NHANES III Dietary Interview Component Dietary interviews were administered to all examinees by a trained dietary interviewer in the mobile examination center (MEC). Respondents reported all foods and beverages consumed except plain drinking water (i.e., not bottled) for the previous 24-hour time period (midnight to midnight). An automated, microcomputer-based dietary interview and coding system known as the NHANES III Dietary Data Collection (DDC) System was used to collect all NHANES III dietary recall data. The DDC system was developed for use in the survey by the University of Minnesota's Nutrition Coordinating Center (NCC). The dietary interviews were conducted in English and Spanish by bilingual dietary interviewers in a private room to ensure confidentiality. Proxy respondents were permitted for infants and children aged two months through five years and for other respondents who were unable to report on their own. Children aged six to 11 years were permitted to report their own intake if the interviewer deemed it acceptable and appropriate, but many interviewers for respondents in this age category were completed by proxy or with the child and a proxy. The dietary interviewers contacted other information sources such as care providers and schools to obtain complete dietary intake data for respondents. The primary source of food composition data for NHANES III is the U.S. Department of Agriculture (USDA) Survey Nutrient Database; two nutrient files were provided by USDA for use in NHANES III (USDA 1993, 1995). Each USDA file contained food composition values that were appropriate for the time period during which the NHANES III data were collected. Additionally, food composition data for a small number of herbs and spices were obtained from NCC (NCC, 1996). The DDC system's foods database was designed specifically to handle time-related changes in food descriptions, food amounts, and recipes; updated information was applied retrospectively to data collected in the early part of NHANES III. As was mentioned earlier, two USDA food composition databases were used to assign nutrient values to the NHANES III dietary recalls (USDA 1993; USDA, 1995). In addition to data changes that occurred in the nutrient values of foods due to food product reformulations, recipe changes, and so forth, the U.S. marketplace underwent tremendous growth and change as new food product lines were introduced and new food components were added to the food supply (e.g., fat substitutes and artificial sweeteners). The impact of these and other changes may require additional analysis. Dietary recall interviews were edited by the interviewers to ensure that they were as complete as possible. NCHS completed all final editing and determinations regarding the completeness and reliability of the dietary recalls. Analysts should note that the data reported are self-reported data. Extreme values were verified. Information pertaining to the use of nutritional supplements and antacids was reported separately during the Household Adult and Household Youth Interviews. A number of quality-control monitoring techniques were employed during the survey. The techniques for monitoring the Dietary Interview component included observations of actual dietary interviews and reviews of audiotape interviews by NCHS and contractor staff. In addition, the dietary interviewers worked in two-person teams; there was one team in each MEC. The dietary interviewers performed 10-percent cross-check reviews of their partners' work using printed recall reports. Finally, newsletters, field memoranda, telephone calls, and staff retraining sessions were other methods used to maintain quality control during the survey. Refer to the NHANES III Dietary Interviewer's Training Manual for the dietary interview protocol (U.S. DHHS, 1996b). Analysts are encouraged to use six years of survey data in their analyses. The reliability of estimates is improved when larger sample sizes are used. For more detailed information, see the Analytic and Reporting Guidelines for NHANES III (U.S. DHHS, 1996b). In addition, MEC final examination weights (WTPFEX6) should be used when analyzing the total nutrient intake data and related questionnaire data in this file. For more information on the use of sample weights in NHANES III data analysis, refer to the NHANES III Analytic and Reporting Guidelines (U.S. DHHS, 1996b). NHANES III Total Nutrient Intakes and Foods Data Files NCHS prepared 4 datasets that are based on the 24-hr dietary recall interview. Total nutrient intakes were reported in the NHANES III Examination Data file (Catalog 76200). Three foods files were prepared; three files are found in Catalog 76700: NHANES III Individual Foods Data File from the Dietary Recall; NHANES III Combination Foods Data File from the Dietary Recall; NHANES III (Variable) Ingredients Data File from the Dietary Recall. Documentation was prepared for each of the foods data files. Data users are encouraged to review all of the documentation prior to using the data files. Look-up Tables for the NHANES III Foods Data Files Textual descriptions for several NHANES III Foods Data File numeric code variables are located in an Appendix section that accompanies the Foods Data Files. The Appendix files are referred to as "look-up" tables throughout the data file documentation for the Foods Data Files. Computer code is provided so that data users can merge the foods data files with the information in the Appendix/look-up tables. INDIVIDUAL FOODS FILE The NHANES III Individual Foods File (IFF) is comprised of records. Each IFF record includes a meal number (DRPMN), a food number (DRPFN), and a component number (DRPCN). The IFF was sorted by case, meal number, food number (within meals), and component number (within foods). Meals