Abstract
Neighborhood walkability—features of the built environment that promote pedestrian activity—has been associated with greater physical activity and lower body mass index (BMI; calculated as weight (kg)/height (m)2) among neighborhood residents. However, much of the literature has been cross-sectional and only a few cohort studies have assessed neighborhood features throughout follow-up. Using data from the Reasons for Geographic and Racial Differences in Stroke Study (2003–2016) and a neighborhood walkability index (NWI) measured annually during follow-up, we assessed whether the cumulative experience of neighborhood walkability (NWI-years) predicted BMI and waist circumference after approximately 10 years of follow-up, controlling for these anthropometric measures at enrollment. Analyses were adjusted for individual-level sociodemographic covariates and the cumulative experience of neighborhood poverty rate and neighborhood greenspace coverage. Almost a third (29%) of participants changed address at least once during follow-up. The first change of residence, on average, brought the participants to neighborhoods with higher home values and lower NWI scores than their originating neighborhoods. Compared with those having experienced the lowest quartile of cumulative NWI-years, those who experienced the highest quartile had 0.83 lower BMI (95% confidence interval, –1.5, −0.16) and 1.07-cm smaller waist circumference (95% confidence interval, –1.96, –0.19) at follow-up. These analyses provide additional longitudinal evidence that residential neighborhood features that support pedestrian activity are associated with lower adiposity.
Keywords: body weight, body mass index; neighborhood walkability; waist circumference
Abbreviations
- BMI
body mass index
- CI
confidence interval
- MESA
Multi-Ethnic Study of Atherosclerosis
- NWI
neighborhood walkability index
- REGARDS
Reasons for Geographic and Racial Differences in Stroke
- WC
waist circumference
Neighborhood walkability refers to features of the built environment that promote pedestrian activity and independence from private automobiles. Two of the indicators of built form that support pedestrian activity are higher population density and a higher density of establishments that support daily or weekly commercial needs (e.g., grocery stores, banks, churches, civic institutions) (1). Higher residential neighborhood walkability has been associated with more walking, higher overall physical activity levels, lower body mass index (BMI), lower incidence of diabetes, and improved glycemic control among residents (2–9). A majority of the extant literature on neighborhood walkability has been based on cross-sectional studies and has focused on single cities or regions (10, 11). In a salutary development for this field of study, more longitudinal studies have become available over the past decade (12). However, many of the prospective cohort analyses of BMI or physical activity assessed urban form or walkability once or only a few times during follow-up (13). For example, prior prospective studies, such as recent analyses in the Reasons for Geographic and Racial Differences in Stroke (REGARDS) Study and those in the Black Women’s Health Study, Harvard Alumni Study, and analyses of participants in the Multi-Ethnic Study of Atherosclerosis (MESA) who moved residence, used data on walkability at 1 or 2 time points (14–22). The limited availability of time-varying data on urban form or walkability limits analyses of the cumulative effect of exposure to neighborhood walkability or the effects of changes in neighborhood walkability due to residential moves or changes in the environments around people.
This longitudinal research has other limitations. Some studies relating changes in BMI to changes in neighborhood walkability have examined relatively short durations between observations; for instance, in a recent study of military veterans, walkability and BMI data were collected annually; in MESA, observations were spaced out roughly every 2 to 4 years (3, 4, 13). Because walking is a moderate-intensity activity, it is unclear that changes in walkability between annual or closely spaced observations will yield changes in weight observable in the time between observations. Moreover, few studies have assessed waist circumference (WC) as an outcome, which, because of changes in height and body composition in older populations, may be a better measure of adiposity than BMI (3, 22). Last, not all longitudinal studies include those who changed residence (movers) as well as those who did not (nonmovers) or measure the post-move environments for those who move (5, 14–16).
Here, we address the limitations of previous research by estimating an annual walkability score during the 10-year follow-up of a large, racially and geographically diverse cohort. Analyses assessed whether the cumulative experience of neighborhood walkability was associated with weight and WC at the end of follow-up, conditional on weight and WC measured at the beginning of cohort follow-up, with stratification by residential relocation during the follow-up period.
METHODS
REGARDS cohort
The recruitment and follow-up of the REGARDS cohort have been described extensively elsewhere (22–24). Briefly, between January 2003 and October 30, 2007, a total of 239 participants between the ages of 45 and 98 years (mean, 64.86 years) were recruited from a commercially available national list, with oversampling of Black participants and those living in the “Stroke Belt,” a US region with elevated stroke mortality (25). Study participants completed a computer-assisted telephone interview and took part in an in-home physical examination.
Between 2013 and 2016, approximately 10 years after baseline assessment, a second in-person examination was conducted. At the time of the second visit, 21.8% of the cohort participants had died and an additional 24.7% of participants declined to take part in the assessment. Thus, in total, 15,521 participants (53.5% of the cohort) completed the follow-up in-person examination (26). Height, weight, and WC were measured at both home visits (27). For the analyses conducted here, BMI values (calculated as weight (kg) divided by height (m)2) were considered outliers if they were less than 12 or greater than 70 (n = 17). WC values were considered outliers if they were 56.2 cm or less for women and 55.2 cm or less for men (the 0th percentile observed in the 2011–2014 National Health and Nutrition Examination Survey), or at least 180.4 cm for women and 202 cm for men, the 99th percentile seen in a case series of patients who had had bariatric surgery (28). Participants with outlying values for BMI (n = 17) or WC (n = 1) were removed from respective analyses of BMI, weight, and WC.
