Are secondary data sources on the neighbourhood food environment accurate? Case-study in Glasgow, UK
Introduction
There has been increasing interest recently in the role of the neighbourhood food environment as an independent predictor of food choice, and thus of diet quality, in high income nations (McKinnon et al., 2009, Cummins and Macintyre, 2006). Studies have generally tended to measure exposure to food environments through the use of data on the location of retail outlets, which are used to calculate measures of accessibility to food outlets in terms of time, distance and density. For pragmatic reasons, the majority of these studies have relied on pre-existing data obtained from public, commercial and online sources rather than using ‘on the ground’ field observation methods (which are time and cost intensive) to characterise environmental exposures (Paquet et al., 2008, Wang et al., 2006).
However, studies of neighbourhood food environments and diet depend on the validity of the data sources used to enumerate the distribution and characteristics of food outlets. Investigations of the validity of commonly used food retail business directories in the USA have demonstrated problems with completeness and the retention of outdated information (Lanvin, 1998) and errors in reporting of store location and attributes (Schwartz, 1987). Assessments of the validity of such data in health research are very rare, with only one published study investigating agreement between published or online commercial sources and field observations. In that study, in Montreal, the authors reported that even though the level of agreement was not perfect, there was no evidence of significant levels of systematic error (Paquet et al., 2008). Thus little is known about the validity of secondary data sources in studies of environmental determinants of diet. The aim of the study reported here is to assess the validity of secondary data on the location of food outlets by quantifying agreement between a local government listing of food premises and direct field observation, and whether agreement varies by neighbourhood deprivation, and over time.
Section snippets
Methods
Data were collected as part of two large surveys of food price and availability in Glasgow during July, August and September of 1997 and 2007. Full details of the methods and substantive findings on neighbourhood differences in food price and availability based on the 1997 survey have been published elsewhere (Cummins and Macintyre, 2002). The methods employed in the 1997 survey were replicated in 2007 in order to ensure that consistent methods were used. Briefly, our sample frame at both time
Results
Table 1 shows the percentage agreement between the list and field observation for all the sampled stores, and by neighbourhood deprivation, in 1997 and 2007. In 1997, 86.8% of identified food stores were confirmed as currently trading, and in 2007, 88.0%. In both years, stores were more common in deprived areas. Agreement ranged from 86.6% to 87.5% in 1997, and 90.3% to 87.3% in 2007. There was a negative stepwise relationship between neighbourhood deprivation and percentage agreement – as
Discussion
Our data suggest that the overall level of agreement between publicly available secondary data on food outlets and observations ‘on the ground’ was high, but imperfect. Errors were slightly more common in more deprived neighbourhoods, and this relationship between error rate and deprivation may have slightly increased between 1997 and 2007 despite overall improvements in validity. Compared to Paquet et al. (2008) the lower positive predictive values reported here may be due to omission of false
Conclusions
This study represents the first attempt to assess the validity of secondary data sources of the neighbourhood food environment outside of Canada, and the first to investigate changes in validity over time. Our results suggest that the Public Register of Food Premises may be a useful for characterising the neighbourhood food environment in the UK, though it is not perfect and may underestimate exposure-effect associations. The data presented here is likely to be in the upper bounds of validity,
Conflicts of interest statement
The authors declare that there are no conflicts of interest.
Acknowledgments
This research was funded by The Leverhulme Trust and the UK Medical Research Council, WBS U.1300.00.006. Steven Cummins is supported by a National Institute of Health Research Fellowship. We are grateful to Kate Campbell for coordinating the data collection in shops in 2007, and the field workers for collecting the data.
References (10)
- et al.
Measures of the food environment: a compilation of the literature 1990–2007
Am. J. Prev. Med.
(2009) - et al.
Deprivation and Health in Scotland
(1991) - et al.
The price and availability of food in Glasgow: a systematic study of an urban foodscape
Urban Stud.
(2002) - et al.
Food environments and obesity—neighbourhood or nation?
Int. J. Epidemiol.
(2006) A clash of the titans. Comparing America's most comprehensive business directories
Database
(1998)
Cited by (59)
Measures of the food environment: A systematic review of the field, 2007–2015
2017, Health and PlaceFood availability/convenience and obesity
2014, Advances in NutritionBusiness List vs Ground Observation for Measuring aFood Environment: Saving Time or Waste of Time(orWorse)?
2013, Journal of the Academy of Nutrition and DieteticsValidity of secondary retail food outlet data: A systematic review
2013, American Journal of Preventive MedicineCitation Excerpt :Few studies13,18,24 elaborated on how secondary data were entered and edited or how duplicate or possible duplicate retail food outlets were eliminated or combined. Only one study27 did not use GIS software or other digital mapping systems to map the location of retail food outlets ascertained from at least one of the secondary data sources examined. Nevertheless, only three studies12,18,28 reported the number of retail food outlets ascertained via secondary data sources that were successfully geocoded (i.e., located the associated latitude and longitude).
Objective food environments and health outcomes
2013, American Journal of Preventive Medicine