are comprised of foods. Foods are comprised of one or more components. Most components in the IFF are foods. There are some ingredient-type components (salt, water, corn meal, etc.) in the IFF. Components were either eaten alone or in combination with other foods. The term "component foods" may be used for most of the components in the IFF. Components may have ingredient records associated with them. Ingredient information is reported separately in the NHANES III Variable Ingredients File. The IFF contains information on all component level foods and beverages reported by examinees whose dietary recalls have a final dietary recall status code (DRPSTAT) equal to 1, 2, or 5; the documentation for this file includes an explanation of DRPSTAT values. Partial foods data are reported in the IFF for examinees with incomplete recalls (DRPSTAT=2) and for nursing infants and children (DRPSTAT=5). If total dietary intake information is required for data analysis, only examinees with DRPSTAT=1 should be selected for analysis. The IFF can be linked to the other NHANES III foods files by case, meal number (DRPMN), food number (DRPFN), component number (DRPCN), and ingredient number (DRPIN). Multi-component or combination foods have a combination foods flag (DRPCFF) value equal to 1 in the IFF; combination foods are described in the Combination Foods File. Some component foods in the IFF have variable ingredients (DRPVIF=1); the Variable Ingredients File contains information about these ingredients. The Appendix section of the IFF file documentation contains a series of tables that illustrate the linkages between the NHANES III foods files. Additionally, data users should refer to the documentation sections for each of the NHANES III foods files to learn more about the content and uses of these data. Coding Foods Reported During NHANES III An underlying principle of NHANES III Dietary Data Collection (DDC) System database maintenance was to use U.S. Department of Agriculture (USDA) Survey Nutrient Data Base (SNDB) files as the primary data sources for food codes (Codebook File), recipes (Recipe File), and nutrient data (Nutrient File). Hispanic HANES was the first HANES to use SNDB data bases exclusively to code and report dietary findings. Continued use of the SNDB data bases during NHANES III served to maintain consistency with Hispanic HANES (HHANES) data for nutrition monitoring purposes. Most of the foods and beverages reported during NHANES III are coded using USDA SNDB food codes (hereafter referred to as "USDA food codes"). A small number of non-USDA food codes are included in the NHANES III data release files because there were no USDA food codes for spices and certain recipe ingredients. All component food codes DRPFCODE)reported in the IFF have text descriptions in the look-up table called "Codebook". Many of the component foods reported during NHANES III were coded using the food code that USDA would use in its food consumption surveys. For some foods reported during NHANES III, however, the coding method used was not based on the USDA code for the following reasons: 1. The DDC system foods database was more specific than the USDA database with respect to recipe ingredient specification. Ingredient information was used to compute the nutrient content of recipe foods reported in NHANES III. 2. A brand name product was not in the USDA database. The DDC system included more than 6,000 brand names in approximately 30 food categories; the USDA database contains fewer brand name products. NCC assigned USDA food codes to all brand-name foods in the DDC system. The coding decisions for brand-name foods were based upon USDA, NCC, and manufacturer information. 3. The brand name was in the USDA database, but the University of Minnesota Nutrition Coordinating Center (NCC) coded it differently from USDA. For example, NCC and USDA used different criteria to code brand-name cookies, salad dressings, and crackers; NCC nutrient criteria were used to assign USDA food codes to commercial products in these food groups. 4. An NCC recipe was used instead of the USDA recipe. 5. The food was not in the USDA database. Food Type Categories All component foods in the IFF are divided into two food type categories as denoted by the variable "DRPREC." The food category determines the method used to code foods and assign nutrient values to foods. The first food category type is "elemental" foods. Elemental foods include milk, fresh fruits and vegetables, ready-to-eat breakfast cereal, sweeteners, and fats and oils. Some mixture foods also are classified as elemental foods in the DDC system foods database. USDA food codes were assigned to elemental foods. The USDA food code for elemental foods (DRPFCODE) has a direct link to the USDA SNDB Nutrient File that was used to assign nutrient values to all elemental foods. The second category of foods is "recipe" foods. Recipe foods are denoted by DRPREC=1. The survey files for recipe foods contain ingredient records; the ingredient records for each recipe food are linked together by a USDA food code. The USDA food codes for recipe foods (DRPFCODE) are reported in the IFF. The nutrient values for recipe foods were calculated using recipe ingredient nutrient values found in the USDA SNDB Nutrient Files provided for use in NHANES III. The USDA Nutrient Files for NHANES III are slightly different from the standard public release USDA SNDB Nutrient Files because special food codes for recipe ingredients (usually denoted by food codes that begin with numbers "00" were added to the USDA file at NCHS's request. The nutrient values for recipe ingredients were summed to produce the nutrient values for all recipe foods in the IFF. A USDA food code was assigned to recipe foods as a means of linking the recipe