Health & Place 30 (2014) 13–19

Contents lists available at ScienceDirect

Health & Place journal homepage: www.elsevier.com/locate/healthplace

Social disorder, physical activity and adiposity in Mexican adults: Evidence from a longitudinal study Luis Ortiz-Hernández a,b,n, Ian Janssen c,1 a

School of Kinesiology and Health Studies, Queen's University, Address: 28 Division St., Kingston, Ontario, Canada K7L 3N6 Departamento de Atención a la Salud, Universidad Autónoma Metropolitana Xochimilco, Address: Calz. del Hueso 1100, Col. Villa Quietud, Del. Coyoacán, D.F., México 04960, Mexico c Department of Public Health Sciences, and School of Kinesiology and Health Studies, Queen's University, Address: 28 Division St., Kingston, Ontario, Canada K7L 3N6 b

art ic l e i nf o

a b s t r a c t

Article history: Received 12 September 2013 Received in revised form 29 July 2014 Accepted 1 August 2014

This study analyzed the prospective relationship of community social disorder with sedentary behavior, sport participation, and adiposity in Mexican adults from the National Mexican Family Life Survey (MxFLS). The sample included 8307 adults (aged Z20 years) from 145 communities. During a three-year follow-up, participants from communities with high social disorder had a 1.36 cm larger increase in waist circumference than participants from communities with low social disorder. However, there were no differences in body mass index, television, or sport participation. These findings emphasize the need to promote healthy social environments in local communities. & 2014 Elsevier Ltd. All rights reserved.

Keywords: Social disorder Obesity Waist circumference Physical activity Sport

1. Introduction It is well recognized that there is a global obesity epidemic. Mexico has one of the highest prevalences of obesity in the world, i.e. 32.8% of adults were obese in 2012 (Food and Agriculture Organization of the United Nations (FAO) (2013)). Because of its consequences to the individual, family, and society, it is necessary to identify and intervene upon the modifiable determinants of obesity. The physical and social characteristics of the community environment are recognized determinants of obesity (Kirk et al., 2010; Leal and Chaix, 2011). Social characteristics of communities such as social disorder and safety are understudied aspects of the community environment (Leal and Chaix, 2011; Papas et al., 2007). Social disorder refers to the existence of incivilities such as vandalism, graffiti, vacant lots, public intoxication, litter, and dilapidated buildings. The concept of social disorder is related with in the “broken windows” theory (Wilson and Kelling, 1982),

n Corresponding author at: Departamento de Atención a la Salud, Universidad Autónoma Metropolitana Xochimilco, Address: Calz. del Hueso 1100, Col. Villa Quietud, Del. Coyoacán, D.F., México 04960, Mexico. E-mail addresses: [email protected] (L. Ortiz-Hernández), [email protected] (I. Janssen). 1 Tel.: þ613 533 6000x78631.

http://dx.doi.org/10.1016/j.healthplace.2014.08.001 1353-8292/& 2014 Elsevier Ltd. All rights reserved.

which posits that the presence of incivilities implies that many community members do not enforce the informal rules of order and suggests that there is lowered sense of mutual regard and civility (Wilson and Kelling, 1982). These environmental cues create an atmosphere that is conducive for mischievous behaviors and crime (Herbert, 1993). Social disorder could influence obesity through a physical activity and/or chronic stress pathways. People who live in communities with social disorder are often fearful of being victimized, often do not trust their neighbors, and often withdrawal from their community (Herbert, 1993). Therefore, it can be expected that they will spend less time in outdoor physical activities. Living in constant fear could trigger the chronic stress response, which is characterized by elevated cortisol secretion and changes in the production of several sex and growth hormones that are associated with increased accumulation of body fat, mainly in the abdominal region (Bjorntorp, 2001). An increase of the glucocorticoid sex hormone can also trigger overeating and weight gain (Bjorntorp, 2001). With one exception (Mujahid et al., 2008), previous studies with adults have observed a positive association between social disorder (Burdette and Hill, 2008; Christian et al., 2011; Ellaway et al., 2005; Glass et al., 2006; Poortinga, 2006; Stafford et al., 2007) and perceived crime (Fish et al., 2010) with anthropometric indicators of obesity. The association with physical activity is

