The following technical issues were considered critical to the development of the Healthy Eating Index:
Serving sizes used to compute the Index scores were intended to be as consistent as possible with those presented in the Food Guide Pyramid. Identification of the key underlying commodities contained in various foods and determination of appropriate serving size conversion factors provided a basis for the approach used.
Serving amounts for breads and grains were determined according to an "equivalent flour" approach. For example, the Food Guide Pyramid designates a slice of bread as one serving. A typical slice of bread contains 17 grams of flour. The number of servings for any bread on the Index was calculated based on the number of grams of flour it contained divided by the number of grams of flour contained in a slice of bread (17). Similar approaches were applied to pastas and cereal grains.
The Food Guide Pyramid counts 1/2 cup of most cooked vegetables, 1 cup of most raw leafy vegetables, and 1/2 cup of most raw nonleafy chopped vegetables as single servings. Different vegetables have different densities, resulting in different gram weights per cup or half-cup measures. Different gram/serving size factors were used to calculate the index for most vegetables, based on the weight of a cup or half-cup of the relevant commodities.
Fruits were treated similarly to vegetables. Gram/serving size factors were developed for each fruit based on the weights of the various fruit amounts.
Serving amounts for various kinds of milk and milk products were calculated based on the grams of nonfat milk solids contained in a food divided by the amount of grams of nonfat milk solids contained in 1 cup of milk (the serving size specified for milk in the Food Guide Pyramid). To determine a serving size for different cheeses, the weight of all milk products in a cheese was totaled and then divided by the conversion factor used in the Food Guide Pyramid for cheese.
Serving sizes of meats are specified in the Food Guide Pyramid in terms of 2 - 3 ounces of lean meat. The Index calculations assume a serving size of 2.5 ounces for meats. Serving size conversion factors for meats were based on the amount of fat-free meat commodity included in the database for various foods. Conversion factors for converting grams of nuts and peanut butter to serving sizes were developed according to those specified in the Food Guide Pyramid. The gram conversion factors developed were based on the weight of these quantities.
In calculating the Index, it was necessary to assign the foods in "mixtures" to their constituent food groups in the appropriate amounts. Pizza, for instance, may make significant contributions to several different food groups, including grains, vegetables, milk, and meat.
The approach used was a straight-forward extension of the approach used to estimate serving sizes. Commodity compositions of foods were identified. Once identified, commodities were assigned to appropriate food groups, based on the gram/serving size factor calculated.
Prior to scoring the first five components of the Index, it was necessary to determine the recommended numbers of servings by food group for each of the individuals on the data file. The food energy RDAs for some age/gender combinations were different from the three levels of energy intake presented in the Food Guide Pyramid. Interpolation techniques were used to estimate the required number of servings for age/gender combinations not addressed in the Food Guide Pyramid. Food servings specified in the Food Guide Pyramid for three food energy RDA levels were used as a basis for predicting comparable food servings at other food energy levels for each food group.
Two issues arose in taking this approach. Children 1 - 3 years of age have a food energy RDA less than the lowest calorie level in the Food Guide Pyramid. Extrapolation of the Food Guide Pyramid's recommended number of servings to a lower calorie level would result in smaller numbers of servings than the minimums shown. The following statement from the Food Guide Pyramid provided a basis for the technical approach taken to address the issue:
Preschool children need the same variety of foods as older family members do, but may need less than 1,600 calories. For fewer calories, they can eat smaller servings.
The number of servings for children 1 - 3 years of age was thereby held constant at the minimums shown in the Pyramid, but the serving sizes were reduced proportionately.
Similarly, males 15 - 50 years of age have food energy RDAs slightly higher than the highest calorie level in the Food Guide Pyramid. Simple extrapolation would result in larger numbers of servings than the maximum numbers shown. The text of the Food Guide Pyramid provides no guidance regarding the adjustment of numbers of servings or serving sizes to accommodate higher food energy levels. Rather than exceeding recommended serving sizes, it was decided that food servings would be truncated at the maximums shown in the Food Guide Pyramid. Preliminary analysis indicated that if serving sizes had been slightly increased, the results obtained from the Index would not have been significantly different.
The food coding structure used to compute the Index was based on USDA's coding structure for data in the 1989 and 1990 CSFII. In an effort to simplify coding, items which were similar but coded separately in CSFII were grouped together for the purposes of this study. The following principles guided the coding decisions made:
In an effort to facilitate coding, food mixtures were broken down to their constituent components. Only component foods present in substantial quantities were included in the variety calculations. A threshold of 1/2 a Food Guide Pyramid serving was used. Food components contributing less than 1/2 a Food Guide Pyramid serving were not computed in the Index variety score. It is possible that variety scores for some individuals were slightly underestimated by this approach. Several servings in a food group consumed in less than 1/2 serving amounts throughout a day could exceed the 1/2 serving threshold when added together. The potential effects of this underestimation on the final results is believed to be extremely small.
A second simplifying convention used was to assume that food mixtures containing two or more components from the same food group (e.g., mixed vegetables) could be reasonably allocated, equally, to two codes representing the components present in the highest proportions.
Foods often principally fall within one group but contain small amounts of other groups. Bread, for example, is mainly a grain but also contains small amounts of milk and egg products. An approach considered for this study was to exclude the "incidental" food group contributions from the computations. This approach was rejected based on the following rationale:
It was decided that all contributions to various food groups would be counted in computing the Index with no minimum size cutoff values imposed. The following examples illustrate some of the implications of this approach:
In developing the Index, consideration was given to including a component to address food energy intake. Obesity is a significant health problem in this country. Nevertheless, it was decided that the inclusion of physical measures of appropriate body weight, such as a body mass index (BMI) or conformance to standard weight-for-height tables, would be inappropriate since they are influenced by factors, such as physical activity, unrelated to people's eating patterns. Use of a measure based on food energy in relation to the RDA was also rejected for this study, as preliminary tabulations of the data indicated that these measures were not highly correlated with physical measures of obesity. A comparison of the BMI and caloric intake is presented in Table 10.
BMI values from 24 to 27 in women or 25 to 27 in men indicate overweight; those over 27 indicate obesity. In Table 10, persons with a BMI of greater than 30 have a level of energy intake similar to individuals with a BMI of 20 or less. Thus, BMI is not highly correlated with caloric intake.