As food consumption was reported from the 2011–12 National Nutrition and Physical Activity Survey (NNPAS) component of the 2011–13 Australian Health Survey (AHS), FSANZ developed new nutrient profiles for reported foods and beverages that were not present in the core dataset.
As with the development of the core dataset, a range of techniques were used to generate these nutrient profiles. The most common techniques were:
- Recipe calculations
- Modifying an existing nutrient profile using label data and imputation
- Creating 'not further defined' lines
However, other techniques such as borrowing data were also used. This process saw the core database expand from 2,200 foods to over 5,600 foods in the AUSNUT 2011–13 database.
Recipe calculations were the most common technique used for generating nutrient profiles for foods and beverages reported during the NNPAS.
In AUSNUT 2011–13, recipe calculations were used to generate nutrient data for:
- home prepared and commercial meal-type mixed dishes such as pasta meals, stir-fries, casseroles
- takeaway and fast food products such as burgers, sushi, sandwiches etc
- cakes, biscuits and slices
- prepared beverages such as coffees and smoothies
- cooked meats, eggs, fish and vegetables with added oils
- some processed foods where no suitable analytical data were available
- 'not further defined' foods and beverages.
To undertake recipe calculations FSANZ needs information on:
- typical recipes and cooking practices used in Australia, including information on the types of ingredients used, their relative proportions and the most common preparation and cooking processes
- label ingredient and Nutrition Information Panel (NIP) data for processed foods
- the nutrient content of each of the ingredient foods (derived from the core dataset)
- how much moisture a food will lose or gain during cooking (weight change factor)
- the proportion of each vitamin and mineral that is lost in each ingredient during food preparation (retention factors).
The process used to carry out the recipe calculations for AUSNUT 2011–13 is comparable to that used internationally for similar purposes (Charrondiere et al. 2011).
The AUSNUT 2011–13 recipes are based on information found in a range of common Australian cook books and recipe websites, known commercial kitchen preparation procedures and product preparation instructions, gross composition data, and cooking and preparation practices observed during the time period of the NNPAS. Where analytical data was available for a similar food, the analytical data could also be used as a guide to the likely final composition of the recipe food.
Recipes created for processed products were developed using label ingredient lists. Using this approach, the amount of each major ingredient was adjusted so that the final nutrient profile was similar to the nutrient data presented on the product's NIP. It should be noted that this process does not generally take into account the use of food additives that do not contribute to the nutrient content of the food and the recipe generated may not reflect the exact formulation of the product available for sale.
Recipes were also used to create nutrient profiles for a group of ingredient foods that were used in subsequent recipe calculations. For example, in all home prepared casseroles containing vegetables, a single record of Mixed vegetables for use in home prepared casseroles was used in these recipes, and its nutrient profile was generated using a recipe based on the frequency of consumption of all vegetables added to casseroles, as reported in the NNPAS. By doing this, the number of new nutrient profiles that needed to be created was greatly reduced.
The recipes for home prepared foods do not take into account the use of salt during cooking, as this discretionary salt use was not captured in the NNPAS. Therefore, sodium values for these foods are likely to be lower than may occur in practice if these foods had salt added during preparation.
For a complete list of recipes used in AUSNUT 2011–13, refer to the AUSNUT 2011–13 Food Recipe File.
The retention factors and weight change factors were reviewed and updated where necessary before commencing the recipe calculation process. Generally, retention factors published by the USDA (USDA, 2003) were used, although it was necessary to expand these to include factors for iodine, selenium and vitamins B6, B12 and E. Weight change factors developed for the 1995 National Nutrition Survey (NNS) were used. Factors were adjusted for some foods until the moisture value in the recipe food was close to the moisture value of a similar analysed food where available. Due to the limited evidence base for these factors, they should be regarded as indicative only. For more information on the review of weight change factors and retention factors undertaken by FSANZ, refer to the Review of factors used in recipe calculations. For a complete list of retention factors used in AUSNUT 2011–13, refer to the AUSNUT 2011–13 Food Retention Factor File.