ingredient records. The USDA food code that is used to report the recipe food should be used only to provide a basic food description for the food and was not used to assign nutrient values to recipe foods. Further, a USDA food code may be used to code more than one type of recipe food; this was because the DDC system included more food description options. All food codes (DRPFCODE) used in NHANES III have text descriptions associated with them in a look-up table called "Codebook". To summarize, the distinctions between elemental foods and recipe foods are: 1) recipe foods have ingredient records associated with them and elemental foods do not; 2) recipe ingredient information was used to compute the nutrient values of recipe foods reported during the survey. Using Ingredient Information to Calculate Nutrient Values of Foods Recipe foods have ingredient records associated with them in the comprehensive DDC System output files. Many of the ingredients used to prepare recipe foods were "variable" ingredients meaning that respondents could specify the types of ingredients that were used to prepare the foods they ate. The variable ingredient flag (variable name: DRPVIF) denotes the recipe foods that had variable ingredients. The ability to vary the types of ingredients that were used to prepare recipe foods is important because the nutrient values for recipe foods that have a particular food code (DRPFCODE) can have a range of values rather than a single nutrient profile. To illustrate, take the example of a homemade macaroni and cheese casserole. There were two variable ingredient probes in the DDC system for this entry. One probe pertained to the type of cheese used, and the second probe was for the type of milk used in the recipe. Assume that the same basic recipe was used for this dish. If one respondent used low-fat cheddar cheese and skim milk, and a second respondent used regular- fat cheddar cheese and whole milk, the nutrient content of the two dishes would differ because two major recipe ingredients had different nutrient values. A second example would be for a commercial food prepared at home. Many commercial foods were defined as recipe foods in the DDC system so that specific information about the ingredients used to prepare commercial foods could be ascertained. One example of a commercial food product with variable ingredient probes was commercial breaded chicken that was purchased in frozen form and fried at home. The DDC system probes included the type of fat used to fry the chicken and a probe for the addition of salt during food preparation. A second example of a commercial product with variable ingredients was a brand-name meal replacement beverage that was reconstituted with fluid milk; the type of milk used to prepare the beverage was a variable ingredient. These examples illustrate how preparation ingredients produce variations in the nutrient content of the prepared foods. Notes to Analysts Ordinarily, respondents were not asked to report plain drinking water during the dietary interview because a separate set of questions addressed plain drinking water consumption. Plain drinking water was a component of certain foods, however. This occurred when foods were diluted with extra water or when modified recipes were entered by individual components that included plain drinking water. In these instances, drinking water was included in the file as a component record, and the nutrients contributed from the drinking water were included in the IFF and the Dietary Recall Total Nutrient Intakes portion of the Examination File. Respondents also were not asked to quantify the amount of salt added during food preparation or at the table. A separate set of questions was administered to determine categories of salt use at the table. This information was reported in the Dietary Recall Total Nutrient Intakes portion of the Examination File. Salt appears in the IFF as a component record for some foods that were reported as having modified recipes. If a food was entered by ingredient-type food components that included salt, a component record for salt was included in the IFF. (Note: Also refer to the documentation for the Combination Foods File.) Food Descriptions 1. Brand-name foods The DDC system foods database contains more than 6,000 brand-name foods. DDC system brand-name products are grouped into more than 30 food categories and include commercial frozen entrees, "fast food" restaurant menu items, ready-to- eat breakfast cereals, candy, fats and margarine, and juice drink beverages. The brand name foods in the IFF have a USDA food code and a numeric brand product code (DRPCOMM); DRPCOMM is linked to a look-up table called "Brands". 2. Generic foods Generic foods in the IFF have USDA food codes assigned to them; the USDA food codes are linked to a food code description in the look-up table called "Codebook". Many generic foods have expanded food descriptions in the IFF. The food identification code (DRPFID) variable is linked to an expanded food description; the look-up table "IDCODE" contains the text descriptions for DRPFID. Two examples are provided to illustrate the use of food identification codes (DRPFID). The first example is trout. The DDC system probes for trout included several types of trout -- rainbow, brown, speckled, and so forth. The USDA Codebook does not distinguish among types of trout but uses the same food code for all varieties of trout. The food identification codes (DRPFID) in the IFF can be used to distinguish between different types of trout that were reported in the Survey. In this example, if a respondent reported eating rainbow trout, the DRPFID would be a specific code for rainbow trout. A second example relates to cuts of meat such as beefsteak. The DDC system probes included the cut