Geospatial analyses
Participant-reported addresses at enrollment and during follow-up were supplemented with address data purchased from LexisNexis (New York, New York). The data sources were combined to identify the address of residence at enrollment and in January of each year of follow-up, and the addresses were geocoded using ArcGIS Business Analyst Desktop 10.5.1 with ESRI 2016 Business Analyst Data (ESRI, Redlands, California). Radial buffers of 1 km and 5 km were defined around the longitude and latitude coordinates of participant addresses and characterized for built and social environment characteristics. Figure 1 shows the spatial distribution of the REGARDS cohort at enrollment with 5-km buffers around the residences depicted, with the buffer boundaries of overlapping buffers dissolved. The 1-km radii were used because prior work found associations between walkability measured at this scale and physical activity and BMI, and the larger size was chosen as being perhaps more appropriate for more rural environments and to support sensitivity analyses (1, 2, 29). Land area within the buffers was calculated using the areal hydrography layer from the US Census Bureau 2016 TIGER Line Geodatabase files (30) representing 2010 US Census geographies. All spatial calculations were completed using ArcGIS Desktop, version 10.3. In total, 29,897 REGARDS participants had valid baseline BMI data and addresses at enrollment that could be geocoded.
Figure 1.
Spatial distribution of Reasons for Geographic and Racial Differences in Stroke Study (REGARDS) participants at enrollment, United States, 2003–2007. The map displays, in pink, 5-km radial buffers around the residence of REGARDS participants at enrollment in the cohort. The borders of overlapping radial buffers have been dissolved to create agglomerated areas. The map scale is 1:26,500,000.
The National Establishment Time Series (NETS) data set for 1990 to 2014 was used to identify walkable destinations, retail, commercial, institutional, and not-for-profit establishments that contribute to neighborhood vibrancy and allow residents to meet daily needs without driving (1, 29). The count of walkable destinations within each radial buffer was divided by the land area of the buffer to measure the density of walkable destinations for each year of follow-up. For years of follow-up after 2014, the values for 2014 were carried forward through 2016.
Areal-weighted interpolation of harmonized tract-level data from the 2000 and 2010 US population census, published in the Longitudinal Tract Database (Brown University, Providence, Rhode Island), were used to estimate the population density of the radial buffers (31–33). Similarly, Longitudinal Tract Database data from the 2000 population census and the 2008–2012 American Community Survey were used to estimate the percentage of the population in each buffer living in poverty, median household income, and median home value in 2000 and 2012. Because green space may be associated with physical activity and lower walkability as measured by population density and destinations that serve daily needs, the National Land Cover Database data from 2001, 2006, and 2011 were used to estimate the proportion of each radial buffer with green landcover. For each measure, linear interpolations were applied to estimate values for intervening years, and values from the final available year of data were carried forward for subsequent years (3, 34, 35).
The annual density of walkable destinations and estimated population density for each radial buffer were z scored and the 2 z-scored values were summed to create a neighborhood walkability index (NWI) (1). To place the NWI scores for the participants in a national context, density of walkable destinations and population density for 2010 were calculated for each census tract in the continental United States that had more than 100 residents. The means and standard deviations (37.03/km2 (117.76) and 2,029.42/km2 (4,543.81) for density of walkable destinations and population density, respectively) for these tracts were then used to calculate the z scores annually for REGARDS participants’ radial buffers.
Statistical analyses
The 29,897 REGARDS participants who had valid enrollment data for BMI and whose addresses could be geocoded were included in cross-sectional analyses of baseline neighborhood and anthropometric data and in analyses of changes in NWI between the baseline address and last address observed during follow-up. Data from the 14,127 participants who took part in the second in-person examination that occurred approximately 10 years after enrollment and who had valid BMI data recorded at the examination were analyzed for longitudinal changes in anthropometric measures. Descriptive statistics were calculated for NWI scores for US tracts and for the 1-km radial buffers of REGARDS participants at enrollment, and for BMI, weight, and WC at enrollment. To investigate possible differences in those who were followed up via the in-home examination and those who were not, these descriptive data were calculated for all REGARDS participants at enrollment and separately for those who participated in the in-home follow-up. We used t tests to determine whether enrollment and follow-up NWI, BMI, and WC data significantly differed for participants who moved versus those who did not. For those who took part in the second in-person examination and moved during follow-up, NWI scores, poverty rates, median household income, and median home value were calculated for the residential neighborhood as of January prior to the move and for the residential neighborhood as of January after the move. Building upon the work of Long et al. (26), inverse probability weights for successful follow-up were calculated using individual-level data and incorporating 1-km radial buffer data at enrollment for NWI score, the poverty rate, the proportion of the radial buffer covered by green space, and the median home value. Inverse probability weights were calculated on the basis of the initial cohort of 29,897 enrolled participants with valid anthropometric data and an address that could be geocoded. A total of 14,127 participants from the initial cohort were followed up at the second in-person examination. These weights were incorporated into the statistical analyses to account for possible bias due to selection into the group with follow-up for anthropometric outcomes.
For each participant, the cumulative experience of neighborhood NWI (NWI-years) was calculated by summing the annual NWI score for 1-km buffers across each year of follow-up. As with the NWI-years measure, the cumulative experience of proportion of the radial buffer covered by green space (% green space-years) and proportion of residents in the radial buffer living in poverty (% poverty-years) were calculated. The cumulative NWI-years measure had a very skewed distribution; therefore, total NWI-years was then categorized into quartiles defined by the distribution of NWI-years in the entire REGARDS population.
