Background Walking is an important component of physical activity. It is commonly opined that the ‘walkability’ of someone’s environment encourages walking behaviour, but the evidence base for this is limited and there is no consensus on how to measure walkability. We have undertaken a small, proof-of-concept, cross-sectional observational study, the objectives of which were two-fold. Firstly, to examine the association between walkability of the area around participants’ homes and different aspects of physical activity as well as anthropometric measurements. Secondly we aimed to critique, test and extend existing methods of measuring walkability with GIS methods, in preparation for a larger multi-centre study.

Methods Using GIS software we built a three-domain measure of walkability based on topography (the area accessible by walking points given road and footpath coverage), local amenities available in a short walk, and availability of public transit points. These measures were calculated for the homes of 1779 cohort study participants aged 45–83 years. Response variables were self-reported sports activities [h/week], time spent walking or cycling, as well as walking and cycling as categorised as never, rarely, sometimes, often or mostly. We also captured BMI and waist circumference measured at follow-up. We estimated the association between walkability and our response variables using linear regression models that were adjusted for age and education.

Results We found few significant associations between any of our response variables and any of the dimensions of walkability. A very weak positive association between time spent walking and cycling with proximity of transit points was observed, however the effect sizes involved amounted to fewer than a few minutes exercise per week between high and low transit point density (ß 0.024, 95% Confidence Interval 0.006–0.041 for transit points within 500 m distance).

Discussion We did not see a convincing association between walkability and our response variables in this study. This could be because there was insufficient heterogeneity in either our neighbourhoods or participants, an issue of statistical power, or because there really is no association. However we did demonstrate that we can get a sophisticated, sensitive and reproducible measure of walkability of environments for use in population based research. We will apply methods used here (slightly refined) in a large study involving six different cities across Germany. Our presentation will focus on the methodological aspects of this study and how we tackled some of the measurement issues and discuss the refinements we propose to use in the next phase.