14

L. Ortiz-Hernández, I. Janssen / Health & Place 30 (2014) 13–19

inconsistent (Foster and Giles-Corti, 2008); in some studies the relationship was negative (Glass et al., 2006; Stafford et al., 2007), in others there was no relationship (Burdette and Hill, 2008; Poortinga, 2006; van Lenthe et al., 2005), and in others there was a positive relationship (Poortinga, 2006). Findings from two studies support that social disorder influences obesity through the chronic stress pathway (Burdette and Hill, 2008; Stafford et al., 2007). Previous research examining the relationship between social disorder and obesity in adults has several common limitations. Only cross-sectional designs have been used to analyze the association of social disorder with weight (Leal and Chaix, 2011), which do not permit the establishment of causal relationships. In most studies participants came from a single urban area (Feng et al., 2010; Glass et al., 2006; Mujahid et al., 2008), which makes it difficult to generalize the findings. Some research was based on self-reported anthropometry (Burdette and Hill, 2008; Christian et al., 2011; Ellaway et al., 2005), which can produce biased estimates because the under- or over-reporting of weight and height are not random (Bostrom and Diderichsen, 1997). It is relevant to assess different indicators of adiposity because they can be interpreted as a proxy of different biological mechanisms. Body mass index (BMI) is considered an indicator of total adiposity, while higher values of waist circumference (WC) have been considered an indicator of chronic stress (Bjorntorp, 2001). While participation in physical activity has often been considered, existing studies have not considered sedentary behaviors (Foster and Giles-Corti, 2008; Humpel et al., 2002). The theory of social disorder proposes that people living in disordered communities can experience withdrawal from their communities (Herbert, 1993; Wilson and Kelling, 1982), therefore, it can be expected that they will spend more time in indoor activities, like watch television. Finally, all research about this topic has been done in highincome countries (Burdette and Hill, 2008; Christian et al., 2011; Ellaway et al., 2005; Glass et al., 2006; Mujahid et al., 2008). It has been suggested that the effects of social disorder could be more significant in Latin America than in high-income countries because insecurity and crime are more frequent (Pion-Berlin and Trinkunas, 2011). For example, during 2012 48.0% of Mexicans thought that in their neighborhoods robberies happen very or quite frequently, and 66.7% reported existence of alcohol consumption on the streets; while in 2011 in the United States of America the figures were 11.3% and 15.5%, respectively (World Values Survey, 2010–2014). This problem has worsened in the last three decades because in addition to ordinary crimes, other serious problems have developed, for example, the establishment of national or transnational criminal organizations (for drug dealing and/or kidnaping) and the expansion of “no-go” zones (Pion-Berlin and Trinkunas, 2011). This situation has created an almost generalized ambience of fear and anxiety. In this way, although the direction of the association between social disorder and weight could be the same in high-income countries and in Latin American ones, in the later the association could be stronger. At the same time, because of higher levels of crime and social disorder, it could be expected that Latin American residents are more familiar with an unsecure environment. This familiarity could result in no effect of social disorder on adiposity or only at high levels of social disorder. The aim of this study was to examine the prospective relationship between community social disorder and adiposity indicators in Mexican adults. We addressed several of the key gaps and limitations in the existing literature by examining a large representative sample of Mexican adults followed in a panel study. We hypothesized that the gain in BMI and WC would be higher among people living in communities with social disorder and that they

spend less time engaging in recreational physical activity and more time engaging in sedentary behavior.

2. Material and methods 2.1. Study sample This study was based on adult (aged Z20 years at baseline) participants from the 2002 and 2005 cycles of the longitudinal National Mexican Family Life Survey (MxFLS) (Rubalcava and Teruel, 2006, 2007). In the MxFLS a representative sample of Mexican households was selected using a stratified, multistage, probabilistic design. All residents from selected households were invited to participate. The 2002 cycle included 8440 households from 150 communities; approximately 90% of these households from 145 communities were followed in 2005. The rural localities had 2500 inhabitants or less. In urban areas the sampling was carried out using census tracts (in Mexico they are called Áreas Geoestadísticas Básicas or AGEBs) as primary unit. Each urban AGEB can include from 1 to 50 blocks where on average live 2000 people. The MxFLS consisted of an in-person interview and physical measures obtained at participants' homes, in-person interviews completed by community leaders (e.g., mayor) and service providers (e.g., education, health care), and interviewers' assessments of the communities. In the 2002 cycle, 20,047 people aged 20 years or older were members of the households included in the MxFLS. Participants with physical and mental disabilities were excluded (n ¼218). An additional 5622 participants were excluded from the television and sport and recreation analyses because they did not respond to that section of the questionnaire (n ¼5188) or because they were missing data for one or more of the covariates (n¼ 474). The main reason for this missing data is that household members were not present at the moment of fieldwork. An additional 942 participants were excluded from the obesity analyses because they were missing the anthropometric measures (n ¼246) or had extreme values that were considered invalid (BMI r10.0 or Z58.0 kg/m2, WC Z181.0 cm, n ¼373), or because they were pregnant or breastfeeding (n ¼ 323). Thus, the final sample sizes for the analyses based on the 2002 cycle were 14,167 for television and sport and recreation and 13,225 for BMI and WC. For the longitudinal analyses, 4142 were either lost to follow-up or had incomplete television, sport and recreation or covariate data in 2005 and 4918 were either lost to follow-up or had incomplete obesity or covariate data in 2005. Thus, the longitudinal analyses were based on 9755 participants for television and sport and recreation and 8307 participants for BMI and WC. In comparison with people with television and sport and recreation participation data, among those without them was higher the frequency of men, 65 year and older, single, those with higher education, members of households with high-income or headed by a person with high level of education, that lived in metropolitan areas, the west-central region, or in high social disorder communities (see Appendix 1 in the Supplementary data). The follow-up rate was lower amongst men, those with a higher education and income, 65 year and older, those who were single, and residents of metropolitan areas, of the central region, or in low social disorder communities (see Table 1). 2.2. Social disorder Potential social disorder measures came from the interviews completed by community leaders and from the assessments completed by the interviewers. Because the repeatability of the data collected from the community leaders was low