To account for possible differences in height between groups of participants, BMI scores were used for analyses that compared anthropometric values between groups of participants at cohort entry or follow-up. Because there are often decreases in height with aging, which causes BMI values to increase even if weight has remained static, weight and changes in weight were analyzed when describing changes in anthropometric values within individuals.
A longitudinal analytical strategy was used to estimate weight at follow-up as predicted by quartiles of NWI-years, conditional on weight at enrollment and covariates. The analyses were implemented as generalized estimating equation linear regression models with clustering on county of residence at follow-up and incorporation of inverse probability weights for follow-up. Additional covariates for participant age, height, and income at follow-up; education attainment at enrollment; race; sex; duration of follow-up; % poverty-years; and % green space-years were included in the models. Because the issue of residential moves has been understudied, stratified analyses were also conducted among those who moved residences during follow-up and those who stayed in the same residence throughout. Additionally, because REGARDS sampled on Black and White race and was designed to assess the health of older adults, stratified analyses were also conducted by race and then also by age, with those aged 65 years or older at enrollment placed in 1 stratum and those younger than 65 years in the other stratum. To avoid calculating a very large number of P values, differences in associations between NWI-years and weight across strata were judged by assessing whether the 95% confidence intervals (CIs) around the stratum-specific point estimates of associations in 1 stratum (e.g., White participants) covered the stratum-specific point estimates of associations generated for the other stratum (e.g., Black participants). Analyses were conducted using 1-km buffer measures and then 5-km buffer measures. All analyses were repeated for WC at follow-up.
RESULTS
For the entire REGARDS participant population with valid BMI data at enrollment and addresses that could be geocoded (n = 29,897), the mean BMI was 29.32, weight was 84.32 kg, and WC (n = 29,443) was 95.23 cm. For those who took part in the second in-home examination 10 years later, BMI (n = 14,127) at enrollment was 29.28, weight was 84.40 kg, and WC (n = 13,989) was 95.20 cm. Between the first and second home examination, weight decreased by a within-person average of 1.83 kg and WC increased by a within-person average of 2.45 cm.
The mean NWI scores for REGARDS participants at enrollment and follow-up were lower, with a narrower interquartile range, compared with NWI scores for US Census tracts in 2010 (Table 1). REGARDS participants who took part in the in-home examination 10 years after enrollment had very similar NWI scores at enrollment as the overall REGARDS participant population (Table 1). Among those who were followed up for anthropometric measures, there was a small mean within-person decrease in NWI of 0.009 units (interquartile range, − 0.026, 0.032) and the Spearman rank correlation between NWI at enrollment and at follow-up was 0.89. The Spearman rank correlation between NWI at enrollment and the sum of NWI across years of follow-up (NWI-years) was 0.92; among nonmovers and movers, the Spearman rank correlations were 0.96 and 0.81, respectively.
Table 1.
Neighborhood Walkability Index Scores for All Tracts in the United States and for Participants in the Reasons for Geographic and Racial Differences in Stroke Study at Enrollment (2003–2007) and Follow-up (2013–2016), United States
| NWI Score | Mean | Median (IQR) | Minimum | Maximum |
|---|---|---|---|---|
| 2010 NWI scores for US tracts with population > 100 | 0.00 | −0.44 (−0.72, −0.01) | −0.76 | 69.71 |
| NWI score at enrollment for all REGARDS participants (1 km) | −0.24 | −0.46 (−0.67, −0.15) | −0.76 | 26.78 |
| NWI score at enrollment for those followed up for anthropometric measures (1 km) | −0.28 | −0.49 (−0.68, −0.19) | −0.76 | 17.79 |
| NWI score at follow-up for those followed up for anthropometric measures (1 km) | −0.28 | −0.49 (−0.69, −0.22) | −0.76 | 17.47 |
Abbreviations: IQR, interquartile range; NWI, Neighborhood Walkability Index; REGARDS, Reasons for Geographic and Racial Differences in Stroke.
Among all REGARDS participants, 25% changed residences at least once during follow-up, and among those who were followed up for anthropometric outcomes, 29% changed residence at least once. In the entire population, and among those who were followed up for anthropometric outcomes, the participants who moved had a higher NWI score at enrollment than those who did not move during follow-up, but movers and nonmovers had similar NWI scores at the end of follow-up (Table 2). Among those who moved, the first change of residence, on average, brought the participants to neighborhoods with lower NWI scores, lower poverty rates, higher median household incomes, and higher home values than in their originating neighborhoods (Table 3). Participants who made additional residential moves after this first change of residence moved to neighborhoods with similar NWI and socioeconomic conditions as the destination neighborhoods for their first move (data not shown).
Table 2.
