|
Citation
|
Braun L, Rodriguez D, Gordon-Larsen P, Song Y, Meyer K, Lewis C. J. Transp. Health 2015; 2(Suppl): S16-S17. |
|
Abstract
|
Background
While many studies have found measures of the built environment to be associated with active transportation, most have relied on cross-sectional research designs that are particularly vulnerable to residential self-selection. This study contributes longitudinal evidence based on changes in walking behavior following residential relocation.
Methods
This analysis used data from the Coronary Artery Risk Development in Young Adults (CARDIA) study, a population-based prospective epidemiologic study that began in 1985-1986 with 5,115 participants in Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota; and Oakland, California. The sample consisted of 1,063 respondents who moved residential locations between the year 15 (2000-2001) and year 20 (2005-2006) exams. The primary exposure variable was a walkability index derived from population density, land use, street connectivity, and local parks data, measured at the pre- and post-move residential locations. Physical activity from walking was operationalized as a categorical measure of participation and continuous measures of frequency and exercise units. Random effects and fixed effects models were used to estimate the associations between change in the walkability index and change in each measure of walking behavior, adjusting for sociodemographic and health covariates.
Results
In the random effects models, the walkability index was positively associated with the odds of engaging in any walking activity and with walking frequency and exercise units. However, no significant associations between walkability and walking behavior were observed in the fixed effects models. Hausman tests suggested that the random effects estimates were biased for walking participation and exercise units, favoring the use of fixed effects.
Conclusions
The built environment was associated with walking activity in random effects but not fixed effects models. Approaches that base estimates on variation between units (e.g., cross-sectional and random effects regression) may be subject to bias from unmeasured confounding, such as that arising from residential self-selection.
Language: en |