L. Ortiz-Hernández, I. Janssen / Health & Place 30 (2014) 13–19

Table 1 Descriptive characteristics and follow-up rate of the study sample. Mexican adults followed in the MxFLS 2002–2005. Characteristic

Full sample in 2002, (%)a

Follow-up rate, (%)

(n ¼14,167)

Follow-up sample in 2002, (%)a (n¼9755)

Sex Men Women

42.5 57.5

39.5 60.5

64.8 72.4

Age (y) 20–39 40–64 Z 65

53.7 36.4 9.9

51.8 39.4 8.8

67.0 74.2 62.1

Marital status Single Married or common-law Separated or divorced Widowed

17.8 71.9 4.7 5.6

15.1 74.8 4.9 5.2

59.1 71.8 70.6 64.9

Education of head of household Preschool or no education Elementary Middle High school College or graduate

14.1 46.0 20.3 9.5 10.1

14.3 47.8 20.1 8.8 9.0

68.4 71.8 68.3 64.2 60.4

Size of municipality Metropolitan areas (4 100,000) Urban areas (15,001–100,000) Small urban areas (2500–15,000) Rural ( o 2500)

49.0 10.3 17.6 23.1

43.6 11.1 18.4 26.9

60.5 69.6 72.6 75.4

Region of country Southeast Central West-central Northeast Northwest

21.8 34.8 22.3 11.9 9.2

24.5 30.9 23.6 11.2 9.8

74.2 64.8 71.8 66.5 68.2

Social disorder Low High

53.5 46.3

45.3 54.7

62.7 76.5

a

The percentages (%) are weighted proportions.

(i.e., correlations of 0.04–0.14 between 2002 and 2005 responses), we excluded these data. Through direct assessments, the interviewers rated the social disorder of the communities by responding to the following three questions: (a) “Does the community have abandoned buildings?”, (b) “How many walls in the community are painted with graffiti?”, and (c) “Does the community have abandoned or ransacked cars?” The response options for each question were: “none”, “very few”, “some”, and “many”. The responses in the 2002 and 2005 cycles for these three items were correlated (Spearman r ¼0.22, 0.52, and 0.22, respectively, p o0.010). Furthermore, in the 2005 cycle the interviewers completed two assessments for 138 communities, and the concordance between the assessments was good (kappa statistic ¼ 0.71, 0.73, and 0.72, respectively, p o0.001). Principal components analysis revealed that the three interview assessed social disorder items formed a single factor irrespective of whether the analysis used the communities or participants as observation units. The solution that used the participants as the observation unit was used to create a social disorder factor score. This factor score explained the 61.2% of the variance (Eigenvalue ¼1.83, Cronbach's alpha ¼0.77). The weights for the abandoned buildings, graffiti, and abandoned cars items were 0.56, 0.59, and 0.58, respectively. We decided to create categories of this variable because a high proportion of participants lived in communities with low scores of social disorder; for

15

example in the 2002 cycle 38.9% of participants inhabited in communities with a score of zero. The mean of the factor score was used as the cut-off to distinguish between participants who lived in communities with a high or low social disorder. Tertiles could not be used to define the cut-offs because most urban areas were in the highest tertile, which was problematic for the regression analyses. 2.3. Outcomes The outcomes were time spent watching television, participation in sports and recreation, BMI, and WC. To assess television, participants were first asked “During the past week, did you watch television?”, for those responding “yes”, they were asked “From Monday through Sunday of the past week, how many hours did you watch television”. Time watching television was reported in hours and converted to minutes per day for analyses purposes. Values 4720 min/day (12 h/day) were truncated. Similar questions were used to assess the participation in sport and recreational activities (“During the past week did you make or participate in sports, cultural, or entertainment activities outside your household?”); for this variable the time was expressed as minutes per week and the values 42400 min/week (equivalent to 8 h/day for 5 days) were truncated. Using standardized techniques (Lohman et al., 1988), a trained anthropometrist or nurse measured participants' weight, height, and WC. Weight was measured with a digital scale (Tanita) of the nearest 0.1 kg and height was measured to the nearest 0.1 cm with a SECA stadiometer. The BMI was calculated as weight/height2. WC was measured at the midpoint between the last rib and the iliac crest. 2.4. Covariates Several characteristics of the participants, their families, and communities were considered as covariates. Participant characteristics were: sex, age (continuous), marital status (single, married or common law, separated or divorced, and widowed), education (preschool or no education, elementary, middle school, high school, and college), occupation (student, homemaker, retired and unemployed, professionals and technicians, salesmen and office workers, industrial workers, and agricultural workers), smoking (smoker or non-smoker), and the number of reported health issues (diabetes, hypertension, heart disease, cancer, arthritis, gastric ulcer and migraine). Family characteristics consisted of the family income (divided by quartiles) and the education of the head of the household (with the same classification as the education of the respondent). Community characteristics consisted of the population size (metropolitan areas 4100,000 inhabitants, urban areas 15,001–100,000, small urban areas 2500–15,000, and rural areas o2500) and geographic region (southeast, central, west-central, northeast, and northwest). The month of the year when the fieldwork was carried out was considered as a proxy of seasonal variation. 2.5. Statistical analysis Cross-sectional analyses were done in STATA version 11 and took into account the complex survey design and sample weights. Regression analyses were done utilizing the multilevel modeling for repeated measures procedures (i.e. linear mixed-effect models for repeated measures). The first step of the multi-level modeling was to estimate the proportion of the variance in the outcomes attributed to the community-level. To do this, we included the community as a random term in order to calculate the intra-class correlation. This revealed that the variability attributed to