Neighborhood Walkability Index Scores and Anthropometric Measures of Participants in the Reasons for Geographic and Racial Differences in Stroke Study Who Did and Did Not Move During Follow-Up, United States, 2003–2016
| All Participants With Geocoded Addresses and Measurement at Baseline | Those With Geocoded Addresses and Followed Up With Measurement at Baseline and Follow-up | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Did Not Move | Moved | Did Not Move | Moved | |||||||||||||
| Measure | No. | Mean | Median | P Value | No. | Mean | Median | P Value | No. | Mean | Median | P Value | No. | Mean | Median | P Value |
| Neighborhood Walkability Index | ||||||||||||||||
| Baseline | 22,206 | −0.25 | −0.47 | 7,381 | −0.20a | −0.42 | 10,018 | −0.29 | −0.50 | 4,109 | −0.24a | −0.45 | ||||
| End of follow-up | −0.24 | −0.46 | −0.25 | −0.47 | −0.28 | −0.48 | −0.29 | −0.50 | ||||||||
| Within-person change | 0.01 | <0.001b | −0.05 | <0.001b | 0.01 | <0.001b | −0.05 | <0.001b | ||||||||
| Measured BMIc | ||||||||||||||||
| Baseline | 22,206 | 29.37 | 28.38 | 7,381 | 29.16a | 28.20 | 10,018 | 29.38 | 28.41 | 4,109 | 29.05a | 28.13 | ||||
| End of follow-up | 29.09 | 28.15 | 28.97 | 28.04 | ||||||||||||
| Within-person change in BMI | −0.29 | <0.001b | −0.07 | 0.18b | ||||||||||||
| Within-person change in weight (kg) | −1.96 | <0.001b | −1.53 | <0.001b | ||||||||||||
| Waist circumference (cm) | ||||||||||||||||
| Baseline | 22,098 | 96.16 | 95.25 | 7,345 | 95.71a | 94.62 | 9,919 | 95.22 | 93.98 | 4,070 | 95.12 | 93.98 | ||||
| End of follow-up | 97.56 | 96.52 | 97.80 | 96.52 | ||||||||||||
| Within-person change | 2.33 | <0.001b | 2.69 | <0.001b | ||||||||||||
Abbreviation: BMI, body mass index.
a P < 0.05 by t test comparing participants who moved with those who did not move.
b By paired t test comparing baseline and follow-up data within participants.
c Calculated as weight (kg)/height (m)2.
Table 3.
Neighborhood Characteristics of Participants in the Reasons for Geographic and Racial Differences in Stroke Study Followed Up for Anthropometric Outcomes Who Moved Residences at Enrollment, in the January Before Moving, and at Their Destination Neighborhood, United States, 2003–2016
| Neighborhood Characteristic | At Enrollment | In January Prior to Change in Residence | At Destination Neighborhood | |||
|---|---|---|---|---|---|---|
| Mean | Median | Mean | Median | Mean | Median | |
| NWI | −0.23 | −0.44 | −0.21 | −0.41 | −0.28 | −0.51 |
| Poverty, % | 0.18 | 0.16 | 0.19 | 0.17 | 0.16 | 0.13 |
| Home value, $ | 162,289 | 125,937 | 175,708 | 133,474 | 186,170 | 150,827 |
| Annual household income, $ | 51,598 | 46,403 | 50,262 | 44,806 | 54,581 | 50,665 |
Abbreviation: NWI, Neighborhood Walkability Index.
Higher quartile categories of NWI-years experienced within 1 km of the home during follow-up were associated with lower body weight at follow-up, after adjustment for weight and height at enrollment and other individual- and neighborhood-level covariates (Table 4). Comparing those who did and did not move during follow-up, the associations observed between NWI-year quartiles 3 and 4 and weight were numerically larger for movers. However, the confidence intervals observed for movers covered the point estimates observed among nonmovers, indicating that the differences in the point estimates were not statistically significant. The associations between quartiles of NWI-years and weight at follow-up was similar for White people and Black people.
Table 4.
Differences in Weight and Waist Circumference at Follow-up by Quartile of Cumulative Neighborhood Walkability Index Score for 1 km Around Residence, Among Participants in the Reasons for Geographic and Racial Differences in Stroke Study, United States, 2003–2016
| Neighborhood Walkability | Weight at Follow-up, kg | Waist Circumference at Follow-up, cm | ||||
|---|---|---|---|---|---|---|
| Difference in Weight | 95% CI | P Value | Difference in Waist Circumference | 95% CI | P Value | |
| Entire Population | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.69 | −1.19, −0.20 | 0.01 | −0.35 | −0.91, 0.21 | 0.22 |
| Quartile 3 | −0.68 | −1.32, −0.04 | 0.04 | −0.59 | −1.30, 0.12 | 0.1 |
| Quartile 4 | −0.83 | −1.50, −0.16 | 0.02 | −1.07 | −1.96, −0.19 | 0.02 |
| P for trend | 0.06 | 0.02 | ||||
| Participants 65 Years or Older at Enrollment | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.67 | −1.35, 0.01 | 0.05 | −0.29 | −1.10, 0.53 | 0.49 |
| Quartile 3 | −0.91 | −1.69, −0.13 | 0.02 | −0.45 | −1.45, 0.54 | 0.37 |
| Quartile 4 | −0.82 | −1.78, 0.14 | 0.09 | −0.99 | −2.29, 0.30 | 0.13 |
| P for trend | 0.15 | 0.15 | ||||
| Participants 64 Years or Younger at Enrollment | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.65 | −1.40, 0.09 | 0.08 | −0.34 | −1.10, 0.42 | 0.38 |
| Quartile 3 | −0.36 | −1.29, 0.57 | 0.44 | −0.56 | −1.51, 0.40 | 0.25 |
| Quartile 4 | −0.70 | −1.61, 0.22 | 0.13 | −0.85 | −1.84, 0.15 | 0.1 |
| P for trend | 0.31 | 0.10 | ||||
| Among Black People | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.60 | −1.60, 0.41 | 0.24 | −0.62 | −1.67, 0.44 | 0.25 |
| Quartile 3 | −0.79 | −2.06, 0.48 | 0.22 | −1.06 | −2.33, 0.22 | 0.10 |
| Quartile 4 | −0.79 | −2.01, 0.43 | 0.20 | −1.80 | −3.32, −0.28 | 0.02 |
| P for trend | 0.33 | 0.03 | ||||
| Among White People | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.77 | −1.37, −0.17 | 0.01 | −0.35 | −1.02, 0.32 | 0.31 |
| Quartile 3 | −0.56 | −1.31, 0.19 | 0.14 | −0.39 | −1.27, 0.50 | 0.39 |
| Quartile 4 | −0.87 | −1.72, −0.02 | 0.04 | −0.46 | −1.44, 0.53 | 0.37 |
| P for trend | 0.09 | 0.37 | ||||
| Participants Who Did Not Move During Follow-up | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.69 | −1.30, −0.08 | 0.03 | −0.06 | −0.79, 0.66 | 0.87 |
| Quartile 3 | −0.54 | −1.31, 0.23 | 0.17 | −0.20 | −1.06, 0.66 | 0.65 |
| Quartile 4 | −0.90 | −1.74, −0.06 | 0.04 | −1.02 | −2.10, 0.07 | 0.07 |
| P for trend | 0.13 | 0.07 | ||||
| Participants Who Moved During Follow-up | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.89 | −1.75, −0.04 | 0.04 | −1.06 | −1.97, −0.15 | 0.02 |
| Quartile 3 | −1.37 | −2.47, −0.26 | 0.02 | −1.51 | −2.80, −0.23 | 0.02 |
| Quartile 4 | −1.16 | −2.47, 0.14 | 0.08 | −1.34 | −2.66, −0.03 | 0.05 |
| P for trend | 0.08 | 0.04 | ||||
Abbreviations: CI, confidence interval; NWI, Neighborhood Walkability Index.
