Jahns, Lisa (2004). Methodological explorations in intra-individual differences in children's nutrient intake. Master's thesis / Doctoral dissertation.
This dissertation consists of three parts: (1) Within- and between-person differences in variation of nutrient intakes of Russian and US children; (2) The use of external within-person variance estimates to adjust nutrient intake distributions; and (3) Physical activity and energy intake as determinants of overweight among Russian children. The Russian data is derived from the Russia Longitudinal Monitoring Survey (RLMS) in 1995, 1996, and 2002. The data on US children is from the Continuing Survey of Food Intakes by individuals (CSFII) in 1996. Dietary data was collected using interviewer-administered 24 hr recalls in both countries. Mean energy intake was 11% lower in Russian (7511 ± 121 kJ) than US (8484 ± 152 kJ) children in 1996. The within-person variances were almost always greater in the US, reflecting a greater day-to-day dietary diversity among US children as compared to Russian children of the same age. Similarly, the between-person variation was also nearly always greater in the US, reflecting a larger spread of usual intakes between children within the US. The ratios of within- to between-individual variance were similar (Russia: range 0.9-1.6, median 1.2, US: range 1.4-1.7, median 1.3) and were almost always greater than 1.0. Analysis of the prevalence of inadequate intake of an example nutrient (vitamin C), using external variance estimates to adjust the usual intake distribution showed that when the naive distribution is used, the resulting prevalence estimate is inflated by 100-1300%. When the intake distribution is adjusted using either US or Russian children's variance estimates the estimate is much less biased. The prevalence of overweight among Russian children decreased by 25% between 1995 and 2002, with a significant decline among high income children. There was a direct relationship between income and energy intake, and an inverse relationship between income and physical activity. Only hours spent in moderate physical activity was protective. In conclusion, the results suggest that in moderately large samples, adjusting distributions with appropriate external variance estimates results in more reliable prevalence estimates than using one-day data. It is important to examine effects of income on diet and activity to understand how income relates to obesity.