Background Studies of neighbourhood walkability and body mass index (BMI) have shown mixed results, possibly due to biases from self-reported outcomes or differential effects across age groups. Our objective was to examine relationships between walkability and objectively measured BMI in various age groups, in a nationally representative population.
Methods The study population came from the 2007–2011 Canadian Health Measures Survey, a cross-sectional survey of a nationally representative Canadian population. In our covariate-adjusted analyses, we included survey respondents aged 6–79 who were not pregnant, did not live in rural areas, were not missing data and were not thin/underweight. We used objectively measured height and weight to calculate BMI among adults aged 18–79 and zBMI among children aged 6–17. We categorised respondents into walkability quintiles based on their residential Street Smart Walk Score values. We performed linear regression to estimate differences between walkability quintiles in BMI and zBMI. We analysed adults and children overall; age subgroups 6–11, 12–17, 18–29, 30–44, 45–64 and 65–79; and sex subgroups.
Results The covariate-adjusted models included 9265 respondents overall. After adjustment, differences between walkability quintiles in BMI and zBMI were small and not statistically significant, except for males aged 6–17 in the second-highest walkability quintile who had significantly lower zBMIs than those in the lowest quintile.
Conclusion After accounting for confounding factors, we did not find evidence of a relationship between walkability and BMI in children or adults overall, or in any age subgroup with sexes combined. However, post hoc analysis by sex suggested males aged 6–17 in more walkable areas may have lower zBMIs.
- preventive medicine
- public health
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
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.
Strengths and limitations of this study
This study uses objective measures of walkability and BMI to examine how the relationship between walkability and BMI differs between different age groups.
This study analyses a large nationally representative Canadian population sample, which allows for stratification by age and adjustment for numerous sociodemographic variables.
Differences between study participants in variables such as caloric intake, amount of time spent in neighbourhood of residence and preference for living in a more walkable neighbourhood were not accounted for.
Residents of rural areas and individuals aged younger than 6 years or older than 79 years are not included in the study, so results may not be generalisable to these populations.
Body mass index (BMI), a measure of body fat or adiposity, is an important risk factor for many chronic diseases.1 2 BMIs from 25.0 to 29.9 kg/m2 are classified as overweight and 30.0 kg/m2 and over are classified as obese.1 3 People with BMIs of 25.0 kg/m2 and over are at increased risk of common and serious chronic illnesses, such as cardiovascular disease, type 2 diabetes and certain cancers.1 3 The prevalence of overweight and obesity, as measured through elevated BMI, has increased across the globe in recent decades.3 4 Elevated BMI is common in the Canadian population; over half of adults and over a quarter of children have BMIs of 25 kg/m2 or more.5 6
Health interventions that focus on changing individual behaviours have fallen short of reducing the high prevalence of overweight and obesity, which has prompted a focus on the environmental determinants of BMI.7 In recent years, public health has focused on walkability, a measure of neighbourhood design that includes residential density, proximity to stores and services, and intersection density.7 The underlying hypothesis is that walkable neighbourhoods encourage walking for transportation and other types of physical activity that contribute to reducing adiposity and lowering BMI.8
Despite the enthusiasm around walkability, research into the relationship between walkability and BMI or overweight and obesity has been mixed, with some studies showing a relationship and others showing minimal effects.9 10 To some extent, this may be the result of differences in analytical methods or in measurement of study variables, such as BMI assessment method. Most studies have relied on self-reported height and weight to estimate BMI, which have been shown to be less accurate than direct measures.4 11 People tend to overestimate height and underestimate weight, resulting in an artificially low BMI.4 11 Failure to account for age differences in the relationship between walkability and BMI may also have contributed to the mixed results. Different studies have focused on specific age groups, such as children, youth, working-age adults and seniors; however, all of these age groups have not been analysed in a single study of walkability and BMI, where differential effects by age could be examined. In addition, many earlier studies of walkability focused on only one or two cities, which may not have been representative of a broader population. Therefore, the objective of our study was to examine associations between walkability and objectively measured BMI and compare associations across a range of age groups, in a nationally representative population.
This study was a cross-sectional survey that used cycle 1 (2007–2009) and cycle 2 (2009–2011) of the Canadian Health Measures Survey (CHMS).12 13 The CHMS is a national health survey that collects both self-reported and direct measures. It uses a multistage sampling design stratified by geographical region, age and sex. The survey covers 96% of Canadians, but excludes people in institutions, full-time Canadian Armed Forces members, people living on reserves or other Aboriginal settlements, certain remote regions and the three Canadian territories.
