Background Neighbourhood characteristics may contribute to differences in physical inactivity.
Purpose To evaluate whether the availability of sports facilities helps explain the differences in physical inactivity according to the economic context of the neighbourhood.
Methods 6607 participants representative of the population aged 16–74 years, resident in Madrid (Spain) in 2005, were analysed. Using ORs calculated by multilevel logistic regression, the association between per capita income of the neighbourhood of residence and physical inactivity was estimated, after adjusting for age, population density, individual socioeconomic characteristics and the availability of green spaces. The analysis was repeated after further adjustment for the availability of sports facilities to determine if this reduced the magnitude of the association.
Results Residents in the neighbourhoods with the lowest per capita income had the highest OR for the prevalence of physical inactivity. In participants aged 16–49 years, after adjusting for the availability of sports facilities, the magnitude of the OR in the poorest neighbourhoods with respect to the richest neighbourhoods increased in men (from 2.22 to 2.35) and declined by 13% in women (from 2.13 to 1.98). In contrast, in the population aged 50–74 years, this adjustment reduced the magnitude of the OR by 21% in men (from 2.00 to 1.80) and by 53% in women (from 2.03 to 1.48).
Conclusions The poorest neighbourhoods show the highest prevalence of physical inactivity. The availability of sports facilities explains an important part of this excess prevalence in participants aged 50–74 years, but not in younger individuals.
- Environmental Health
- Social Epidemiology
- Physical Activity
Statistics from Altmetric.com
There is ample empirical evidence supporting the inverse relationship between physical inactivity and health. The lack of physical activity increases the risk of suffering different types of chronic diseases like cancer, cardiovascular diseases, diabetes and obesity.1–4 Accordingly, many studies have investigated the importance of different factors as possible determinants of physical inactivity. Among these possible factors are characteristics of the areas where people live.
Socioeconomic context is one characteristic for which there is wide consensus about its relationship with physical inactivity. Specifically, people who live in areas with worse socioeconomic indicators have a greater frequency of physical inactivity, regardless of their individual socioeconomic circumstances.5–9 However, there is considerable uncertainty about the mechanisms by which socioeconomic context influences the frequency of physical inactivity.
Some authors have suggested that the higher prevalence of physical inactivity in more depressed socioeconomic areas could be owing to less availability of sports and recreational facilities.10–12 However, the research on the availability of physical activity resources by area wise socioeconomic status has yielded inconsistent results. While some studies have found that sports facilities were less accessible in more socioeconomically deprived areas,11–14 others have observed greater access to sports facilities in the more deprived areas.6 ,15 Furthermore, other studies have not found any relationship between socioeconomic context and the availability of these types of facilities.9 ,16
The findings on the relationship between the availability of sports facilities and physical inactivity are also inconsistent. Whereas various studies suggest that the probability of physical inactivity is lower in individuals who have greater access to sports facilities,17–20 others have not found this relationship.21–25 Finally, with the exception of two investigations carried out in Holland9 and Spain,8 no other studies have analysed the role of the availability of sports facilities in the relationship between area socioeconomic context and physical inactivity.
The results of the Spanish investigation should be interpreted with caution, however, because the province was used as the unit of aggregation for the data on socioeconomic context. There is wide demographic variation between Spanish provinces, with the population ranging from 95 000 inhabitants in the least populated province to 6.5 million in the most populated, and with a population density of 9–800 inhabitants/km2, respectively. These figures suggest that there may be considerable heterogeneity within some provinces with regard to socioeconomic situation and the availability of infrastructures. Thus, in interpreting the results of the Spanish study, one cannot rule out a possible underestimation of the association between socioeconomic context and physical activity, or of the effect of the availability of sports facilities on this association.
To overcome the aforementioned limitations, the present study aims to test this association in the neighbourhoods of the city of Madrid, the capital of Spain. The specific objectives are twofold: to evaluate the magnitude of the association between the economic context of the neighbourhood of residence and physical inactivity of its residents and to evaluate whether the availability of sports facilities in the district helps to explain this association.