Similar to the results for weight at follow-up, higher quartiles of cumulative NWI were associated with smaller WCs at follow-up, after adjustment for WC and height at enrollment. The associations between NWI-year quartiles and WC were larger among Black people than White people and were only statistically significant among Black people. Comparing movers and nonmovers, the association observed between NWI-year quartile 4 and WC was numerically larger for movers. However, the CIs around the association observed for movers covered the point estimate of the association observed among nonmovers, indicating that the differences in the point estimates were not statistically significant.
NWI-years within 5 km of the home was not associated with weight at follow-up but was inversely associated with WC, with the strongest associations observed among movers and among Black participants (Table 5). However, CIs for estimates among Black participants covered the point estimates observed among White participants, and the CIs for estimates among movers covered the point estimates among nonmovers, indicating that stratum-specific point estimates were not statistically significantly different from each other.
Table 5.
Differences in Weight and Waist Circumference at Follow-Up by Quartile of Cumulative Neighborhood Walkability Index Score for 5 km Around Residence, Among Participants in the Reasons for Geographic and Racial Differences in Stroke Study, United States, 2003–2016
| Neighborhood Walkability | Weight at Follow-up, kg | Waist Circumference at Follow-up, cm | ||||
|---|---|---|---|---|---|---|
| Difference in Weight | 95% CI | P Value | Difference in Weight | 95% CI | P Value | |
| Entire Population | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.21 | −0.76, 0.34 | 0.46 | −0.34 | −0.93, 0.26 | 0.27 |
| Quartile 3 | −0.34 | −1.07, 0.39 | 0.36 | −0.87 | −1.70, −0.04 | 0.04 |
| Quartile 4 | −0.68 | −1.54, 0.19 | 0.13 | −1.34 | −2.50, −0.18 | 0.02 |
| P for trend | 0.13 | 0.02 | ||||
| Participants 65 Years or Older at Enrollment | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.04 | −0.83, 0.76 | 0.93 | −0.06 | −0.92, 0.80 | 0.89 |
| Quartile 3 | −0.20 | −1.22, 0.83 | 0.71 | −1.04 | −2.27, 0.19 | 0.1 |
| Quartile 4 | −0.54 | −1.71, 0.63 | 0.37 | −1.39 | −2.97, 0.18 | 0.08 |
| P for trend | 0.31 | 0.04 | ||||
| Participants 64 Years or Younger at Enrollment | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.31 | −1.09, 0.48 | 0.44 | −0.63 | −1.43, 0.16 | 0.12 |
| Quartile 3 | −0.40 | −1.47, 0.68 | 0.47 | −0.64 | −1.73, 0.45 | 0.25 |
| Quartile 4 | −0.66 | −1.88, 0.57 | 0.29 | −1.02 | −2.32, 0.29 | 0.13 |
| P for trend | 0.32 | 0.17 | ||||
| Among Black People | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.29 | −1.43, 0.85 | 0.61 | −0.91 | −2.14, 0.33 | 0.15 |
| Quartile 3 | −0.73 | −2.16, 0.70 | 0.32 | −1.66 | −3.21, −0.11 | 0.04 |
| Quartile 4 | −0.80 | −2.35, 0.75 | 0.31 | −2.17 | −4.00, −0.34 | 0.02 |
| P for trend | 0.28 | 0.02 | ||||
| Among White People | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.19 | −0.78, 0.40 | 0.53 | −0.18 | −0.87, 0.50 | 0.60 |
| Quartile 3 | 0.00 | −0.86, 0.86 | 1.00 | −0.45 | −1.48, 0.59 | 0.40 |
| Quartile 4 | −0.74 | −1.71, 0.22 | 0.13 | −0.73 | −1.98, 0.53 | 0.26 |
| P for trend | 0.24 | 0.25 | ||||
| Participants Who Did Not Move During Follow-up | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.48 | −1.19, 0.24 | 0.19 | −0.13 | −0.94, 0.69 | 0.76 |
| Quartile 3 | −0.81 | −1.78, 0.17 | 0.10 | −0.72 | −1.84, 0.40 | 0.21 |
| Quartile 4 | −1.27 | −2.41, −0.14 | 0.03 | −1.28 | −2.81, 0.25 | 0.10 |
| P for trend | 0.03 | 0.06 | ||||
| Participants Who Moved During Follow-up | ||||||
| NWI-years | ||||||
| Quartile 1 | 0 | Referent | 0 | Referent | ||
| Quartile 2 | −0.05 | −0.92, 0.83 | 0.92 | −1.00 | −1.95, −0.04 | 0.04 |
| Quartile 3 | 0.04 | −1.09, 1.18 | 0.94 | −1.23 | −2.45, −0.02 | 0.05 |
| Quartile 4 | −0.18 | −1.68, 1.31 | 0.81 | −1.61 | −3.03, −0.20 | 0.03 |
| P for trend | 0.85 | 0.03 | ||||
Abbreviations: CI, confidence interval; NWI, Neighborhood Walkability Index.
