Estimating adult equivalent scale for nutrition data using polynomial spline model.

摘要

Background. Household food inventory (HFI) data is a good approach to assess food availability for households, and has been recognized as a relevant and practical mean to study and/or influence residents' dietary intake behavior. However, there are very limited resources to study HFI at individual level instead of household level. Currently existing methods like per-capita approach and calorie-based scale have numerous shortcomings. Intuitively, per-capita method can only incorporate household size effects into consideration; and calorie-based scale, which is more developed method than per-capita method, can incorporate both household size and composition effects. However, calorie-based scale is still limited by discontinuity and incapability to differentiate different nutrients. The proposed method, adult equivalent scale (AES), was developed using polynomial spline model, which intends to improve the existing models.;Methods. The major goal of the paper is to construct a unique continuous AES for each of nutrient of interest (total calorie, carbohydrates, protein, fat, saturated fatty acid (SFA), dietary fiber, Vitamin D, Vitamin C, calcium, sodium, potassium, total and added sugars) based on national level dietary data. Number of model candidates was selected based on the standardized mean dietary intake data, and then model fits were assessed graphically and non-graphically using AIC, BIC and R2. The model fit of the finalized model was further assessed using cross validation and AES profiles.;Results. The resulting AES was normative and hence is easy to apply. The continuous characteristic of the suggested AES allows more sensitive analysis referring at each specific age and gender combination rather than a block of age groups. Graphical and non-graphical assessments including cross validation suggested that unique AES profile will be needed for different combination of nutrient type and gender. Per capita method lacks the ability to incorporate household composition or nutritional uniqueness; while calorie-based scale could not incorporate nutritional uniqueness and was discontinuous, which was insensitive to children of different ages.;Conclusions. The generated AES is continuous and unique for each nutrient type and gender combination. Comparing to the per-capita or calorie based scales, it better represents the nutritional availability for each child with specific age and gender. However, the generated AES is based on American national data, and hence may have limited generalizability to other populations.