# Developing Scoring Algorithms

We developed scoring procedures to convert screener responses to estimates of individual dietary intake for fruits and vegetables (cup equivalents), dairy (cup equivalents), added sugars (tsp), whole grains (ounce equivalents), fiber (g), and calcium (mg) using the What We Eat in America 24-hour dietary recall data from the 2003-2006 NHANES. The following equations were estimated in the NHANES 2003-2006, using SAS PROC REG.

### For cup equivalents of fruits and vegetables:

E ([Fruits and Veg]1/2) = b0+ b1([NFG1P1 + NFG2P2+... + NFG10P10]1/2)

Cup equivalents of fruits and vegetables were square-root-transformed to approximate normality; NFGk is the usual number of times per day an individual consumed food group k; Pk is the median portion size in fruit and vegetable cup equivalents of group k; and k indexes the ten fruit and vegetable food groups. We calculated weighted least-squares estimates of the regression coefficients b0 and b1 in the NHANES 2003-2006 samples, stratifying by sex and age group (ages 2-17 years; 18+), and excluding extreme exposure values.

### For cup equivalents of dairy:

E ([Dairy]1/2) = b0+ b1([NFG1P1+ NFG2P2+...+ NFG4P4] 1/2)

Cup equivalents of dairy were square-root-transformed to approximate normality; NFGk is the usual number of times per day an individual consumed food group k; Pk is the median portion size in dairy cup equivalents of group k; and k indexes the four dairy food groups. We calculated weighted least-squares estimates of the regression coefficients b0 and b1 on the NHANES 2003-2006 samples, stratifying by sex and age group (ages 2-17 years; 18+), and excluding extreme exposure values.

### For teaspoons of added sugars, not including cereals:

E ([Added Sugars].33) = b0+ b1([NFG1P1+ NFG2P2+...+ NFG7P 7].33)

Teaspoons of added sugars were cube-root-transformed to approximate normality; NFGk is the usual number of times per day that an individual consumed food group k; Pk is the median portion size in added sugars teaspoons of group k; and k indexes the six or seven added sugars food groups.

The item sugar in coffee/tea was not asked of children younger than age 12 years in NHANES. Thus, separate algorithms, not including the sugar in coffee/tea item, were estimated for children less than age 12 years.

We calculated weighted least-squares estimates of the regression coefficients b0 and b1 in the NHANES 2003-2006 samples, stratifying by sex and age group (ages 2-11 years; 12-17; 18+), and excluding extreme exposure values.

Users of the DSQ may prefer to include the sugar in coffee/tea item for children. If so, another algorithm is available that provides regression coefficients for all children and adolescents (ages 2-17 years).

We estimated a second version of added sugars that include cereals.

### For teaspoons of added sugars including cereals:

E ([Added Sugars].33) = b0+ b1([NFG1P1+ NFG2P2+...+ NFG13P13].33)

Teaspoons of added sugars were cube-root-transformed to approximate normality; NFGk is the usual number of times per day that an individual consumed food group k; Pk is the median portion size in added sugars teaspoons of group k; and k indexes the twelve or thirteen added sugars food groups: hot cereals (2 groups), cold cereals (4 groups), and the six or seven added sugars food groups.

The item sugar in coffee/tea was not asked of children younger than age 12 years in NHANES. Thus, separate algorithms, not including the sugar in coffee/tea item, were estimated for children younger than age 12 years.

We calculated weighted least-squares estimates of the regression coefficients b0 and b1 in the NHANES 2003-2006 samples, stratifying by sex and age group (ages 2-11 years; 12-17; 18+), and excluding extreme exposure values.

Users of the DSQ may prefer to include the sugar in coffee/tea item for children. If so, another algorithm is available that provides regression coefficients for all children and adolescents (ages 2-17 years).

### For teaspoons of added sugars from sugar-sweetened beverages:

The dependent variable added sugars from sugar-sweetened beverages included the following beverages: sodas, fruitades/sports drinks, and sugar in coffee/tea. Teaspoons of added sugars were square-root-transformed to approximate normality; NFGk is the usual number of times per day that an individual consumed food group k; Pk is the median portion size in added sugars teaspoons of group k; and k indexes the two or three sugar sweetened beverage food groups.

The item sugar in coffee/tea was not asked of children younger than age 12 years in NHANES. Thus, separate algorithms, not including the sugar in coffee/tea item, were estimated for children less than age 12 years.

We calculated weighted least-squares estimates of the regression coefficients b0 and b1 in the NHANES 2003-2006 samples, stratifying by sex and age group (ages 2-11 years; 12-17; 18+), and excluding extreme exposure values.

Users of the DSQ may prefer to include the sugar in coffee/tea item for children. If so, another algorithm is available that provides regression coefficients for all children and adolescents (ages 2-17 years).

### For ounce-equivalents of whole grains:

E ([Whole Grains] 1/2) = b0 + b1 (NFG1P1) + b2(NFG2P2 )+ ... + b9(NFG9P9)

Ounce-equivalents of whole grains were square-root-transformed to approximate normality; NFGk is the usual number of times per day an individual consumed food group k; Pk is the median portion size in grams of group k; and k indexes the nine whole grain food groups: hot cereals (2 groups), cold cereals (4 groups), brown rice, whole grain bread, and popcorn.