DISCUSSION
These analyses provide additional longitudinal evidence that residential neighborhood features that support pedestrian activity are associated with lower adiposity. This research builds on a limited body of prospective research on neighborhood walkability and is among the first to assess the link between the cumulative experience of neighborhood walkability and weight and WC at the end of follow-up. Much of the prior work assessed neighborhood features at 1 or 2 time points during cohort follow-up (14–22). A recent study that used annual measures of neighborhood walkability from 2009 to 2014 found small reductions in BMI associated with increases in neighborhood walkability (13). A study from MESA included repeated measures of built environments through time and observed effects of changes in built environments on utilitarian walking; another MESA study found similar effects on BMI (3, 4). Along with MESA, the study reported here is one of few longitudinal studies with repeated measures of built environments and a decade of follow-up.
The overall average experience of the REGARDS cohort between enrollment and follow-up was of weight loss and an increase in WC. Such changes in anthropometric measures are typical of individuals in this age group and reflect loss of muscle and bone mass (sarcopenia) and gain in central adiposity (36). The average differences in weight at follow-up across quartiles of NWI-years are modest: less than 1 kg. These modest differences may reflect 2 offsetting effects: by promoting walking, neighborhood walkability may lead to improved muscle quality (37–39) and maintenance of muscle mass and weight during aging, while also promoting energy expenditure and smaller gains in adipose tissue, thereby reducing weight (2, 4, 23). Thus, at older ages, neighborhood walkability may have only a modest association with total weight but a quite healthful effect on overall body composition. This interpretation is supported by the association between higher neighborhood walkability and lower WC, a more direct measure of central adiposity than body weight.
Critiques of prior cross-sectional studies of neighborhood walkability have focused on the possibility of self-selection for neighborhoods with greater walkability among those who value walking for exercise, leisure, and/or running errands (40, 41). For participants who move residences, few past long-term studies were able to measure the walkability of the destination neighborhood at the time of the move (13, 21). The baseline and follow-up survey data in REGARDS do not allow for analyses of participants’ motivations for moving residences. But the neighborhood-level data show that those who moved during follow-up originally lived in neighborhoods that had significantly higher NWI scores than those who did not change residences. Furthermore, moves brought participants to neighborhoods that had higher median household incomes, higher home prices, lower poverty rates, and lower NWI scores. These analyses add to our understanding of the changes in neighborhood environments experienced by participants upon moving residence and suggest that economic factors are associated with choices of destination neighborhoods. Within the REGARDS sample, the associations between the experience of high cumulative NWI and BMI and WC at follow-up, approximately 10 years after enrollment, do not appear to vary by residential move status. The 95% CIs around the estimates of the differences in BMI and WC at follow-up, comparing those in the fourth quartile of NWI-years with those in the first quartile, among movers, covered the point estimate for the same associations among nonmovers.
This study has a number of strengths, including the large sample recruited over a wide geographic range (Figure 1), that the cohort includes a large proportion of Black participants, and the availability of residential address data that allowed for the calculation of cumulative neighborhood-exposure measures. Taken together with other recent longitudinal studies, our results increase confidence that neighborhood walkability promotes physical activity and maintenance of a healthy weight (3, 4, 13). Inclusion of analysis of WC is also a strength; whereas BMI and body weight are widely used outcomes in research on the built environment and health, studies relating walkability to WC are less common, and those with a longitudinal design even rarer.
That said, the neighborhood measures we used have some limitations. First, due to limitations in the National Establishment Time Series (NETS) license, data were only available from 1990 to 2014, and census population counts were only available at 3 time points. Thus, extrapolation of data to years not covered by the geospatial data was required to create annual estimates, an approach that has been used previously (3, 33, 34). The NWI used here also does not include a measure of street connectivity or transit access, characteristics of urban form believed to influence pedestrian activity. However, population density and destination density are expected to be highly correlated with street connectivity and transit availability. Last, the NWI does not measure experiential aspects of the pedestrian environment, such as concerns about safety from crime and traffic, aesthetic qualities, and social interactions (7, 42, 43). However, such experiential qualities are expected to be correlated with neighborhood walkability, socioeconomic status, and green space, which are controlled for in our analyses (43, 44).
In conclusion, within the REGARDS cohort, the overall average experience of the participants between enrollment and follow-up 10 years later was of weight loss and an increase in WC. Within this context, the experience of higher cumulative neighborhood walkability during follow-up was beneficially associated with adiposity at follow-up and may positively influence overall health in older adults.