16

L. Ortiz-Hernández, I. Janssen / Health & Place 30 (2014) 13–19

differences among communities was 10.0% for television, 2.0% for sport and recreation, 6.1% for BMI, and 7.8% for WC. The likelihood ratio test revealed significant differences (p o0.050) between the models without and with the community as a random term for all four outcomes. The second step was to select the appropriate variance and covariance structure for the residuals associated with repeated measures. The options we selected at this step were based on: (1) whether the variance in the outcome was or was not constant in the 2002 and 2005 cycles, and (2) whether the repeated measures were correlated. The intraclass correlation coefficient between the two repeated measures was 0.29 for television, 0.10 for sport and recreation, 0.79 for BMI, and 0.62 for WC. For all four outcomes the models with the identity covariance matrix had the lowest likelihood ratio and therefore this variance–covariance structure was used. All models included a time X social disorder interaction term and were adjusted for the covariates. Height was also included as a covariate in the BMI and WC models. Only covariates with a p o0.10 were retained in the models. There is no consensus about the best scale to measure social disorder or incivilities because in some occasions have been treated as continuous variables (Burdette and Hill, 2008; Mujahid et al., 2008; Stafford et al., 2007) but others treat them as categorical (Ellaway et al., 2005; Glass et al., 2006; Poortinga, 2006). For this reason different models were estimated to analyze the social disorder as continuous or dichotomous variable. As the social disorder can be specially high and problematic in urban zones, a subsidiary analysis was done using data of participants from those localities. To assess in what extent attrition could affect our estimations, multivariate imputation using chained equations (MICE) was carried out (Lee and Carlin, 2010). The MICE method does not assume a multivariate normal distribution; therefore it can impute values to different types of variables. In five simulations the variables with more missing values (education, occupation and health issues of participants, education of head of the household, and family income) were filled using the rest of variables as predictors in fully conditional specification equations. In theses imputation models the outcome variables, social disorder and other covariates were introduced as predictors. There is no consensus if the missing values in outcome variables should be filled with imputations because similar results are obtained when subjects with missing values in outcome variables are excluded than when they are incorporated, however, the standard errors are lower when they are omitted (White et al., 2011). We did two sets of simulations: (1) models with multiply imputed data on all individuals predicting missing values for outcome variables and covariates, and (2) models restricted to individuals with observed values in outcome variables and predicting missing values for covariates only. As it has been reported (White et al., 2011), the results in both sets were similar; therefore we only report the results for the second one.

3. Results The socio-demographic characteristics of the sample in 2002 and 2005 are in Table 1. More than half of the participants were women, aged 20–39 years, married, and lived in a family in which the head of the household had less than a middle school education. Almost half of the sample resided in metropolitan areas and the central region of Mexico. Table 2 provides a description of the social disorder measures that were observed in the communities in 2002 and 2005. In 2002, approximately one third of the communities had abandoned buildings and abandoned or ransacked cars, and approximately

Table 2 Distribution of the communities according to the social disorder measures (N ¼ 145) at the MxFLS 2002–2005. Social disorder measure

Year

Abandoned buildings Graffiti painted walls Abandoned or ransacked cars

2002 2005 2002 2005 2002 2005

Rating of social disorder measure, (%) None

Very few

Few

Some

Many

66.9 53.1 49.7 38.6 69.0 58.6

16.5 34.5 23.4 36.6 17.2 30.3

9.7 8.9 6.9 8.9 7.6 5.5

6.2 2.8 14.5 12.4 5.5 5.5

0.7 0.7 5.5 3.5 0.7 0.0

Table 3 Average television, sport, body mass index (BMI), and waist circumference (WC) values of the study participants. Mexican adults followed in the MxFLS 2002–2005.

Men þWomen Television, min/day Sport, min/week BMI, kg/m2 WC, cm Men Television, min/day Sport, min/week BMI, kg/m2 WC, cm Women Television, min/day Sport, min/week BMI, kg/m2 WC, cm

Full sample in 2002 (n)

Follow-up sample in 2002 (n)

Follow-up sample in 2005 (n)

(14,167) 89.8 50.5 (13,225) 27.5 86.3

(9755) 86.9 43.5 (8307) 27.8 86.6

(9755) 85.0 31.5 (8307) 27.9 91.7

(6124) 85.8 68.2 (5738) 26.9 89.1

(3952) 82.2 58.6 (3352) 27.1 89.9

(3952) 79.5 38.4 (3352) 27.2 92.9

(8043) 92.7 37.5 (7487) 28.0 84.1

(5803) 90.1 33.8 (4955) 28.3 84.5

(5803) 88.6 27.1 (4955) 28.4 91.0

MxFLS, National Mexican Family Life Survey. The estimates are weighted means.

