Characterizing children's eating patterns: does the choice of eating occasion definition matter?

BACKGROUND: Recommendations to define eating occasions (EO) currently exist for research in adults, but not for children or adolescents. We examined how varying EO definitions affect the characterization of eating patterns in children and adolescents. METHODS: Cross-sectional dietary data collected using a 24-h recall data during the 2011-12 Australian National Nutrition and Physical Activity Survey (1364 boys and 1337 girls aged 2-18 years) were analyzed. Eight definitions were applied: participant-identified, time-of-day, and 6 neutral definitions (EO separated by 15- or 60-min and/or an additional energy criterion of 21 or 210 kJ). Frequency of and total energy intake from meals, snacks, and all EO were estimated. F tests stratified by gender and age-group, were used to assess differences between definitions. Agreement between definitions of meal and snack frequencies was assessed using intraclass correlation coefficients (ICC). Linear regression was used to estimate the proportion of variance in total energy intake (kJ) and BMI z-score predicted by each definition. RESULTS: Mean frequencies of meals and snacks differed between the participant-identified and time-of-day definitions, in boys and girls and for all age groups (P < 0.01). Across the six neutral definitions, there were differences between mean frequencies of EO with the largest mean difference observed for children aged 2-3 y (boys: 2.3, girls: 2.5; P < 0.003). Between the participant-identified and time-of-day definitions, there was good agreement for frequencies of snacks (ICC for both genders: 0.93) but not meals (boys: 0.36; girls: 0.38). The 15-min time interval plus 210 kJ definition of an EO consistently predicted the most variance in total energy intake (R2 range = 8.1-34.8). Definitions that delineated meals and snacks better predicted variance in BMI z-score, when compared to the neutral definitions. CONCLUSIONS: How eating patterns are characterized vary depending on the EO definitions employed, particularly in young children. Variance in total energy intake was best predicted by a variation of the neutral definition whereas definitions that delineated meals and snacks performed better in relation to predicting BMI variance. Further international research that compares EO definitions in children will help inform a standard approach.