'Reliable' measures of the food environment have been described as the 'foundation' of research that will help to inform obesity related policy . This research has highlighted that despite some difficulties, secondary data sources available in the UK provide an accurate picture of the urban/rural, high/low socio-economic status food environment, and that desk based classification is an acceptable alternative to fieldwork.
Comparison of the foodscape according to geographic area
There is little information about the UK foodscape across urban/rural and SES divides. Despite the fact that all areas were approximately the same geographic size, the area that had the greatest number of food outlets was the urban low SES area (n = 210) and the area with the least number of outlets was the rural high SES (n = 19). The population sizes of the areas varied with a population of 1401 in the urban high SES area and 5024 in the urban mixed SES area. The observations and descriptive results during fieldwork highlighted the differences in the areas. There was clustering of outlets in the urban low SES area, and a lack of residential housing while the rural areas had outlets dispersed throughout. Likewise, Macdonald et al. , found that the least deprived areas of Glasgow were the least well served by food outlets and the most deprived were the best. However, despite the fact that these findings also hold true for high and low SES areas of North East England, Macdonald et al.  concluded that their least deprived areas did not necessarily have better access to food outlets and vice versa. In this study and in the work of Macdonald et al.  urban low SES areas having the greatest concentrations of food outlets does not necessarily mean that individuals in urban low SES areas had the best overall access to a healthy diet, or to wide variety of food types; future research will seek to examine these relationships further. In an earlier study also in Newcastle, White et al.  concluded 'there are inequalities in retail provision that are geographically patterned, but these are not necessarily all "bad"'. This statement holds true for other areas of North East England explored in the present study.
Cummins et al.  reported that the number of McDonald's fast food outlets was highest in the most deprived areas of England and Scotland compared to the least deprived. In the present study only three out of the 438 outlets investigated were classified as fast food outlets (not necessarily McDonalds but similar; one in the urban mixed SES area and two in the urban low SES area). There were no fast food outlets found in either of the high SES areas. In Glasgow, Cummins and Macintyre  reported that the most deprived areas had the most frozen food and discount food outlets. Similar results were found in the present study where the highest number of these outlet types was reported in the urban low SES area although none were found in the rural low SES area. In a recent review Beaulac et al. [5, 14] concluded that 'there is little evidence that socioeconomically deprived areas of the UK are systematically disadvantaged by food deserts.' This study does not provide evidence that low SES areas are more at risk of food deserts than other areas as they have more food outlets compared to the higher SES areas.
Rural areas may be more disadvantaged compared to urban areas when it comes to food availability, as the rural areas have fewer food outlets that are dispersed. Availability or 'potential access' to food has both social and geographic elements. Sharkey  suggested that changes in the food environment such as price, variation and quality of food, and the size and numbers of food outlets are more likely to have negative effects on rural areas in parts of the U.S. In the present study, food outlets in rural areas (especially the mixed SES rural area) were more spread out, which may have an effect on where residents of rural areas purchase their food. As found elsewhere, this study also observed food outlets in urban areas to be more clustered. Smith et al.  found that people who live in the most deprived areas of Scotland had the shortest travelling distance to the nearest food store compared to the least deprived areas. The present study reported that the low SES areas had more outlets than the high SES counterparts. However the type of outlet differed between areas.
Number of outlet types
The classification of the types of outlet differed between urban and rural areas and between SES of areas. In every study area the highest number of outlets could be defined as places where food is consumed away from the home such as takeaways, traditional restaurants or pubs/bars, rather than supermarkets or stores where foods can be purchased and prepared at home. This supports longitudinal work in Northumberland which also observed a high number of outlets of 'foods for consumption away from the home' . The highest frequencies of outlets in all areas are probably more likely to sell higher calorie beverages and higher calorie and fat foods compared to what one might consume at home [3, 25]. It is acknowledged that the relationship between an individual's food intake and the wider food environment is 'complex' [6, 25] and merits further investigation.
Validation of secondary sources: local authority list compared with fieldwork
The results of this study, covering a range of geographic areas, supports previous work in an inner city area by Lake et al. , which highlighted that there were a number of outlets in each study area that were neither present in the field or on the Local Authority data list. Those not on the Local Authority food outlet list could only be identified and classified in the fieldwork and vice versa. These findings suggest that in order to obtain the most accurate picture of the food environment, fieldwork must be conducted so that any changes or closures to outlets can be noted. However, it is likely that the food environment changes rapidly due to new food outlets opening or closing and so fieldwork would need to be carried out regularly if the food environment was to be studied over a longer period of time. In order to calculate how accurate the Local Authority's food outlet data actually was, sensitivity and PPV analyses were used. Although none of the areas had a PPV of 100%, all were fairly high (PPV > 78%), which suggested that the Local Authority food outlet data would be an acceptable alternative to fieldwork. The highest PPV was in the urban low SES area (PPV = 87%). These PPVs are slightly lower than the PPV of 91.5% obtained in inner city Newcastle by Lake et al.  using the same methods. In the current study both urban study areas had a PPV value of 81% and 87%. The sensitivities between the Local Authority food outlet data and the fieldwork were considered to be 'moderate' to 'excellent' in all study areas, which also suggested the Local Authority data is likely to be a satisfactory representation of the food environment. It is interesting to note the rural low SES area had a high PPV (86%), yet a 'moderate' sensitivity, which reflects the large number of missing outlets on the Local Authority list that were present in reality. It also serves to emphasise the importance of measuring both PPV and sensitivity. Given the extra time and resources required to conduct fieldwork (particularly in large rural areas), this paper argues that the use of Local Authority collected secondary data is a viable option, even in rural locales. This is in agreement with previous work , although secondary data sources (commercial as well as Local Authority) should always be used with caution and with appreciation for their potential limitations [8, 10].