The CHMS involves an in-person questionnaire conducted at participants’ households, followed by direct measures taken at mobile clinics within 50 km of their households.12 13
We combined CHMS cycles 1 and 2 to increase sample size. The combined response rate for people who completed both the household questionnaire and visit to the mobile examination centre was 51.7% in the 2007–2009 CHMS and 55.5% in the 2009–2011 CHMS. The response rate averaged across both cycles was 53.6%. Respondents were weighted to be nationally representative. Both cycles used the same data collection methods and included respondents with similar characteristics.12 13 Cycle 1 included people aged 6–79, while cycle 2 included those aged 3–79; for comparability, we excluded children 3–5 years old from cycle 2 before combining the two cycles. When calculating walkability, we used points within census-defined geographical areas as proxies for residential locations; however, these geographical areas were very large in rural areas and therefore poor proxies for residential locations, so we excluded respondents living in rural areas. Non-rural dwelling was defined as continuous built-up areas of 1000 people or more, with at least 400 people/km2.14 We also excluded women who were pregnant or missing data on pregnancy because elevated BMI due to pregnancy is not necessarily an indicator of overweight or obesity. We excluded children under 18 who were missing birth dates, as these were needed to calculate BMI-for-age z-scores, and excluded any respondents missing directly measured height or weight, as these were needed to calculate BMIs and BMI z-scores. After these exclusions, our population sample included people aged 6–79, living in non-rural areas, not pregnant, and not missing data on pregnancy, height, weight or birth date if under 18 (figure 1). This is the population described in table 1 and analysed in the unadjusted analyses. Our adjusted analyses also excluded people missing any of the covariates in the models, as well as people classified as thin or underweight (BMI z-score >2 SD below the mean for children, BMI <18.5 kg/m2 for adults). This was done because BMIs low enough to be classified thin or underweight are associated with unique health problems, such as chronic respiratory diseases,15 that are not hypothesised to be related to walkability.
Our walkability indicator was the Street Smart Walk Score metric (henceforth referred to as Walk Score).16 It is a validated metric that ranges from 0 (low walkability) to 100 (high walkability).17 18 It is based on the number, variety and proximity of different neighbourhood amenities such as restaurants/bars, parks and schools, as well as street connectivity. The website https://www.redfin.ca/how-walk-score-works contains more information on the Walk Score methodology. In 2014, we obtained Walk Score values for latitude/longitude coordinates within the census dissemination areas in which CHMS respondents lived. Dissemination areas are designed to cover areas with 400–700 people.19 In non-rural areas, this is small enough to approximate the locations of respondents’ residences. We matched dissemination areas and their corresponding Walk Score values to respondents by their postal codes using Statistics Canada’s Postal Code Conversion File.20
Body mass index
The CHMS measured participants’ weights using a digital scale and their heights using a stadiometer.
For respondents aged 18–79, we calculated BMI as weight in kilograms divided by squared height in metres. For respondents aged 6–17, we used a tool from WHO that calculates BMI-for-age z-score.21
We divided respondents into quintiles based on their survey-weighted Walk Score values. We also log transformed adult BMI to correct for its unequal variability at different Walk Score values, as recommended by a Box-Cox transformation of BMI. We examined the degree of correlation between respondents in the same dissemination areas using the intraclass correlation coefficient. We built both unadjusted and covariate-adjusted linear regression models. Previous studies have shown that age interacts with walkability,10 22 so we performed subgroup analyses within the following age strata: 6–11, 12–17, 18–29, 30–44, 45–64 and 65–79. These groupings reflected our hypothesis that walkability may have varying effects on the following life stages: child, youth, young adult, early-middle-aged adult, late-middle-aged adult and older retired/semiretired adult. We also performed post hoc subgroup analyses of males and females separately, as sex has also been shown to interact with walkability.23 For age groups 6–11 and 12–17, we modelled BMI z-score, adjusting for age, sex, cultural/racial origin, immigration within the past 10 years, household income, fruit/vegetable consumption and survey cycle. For age groups 18–29, 30–44, 45–64 and 65–79, we modelled log(BMI), adjusting for all variables listed above, plus marital status, smoking and leisure physical activity. We identified these variables as the most important confounders based on earlier research.2 23–28 After modelling log(BMI), we reverse transformed the mean predicted log(BMI) values and their CIs to obtain the average predicted BMI in each Walk Score quintile. In all analyses, we used the bootstrap survey weights that were provided with the CHMS data to account for the complex survey design. We performed all analyses using SAS V.9.4.
Patient and public involvement
Participant data are from a survey previously conducted by Statistics Canada. All participant identifiers had been removed from the data, so it was not possible to involve participants in the design of the study or dissemination of the results.
