Revista Brasileira de Epidemiologia (Sep 2007)
Redução de lista de alimentos para questionário de freqüência alimentar: questões metodológicas na construção Food list reduction for a food frequency questionnaire: methodological issues
Abstract
OBJETIVO: Reduzir lista de alimentos de um questionário de freqüência alimentar semiquantitativo através de diversas estratégias metodológicas, mantendo-se a capacidade de análise do consumo de energia, proteínas, glicídios, lipídeos, cálcio e ferro. MÉTODOS: Utilizou-se um questionário de freqüência semiquantitativo com 80 itens alimentares respondidos por 224 meninas (12-19 anos). Reduziu-se a lista por regressão linear múltipla, freqüência de consumo e composição nutricional dos alimentos. Analisou-se por teste t de Student, sensibilidade de distribuição das adolescentes por quartis de consumo, coeficiente Kappa e modelos de regressão. RESULTADOS: Os alimentos incluídos apresentaram coeficiente de regressão positivo, com valor de F significativo (POBJECTIVE: To reduce the food list of a semi-quantitative food frequency questionnaire, through several methodological strategies, maintaining the ability to assess energy, proteins, glycids, lipids, calcium, and iron consumption. METHODS: A semi-quantitative 80-item food frequency questionnaire, answered by 224 female adolescents (aged 12-19 years) was used. The list was reduced through multiple linear regressions, frequency of consumption and food nutritional composition. It was assessed through Student's t-test, sensitivity of adolescents' distribution according to consumption quartiles, Kappa coefficient and regression models. RESULTS: Food included in the analysis showed a positive regression coefficient, with a significant F value (P<0.001). The models' total variances were over 0.87, and some food types did not show plausible models. Thirty-two food types were selected among the models, and 37 among the most commonly consumed ones. Both criteria included some different foods and, when these were merged, 40 food types formed the abridged list. Consumption averages did not vary between the complete list and the reduced one. Sensitivity analyses ranged from 85% to 100%. The agreement level was significant for all diet factors. Regression models in the reduced list kept total variances over 0.85 and similar food types as compared to the previous ones. CONCLUSION: The established criteria were appropriate to reduce the food list from 80 to 41 items, keeping the predictive ability and food intake assessment of the components investigated in this population group.
Keywords
