Background Evidence of a relationship between residential retail food environments and diet-related outcomes is inconsistent. One reason for this may be that food environments are typically defined in terms of the absolute number of particular store types in an area, whereas a measure of the relative number of healthy and unhealthy stores may be more appropriate.

Methods Using cross-sectional data from the VicLANES study conducted in Melbourne, Australia, multilevel logistic regression analysis was used to estimate the independent associations between absolute measures (numbers of healthy and unhealthy stores) and a relative measure (relative density of healthy stores) of the food environment, and self-reported variety of fruit and vegetable purchasing in local households. Purchasing behaviour was measured as the odds of purchasing above the median level of fruit and vegetables.

Results Compared to households in areas where healthy food stores made up no more than 10% of all healthy and unhealthy stores, households in areas with 10.1–15.0% healthy food stores and >15% healthy stores had increased odds of healthier purchasing (OR=1.48 (95% CI 1.12 to 1.96) and OR=1.45 (95% CI 1.09 to 1.91), respectively). There was less evidence of an association between absolute numbers of healthy or unhealthy stores and fruit and vegetable purchasing.

Conclusions We found strong evidence of healthier fruit and vegetable purchasing in households located in areas where the proportion of food stores that were healthy was greater. Policies aimed at improving the balance between healthy and unhealthy stores within areas may therefore be effective in promoting greater consumption of fruit and vegetables.