The first decision regarding nutrients is whether to include negative only, positive only, or both negative and positive nutrients. The purpose and specific goals of the system may influence this decision, as discussed previously. The decision of which nutrients to include should be based on science and dietary recommendations. The scientific evidence for nutrients was described in Sections 2.4.1 and 2.4.2. Another approach to be considered is including food components or groupings, such as the percentage of whole grains or fruits and vegetables in the food.
Some decisions are needed with regard to the representation of the nutrient. For example, saturated fat may be represented in an absolute amount (grams), as a percentage of energy, or as a percentage of fat. Saturated fat as a percentage of total fat could be useful as an indicator of fat quality, in particular, for foods high in unsaturated fats such as nuts. Trans fat has most often been represented in absolute amounts, but some systems have used percentage of energy. Some systems have proposed combining saturated and trans fat. An alternative could be to use the amount of solid fats, a component in the HEI.
FOP systems may use specific foods or food groupings instead of or in addition to nutrients. For example, a system may include nuts as an indicator of a "good" unsaturated fat. One system that we reviewed, the FSA Ofcom model, uses the percentage of nuts, fruits, and vegetables as a positive component in a score (Scarborough, Boxer et al., 2007). This particular system did not include the positive nutrients that are in these foods as criteria (unsaturated fat or vitamins) but did include other positive nutrients such as fiber and protein.
Collinearity of nutrients and foods could be a consideration in the decision. Many nutrients are correlated, and inclusion of both correlated nutrients may not add predictability of the algorithm score to ranking of an overall index of dietary quality. For example, total fat and saturated fat are correlated, so including both in an algorithm may not improve the ability of the score to explain diet quality. Likewise, vitamin D is found in calcium-rich foods; thus, including both may not be necessary. This suggests use of a statistical approach to select nutrients. The scientific approach to selecting nutrients is critical; however, the science is less clear on which "positive" nutrients are most important to include in a nutritional profile of foods compared with which "negative" nutrients to include.