ACKNOWLEDGMENTS
Author affiliations: Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, New York, United States (Andrew G. Rundle, James W. Quinn); Columbia Population Research Center, Columbia University, New York, New York, United States (Kathryn M. Neckerman); Department of Biostatistics, University of Alabama School of Public Health, University of Alabama, Birmingham, Alabama, United States (Suzanne E. Judd); School of Kinesiology, University of Michigan, Ann Arbor, Michigan, United States (Natalie Colabianchi); Urban Health Collaborative, Dornsife School of Public Health, Drexel University, Philadelphia, Pennsylvania, United States (Kari A. Moore); and Epidemiology and Biostatistics, Dornsife School of Public Health, Drexel University, Philadelphia, Pennsylvania, United States (Jana A. Hirsch, Gina S. Lovasi).
The REGARDS research project is supported by cooperative agreement U01 NS041588 cofunded by the National Institute of Neurological Disorders and Stroke (NINDS) and the National Institute on Aging (NIA), National Institutes of Health, Department of Health and Human Services. The integration and analyses of geographic data were supported by NIA (grants 1R01AG049970 and 3R01AG049970-04S1) and NINDS (grant R01 NS092706).
The data from this work are available upon application to the REGARDS study. Information on procedures for proposing analyses, papers, and ancillary studies can be found at https://www.uab.edu/soph/regardsstudy/researchers.
The content is solely the responsibility of the authors and does not necessarily represent the official views of the NINDS or the NIA.
Conflict of interest: none declared.
REFERENCES
- 1. Rundle AG, Chen Y, Quinn JW, et al. Development of a neighborhood walkability index for studying neighborhood physical activity contexts in communities across the U.S. over the past three decades. J Urban Health. 2019;96(4):583–590. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Rundle AG, Sheehan DM, Quinn JW, et al. Using GPS data to study neighborhood walkability and physical activity. Am J Prev Med. 2016;50(3):e65–e72. [DOI] [PubMed] [Google Scholar]
- 3. Hirsch JA, Moore KA, Barrientos-Gutierrez T, et al. Built environment change and change in BMI and waist circumference: Multi-Ethnic Study of Atherosclerosis. Obesity (Silver Spring). 2014;22(11):2450–2457. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Hirsch JA, Moore KA, Clarke PJ, et al. Changes in the built environment and changes in the amount of walking over time: longitudinal results from the Multi-Ethnic Study of Atherosclerosis. Am J Epidemiol. 2014;180(8):799–809. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Tabaei BP, Rundle AG, Wu WY, et al. Associations of residential socioeconomic, food, and built environments with glycemic control in persons with diabetes in New York City from 2007–2013. Am J Epidemiol. 2018;187(4):736–745. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Creatore MI, Glazier RH, Moineddin R, et al. Association of neighborhood walkability with change in overweight, obesity, and diabetes. JAMA. 2016;315(20):2211–2220. [DOI] [PubMed] [Google Scholar]
- 7. Freeman L, Neckerman K, Schwartz-Soicher O, et al. Neighborhood walkability and active travel (walking and cycling) in New York City. J Urban Health. 2013;90(4):575–585. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Sallis JF, Cerin E, Conway TL, et al. Physical activity in relation to urban environments in 14 cities worldwide: a cross-sectional study. Lancet. 2016;387(10034):2207–2217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Frank LD, Andresen MA, Schmid TL. Obesity relationships with community design, physical activity, and time spent in cars. Am J Prev Med. 2004;27(2):87–96. [DOI] [PubMed] [Google Scholar]
- 10. Cerin E, Nathan A, Cauwenberg J, et al. The neighbourhood physical environment and active travel in older adults: a systematic review and meta-analysis. Int J Behav Nutr Phys Act. 2017;14(1):15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Dixon BN, Ugwoaba UA, Brockmann AN, et al. Associations between the built environment and dietary intake, physical activity, and obesity: a scoping review of reviews. Obes Rev. 2021;22(4):e13171. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Chandrabose M, Rachele JN, Gunn L, et al. Built environment and cardio-metabolic health: systematic review and meta-analysis of longitudinal studies. Obes Rev. 2019;20(1):41–54. [DOI] [PubMed] [Google Scholar]
- 13. Tarlov E, Silva A, Wing C, et al. Neighborhood walkability and BMI change: a national study of veterans in large urban areas. Obesity (Silver Spring). 2020;28(1):46–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Coogan PF, White LF, Adler TJ, et al. Prospective study of urban form and physical activity in the Black Women's Health Study. Am J Epidemiol. 2009;170(9):1105–1117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Coogan PF, White LF, Evans SR, et al. Longitudinal assessment of urban form and weight gain in African-American women. Am J Prev Med. 2011;40(4):411–418. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Lee IM, Ewing R, Sesso HD. The built environment and physical activity levels: the Harvard Alumni Health Study. Am J Prev Med. 2009;37(4):293–298. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Berry TR, Spence JC, Blanchard C, et al. Changes in BMI over 6 years: the role of demographic and neighborhood characteristics. Int J Obes (Lond). 2010;34(8):1275–1283. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Wasfi RA, Dasgupta K, Orpana H, et al. Neighborhood walkability and body mass index trajectories: longitudinal study of Canadians. Am J Public Health. 2016;106(5):934–940. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Chandrabose M, Cerin E, Mavoa S, et al. Neighborhood walkability and 12-year changes in cardio-metabolic risk: the mediating role of physical activity. Int J Behav Nutr Phys Act. 2019;16(1):86. