Residential moves, neighbourhood walkability, and physical activity: a longitudinal pilot study in Ontario Canada

Patricia A. Collins, Justin Tait, Allan Fein, James R. Dunn

Numerous cross-sectional studies have consistently demonstrated an association between attributes of urban form or 'walkability' and individual- and population-level physical activity (PA) patterns. However, in the absence of longitudinal research, the self-selection problem undermines the claim that a walkable built form produces more physically active people. Through a longitudinal pilot study of 'imminent movers' in Ontario using a quasi-experimental approach, we sought to examine the feasibility of longitudinal methods that would produce stronger evidence for a causal relationship between the built environment and PA levels. Participants were recruited using publicly available real estate listings. Successful recruits were sent a PA diary to track their activity for a week, and were also scheduled for a 45-min phone interview that collected demographic details, neighbourhood perceptions and self-efficacy for walking, and verified the PA diary. Following their move, participants were given the same tasks and then sorted into groups based on changes in their neighbourhood walkability (measured with Walk Score) from baseline to follow-up. There were challenges in recruiting a sufficient number of participants and counter-factuals to examine the relationship between changes in walkability and PA. Our limited sample showed a substantial decrease in Walk Score over the entire sample, from an average of 45.8 to 30.6, with most participants moving to less walkable areas. From baseline to follow-up, the largest declines in reported self-efficacy for walking were to grocery stores, banks, and for entertainment. For the entire sample, utilitarian PA decreased, while recreational and job-related PA increased. This pilot study highlighted the methodological challenges involved in collecting quasi-experimental evidence on the effect of walkable environments on PA. Additionally, the low sample size and the tendency for most participants to move to less walkable areas meant there were insufficient counter-factuals for study of the effect of walkability on PA. Despite these challenges, we saw important changes in self-efficacy for walking that were commensurate with changes to the built environment. In sum, while longitudinal research on health and the built environment is urgently needed, recruiting an adequate sample size for a quasi-experimental study such as this is extremely challenging.