There were 9425 people aged 6–79, living in non-rural areas, not pregnant, and not missing data on pregnancy, height, weight or birth date if under 18 in our sample. Table 1 shows the sociodemographic characteristics of the overall sample, as well as within each Walk Score quintile. The population analysed in our multivariable analysis, which also excluded people who were thin/underweight or missing covariates, was 9265, including 3098 children aged 6–17 and 6167 adults aged 18–79 (figure 1). The intraclass correlation coefficient for respondents with the same dissemination area was 0.05, indicating low correlation; we, therefore, did not account for clustering by dissemination area in the analyses.
Among those aged 6–17 years overall, children in each of the higher Walk Score quintiles had slightly lower average BMI z-scores than those in the lowest Walk Score quintile; however, none of these differences were statistically significant, both before and after adjusting for covariates (table 2). In the unadjusted analysis, respondents 6–11 years old in the third Walk Score quintile had significantly lower BMI z-scores than those in the lowest Walk Score quintile, on average; however, this association was no longer significant after adjusting for age, sex, cultural/racial origin, immigration, household income and fruit/vegetable consumption (table 2). In the subgroup analysis by sex, males aged 6–17 in higher Walk Score quintiles had lower average BMI z-scores, a difference not observed among females (table 3). The difference among males in the fourth quintile was statistically significant and remained significant after adjusting for age, cultural/racial origin, immigration, household income and fruit/vegetable consumption. There were no other significant associations between any Walk Score quintiles among children under age 18, both before and after adjusting for covariates.
In the unadjusted analysis of all adults aged 18–79, people in the highest Walk Score quintile had significantly lower log(BMI) values than those in the lowest quintile (table 4). The highest quintile also had significantly lower log(BMI) values among the subgroup of adults aged 18–29, and among the male and female subgroups (tables 4 and 5). However, all of these associations were no longer significant after adjusting for age, sex, cultural/racial origin, immigration, household income, marital status, smoking, fruit/vegetable consumption and leisure physical activity. While average log(BMI)s were slightly lower in the highest Walk Score quintiles in all other comparisons among adults, the differences were not statistically significant, both before and after covariate adjustment.
After adjusting for relevant covariates, our study did not identify a significant relationship between neighbourhood walkability and objectively measured BMI in the overall study population or in any of the age subgroups, although, while not statistically significant, average BMI was slightly lower in the highest walkability quintile among adults overall and in all adult subgroups. In addition, in our post hoc analysis of sex subgroups, there was evidence that males aged 6–17 living in higher Walk Score quintiles had lower BMI z-scores than those in the lowest quintile, on average.
A number of earlier studies that examined the built environment and body mass showed null results. A longitudinal study of Australian adults did not find an association between walkability and self-reported weight change.8 An American longitudinal study saw no significant change in objectively measured BMI associated with a 1 SD change in walkability index.27 Conversely, several previous studies have identified statistically significant relationships. A study of adults aged 30–64 in Southern Ontario, Canada, found that people in the two highest quintiles of walkability had no change in prevalence of self-reported overweight and obesity over 11 years, while people in the lowest three quintiles had increased prevalence of overweight and obesity over this time period.26 Another study in Ontario found that people in the highest walkability quintile had lower self-reported BMIs than people in the lower quintiles.29 Ontario is a province of Canada that includes over one-third of Canada’s population, so the aforementioned positive results are based on populations similar to our study population. Perhaps our discordant results are due to our use of objective measures of BMI, rather than self-report. Self-reported BMI values are prone to biases,4 so it is possible that mitigating these biases resulted in our non-significant results.
Findings from other Canadian studies were more mixed. Pouliou and colleagues found an association between walkability and self-reported BMI in Vancouver, but not in Toronto.30 Glazier and colleagues found that people in the lowest walkability quintile had higher odds of self-reported overweight and obesity combined, but did not have higher odds of obesity alone.31 A longitudinal study by Wasfi and colleagues found that men who moved to more walkable areas had decreased BMI trajectories, while men who moved to less walkable areas had increased BMI trajectories; however, no relationship between walkability and BMI was found among women.23 After adjusting for confounders, our study did not show a significant relationship between walkability and BMI among adult males or females; however, we did find evidence of a lower average BMI z-score among male children in higher Walk Score quintiles, which did not show up among female children. Firm conclusions cannot be drawn from this finding, as it was the result of a post hoc subgroup analysis. However, future research should further explore how the relationship between walkability and BMI z-score may differ between males and females.
