A stratified random selection of small, independent food stores drawn from a single industry category in a single commercial database did not yield a homogenous group of small food stores. Instead, the sample yielded a heterogeneous group of stores in terms of nutritious food options: some stores provided many nutritious food options and fresh fruits and vegetables, but most provided a limited variety of nutritious food items and produce. Store attributes (number of employees and sales volume) listed in the commercial database did not distinguish store type as well as the in-store survey and census data did. These findings reinforce those of previous studies that found significant discrepancies between store categorizations from secondary food retail databases and field observations (22
) and suggest that database imprecision may introduce error or bias or both into public health and epidemiological research.
Commercial databases may not identify food stores in more deprived neighborhoods as accurately as they do in less deprived neighborhoods (25
). This was not the case in our study. However, convenience stores (limited availability of nutritious foods) tended to be in more densely populated census tracts, and grocery stores (better availability of nutritious food) tended to be in tracts that had a higher percentage of whites. Convenience and specialty stores were found in tracts that had a higher average percentage of Asians. Powell et al also found differences in agreement on census tract race/ethnicity between field observations and proprietary database information for grocery stores in Chicago (24
), corroborating evidence that discrepancies in measures of community nutrition environments do not vary randomly among all neighborhoods. Store visits may be necessary to obtain a more accurate understanding of the availability of nutritious food.
Our results show discrepancies between a commercial database and surveyed characterizations of store types across neighborhoods, thereby complicating efforts to quantify the availability of nutritious food in large areas by using commercial databases. Improving the availability of nutritious food items and fresh foods at small grocery, convenience, and specialty food stores is a promising approach for improving community nutrition environments in underserved communities (26
). The national Healthy Food Financing Initiative allocated more than $400 million in 2011 to fund local, state, and regional collaborations that expand access to nutritious foods (27
). Now that funding is available to support community nutrition environments, it is essential to identify accurately high-need areas that should be prioritized for intervention.
Our study had several limitations. The survey assessed the availability of nutritious food in each store but did not evaluate price or accessibility, such as proximity to public transportation, which could affect the ability of some people to access nutritious foods. We did not compare the availability of nutritious foods with energy-dense and snack foods, which are associated with body mass index (28
) and fruit and vegetable intake (29
), nor did we examine the proximity of each store to other food stores. This study used data that are not temporally consistent. Socioeconomic and demographic data were from the 2000 US Census, commercial data were from 2008, and surveys were conducted in 2009. The 1-year lag between the collection of data obtained from the database and the administration of the surveys may have contributed to our inability to locate 15% of the stores selected from the database, but other field validation studies of food stores have found similar rates of database overcounts (22
). Our study results may not be generalizable to other areas. Each county in this study has a higher median household income than that of California and the United States.
Our study had several strengths. It is the first to compare data from in-store surveys of nutritious food availability at small food stores with data from a commercial database and data on socioeconomic and demographic characteristics. It demonstrates the use of a multidimensional approach to evaluate variability in community nutrition environments (30
) by considering both the location and context of food stores and the food products offered.
The variables in a commonly used commercial database do not accurately correspond to the variables public health and epidemiology researchers are interested in, namely indicators of the availability of nutritious and fresh food. Industry classification for small food stores varies. Although conducting in-store surveys requires more time and resources than collecting information from a database, surveys may be necessary to assess accurately the food environment and identify where improved availability of nutritious food is most needed.