of steak reported -- sirloin, round, tenderloin, and so forth. In summary, the food identification codes often provide more specific, descriptive information for foods that have the same USDA food code. Food Amount Information The DDC system's computer data entry screens usually displayed several options for entering food amount data. In general, the DDC system's food amount options corresponded to the food amount options listed in the USDA SNDB Codebook. In addition to weight and volume options, many foods could be quantified by means of "food specific units" (FSUs). For foods such as whole chicken parts, pork chops, commercial sliced bread, sliced luncheon meats, and so forth, the FSU was the preferred method for quantifying such foods because their dimensions were difficult to estimate. All DDC system food amount entries, including the food models, volume amounts, and FSUs, were converted into gram weights automatically during final data processing and preparation. All food amounts in the IFF were reported as grams of food eaten. Unusually large amounts of food were verified during the dietary interview. The DDC system's data quality control features included a "maximum amount check verification screen" for each food item. This screen appeared whenever large food-specific amounts of food and beverages were entered during the interview. Interviewers were required to verify that the amount of food or beverage reported was correct. Default Selections for Foods and Food Amounts 1. Default Selections for Foods The DDC system was designed to collect specific information about foods, yet respondents' knowledge about the foods they ate varied. When respondents were unable to provide specific information about the foods they ate, the dietary interviewers used the DDC system's default selection options to complete data entry for foods reported during the survey. The DDC system had default selection options for the type of food, ingredients used to prepare foods, and food preparation methods. DDC system default options were available for many home-prepared and commercially prepared foods. When the origin of the food (i.e., commercially prepared or homemade) was unknown, a system default option "unknown as to whether commercially prepared or homemade" was selected by the interviewer. Default selections also were available for food preparation methods and the ingredients used to prepare foods. The DDC system's default food selections have USDA food codes associated with them that are linked to the USDA SNDB files described earlier. 2. Default Food Amounts Some foods were not quantified at the time of the dietary interview for by a number of reasons. Example #1: Food amounts were known but were reported using an amount option that was not available to the interviewer at the time of the interview. Therefore, an amount could not be entered using the DDC system. The interviewer noted the amount description provided by the respondent. NCHS and USDA staff completed the research required to quantify these foods. New food amount options were added to the DDC system throughout the survey. Example #2: The respondent was unable to quantify the amount of food consumed, but the food was from a small list of foods for which the interviewers were permitted to calculate a default amount. The dietary interviewer initially "flagged" the food amount as having an unknown amount. All information provided by the respondent that could be used to calculate a default amount was recorded by the interviewer. During the interviewer's edit, amounts of certain foods, including sandwich condiments, catsup and barbecue sauce on meat, coffee creamer, butter and margarine added to bread, and milk added to beverages and cereal, were calculated. NCHS reviewed the interviewers' calculations to verify that the calculation was performed correctly. Example #3: The amount of food consumed was unknown, and no default amount standard existed for the food. This problem was most common when the recall involved infants and young children who attended day care or school on the day of the recall. The interviewers were instructed to flag the food as having an unknown amount. In the meantime, the dietary interviewers attempted to obtain information from day care providers, schools, etc. If the amount could not be entered, NCHS assigned a default food amount. The default food amounts usually were based on a "not further specified amount" for a similar food in the USDA SNDB Codebook. NCHS developed editing guidelines that were used to assign food amounts to many types of foods. Examples #2 and #3 describe situations in which food was consumed but for which the respondents could not quantify the food. In both instances, the amount consumed was entered initially into the DDC system as an "unknown amount." A food amount was assigned later. The default amount flag (DRPCAUF) in the IFF denotes the foods described in examples #2 and #3 that had default amounts assigned; if DRPCAUF=1, a default amount was assigned to the component food. Food Preparation Information The IFF includes information on food preparation methods and ingredients used to prepare foods. The interview probes for food preparation methods varied according to the type of foods reported. For example, the probes for vegetables usually began with the name of the vegetable and whether it was eaten raw or if it was cooked from fresh, frozen, canned, etc. If the vegetable was cooked, the cooking method and use of preparation ingredients, such as marinades, fat, and salt, were ascertained. The interview probes for mixed dishes began with a brand name probe or a question to ascertain whether the dish was commercially prepared or homemade; the questions that followed dealt with cooking methods, preparation ingredients, and