half had graffiti. The three social disorder measures deteriorated between 2002 and 2005 as a smaller proportion of the communities were rated in the “none” category in 2005. A description of the television, sport and recreation, BMI, and WC data is provided in Table 3. In 2002, the average time spent watching television was 89.8 min/day and the average BMI was 27.5 kg/m2. These values did not change over the 3-year follow-up, irrespective of gender. Conversely, time spent in sport and recreation decreased from 58.6 min/week in 2002 to 38.4 min/week in 2005, while the average WC increased from 86.6 cm to 91.7 cm. The reduction in sport and recreation was greater in men than in women (  20.2 min/week vs.  6.7 min/week), whereas the increase in WC was smaller in men (3.0 cm vs. 6.5 cm). Differences in television, sport and recreation, BMI, and WC between participants residing in communities with high and low social disorder are presented in Table 4. In 2002 and 2005, men and women living in high social disorder communities spent more time watching television than men and women living in low social disorder communities (p o0.010). These differences ranged from 17 min/day to 29 min/day. In 2002 among men and in 2005 in women, sport and recreation participation was higher in participants residing in high social disorder communities. There were no differences in BMI between high and low social disorder communities irrespective of sex and year. In 2002 there were not differences in WC between high and low social disorder

L. Ortiz-Hernández, I. Janssen / Health & Place 30 (2014) 13–19

17

Table 4 Television, sport, body mass index (BMI), and waist circumference (WC) of participants according to social disorder of their community. Mexican adults followed in the MxFLS 2002–2005. Sex and outcome measure

2002

2005

Low social disorder

High social disorder

p

Low social disorder

High social disorder

p

Men Television, min/day Sport, min/week BMI, kg/m2 WC, cm

71.4 47.9 27.0 89.7

95.2 71.6 27.1 90.1

0.000 0.031 0.769 0.638

66.7 31.6 27.0 91.5

91.1 44.5 27.4 94.2

0.000 0.093 0.295 0.003

Women Television, min/day Sport, min/week BMI, kg/m2 WC, cm

81.9 33.5 28.2 84.5

99.2 34.2 28.5 84.5

0.003 0.944 0.397 0.944

72.4 15.3 28.2 89.9

101.4 36.5 28.6 91.9

0.000 0.009 0.148 0.020

The estimates are weighted means. MxFLS, National Mexican Family Life Survey

Table 5 Multilevel linear regression models of changes in television, body mass index (BMI) and waist circumference (WC) among Mexican adults followed in the MxFLS 2002–2005. Sport and recreation (min/week)

Television (min/day)

BMI (kg/m2)

β

β

β

p

p

WC (cm) p

β

p

Complete-case analysis (n) Social disorder as dichotomous variable Survey cycle (05 vs. 02) Social disorder Cycle X social disorder

 1.83 4.37  7.40

0.638 0.267 0.107

1.87 0.10 0.79

0.206 0.961 0.739

0.17  0.01 0.06

0.011 0.911 0.550

3.66  0.07 1.36

0.000 0.761 0.000

Social disorder as continuous variable Survey cycle (05 vs. 02) Social disorder Cycle X social disorder

 1.84 0.15  1.99

0.654 0.916 0.254

 0.90 0.26 1.05

0.632 0.700 0.164

0.13  0.02 0.01

0.181 0.640 0.831

4.50  0.18 0.12

0.000 0.041 0.243

Imputation (n) Social disorder as dichotomous variable Survey cycle (05 vs. 02) Social disorder Cycle X social disorder Social disorder as continuous variable Survey cycle (05 vs. 02) Social disorder Cycle X social disorder

(9755)

(9755)

(10,619)

(8307)

(10,619)

(8307)

(10,566)

(10,566)

 4.27  0.62  1.23

0.226 0.863 0.770

0.08  0.51 0.94

0.968 0.812 0.697

0.11  0.05 0.05

0.231 0.602 0.670

4.15  0.11 1.06

0.000 0.663 0.000

 2.15 0.16  1.92

0.554 0.981 0.184

 0.79 0.00 0.93

0.693 0.999 0.250

0.15  0.02  0.01

0.122 0.543 0.716

4.56  0.13 0.04

0.000 0.104 0.678

Repeated measures at level-1, participants at level-2, and communities at level-3. MxFLS, National Mexican Family Life Survey; β, regression coefficient for fixed effects. Models adjusted by sex, age, marital status, education, occupation, smoking, and health issues of participants; family income, education of head of household, geographic region, size of municipality, and month when the fieldwork was carried out. The height of participant was included in the models for WC and BMI.

communities. However, in 2005 the average WC was 2.0 cm (women) to 2.7 cm (men) higher in participants residing in high social disorder communities than in participants residing in low social disorder communities. The socio-demographic characteristics are further described according to the social disorder of the community in Appendix 2 (see Supplementary data). The results of the multilevel models appear in Table 5. In the complete-case analysis, after adjusting for covariates, the time by social disorder (as dichotomous variable) interaction term was not a significant predictor of television, sport and recreation or BMI; but it was a significant predictor of WC (β ¼1.36, po 0.000). The marginal means for this interaction was plotted in Fig. 1. Although WC increased in both low and high social disorder communities between 2002 and 2005, the magnitude of the change was higher in participants that lived communities with high social disorder. Social disorder did not predict any outcome when it was introduced in the models as a continuous variable. The same patterns were observed when the analysis was done with imputed data. In

Fig. 1. Changes in waist circumference among according to level social disorder in their communities. Mexican adults followed in the MxFLS 2002–2005 Marginal means adjusted by covariates and estimated by multilevel modeling are graphed (see Table 5). MxFLS, National Mexican Family Life Survey.

these simulations the effect of social disorder as a dichotomous variables on WC was confirmed, but the size of effect was smaller (β ¼ 1.06, po 0.000).

