| Food Composition Data: A User's Perspective (1987) |
|The uses of food composition data|
|Using food composition data to communicate nutrition to the consumer|
Nutrient and food constituent data sources
A major consideration in developing NUTREDFO was to obtain the most accurate nutrient and food constituent values possible. For those nutrients in foods for which analytical values existed the objective was to determine the most appropriate original source of data. For nutrient values for which analytical data did not exist, the most legitimate criteria for imputation had to be identified.
Table 1. NUTREDFO system functions for analytical testing
1. List nutrient and food constituent levels for any food both in the permanent and temporary data file (this function lists any of the 26 nutrients, food constituents, and serving size specified by the user).
2. List foods in rank order by nutrient or food constituent based on contribution of that nutrient or constituent per serving of the food item.
3. List foods in rank order for any specified nutrient based on the percentage of standard for that nutrient. Any one of three standards can be used in this function. including the 1980 RDA , Single-value Nutrient Allowances per 101)0 Kilocalories  or the US Recommended Daily Allowance .
4. List foods in rank order for a user-specified nutrient based on its Index of Nutritional Quality (INQ) . This rank-ordering uses the same standards as in those in no.3 above.
5. Calculate the mean, standard deviation of the mean, and minimum/maximum range for selected nutrients in user-specified groups of foods.
6. Allow the user to temporarily change serving sizes and nutrient and food constituent levels in both the permanent and temporary data files. This function can be used to examine the effects of changes in nutrient levels on selected foods, but does not jeopardize the security of data in either file.
7. Calculate the nutrient value of menus. Menus can be for individual eating occasions, individual days, and for a group of days.
8. Calculate the Index of Nutritional Quality; percentage of standard for nutrient totals, percentage of calories from total protein, carbobydrate, fat, and alcohol; and calculate values for each nutrient and food constituent per 1000 kilocalories for each eating occasion and each day.
9. Calculate the mean, standard deviation, and minimum/maximum range for each nutrient and food constituent in a multi-day grouping of menus.
USDA was considered the most appropriate original source of data for raw or cooked foods. Data values available in recently revised sections of Handbook No.8[2, 14, 22, 23,30,41, 43,44,57,58,59, 60] were used preferentially, and then provisional food composition tables developed by USDA [13, 25, 26, 45] and values in the 1963 edition of Handbook No.8 . All values were verified with the original source. Whenever possible, analytical iron values from Iron Content of Foods  were used for items not listed in revised Handbook No.8 sections.
USDA computerized data tapes [64,65,67] were used to supply data for missing values in Handbook No.8 printed tables. A recent article by Hepburn  discussed these data tapes. In some cases, particularly for vitamins B6, and B12, published or computerized values had to be recalculated because retention and cooking yield factors had been inconsistently applied .
Journal articles by USDA research specialists [3,4,5,6,9,16,17,18,19,20, 21,37,38,47, 53, 55,56, 72] and other USDA publications [42,48,52] were also used as sources for data for specific nutrients and food constituents not available in Handbook No. 8, the provisional tables, or the data tapes. These nutrients include fatty acids, cholesterol, zinc, folacin, and sugars in ready-to-eat and granola cereals. Occasionally, food composition sources [8,49,70] not published by USDA provided data where none could otherwise be found.
Nutrient values were imputed when acceptable published sources were not available. Imputations were generally derived through mathematical manipulation or adjustment of data from published sources. The preferred approach was to calculate nutrient values for cooked products using data for raw products and applying USDA retention and yield factors. When this was not possible, data values were assumed from similar products, e. g. values for cooked chard were used to estimate values for cooked romaine. An additional recourse was to consult experts in the field, usually Consumer Nutrition Center specialists who were working on particular food and nutrient data and who had access to industry and other data specialists. Calculations for recipe imputations were based on published USDA procedures as used in Handbook No. 8 .
The NUTREDFO system provides specific information about the source for each nutrient and food constituent value. This documentation enables users to have on-line access to the source of every nutrient value in the data base.