Modifying an existing nutrient profile using imputation and label data
Where a new nutrient profile was needed for a food (or beverage) that was similar to a food with an existing nutrient profile in the core dataset, except for a particular characteristic, a copy of the existing nutrient profile was made. The nutrient profile for the new food was then modified to account for the different characteristic. This approach was used for many of the low or reduced fat, reduced salt, fortified, reduced alcohol or intense sweetened varieties of products consumed during the NNPAS. These characteristics were most commonly modified using imputation and label data.
This technique involved imputing nutrient data from a similar food or beverage in the core dataset. For example, if a respondent reported consuming Bread, from wholemeal flour, commercial, added omega 3 polyunsaturates the core dataset food Bread, from wholemeal flour, commercial would be used as a basis for developing a new nutrient profile and the fatty acid content modified based on the fatty content of the analysed food Bread, from white flour, commercial, added omega 3 polyunsaturates.
This technique involved imputing nutrient data from a similar food or beverage in the core dataset, then updating key nutrient values using NIP data from product labels as appropriate. For example, if a respondent reported consuming Breakfast cereal, whole wheat, biscuit, added vitamins B1, B2 & B3 the core dataset food Breakfast cereal, whole wheat, biscuit, no added sugar & salt was used as a basis for developing a new nutrient profile, and the vitamin B1, B2 and B3 and total sugar and sodium values modified to reflect the values presented in the NIP of commonly consumed brands of whole wheat biscuits with added B vitamins. Other nutrient modifications might also have been necessary, such as adjusting the moisture, protein or carbohydrate contents to account for the addition of sugar.
Or if a respondent reported consuming Yoghurt, berry pieces or flavoured, reduced fat (~1%), the core dataset food Yoghurt, berry pieces or flavoured, regular fat (~3%) was used as a basis for developing a new nutrient profile, and the fat value modified to reflect the value presented on the NIP of commonly consumed brands of reduced fat berry yoghurt. Other nutrient modifications might also have been necessary, such as adjusting the moisture, protein, total sugars, cholesterol, vitamin E and retinol contents to account for the reduction of fat.
Wherever possible, nutrient values were averaged over a number of different brands for similar products, unless the nutrient profile was developed to represent a single brand.
Food & beverage descriptions with undefined characteristics
Undefined nutrient profiles were developed for survey foods and beverages where a respondent was unable to identify the exact food or beverage or cooking method of the food they consumed.
Nutrient data for undefined foods were derived using two approaches:
- Using a recipe approach to develop new nutrient profiles that draw on the profiles of closely related foods, with the recipe proportions weighted to reflect consumption patterns observed in the NNPAS or approximate market share information. For example, the nutrient profile for 'bread, from white flour, fresh, not further defined' drew on nutrient data for all white, fresh, fortified or unfortified breads, weighted according to consumption patterns observed in the NNPAS. Similarly the nutrient profile of 'Apple, red skinned, unpeeled, not further defined' drew on nutrient data for all cultivars of red skinned apples such as red delicious, royal gala, jonathan, fuji and pink lady apples weighted according to their approximate market share.
- Assigning an undefined food a nutrient profile of the most frequently consumed product from the relevant category. For example, many people in the NNPAS reported consuming 'tea', without identifying the type of tea (black vs green vs herbal). In this case the nutrient profile for Tea, regular, black, brewed from leaf or teabags, plain, without milk was used as this was the most frequently consumed type of tea (noting that the addition of milk and sugar is captured separately). This latter approach was more likely to be used for undefined foods eaten by few respondents or where the use of another food was unlikely to have any significant effect on the AHS outcomes.
Due to the weighting techniques used to develop undefined lines, the resulting nutrient data do not fully reflect the nutrient profile of the particular food actually consumed. In addition the weighting assumptions reflect patterns of consumption observed in this NNPAS and may not be appropriate for use in other circumstances.
Other techniques used
Where new nutrient profiles could not be developed using nutrient data from the core dataset, nutrient profiles were developed by:
- borrowing nutrient data from international food composition databases such as the USDA, UK, NZ and Danish food tables—this approach was most commonly used for unusual types of fish, fruits and vegetables and dried herbs and spices
- reproducing nutrient data published in AUSNUT 1999.
In each of these instances it was necessary to develop some individual nutrient values by techniques outlined earlier, as these data sources rarely included all nutrients being reported in the AHS.