18

L. Ortiz-Hernández, I. Janssen / Health & Place 30 (2014) 13–19

In the subsidiary analysis taking into account only the urban communities, the same trends were observed, although the differences were higher. In the complete-case analysis, among participants that lived in high social disorder communities had an increase in WC of 2.81 cm higher than those from low high social disorder communities (p¼0.000, data not shown in tables). In the imputation analysis the time by social disorder (as dichotomous variable) interaction term was a predictor of WC (β ¼1.23, p¼ 0.053). Although the interactions were marginal, the social disorder was associated with increase in television (β for time by social disorder interaction¼10.16, p¼0.070) and BMI (β ¼0.42, p¼0.071).

4. Discussion During a three-year follow-up, community social disorder as dichotomous variable was positively associated with changes in WC among Mexican adults. In the whole population, social disorder did not predict changes in BMI, television, or sport and recreation participation. However, in urban settings social disorder was marginally associated with increasing of BMI and television. In cross-sectional studies made in high-income countries, social disorder as continuous variable was positively associated with obesity (Burdette and Hill, 2008) and with WHR but not with BMI (Stafford et al., 2007). The same positive and linear association was observed when social disorder was measured as categorical variable (Christian et al., 2011; Glass et al., 2006; Poortinga, 2006). Just in one study (Ellaway et al., 2005) the relationship exists albeit not lineal, i.e., there were not differences in the risk of obesity between the first three quintiles of incivilities; but the higher risk of obesity happen in the fourth and fifth quintiles. Although one study did not find a positive association between social disorder and obesity (Mujahid et al., 2008), such work relied on a convenience sample and the assessment of neighborhood social environment was based on the participants' perceptions. Taken together, this evidence suggests that in high-income countries the correlation of social disorder with adiposity is positive and linear; which suggests that such effect starts at low levels of the former variable. In contrast, in all Mexican adults followed in the MxFLS, the community social disorder as dichotomous – but not as continuous variable – predicted changes in WC, but it was not correlated with BMI. In Mexican urban areas a positive association of BMI with social disorder as dichotomous variable was observed too, although the interaction term of social disorder by time was marginal. Therefore, our data suggested that in this middle-income country the association of social disorder with WC is not linear, but have a threshold effect. One possible explanation of the difference in respect the pattern of association is that low levels of incivilities are seen as a regular characteristic in many communities from middle-income countries. In other words, residents from communities with low social disorder are accustomed to their hostile environment and that it does not influence their healthy living behaviors. Therefore, in such countries the negative effect on adiposity is only observed at high levels of social disorder. One of the potential reasons why high social disorder is associated with obesity is because people from such communities spend less time in outdoor physical activities and/or more time in sedentary indoor behaviors. In our study time spent participating in sport and recreation and watching television were not predicted by social disorder in the total sample, while in urban settings social disorder was marginally associated with increasing time devoted to television. Community social disorder has been inconsistently associated with physical activity in past studies; in some instances a negative relationship was observed (Ellaway et al., 2005; Glass et al., 2006) while in others no differences existed (Poortinga, 2006; van Lenthe et al., 2005). These conflictive results

