Food deserts and food swamps in a Brazilian metropolis: comparison of methods to evaluate the community food environment in Belo Horizonte

The aim of this study was to evaluate the community food environment, food deserts and food swamps using several methods in Belo Horizonte City, Minas Gerais State, Brazil. Information on the community food environment was obtained from a comprehensive dataset collected from two secondary sources in the city. Analysis units comprised census tracts and buffers traced from the centroid of census tracts. Per-capita income was used as an explanatory variable. Five different methods, including one developed in Brazil, were used to identify and evaluate food deserts and food swamps throughout the city. The method from Brazil (which employs density of healthy establishments per 10,000 inhabitants) classified 37.7% of census tracts as food deserts. Food deserts were more frequent in the lowest-income tracts. Other methods gave food deserts lower values, and one method 2 (census tract) was similar to that found by the Brazilian method. The highest percentages of census tracts classified as food swamps were obtained using method 2 (mRFEI- percent healthy establishments, with 800 m buffer), with 66.6%, and using method 5 (density of outlets selling ultra-processed foods), with 58.5%. Food swamps were also found to be more frequent in the lowest-income tracts, especially when method 4 (count of unhealthy establishments) was applied. We conclude that most parts of Belo Horizonte provide easy physical access to retailers selling predominantly ultra-processed foods. Food deserts and food swamps were widely found, often in the same areas. The metric of food deserts and food swamps developed in Brazil was the most appropriate, because it considered specificities of the local food environment. In addition, it is necessary to improve existing food desert and wetland metrics. For example, include the income variable in these calculations.