The data were derived from the population interviewed in the 2005 Health Survey of the City of Madrid. In this survey, individuals age 16 and older were selected by two-stage cluster sampling, with stratification by census districts, which constituted the first stage units. The census districts were selected with a probability proportional to their population size, while respondents within each district were chosen by simple random selection. The questionnaire was answered by 7341 persons. As the survey targeted the non-institutionalised population, persons over age 74 were excluded since the probability of being institutionalised after that age is relatively high. The city of Madrid has 128 neighbourhoods, with a median population size in 2005 of 23 340 residents (IQR: 15 981–33 701), and with a median population density of 184 inhabitants/hectare (IQR: 88–263). The median area of the neighbourhoods was 135.81 Ha (IQR: 83.3–149.5).
The question used to collect information on physical inactivity in the survey was: “Which of the following possibilities best describes the frequency with which you do physical activity in your free time?: The response options were: (1) none; (2) moderate activity several times a month (walking, bicycling, gardening, light gym workout, activities involving light effort, etc); (3) moderate activity several times a week; (4) intense activity several times a month (tennis, jogging, bicycling, team sports, swimming, etc) and (5) intense activity several times a week. Based on the responses to this question, a binary variable was constructed to group respondents into two categories: individuals who report doing some type of physical activity (options 2–5), and those who report doing no physical activity in their free time (option 1).
Other individual characteristics in the analysis were age, limiting long-standing illness, social class and highest level of education completed by the person interviewed. Respondents were considered to have a limiting long-standing illness if they had any long-standing illness or disability that limited their activity during the last 12 months. Respondents were assigned to one of the following categories of social class based on the occupation of the head of household: professionals, managers and intermediate professions (I), self-employed workers and those working in the service industry (II), skilled manual workers (III) and unskilled manual workers (IV). Individuals were also assigned to one of four categories based on their educational level: no education or less than primary education; primary education; secondary education and tertiary or university education.
Per capita income was the indicator of wealth in each neighbourhood. This figure was estimated by the Institute of Statistics of the Community of Madrid based on tax records for the year 2000. Neighbourhoods were grouped into quartiles according to per capita income. Rather than analysing per capita income as a continuous variable, we preferred to group this variable in several categories in order to estimate the magnitude of the association in various ranges of per capita income. The pattern of the association was similar whether we used tertiles or quartiles, thus it was finally decided to show the results in quartiles since the association was stronger than when using tertiles. Each respondent was then assigned to an income quartile based on the neighbourhood of residence.
The number of sports facilities per 1000 population was estimated for each neighbourhood. A categorical variable was then calculated based on the quartiles of the distribution of this rate, and each individual was assigned to a quartile according to the neighbourhood of residence. Information on the number of sports facilities in each neighbourhood was obtained from the last National Census of Sports Installations, carried out in 2005. This census includes all collective sports installations, both conventional and non-conventional, as well as the sports facilities contained in each of them. ‘Non-conventional’ sports installations include parks and open public spaces with some type of sports facilities, such as exercise equipment, bicycle paths, jogging paths, etc. Installations used by a single family household were excluded. In this study, for each neighbourhood, we calculated the sum of the total number of indoor sports facilities, such as tennis courts, swimming pools, poly-sports courts, etc, plus the total number of sports facilities designed for outdoor physical activity such as fitness circuits, golf courses, etc.
Two characteristics of the neighbourhood of residence were also taken into account as adjustment variables, given their possible relation with physical inactivity: population density and percentage of green spaces per hectare. In this way all green spaces were included and not only those recorded by the census as having some type of sports facilities. Information on these indicators in each neighbourhood was obtained from the Statistics Department of the City of Madrid.