could be attributed to the differences in the assessments of physical activity and social disorder; therefore, the potential role of physical activity and sedentary behavior as mediators of the social disorder and adiposity relationship cannot be discarded. At the same time, our results and from others authors indicate that physical activity and sedentary behavior are not affected or are affected to a minimum extent by social disorder. Again, the fact that the association with a sedentary behavior was only seen in Mexican urban settings could be attributed that the effect is verified at high levels of social disorder. Chronic stress could also possibly explain the association between higher levels of social disorder and obesity. The persistent fear that could occur when living in a social disordered place could trigger a chronic stress response. This response could promote the accumulation of body fat, especially in the abdomen, due to disruption in secretion of several hormones. In support of this pathway, among the Mexican adults studied here, social disorder was associated with a greater increase in WC (a marker of abdominal obesity) but not BMI (a marker of general obesity). Similarly, in a cross-sectional study of adults from England and Scotland, Stafford et al. found that neighborhood social disorder was directly related with the waist-to-hip ratio but not with BMI (Stafford et al., 2007). Findings from another study (Burdette and Hill, 2008) support the role of physiological and psychological distress as mediators of the relationship between perceived social disorder and obesity. It is interesting to note that in the 2002 cycle of the MxFLS there were no differences in WC between participants from low and high social disorder communities, but these discrepancies appear clearly in 2005. One explanation of these trends is that insecurity worsened during the follow-up of participants in the MxFLS. For example, in Mexico the percentage of households that perceived their state as insecure increased from 44.0% in 2002 to 65.0% in 2010 (Instituto Nacional de Estadística, Geografía e Informática (INEGI), 2011). In our study, the change in WC over three years was 1.36 cm higher in participants from high social disorder communities than in participants from low social disorder communities. In the imputation analysis the difference was significant too but smaller (1.06 cm). To put that change in WC into context, a meta-analysis of prospective studies indicates that each 1 cm increase in WC is associated with a 2% increased risk of developing cardiovascular disease (de Koning et al., 2007). In addition, even costly interventions focused on changing physical activity and diet within obese patients in the primary care setting produce modest reductions in WC (i.e. 1.1 cm) (Ross et al., 2012). Therefore, focused programs and policies aimed at improving social disorder in disadvantaged neighborhoods and communities could potentially ameliorate the increasing rates of obesity. Neighborhoods renewal programs can improve health and wellbeing of residents (Mehdipanah et al., 2013). These programs include government investment to maintain or create public spaces (e.g. parks, community centers or sport schools) or social programs (e.g. employment training). Community participation is a key aspect and can be promoted by partnerships with residents to involve them in the design and implementation of programs related with social issues (e.g. addictions and crime prevention). Likewise, creation of green spaces can produce reduction of crimes and higher perception of safety (Garvin et al., 2013). Key strengths of research reported herein are that data are longitudinal in nature and based on a nationally representative sample. In addition, two directly measured anthropometric indicators of obesity were used. However, there are several limitations that should be mentioned. We relied on self-reported physical activity and sedentary behavior data. These self-reported measures have a low to moderate correlation with objective measures (van Poppel et al., 2010). In the case of the MxFLS, the sport and

L. Ortiz-Hernández, I. Janssen / Health & Place 30 (2014) 13–19

recreation participation item was general and included other activities (i.e. cultural and entertainment activities). Another limitation is that data of other potential mediators, such as diet and psychological distress, was not available. Ideally, our measure of social disorder could be enriched with crime statistics at community level; however in Mexico there are no reliable crime statistics at local level. Because of missing data and attrition, our analytic sample was less than a half of the total sample at baseline. Missing data for television and sport and recreation variables were more prevalent in people from high social disorder communities. In addition, in comparison with participants who were lost to follow-up, those that were followed tended to have higher BMI and WC, and spend less time in sport and recreation activities, while attrition was more frequent among residents of low social disorder. These patterns in attrition and missing values could explain the reduction in the sport and recreation time and increase in WC. Imputation analyses were performed to assess the possible effect of the missing values on the relationship between social disorder and outcomes. The results obtained with the complete-case and imputation analysis were basically the same. However, in the simulations with imputed values the effect size for WC was smaller; and in urban settings a marginal effect for television and BMI was observed. Finally, in our study social disorder was assessed using audits made by the interviewers. Although these assessments had adequate reproducibility, the correlation between cycles was low to moderate. This low correlation could in part be attributed to real increases in the frequency of incivilities in many places in Mexico (Instituto Nacional de Estadística, Geografía e Informática (INEGI), 2011; Pion-Berlin and Trinkunas, 2011). In summary, our findings indicate that social disorder predicts increases in abdominal fat, which is consistent with previous research. However, prospectively, sport and recreation participation and television were not correlated with social disorder. Objective measures of physical activity and sedentary behavior should be included in future studies. Also, the role of diet as a potential mediator should be assessed. Although the specific mechanism that explains the link between social disorder and obesity is unclear, we believe that the evidence linking social disorder and obesity is consistent and emphasizes the need to promote healthy social environments. Acknowledgments We are grateful with National Mexican Family Life Survey technical support staff because of its timely answers to our queries about the Survey's details. We are grateful with the challenging observations made by the reviewers. Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.healthplace.2014. 08.001. References Bjorntorp, P., 2001. Do stress reactions cause abdominal obesity and comorbidities? Obes. Rev. 2, 73–86. Bostrom, G., Diderichsen, F., 1997. Socioeconomic differentials in misclassification of height, weight and body mass index based on questionnaire data. Int. J. Epidemiol. 26, 860–866.

