Confirmatory factor analysis to derive meal's dietary patterns in a representative sample of adults and elderly people living in São Paulo

INTRODUCTION: Factor analysis is usually applied to derive dietary patterns however the factor loading’s cutoffs are arbitrary making difficult the decision of food groups selection per pattern. The objective of this study was to confirm patterns per meal that were suggested in a previous study. Correlations among meal’s dietary patterns were also evaluated.

METHODS: Information of two 24 hours recalls from a cross sectional population based study in the city of São Paulo (ISA Capital 2008) was used to estimate the food groups’ usual intake through the Multiple Source Method by meal. The sample size was 857 adults and elderly. We applied confirmatory factor analysis in Mplususing maximum likelihood estimation with robust standard errors.

RESULTS: We confirmed three dietary patterns for breakfast (healthy, milk with chocolate and snacks), four for lunch (traditional, salads, sweetened juice, junk food) and four for dinner (breakfast style, full meal, salads and snack). The comparative fit index was 0.72 and the Root Mean Square Error of Approximation 0.044 (0.042, 0.046). The healthy breakfast was positively correlated with the salad pattern for lunch and dinner and inversely correlated with traditional lunch and breakfast style and full meal for dinner. We found a positive correlation between traditional lunch with a full meal dinner and inverse with salad pattern for dinner. And finally, we observed a positive correlation between salad pattern for lunch and dinner.    

CONCLUSIONS: Our data support the importance to study meals separately once it allowed recognizing that a healthy pattern for one meal was not correlated to a healthy pattern in another. We observed poor meals, without variety, that could not be labelled as healthy (included only one health food group). It is important to stimulate balanced meals with a variety of foods to avoid deficiencies or excesses.