Men and women were analysed separately given that a previous study showed that the association between socioeconomic context of the area of residence and physical inactivity was stronger in women.8 The association of per capita income with individual and neighbourhood characteristics was evaluated by the χ2 test and by linear regression, respectively. The association between each study variable and physical inactivity was then calculated using the OR estimated by logistic regression. Finally, the association between neighbourhood per capita income and physical inactivity was evaluated. After adjusting for age, population density, limiting long-standing illness, individual socioeconomic characteristics and the percentage of green spaces per hectare, the availability of sports facilities was included to determine if this substantially reduced the magnitude of the association. The percentage change in the OR after inclusion of the sports facilities variable was then estimated according to the formula:
Given the hierarchical structure of the data presented in two levels—individuals within neighbourhoods—and the possible residual correlation between persons within neighbourhoods, the OR for neighbourhood characteristics was estimated using multilevel logit models which included a random effect of the intercept at the origin for each neighbourhood. The models were adjusted using the SAS macro procedure GLIMMIX.26 ,27 A previous investigation found that the relationship between socioeconomic context of the area of residence and physical activity was different in participants aged 50–74 years than in those younger than this age group.22 Accordingly, interaction terms between per capita income and age were added to assess whether the association between per capita income and physical inactivity was dependent on age.
Finally, given that the factors that influence frequent physical activity may be different from those that influence other levels of physical activity, we conducted a sensitivity analysis comparing no physical activity (option 1 of the question on physical activity) with frequent physical activity (options 3–5 considered together).
Table 1 shows the individual and area characteristics according to per capita income of the neighbourhood. The relationship between individual characteristics and per capita neighbourhood income was significant, except in the case of distribution by age. The availability of sports facilities, population density and the availability of green spaces did not show a statistically significant gradient with neighbourhood per capita income. However, the measure of sports facilities per 1000 population was 3.6 in the richest neighbourhoods and 1.9 in the poorest ones.
Table 2 shows the relationship of characteristics of the study participants and characteristics of the neighbourhood with the prevalence of physical inactivity. The OR for physical inactivity prevalence showed an inverse gradient with educational level and social class: the highest OR was seen in men with the lowest educational level and in those belonging to the lowest social class. The neighbourhoods with the lowest population density also showed the highest OR for physical inactivity prevalence in men. The lowest OR for physical inactivity prevalence was seen in the neighbourhoods with least availability of green spaces. In the case of availability of sports facilities, the residents in neighbourhoods belonging to the second quartile of this variable showed the highest OR for the prevalence of physical inactivity.
Table 3 shows the association between per capita income and physical inactivity. After adjusting for age, limitation of activity and population density, men living in neighbourhoods with the lowest per capita income had an OR for physical inactivity 2.76 times higher (95% CI 1.84 to 4.16) than those living in neighbourhoods with the highest per capita income. In women, the OR was 2.62 (95% CI 1.76 to 3.91). After adjusting for socioeconomic characteristics of the study participants, the OR was reduced to 2.06 (95% CI 1.34 to 3.18) in men and to 2.05 (95% CI 1.37 to 3.08) in women. The ORs for the two intermediate quartiles of per capita income were no significant in women after adjusting for the socioeconomic characteristics of study participants. So there was no evidence of a difference between women in these quartiles and those in the highest income quartile. In both men and women, additional adjustment for the availability of green spaces barely modified the association. Adjustment for the availability of sports facilities did not greatly reduce the magnitude of the OR for men, but it did in women: the OR in women residing in neighbourhoods with the lowest per capita income declined to 1.75 (95% CI 1.12 to 2.74). Except in the last model for women in which adjustment was carried out on the availability of sports facilities, the magnitude of the OR in all models showed an inverse gradient with per capita income of the neighbourhoods of residence. However, the CIs of the OR in the three categories other than the reference category overlap. We also performed a sensitivity analysis including the area variables in the models as quantitative variables—number of green spaces and number of sports facilities—but the magnitude and pattern of the association did not change substantially.
Table 4 presents the results of the sensitivity analysis, in which the reference group consists of persons who engage in frequent physical activity rather than just any physical activity. The OR in people residing in neighbourhoods with the lowest per capita income, as compared with the results shown in table 3, was somewhat higher in men and somewhat lower in women. Nonetheless, the pattern of the association was similar: the magnitude of the OR showed an inverse gradient with per capita income in the neighbourhood of residence, except in women after adjusting for the availability of sports facilities.
