Learn More about Scoring Algorithms for Screeners
The general term [glossary term:] "scoring algorithms" may refer to a variety of analytical procedures. We use the term narrowly to describe analytical procedures using a less [glossary term:] biased source of dietary data to re-scale data collected using [glossary term:] short-term instruments, or screeners (see Screeners Profile) developed by the National Cancer Institute (NCI). Scoring algorithms produce [glossary term:] mean intake estimates that are closer to those estimated from [glossary term:] 24-hour dietary recalls (24HRs) than are estimates using raw frequency estimates (see 24-Hour Dietary Recall Profile).
Scoring algorithms that re-scale self-reported information from dietary screeners to estimates of dietary intake are available for many NCI-developed instruments. Until the 2009-2010 [glossary term:] National Health and Nutrition Examination Survey (NHANES), when a screener was administered for the first time, these scoring algorithms were developed using data collected on 24HRs. Because screener data were not available, it was necessary to construct a "pseudo-screener" to develop the screening algorithms. This was done by using some of the information from the 24HRs to reflect the frequency information a screener would have provided:
- First, individual foods reported from the 24HRs in the survey were grouped into food groups that mirrored food groups queried on a screener. For example, for the screener question about "fruits," all fruits were classified into the fruit food group.
- Then, for each food group, the number of mentions of each food group for each individual was computed (and converted to times per day), as a proxy for frequency from the screener.
- Also, because most NCI screeners do not ask about portion size, the 24HRs were used to estimate median portion sizes for each of the food groups for 18 different sex/age groups.
- The count and portion size variables were used as [glossary term:] independent variables in a regression analysis to predict the [glossary term:] dependent variable, [glossary term:] total intake of each dietary component. This regression reflects the relationship between the pseudo-screener and the dependent variable.
All of the information above was used to develop scoring algorithms that allow frequency information from a study to be converted to dietary intake estimates. These scoring algorithms:
- Convert reported frequencies to times per day;
- For screeners that do not ask portion size, estimate portion sizes for each food group, based on the respondent's sex and age; and
- Apply regression equations appropriate to each dietary constituent.
NCI screeners and their scoring algorithms are available on the NCI website to estimate fruits and vegetables (cup equivalents), whole grains (ounce equivalents), dairy (cup equivalents), added sugars (teaspoons), percentage energy from fat, fiber (gm), and calcium (mg).
For More Information
National Cancer Institute, Applied Research. Short dietary assessment instruments.
National Cancer Institute, Applied Research. Dietary screener questionnaire in the NHIS Cancer Control Supplement (CCS) 2010: Data Processing and Scoring Procedures. (Accessed 6/30/14).
Thompson FE, Subar AF, Smith AF, Midthune D, Radimer KL, Kahle LL, Kipnis V. Fruit and vegetable assessment: performance of 2 new short instruments and a food frequency questionnaire. J Am Diet Assoc 2002 Dec;102(12):1764-72. [View Abstract]