*4.1. Policy Implications*

As discussed, these research findings contribute to a growing scientific evidence base, which suggests that greater access to away from home food establishments contributes to unhealthy dietary behaviour, excess body weight and obesity [20,49,50] Although some questions remain, this growing literature has direct links to public health policy through informing 'healthy' neighbourhood design [51], which is increasingly understood by planners as a low-agency, population-level public health intervention [52]. For example, planners in English local governmen<sup>t</sup> are actively encouraged to implement planning laws that limit growth in the fast food sector [53], focussing on areas of perceived need, where levels of obesity are currently high, or where existing access to fast food outlets is sufficient. Internationally, there are examples of similar practice [54] and anecdotal evidence to sugges<sup>t</sup> that implementation of such policies is increasingly commonplace [55]. The effectiveness of many of these policies is ye<sup>t</sup> to be determined.

### *4.2. Methodological Considerations and Limitations*

For this study, food outlet exposure was estimated within residential neighbourhoods, defined as within 1 mile of each participant's home address. This home-based characterisation may overestimate some forms of outlet exposure, particularly if there are physical barriers in the environment not accounted for using the buffer method, and underestimate other forms of exposure, particularly for food outlets accessed from the workplace or while commuting between locations [20]. Additionally, the OS POI data on food outlet locations is not a routinely validated data set. Approximately 60% of the data is reported as 'ground truthed', however given the size of the data, it is not feasible to check each location. We used 2014 data with the assumption that secular food outlet change is relatively slow, and therefore unlikely to result in a significant difference in quintile of exposure between the year the food outlet data was collected and the year the survey data was collected, however this assumption is not based on a validation study for the UK. Additionally, the use of self-report data for estimates of household food spending is an important consideration. Although validation of self-report expenditure has been done in previous work [37], there is no UKLHS validation study for household food expenditure that we are aware of. Household food spend was benchmarked against Living Cost and Food data, however, while these adjusted data are likely appropriate for the type of analysis presented here, they may not provide perfectly accurate population level estimates of household food spend. Also, while we have adjusted for known confounders that were available to us and used routinely in previous research, the role of other unmeasured confounders cannot be ruled out, including car ownership or access to or use of public transportation.

Major strengths of this study include the use of a nationally-representative geographically diverse sample of UK households, the use of objectively measured food outlet data and coherence between the characterisation of the food environment and outcome. Using data from this national survey increases the generalizability of our findings to the UK population. Previous studies have typically used data from geographically-circumscribed samples from study cohorts. Although such samples can be representative of their underlying regional populations, they may be more limited in terms of generalizability to a national context. For the first time in the published literature, we exploited novel household food spend data within this national social and economic panel survey, including information on the amount spent purchasing foods specifically for consumption away from home.