The interaction between per capita income and age was significant (p<0.01 in both men and women). This finding reflects the fact that the association between per capita income and physical inactivity differed depending on age and, consequently, analysis of the two age groups separately yielded different results. As seen in table 5, the availability of sports facilities in men increased the magnitude of OR between per capita income and physical inactivity in the 16–49years age group, but reduced it in those aged 50–74 years. In the latter age group, the estimated OR in persons living in neighbourhoods with the lowest per capita income declined by 21%: the OR decreased from 2.00 to 1.80. In women, the availability of sports facilities reduced the association in both age groups. The magnitude of the association in those living in neighbourhoods with the lowest per capita income was reduced by 13% in the 16–49years age group—the OR declined from 2.13 to 1.98—and by 53% in those aged 50–74 years—from an OR of 2.03 to 1.48.
This is the first multilevel study carried out in a Spanish city to have explored the importance of the availability of sports facilities in explaining the relationship between the economic context of the area of residence and physical inactivity in its inhabitants. The findings show that the neighbourhoods in the city of Madrid with the lowest per capita income have the highest physical inactivity. The availability of sports facilities explains a substantial part of this higher prevalence in participants aged 50–74 years.
In recent decades, many studies have focused on analysing the relationship between socio-economic context and physical inactivity. Most such studies have concluded that persons living in more deprived areas have a higher prevalence of physical inactivity than those residing in richer areas.5–9 Similar to these studies, our findings show that the prevalence of physical inactivity in individuals who live in the poorest neighbourhoods of Madrid is about twice that of persons living in more affluent neighbourhoods, after adjusting for socioeconomic characteristics of the residents.
Some authors have found that greater availability of green spaces in the area of residence is associated with a lower frequency of physical inactivity.28–30 It may be that more deprived areas have fewer green spaces, which would explain the differences in the prevalence of physical inactivity according to economic context of the area of residence. In this study, however, the existence of green spaces does not explain these differences in the prevalence of physical inactivity—first, because there is no relationship between the availability of green spaces and per capita income in the neighbourhood of residence, and second, because greater availability of green spaces is associated with a higher frequency of physical inactivity.
Other authors have attributed geographical differences in the prevalence of physical inactivity to the lack of sports and recreational facilities.10–12 However, while most studies have analysed the relationship between sports facilities and physical inactivity,17–20 very few have examined the possible influence of the availability of these infrastructures on the relationship between economic context of the area and physical inactivity in its residents. In two studies that have investigated this relationship, the availability of sports facilities did not explain the association between economic context of the area of residence and physical inactivity.8 ,9 On the other hand, according to a study conducted by Giles-Corti and Donovan,6 individuals who live in areas with lower socioeconomic level have better spatial access to sports facilities, but are less active than those who live in areas with higher socioeconomic level. In the present study, the same as in others,11 ,12 the availability of sports facilities was lowest in the poorest neighbourhoods. Nonetheless, the number of sports facilities explains a substantial part of the relationship between economic context of the neighbourhood of residence and physical inactivity only in participants aged 50–74 years, especially in women.
A previous study in Spain showed that the frequency of use of swimming pools and gyms in women aged 50–74 years showed a gradient according to provincial wealth: the highest frequency of use was observed in the richest provinces.20 Men aged 50–74 years also made more frequent use of these sports facilities in the richest areas, but the gradient was not statistically significant. This different gender pattern in sports facilities use may be responsible for the findings of this study: in women aged 50–74 years the availability of sports facilities explains 50% of the excess prevalence of physical inactivity in the poorest neighbourhoods, whereas in men it explains only 21%.
On the other hand, the frequency of use of sports facilities in the aforementioned study did not show an association with provincial wealth in persons below 50 years, except for use of gyms in women. This may explain why, in participants below this age in this study, the availability of sports facilities does not explain the excess prevalence of physical inactivity in the poorest neighbourhoods in men and explains barely 12% in women.