so forth (U.S. DHHS, 1996). NCC uses special formulas for certain types of recipe foods called "preparation algorithms." The NCC preparation algorithms were part of the DDC system foods database. NCC preparation algorithms were used to add preparation ingredients such as fat, salt, soy sauce, and breading to meat, fish and poultry, rice, pasta, popcorn, and vegetables that were reported during the survey. Preparation ingredients were added automatically to a base food -- usually a plain version of the food reported. The IFF variable DRPPPDC lists more than 100 food preparation methods. The text descriptions for each of the DRPPPPDC are found in a look-up table called "Prepd"; the linking variable is DRPPPDC. One example was boiled rice prepared with fat and salt. The base ingredient in the recipe was plain, boiled rice. Using the NCC preparation algorithm, 0.5 teaspoon of fat was added per 0.5 cup of rice. The type of fat used to prepare the food was a variable ingredient; the respondent could specify any type of fat -- butter, margarine, a specific type of animal fat, etc. Salt use was ascertained by asking the respondent if the rice was cooked in salted or unsalted water; one-half teaspoon of salt per half-cup of rice was assumed if salted water was used. Ingredients used to prepare foods were variable ingredients and have been included in the Variable Ingredients File. A second example was boiled carrots. Made from fresh carrots, fat and salt were added during preparation. The NCC preparation algorithm adds 0.5 teaspoon of fat and 0.08 teaspoon of salt per half cup of boiled carrots. A third example was breaded fried chicken. The NCC data base contains the information required to convert the portion of chicken reported during the interview into ounces of white or dark meat, with or without the poultry skin. The NCC preparation algorithm added the preparation ingredients specified by the respondent, such as marinade, breading, basting, or frying fat, to a base dish of plain, roasted chicken. Commercially prepared meats, vegetables, and so forth usually were coded using either the USDA food code or ingredient information obtained from manufacturers. Preparation ingredient information for commercial restaurant foods was not reported by most respondents. To summarize, the NCC preparation algorithms automatically added the amounts of preparation ingredients to many of the recipe foods reported during NHANES III. The nutrient values for these foods were computed in a manner similar to other recipe foods using the preparation ingredient information. As with other recipe foods, the component food code provides the closest food code description match to the USDA data base. Food identification codes (DRPFID) are also provided for many foods that have NCC preparations; the text descriptions for these codes provide additional descriptive information about these foods. The NHANES III Nutrient Data Base Each HANES used nutrient data bases that were appropriate for the time period of the survey. NCHS compiled the nutrient data bases used for NHANES I and NHANES II; a variety of data sources were used in the data bases including USDA, selected manufacturer data, and Tulane University food composition data. In 1982, NCHS began using the USDA database exclusively; all foods reported during HHANES were coded using USDA food codes and their corresponding USDA nutrient values. Several new food items commonly eaten by the Hispanic subgroups surveyed during HHANES were added to the USDA database for HHANES. Following HHANES, NCHS staff continued to work closely with the USDA's Agricultural Research Service (ARS) staff to update the USDA database with new foods, food weights, and recipes in preparation for NHANES III (1988-94). Hundreds of new food codes for ethnic foods, including Mexican-American and Asian foods, new commercial products, homemade soups, and bakery products, were added to the USDA data base for NHANES III. NCHS and NCC incorporated new recipes into the DDC system for certain Mexican-American foods; some of the recipe changes were required to meet the data base maintenance requirements of the DDC system but do not appear in the USDA data base. The nutrient data bases for NHANES III were provided to NCHS by ARS and can be purchased from the Department of Commerce, National Technical Information Service (NTIS), Springfield, Virginia. (Please refer to the Introduction to this documentation file for information about contacting NTIS.) The NHANES III, Phase 1 (1988-91) data base consisted of a slightly modified version of the USDA file that was used with the 1989 USDA Continuing Survey of Food Intake by Individuals (CSFII). ARS assigned Release Number 5.1 to the nutrient data base for Phase 1. In addition to Release 5.1, ARS provided revised data for a small number of food items to reflect changes in foods that occurred in 1990 and 1991. ARS provided an updated version of the SNDB for NHANES III, Phase 2 (1991-94). The USDA Survey Nutrient Data Base includes total energy and 29 nutrient data fields. All nutrient and dietary fiber intakes for individuals were calculated using the gram amounts of food consumed and the USDA Survey Nutrient Data Base (SNDB) nutrient values for the food expressed per 100 grams of food. NCHS computed percentages of total food energy intake from all energy sources. The energy conversion factors used were 4 kcal per gram for protein and carbohydrate, 7 kcal per gram for alcohol, and 9 kcal per gram for total fat and fatty acids. The total number of kilocalories from each energy source was divided by the individual energy source's total energy intake and multiplied by 100 to produce the percentage contributions. Information pertaining to specific nutrient variables is found in the Technical Notes.