DEVELOPMENT AND EVALUATION OF THE RELIABILITY OF AN INSTRUMENT FOR ASSESSING THE UNIVERSITY FOOD ENVIRONMENT

• Published in: Annals of nutrition and metabolism 2017-10, Vol.71 (Suppl. 2), p.836
• Main Authors: Franco, Amanda da Silva, de Castro, Inês Rugani Ribeiro, Canella, Daniela, Pereira, Alessandra Pereira, dos Santos, Claudia Roberta Bocca, Rodrigues, Lúcia, Barbosa, Roseane Moreira Sampaio, Tavares, Letícia Ferreira
• Format: Article
• Language: English
• Subjects: Colleges & universities; Correlation analysis; Correlation coefficient; Correlation coefficients; Developing countries; Diet; Eating behavior; Food; Health promotion; LDCs; Mathematical analysis; Nutrition; Reliability analysis
• ISSN: 0250-6807; 1421-9697
• DOI: 10.1159/000480486

Background and objectives: The organizational food environment, including the university, has been recognized as strategic for health promotion, since the foods offered there influence the eating habits of the individuals who attend these spaces. There are few studies on the evaluation of the performance of instruments used to assess university food environment. In developing countries, up to this moment there were no studies. This study proposes an instrument to assess the university food environment and evaluates its reliability in public universities in Brazil. Methods: A checklist was prepared for the assessment of the university food environment, containing the following blocks: characterization of establishments, observation of the environment, availability, price, nutritional information and advertisements of healthy and unhealthy food markers. A convenience sample (n=64) of establishments with different characteristics distributed on seven campuses of three different public universities located in the state of Rio de Janeiro was studied. Data were collected from November 2015 to February 2016 by trained fieldworkers. The reliability of the instrument was examined by inter-observer and test-retest tests. For categorical and counting variables, instrument stability was estimated by calculating percent agreement (PA), kappa and prevalence- and interviewer bias-adjusted kappa (ka). For continuous variables intra-class correlation coefficients (ICC) were calculated. The values of PA, k, ka and ICC were classified according to the criteria of Landis & Koch (1977). Results: 73.5% and 65% of the 204 items evaluated presented a substantial or near perfect agreement, respectively, for the inter-observer test and the test-retest. Considering the ka, both the inter-observer test and the test-retest showed that 93% of the items presented substantial or near-perfect agreement. Of the 176 items also assessed by means of PA, 100% showed substantial or near-perfect agreement for the inter-observer test and the test-retest. Conclusions: The instrument presented excellent performance in the context in which it was applied.