We calculated weighted least-squares estimates of the regression coefficients b0 and bk, k = 0, ..., 9 on the NHANES 2003-2006 samples, stratifying by sex and age group (ages 2-17 years; 18+), and excluding extreme exposure values.

### For grams of fiber:

E ([Fiber] 1/4) = b0 + b1(NFG1P1) + b2(NFG2P2 )+ ... + b29(NFG29P29)

Grams of fiber were quarter-root-transformed to approximate normality; NFGk is the usual number of times per day an individual consumed food group k; Pk is the median portion size in grams of group k; and k indexes the 29 food groups, which includes three hot cereal and four cold cereal variables. Two food groups, tomato sauce and dried beans, were frequently consumed in mixed dishes. Because we wanted to represent only the particular food, we estimated the number of grams of tomatoes and legumes, respectively, in these mixed dishes. For tomato sauce, this was done by multiplying the MPED 1 cup equivalent by 171 (1 MPED = 171 grams). For legumes, this was done by estimating the MPED one cup equivalent for legumes and multiplying by 212 (1 MPED = 212 grams).

The item sugar in coffee/tea was not asked of children less than 12 years of age in the NHANES. Thus, separate algorithms, not including the sugar in coffee/tea item, were estimated for children less than 12 years.

We calculated weighted least-squares estimates of the regression coefficients b0 and bk, k = 0, ... , 29 (or 28, when sugar in coffee/tea was not asked), stratifying by sex and age group (ages 2-11 years, 12-17 years; 18+), and excluding extreme exposure values. We first included all 29 (28) food groups in the regression model. After examining the results, we dropped food groups that failed to attain statistical significance at α = 0.25 to form more parsimonious final models. In the fiber model, milk, fruitades/sports drinks, and salad were dropped for boys ages 2-11 years; sugar in coffee/tea was dropped for boys ages 12-17 years; soda and sugar in coffee/tea were dropped for men; brown rice, milk, ice cream, fruitades/sports drinks, candy, and salad were dropped for girls ages 2-11 years; sugar in coffee/tea was dropped for girls ages 12-17 years; and salad and sugar in coffee/tea were dropped for women.

Users of the DSQ may prefer to include the sugar in coffee/tea item for children. If so, another algorithm is available that provides regression coefficients for all children and adolescents (ages 2-17 years). For children and adolescents combined, salad and candy were excluded for both boys and girls.

### For milligrams of calcium:

E ([Calcium] 1/4 ) = b0 + b1 (NFG1P1) + b2(NFG2P2 )+ ... + b29(NFG29P 29)

Milligrams of calcium were quarter-root-transformed to approximate normality; NFGk is the usual number of times per day an individual consumed food group k; Pk is the median portion size in grams of group k; and k indexes the 29 food groups, which includes three hot cereal and four cold cereal variables. Two food groups, tomato sauce and dried beans, were frequently consumed in mixed dishes. Because we wanted to represent only the particular food, we estimated the number of grams of tomatoes and legumes, respectively, in these mixed dishes. For tomato sauce, this was done by multiplying the MPED 1 cup equivalent by 171 (1 MPED = 171 grams). For legumes, this was done by estimating the MPED one cup equivalent for legumes and multiplying by 212 (1 MPED = 212 grams).

The item sugar in coffee/tea was not asked of children less than 12 years of age in the NHANES. Thus, separate algorithms, not including the sugar in coffee/tea item, were estimated for children less than 12 years.

We calculated weighted least-squares estimates of the regression coefficients b0 and bk, k = 0, ... , 29 (or 28) on the NHANES 2003-2006 samples, stratifying by sex and age group (ages 2-11 years, 12-17 years; 18+), and excluding extreme exposure values. We first included all 29 (28) food groups in the regression model. After examining the results, we dropped food groups that failed to attain statistical significance at α = 0.25 to form more parsimonious final models. In the calcium model, doughnuts, salad, and French fries were dropped for boys ages 2-11; sugar in coffee/tea, fruit, and other potatoes were dropped for boys ages 12-17; popcorn, brown rice, soda, sugar in coffee/tea, and other potatoes were dropped for men; fruitades/sports drinks, candy, fruit, other potatoes, and cookies/cake/pie were dropped for girls ages 2-11; popcorn, candy, other potatoes, French fries, fruitades/sports drinks, and tomato sauce were dropped for girls ages 12-17; and popcorn, brown rice, fruitades/sports drinks, doughnuts, and tomato sauce were dropped for women.

Users of the DSQ may prefer to include the sugar in coffee/tea item for children. If so, another algorithm is available that provides regression coefficients for all children and adolescents (ages 2-17 years). For children and adolescents combined, popcorn, sugar in coffee/tea, fruit, and other potatoes were excluded for boys; popcorn, doughnuts, candy, other potatoes, and fruitades/sports drinks were excluded for girls.