Several reasons may account for the null or small importance of sports facilities in explaining the differences in the prevalence of physical inactivity according to per capita income of the neighbourhood of residence in participants below 50 years of age. In a study conducted by Jilcott et al,21 an association was observed between perceived distance to gyms and physical activity, however, no relationship was found between the number of sports facilities and physical activity. It is possible that, for the same distance, persons of working age perceive sports facilities to be closer if they are located near their place of work than if they are located near their residence.
In any event, these findings have important policy implications. The different results by age suggest that increasing the number of sports facilities in poorer areas may not decrease the prevalence of physical inactivity in the younger population of these areas. Other public health intervention strategies are probably needed in this population group.
Certain considerations with regard to the data must be kept in mind when interpreting our findings. This study focused on physical inactivity in general. The estimates of this measure of inactivity according to individual socioeconomic characteristics are consistent with the existing empirical evidence. However, to better understand the explanatory power and possible importance of sports facilities in the relationship between neighbourhood wealth and physical inactivity, it would have been preferable to evaluate this relationship considering specific types of physical activity.
Likewise, we cannot rule out an information bias when classifying participants as active or inactive. The definition of physical inactivity considered only individuals who were completely inactive, excluding those who did some sporadic physical activity, in order to increase the specificity of the definition. This measurement error could have underestimated the association, since a differential information bias in respondents’ reports of physical activity is unlikely to occur depending on the neighbourhood of residence.
On the other hand, it must be recalled that this is a cross-sectional study, therefore it is necessary to consider the direction of the associations observed. However, it is unlikely that the results are owing to individuals with a higher prevalence of physical inactivity moving to neighbourhoods with lower per capita income.
The use of area-level variables such as number of facilities and greenspaces may not truly represent individual access. It would be better to have individual measures of access to these resources, such as distance to the nearest facility or greenspace or number of these types of resources within a certain distance from home. It is not possible to know how misclassification of study participants with respect to individual access might affect the results of our study.
Likewise, different authors have shown that when statistical analyses are conducted based on geographic areas, the relationship between two variables may differ depending on the level of aggregation of the data.31 ,32 That is, the results may vary according to the scale of the areal unit used. In the present study we conducted an analysis aggregating the data at the district level, and we also found a gradient in the relationship between per capita income of the area of residence and physical inactivity, although the magnitude of the association was smaller (data not shown).
In summary, the availability of sports facilities explains part of the relationship observed between the level of wealth in the neighbourhood of residence and physical inactivity in its inhabitants in the population age 50 and over, especially in women, but it does not explain this relationship in younger persons.
What is already known on this subject
The relationship between socioeconomic context and physical inactivity is well established. However, the mechanisms behind this relationship are uncertain.
Some authors have suggested that this relationship could be owing to the availability of sports facilities, but research on their availability by area socioeconomic status has yielded inconsistent results.
Few studies have investigated the role of the availability of sports facilities in the relationship between area socioeconomic context and physical inactivity.
What this study adds
This is the first multilevel study carried out in a Southern European city that has explored the importance of the availability of resources in explaining the relationship between the economic context and physical inactivity.
The poorest neighbourhoods show the highest prevalence of physical inactivity. The availability of sports facilities explains an important part of this excess prevalence in participants aged 50–74 years, especially in women, but not in younger individuals.
Contributors CP and ER originated and designed the study and coordinated the writing of the article. DAA and BA contributed to the analysis of the data and to the drafting of the paper. JMS and MEC contributed to the interpretation of the results and to the drafting of the paper. DM contributed to the design of the study and to the drafting of the paper. All authors contributed to the final version of the article. All authors have seen and approved the final version. ER is the guarantor.
Funding This work was supported by the Ministry of Science and Innovation the, grant number DEP2009-09502, titled ‘Socioeconomic context, availability of sports infrastructures and physical inactivity’.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.