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Kikuchi H, Nakaya T, Hanibuchi T, et al. Objectively measured neighborhood walkability and change in physical activity in older Japanese adults: a five-year cohort study. Int J Environ Res Public Health. 2018;15(9):1814. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Hirsch JA, Diez Roux AV, Moore KA, et al. Change in walking and body mass index following residential relocation: the Multi-Ethnic Study of Atherosclerosis. Am J Public Health. 2014;104(3):e49–e56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Lang IM, Antonakos CL, Judd SE, et al. A longitudinal examination of objective neighborhood walkability, body mass index, and waist circumference: the REasons for Geographic and Eacial Differences in Stroke Study. Int J Behav Nutr Phys Act. 2022;19(1):17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Twardzik E, Judd S, Bennett A, et al. Walk score and objectively measured physical activity within a national cohort. J Epidemiol Community Health. 2019;73(6):549–556. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Howard VJ, Cushman M, Pulley L, et al. The Reasons for Geographic and Racial Differences in Stroke Study: objectives and design. Neuroepidemiology. 2005;25(3):135–143. [DOI] [PubMed] [Google Scholar]
- 25. Howard G, Howard VJ. Twenty years of progress toward understanding the Stroke Belt. Stroke. 2020;51(3):742–750. [DOI] [PubMed] [Google Scholar]
- 26. Long DL, Howard G, Long DM, et al. An investigation of selection bias in estimating racial disparity in stroke risk factors. Am J Epidemiol. 2019;188(3):587–597. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Kramer H, Gutierrez OM, Judd SE, et al. Waist circumference, body mass index, and ESRD in the REGARDS (Reasons for Geographic and Racial Differences in Stroke) Study. Am J Kidney Dis. 2016;67(1):62–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Wiggermann N, Bradtmiller B, Bunnell S, et al. Anthropometric dimensions of individuals with high body mass index. Hum Factors. 2019;61(8):1277–1296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Hirsch JA, Moore KA, Cahill J, et al. Business data categorization and refinement for application in longitudinal neighborhood health research: a methodology. J Urban Health. 2020;98(2):271–284. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. US Census Bureau . TIGER/Line—Areal Hydrography National Geodatabase, 2016. Data File. https://www2.census.gov/geo/tiger/TGRGDB16/tlgdb_2016_a_us_areawater.gdb.zip. Accessed May 9, 2017.
- 31. Logan J, Stults B, Xu Z. Validating population estimates for harmonized census tract data, 2000–2010. Ann Am Assoc Geogr. 2016;106(5):1013–1029. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Logan J, Xu Z, Stults B. Interpolating US decennial census tract data from as early as 1970 to 2010: a longitudinal tract database. Prof Geogr. 2014;66(3):412–420. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Flowerdew R, Green M. Data integration: statistical methods for transferring data between zonal systems. In: Masser I, Blakemore M, eds. Handling Geographical Information: Methodology and Potential Applications. New York, NY: Longman; 1991:38–54. [Google Scholar]
- 34. Crowder K, South SJ. Spatial dynamics of White flight: the effects of local and extralocal racial conditions on neighborhood out-migration. Am Sociol Rev. 2008;73(5):792–812. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. South SJ, Huang Y, Spring A, et al. Neighborhood attainment over the adult life course. Am Sociol Rev. 2016;81(6):1276–1304. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Wilkinson DJ, Piasecki M, Atherton PJ. The age-related loss of skeletal muscle mass and function: measurement and physiology of muscle fibre atrophy and muscle fibre loss in humans. Ageing Res Rev. 2018;47:123–132. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Yoshiko A, Tomita A, Ando R, et al. Effects of 10-week walking and walking with home-based resistance training on muscle quality, muscle size, and physical functional tests in healthy older individuals. Eur Rev Aging Phys Act. 2018;15(1):13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Murata Y, Nakamura E, Tsukamoto M, et al. Longitudinal study of risk factors for decreased cross-sectional area of psoas major and paraspinal muscle in 1849 individuals. Sci Rep. 2021;11(1):16986. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Hiraoka A, Michitaka K, Kiguchi D, et al. Efficacy of branched-chain amino acid supplementation and walking exercise for preventing sarcopenia in patients with liver cirrhosis. Eur J Gastroenterol Hepatol. 2017;29(12):1416–1423. [DOI] [PubMed] [Google Scholar]
- 40. Frank LD, Saelens BE, Powell KE, et al. Stepping towards causation: do built environments or neighborhood and travel preferences explain physical activity, driving, and obesity? Soc Sci Med. 2007;65(9):1898–1914. [DOI] [PubMed] [Google Scholar]
- 41. Norman GJ, Carlson JA, O'Mara S, et al. Neighborhood preference, walkability and walking in overweight/obese men. Am J Health Behav. 2013;37(2):277–282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Shariff-Marco S, Von Behren J, Reynolds P, et al. Impact of social and built environment factors on body size among breast cancer survivors: the Pathways Study. Cancer Epidemiol Biomarkers Prev. 2017;26(4):505–515. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Neckerman KM, Lovasi GS, Davies S, et al. Disparities in urban neighborhood conditions: evidence from GIS measures and field observation in New York City. J Public Health Policy. 2009;30(Suppl 1):S264–S285. [DOI] [PubMed] [Google Scholar]
- 44. Mooney SJ, Bader MD, Lovasi GS, et al. Validity of an ecometric neighborhood physical disorder measure constructed by virtual street audit. Am J Epidemiol. 2014;180(6):626–635. [DOI] [PMC free article] [PubMed] [Google Scholar]