Studies of walkability and BMI have had less consistent findings than studies of walkability and physical activity, which have shown that people in more walkable areas tend to do more walking for transportation and more physical activity.9 10 32 This may seem counterintuitive, as one might assume that higher physical activity should lower BMI. However, BMI is influenced by many factors of which physical activity is only one.24 The higher physical activity associated with higher walkability may not be enough to reduce BMI by a measurable amount, given the multitude of other determinants of BMI.33 For instance, diet may influence BMI to a greater extent than physical activity.34 According to the Walking Calorie Burn Calculator by Shapesense, a 180-pound person who walks 1 km in 15 min burns only 71 calories.35
Strengths and limitations
Our study has numerous strengths, including its use of objective measures of walkability and BMI, its large Canada-wide study population and its inclusion of many sociodemographic characteristics. There are few studies of walkability that have used objective measures of BMI in a nationally representative population. There are also several limitations that should be considered when interpreting our findings. There may be residual confounding from unmeasured covariates, including differences between quintiles in average caloric intake, the food environment or length of time exposed to residential neighbourhoods.36 We were also unable to account for differences in exposure to non-residential areas, such as workplace neighbourhoods, which may also impact BMI. In addition, there was a time lag between the collection of height and weight data from the CHMS in 2007–2011 and calculation of Walk Score values in 2014. However, major changes in Walk Score quintiles are unlikely to have occurred during this time gap, so we do not expect the time difference to impact the results. Likewise, we do not expect the age of the study data, the oldest of which is from 2007, to impact study findings. The relationship between walkability and health outcomes, such as BMI, is unlikely to have changed since 2007. In addition, this was a cross-sectional study and therefore cannot be used to draw conclusions about a causal relationship. Finally, study results may not be generalisable to children younger than 6 or rural residents.
Despite the mixed results from studies of walkability and BMI, the potential health benefits of increasing walkability should continue to be investigated. Physical activity appears to improve health regardless of whether it results in weight loss.37 38 While elevated BMI is correlated with overall mortality and with several chronic diseases,3 15 it has limitations as a risk factor for poor health. The association between BMI and mortality is stronger among some populations than others. For instance, it is stronger among younger adults than among older adults.1 Other measures of adiposity, such as waist circumference, may be better indicators of metabolic risk than BMI.39 One study found that women in more walkable areas had lower odds of abdominal obesity measured by waist circumference, but no significant difference in overall obesity measured by BMI.40 A longitudinal study found that changes in walkability were associated with certain cardiometabolic risk factors, but not with BMI.27 Future studies of walkability should directly examine the chronic diseases associated with insufficient physical activity, such as type 2 diabetes and cardiovascular disease. Given the established link between physical activity and type 2 diabetes and cardiovascular disease,37 38 highly walkable areas may reduce risk of these diseases by increasing physical activity levels.
Our study did not identify significant associations between neighbourhood walkability and BMI, overall or in any age group, after adjustment for a variety of confounders, although, in our post hoc analysis of sex subgroups, there appeared to be a significant association between walkability and BMI z-score among males aged 6–17. Future studies are needed to explore whether a relationship exists among boys, but not girls. Our mostly non-significant findings may reflect a relatively limited influence of moderately increased physical activity on BMI in the absence of a difference in other factors, such as diet. However, previous research has linked walkability with physical activity, which may have health benefits independent of BMI.37 38 Future studies should investigate the relationship between walkability and diabetes or cardiovascular disease. If well-designed studies identify associations with these common chronic diseases, this will strengthen the evidence for improving overall health by increasing walkability.
Access and support for using the Canadian Health Measures Survey was provided by Statistics Canada’s Research Data Centre Program. Street Smart Walk Score values were provided by Redfin’s Walk Score.
MC and HM are joint senior authors.
Contributors JT, RC, LCR, MC and HM contributed to conceptualising and designing the study, planning analytical methods, and analysing and interpreting statistical output. JT conducted the literature review, statistical analysis and drafted the manuscript. JT, RC, LCR, MC and HM reviewed the manuscript for intellectual content, made revisions as needed and approved the final version for publication. JT, RC, LCR, MC and HM agree to be accountable for this work and to ensure that questions relating to the work are investigated and resolved.
Funding Financial support for this study was provided to all authors by Public Health Ontario through an annual peer-reviewed competitive Project Initiation Fund (PIF) (2013-14-019).
Disclaimer The opinions, results, and conclusions reported in this paper are those of the authors and are independent from the funding source. The study sponsor had no role in the study design; collection, analysis or interpretation of data; report writing or review; approval of the manuscript; or the decision to submit for publication.
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval The Ethics Review Board of Public Health Ontario granted ethics approval after reviewing our study protocol. Statistics Canada’s Research Data Centre granted access to the CHMS after reviewing our application and study protocol.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement Data may be obtained from a third party and are not publicly available.
The Canadian Health Measures Survey data are de-identified participant data, available upon successful application to Statistics Canada's Research Data Centre program: https://www.statcan.gc.ca/eng/rdc/index
Street Smart Walk Score® values are geocoded walkability data available from Redfin's Walk Score: https://www.redfin.ca/how-walk-score-works