19

Burdette, A.M., Hill, T.D., 2008. An examination of processes linking perceived neighborhood disorder and obesity. Soc. Sci. Med. 67, 38–46. Christian, H., Giles-Corti, B., Knuiman, M., Timperio, A., Foster, S., 2011. The influence of the built environment, social environment and health behaviors on body mass index. Results from RESIDE. Prev. Med. 53, 57–60. de Koning, L., Merchant, A.T., Pogue, J., Anand, S.S., 2007. Waist circumference and waist-to-hip ratio as predictors of cardiovascular events: meta-regression analysis of prospective studies. Eur. Heart J. 28, 850–856. Ellaway, A., Macintyre, S., Bonnefoy, X., 2005. Graffiti, greenery, and obesity in adults: secondary analysis of European cross sectional survey. Br. Med. J. 331, 611–612. Feng, J., Glass, T.A., Curriero, F.C., Stewart, W.F., Schwartz, B.S., 2010. The built environment and obesity: a systematic review of the epidemiologic evidence. Health Place 16, 175–190. Fish, J.S., Ettner, S., Ang, A., Brown, A.F., 2010. Association of perceived neighborhood safety on body mass index. Am. J. Public Health 100, 2296–2303. Food and Agriculture Organization of the United Nations (FAO), 2013. The State of Food and Agriculture. FAO, Rome p. 2013. Foster, S., Giles-Corti, B., 2008. The built environment, neighborhood crime and constrained physical activity: an exploration of inconsistent findings. Prev. Med. 47, 241–251. Garvin, E.C., Cannuscio, C.C., Branas, C.C., 2013. Greening vacant lots to reduce violent crime: a randomised controlled trial. Inj. Prev. 19, 198–203. Glass, T.A., Rasmussen, M.D., Schwartz, B.S., 2006. Neighborhoods and obesity in older adults – the Baltimore memory study. Am. J. Prev. Med. 31, 455–463. Herbert, D.T., 1993. Neighborhood incivilities and the study of crime in place. Area 25, 45–54. Humpel, N., Owen, N., Leslie, E., 2002. Environmental factors associated with adults' participation in physical activity: a review. Am. J. Prev. Med. 22, 188–199. Instituto Nacional de Estadística, Geografía e Informática (INEGI), 2011. Encuesta Nacional sobre Inseguridad 2010. INEGI, Aguascalientes, México. Kirk, S.F.L., Penney, T.L., McHugh, T.L.F., 2010. Characterizing the obesogenic environment: the state of the evidence with directions for future research. Obes. Rev. 11, 109–117. Leal, C., Chaix, B., 2011. The influence of geographic life environments on cardiometabolic risk factors: a systematic review, a methodological assessment and a research agenda. Obes. Rev. 12, 217–230. Lee, K.J., Carlin, J.B., 2010. Multiple imputation for missing data: fully conditional specification versus multivariate normal imputation. Am. J. Epidemiol. 171, 624–632. Lohman, T.G., Roche, A.F., Martorell, R., 1988. Anthropometric Standardization Reference Manual. Human Kinetics Books, Champaign. Mehdipanah, R., Malmusi, D., Muntaner, C., Borrell, C., 2013. An evaluation of an urban renewal program and its effects on neighborhood resident's overall wellbeing using concept mapping. Health Place 23, 9–17. Mujahid, M.S., Roux, A.V.D., Shen, M.W., Gowda, D., Sanchez, B., Shea, S., Jacobs, D. R., Jackson, S.A., 2008. Relation between neighborhood environments and obesity in the multi-ethnic study of atherosclerosis. Am. J. Epidemiol. 167, 1349–1357. Papas, M.A., Alberg, A.J., Ewing, R., Helzlsouer, K.J., Gary, T.L., Klassen, A.C., 2007. The built environment and obesity. Epidemiol. Rev. 29, 129–143. Pion-Berlin, D., Trinkunas, H., 2011. Latin America's growing security gap. J. Democr. 22, 39–53. Poortinga, W., 2006. Perceptions of the environment, physical activity, and obesity. Soc. Sci. Med. 63, 2835–2846. Ross, R., Lam, M., Blair, S.N., Church, T.S., Godwin, M., Hotz, S.B., Johnson, A., Katzmarzyk, P.T., Levesque, L., MacDonald, S., 2012. Trial of prevention and reduction of obesity through active living in clinical settings: a randomized controlled trial. Arch. Intern. Med. 172, 414–424. Rubalcava, L., Teruel, G., 2006. User’s Guide for the Mexican Family Life Survey Second Wave, 〈http://www.ennvih-mxfls.org〉, (accessed 01.04.10). Rubalcava, L., Teruel, G., 2007. User’s Guide for the Mexican Family Life Survey First Wave, 〈http://www.ennvih-mxfls.org, (accessed 01.04.10). Stafford, M., Cummins, S., Ellaway, A., Sacker, A., Wiggins, R.D., Macintyre, S., 2007. Pathways to obesity: identifying local, modifiable determinants of physical activity and diet. Soc. Sci. Med. 65, 1882–1897. van Lenthe, F.J., Brug, J., Mackenbach, J.P., 2005. Neighbourhood inequalities in physical inactivity: the role of neighbourhood attractiveness, proximity to local facilities and safety in The Netherlands. Soc. Sci. Med. 60, 763–775. van Poppel, M.N.M., Chinapaw, M.J.M., Mokkink, L.B., van Mechelen, W., Terwee, C. B., 2010. Physical activity questionnaires for adults. A systematic review of measurement properties. Sports Med. 40, 565–600. White, I.R., Royston, P., Wood, A.M., 2011. Multiple imputation using chained equations: issues and guidance for practice. Stat. Med. 30, 377–399. Wilson, J.Q., Kelling, G.L., 1982. Broken windows. Atl. Mon. 249, 11. World Values Survey., 2010–2014. URL: www.worldvaluessurvey.org, (accessed 19.06.14).

Social disorder, physical activity and adiposity in Mexican adults: evidence from a longitudinal study.

This study analyzed the prospective relationship of community social disorder with sedentary behavior, sport participation, and adiposity in Mexican a...
287KB Sizes 2 Downloads 6 Views