exposure to ACP may have resulted in sedentary individuals becoming involved in leisure-time physical activity or may have provided people already engaging in physical activity an incentive to continue doing so.

Our study design may have led to bias in that, because our sampling stratification was based on presence of ACP site, inclusion of respondents who did not engage in moderate to high levels of leisure-time physical activity may have been dependent on their exposure to ACP.^33,34 However, our complex weighting procedure was designed to address this potential bias.²² Another concern is that, because this study involved a telephone survey, people without telephone service may have been underrepresented.^35–37 However, our adequate response rate (for a telephone survey) and our poststratification weighting enabled us to adjust for this potential selection bias.³⁸

Finally, all of our data were self-reported and may have been subject to recall bias if there were differences in reporting associated with exposure to ACP. However, in many but not all cases, recall bias tends to bias associations toward the null hypothesis. Also, demographic, smoking, and alcohol use questions similar to those used in the Recife Physical Activity Survey have been validated in the United States, and similar questions on weight and height have been validated in Brazil.^39–43 In addition, a 12-country validation of the translated physical activity questions used in our study showed that these questions had acceptable properties and recommended them for use in research.⁴⁴

Policy Implications

Several recent publications highlight the need for the type of evaluation reported here. For example, a landmark review published in the United States summarized evaluations of public health strategies in various areas, including physical activity.^11–13 Another landmark report was the World Health Organization's Global Strategy on Diet, Physical Activity and Health, which included recommendations for the development of strategies to increase physical activity at the population level.⁴⁵ A later report, which concluded that there was a need for more evaluation and international dissemination of effective evidence-based strategies designed to promote physical activity, proposed a 6-step framework for disseminating the strategy worldwide.⁴⁶

The US Guide to Community Services classification of physical activity programs includes informational, social and behavioral, and environmental and policy approaches to increasing physical activity levels at the community level.¹¹ ACP represents a unique and new category of health promotion in Latin America for which rigorous evaluation is much needed.¹⁴

In Latin America, the widespread popularity of community physical activity classes indicates the potential for coupling effective components of ACP and those of other effective programs. ACP appears to include aspects of another intervention category: community-wide policies and planning. The blending of several intervention categories into a comprehensive community program for promoting physical activity has been observed in other large Latin American cities as well.^47–49 Further defining the environmental, social, and policy aspects of ACP will be important to understanding the intervention and generalizing its results beyond Latin America.

Conclusions

Our findings suggest that community-level, professionally supervised, and publicly available programs such as ACP are effective in increasing levels of leisure-time physical activity. Evaluation and dissemination of local programs such as ACP in Recife can inform physical activity public health efforts both nationally and globally. If other cities and states in Brazil and even other countries are to effectively adopt programs such as ACP, they should consider the needs of local communities.

Our results also suggest that offering opportunities for transport physical activity should be considered as a possible strategy to increase levels of physical activity among the poor. Lack of physical activity is a pressing public health issue worldwide, especially in developing countries, because of its implications with respect to the growing epidemic of chronic diseases. This study is a positive step toward identifying effective community-level interventions designed to promote physical activity and improve health and well-being. Future researchers focusing on ACP and similar interventions should evaluate the appropriate mix of strategies needed to increase leisure-time and transport physical activity, consider designing interventions and evaluations simultaneously to allow for both baseline and follow-up measurements, consider establishing comparison groups, and carefully track the costs associated with replication and dissemination.