A number of the foods in the core dataset did not contain a value for every nutrient being reported in the AHS and these data 'gaps' needed to be filled, so that nutrient intakes generated from the AHS would not be underestimated. The largest gaps existed for nutrients that had not previously been reported in Australian national nutrition surveys, such as total trans fatty acids, selenium, vitamin B6 and vitamin B12.
A range of techniques were used to fill these gaps for specific nutrients. The most common were imputation and borrowing data. However, other techniques such as estimation and use of label data were also used.
Imputation of values, in the context of filling nutrient data gaps, refers to the assumption that a nutrient value in a closely related food will be the same in the food that has the gap. For example, if there was a data gap for iodine in a particular cut of beef, this could be filled by imputing a value from another cut of beef. Where possible, values were imputed from similar foods within the core dataset or from other FSANZ nutrient databases such as AUSNUT 2007 (FSANZ, 2008), developed to support the 2007 Australian National Children's Nutrition and Physical Activity survey (CSIRO, 2008) or from levels permitted in the Food Standards Code.
Imputation can also be used to assume that some foods contain none of a particular nutrient, based on the knowledge of the composition of the food. For example, the vitamin E content of soft drinks has been imputed as zero. This is because vitamin E is a fat soluble vitamin and soft drinks do not contain fat. Their labels also indicate they don't contain added vitamin E. This approach was also commonly used for filling the data gaps for caffeine, folic acid, retinol and vitamin B12.
Imputation has only been used where FSANZ has confidence in the validity of the assumptions made.
Borrowed data was most commonly used to fill gaps for nutrients that had not previously been reported in Australian national nutrition surveys.
Data was borrowed from major international food composition databases/tables including:
- The Composition of Foods, 6th Summary Edition (Food Standards Agency, 2002)
- Danish Food Composition Databank (Saxholt et al, 2008)
- German Food Tables (Souci et al, 1994)
- Concise New Zealand Food Composition Tables (NZ PFR, 2009)
- United States Department of Agriculture National Nutrient Database for Standard Reference (USDA, 2013)
- Tables of Composition of Australian Aboriginal Foods (Brand-Miller et al, 1993)
Data was also borrowed from scientific literature with relevant research papers identified using standard literature searching techniques.
Borrowed data was selected carefully to ensure they reflected similar production, preparation and fortification practices to those that exist in Australia, wherever possible. Where necessary, values provided were converted to the units and modes of expression being used in the AHS.
A smaller proportion of nutrient gaps were filled using recipe calculations/estimation and label data. Recipe calculations were most commonly used to generate values for mixed foods, while label Nutrition Information Panels (NIP) were most commonly used to fill data gaps for fortified foods and to validate whether any major changes in sodium levels had occurred since the food was last analysed.
These techniques were more commonly used to generate nutrient profiles for additional foods consumed during the AHS and have been described in more detail in the next section.