Accuracy of desk based classification compared to fieldwork (agreement assessment)
In most large studies previously conducted in the UK, classification of the food environment was completed using secondary data sources alone [25, 26]. However, very few studies have tested how reliable these data sources actually are using ground-truthing . Additionally, in the UK, studies have not explored rural environments.
The classification tool used in this study was developed in the North East of England . It is perhaps unsurprising that classification of food outlet types is more straightforward in the field than it is from the desk with only limited information. Although most of the outlets could be classified from their name (especially well known chains such as Tesco or WH Smith), some were more difficult (for example 'Lou Lous' food outlet, which was a 'traditional café/coffee shop') and required various additional data sources to make a classification. Some outlets were impossible to classify at the desk due to a lack of information. However, despite difficulties, in most areas there was agreement. In Glasgow, Cummins and Macintyre [2, 10] reported that the level of agreement between secondary data and fieldwork observations was 'high but imperfect'. Similarly, in a US study , the level of agreement between a secondary and primary data collection method were found to be fairly high and significant. The researchers concluded that either method could be used with 'reasonable confidence' . Interestingly, using the classification tool in the combined rural areas, significantly more outlets were in agreement than disagreement. This is surprising as many outlets in the rural areas could not be classified using any of the additional internet sources. Reasons for this are unclear, however it is possible that businesses in rural areas are small and may not feel the need to register with commercial directories or have a website of their own (as they only serve a limited number of local residents). Telephoning food outlets may assist in making accurate food outlet classifications; this approach will be used in future studies.
The agreement in each SES area across the urban and rural divides was similar whether it was the high or low SES. A different result from Cummins and Macintyre  who reported that most errors in agreement were made in the deprived neighbourhoods.
Strengths and limitations
This study investigated the food environments of North East England, including urban and rural, as well as high and low SES area comparisons. Definitions of 'urban' and 'rural' were used according to Department for Communities and Local Government guidelines . Whilst these definitions may not be transferable outside the UK, they are certainly an accurate measure of urbanity and rurality within the UK. An assessment of secondary food environment data across these divides in the UK has been absent in the literature to date, yet has been called for [10, 11] and has only recently been published in the US .
The classification tool  was an appropriate way to measure the food environment and to categorise food outlets that were present. It contained enough detail to give each food outlet a very specific category based on the type and manner in which food was sold. A more detailed 78 point version of this tool exists. Use of this more detailed tool may have influenced the reliability of the classifications made both in the field and at the desk. In this study the inter-rater reliability of the tool was not tested, however all the data was collected by one trained researcher. In the urban mixed SES area of Durham city centre however, a major part of the food environment was not classified as it had previously been excluded on the grounds that it was a market, with stalls ordinarily registered to the owners' home addresses. However, this is a permanent indoor market that housed 12 food outlets of various kinds and observations suggested it to be a major source of food retail. Since the fieldwork was conducted, the classification tool has been further developed to include categories for such markets, ice cream and burger vans, in order to help produce a tool that can record the entirety of the food environment.
Whilst the LSOAs selected as study areas in this research were purposively sampled, in part for their geographic convenience, this was necessary as to ensure that the areas selected contained some food outlets within them to field validate. Many potential LSOAs in the surrounding area were devoid of food outlets altogether and were, for this reason, not a candidate for study. It may have been preferable to conduct a random sample of LSOAs for inclusion in the study, however this was not feasible on this occasion. This could be seen as a limitation. Furthermore, whilst it may have been preferable to 'ground truth' more than six LSOAs in total, this was also not possible. With this said, the LSOAs selected for study were as relatively highly populated with food outlets as was possible, allowing for detailed and substantive comparisons between study sites regardless.
Although this study has given an accurate and detailed picture of the food environments within these defined areas in terms of the category of outlets present, we did not examine the types of food and beverages sold within each outlet, nor individual purchasing patterns. Exploration of the price and nutritional profiles of products sold within each outlet in relation to the foodscape is currently being explored by this group and will give an even more detailed representation of the food environment.
Various difficulties arose when obtaining the field data. In the rural mixed SES and rural high SES areas the amount of land covered within the boundaries was too large to reasonably cover on foot, therefore a car had to be used. This was both labour and time intensive, and only serves to strengthen the call for an alternative and accurate source of food outlet data to be identified. In reality, it is difficult for researchers to ground-truth large geographic areas, thus secondary data sets are important resources.