Task 3: Key Concepts About Food Frequency Data Files

 

Food Frequency Questionnaire

A Food Frequency Questionnaire (FFQ) was used in NHANES 2003-2004 and NHANES 2005-2006 to collect information on the frequency with which selected food items were consumed during the past 12 months.  Along with the two 24-hour dietary recall interviews and interview information on dietary supplement use, food security, and dietary behavior, the FFQ completes the dietary assessment portions of the survey. The FFQ was mailed to all participants aged 2 and older who completed at least one 24-hour recall.  The purpose of the FFQ is to be used in conjunction with the 24-hour recall data to model usual intake.

The NHANES Food Frequency Questionnaire asks participants about their consumption of more than 100 different food items.  Specifically, there are 151 frequency questions, which include follow-up questions querying intake of certain foods over two seasons (i.e. summer OR winter versus rest of year).  The FFQ also includes follow-up questions asking about the proportion of the time certain subtypes of the food were eaten over the past 12 months.  In contrast to some other food frequency questionnaires, the NHANES Food Frequency Questionnaire does not query portion size.

A sample FFQ question for soft drink consumption is listed below.  Notice that the question includes an introductory “stem” question to ascertain if the food was consumed in the past 12 months.  If the participant answered “No” they are directed to the next question.  If the participant answers affirmatively, the stem branches to additional questions about seasonal soda consumption and types of soda (e.g., diet soda and caffeine-free soda).  The resulting data files provide frequency information for all these questions which can be collapsed to examine intake for all soda combined, or by type.

 

Example of Food Frequency Questionnaire  

 

Food Frequency Questionnaire Data Files

These data are captured in four files, found on the 2003-2004 Dietary Files page. 

 

 

The FFQ Questionnaire File, which provides the reported “raw” responses to every question on the FFQ, contains one record per person.  Each record contains data for all questions on the FFQ, and data are coded using a separate variable for each question.  The variable naming convention is FFQXXXXX, where XXXXX equals the question number (e.g., FFQ0009C corresponds to Question 9c, which is “How often were these soft drinks, soda, or pop diet or sugar-free?”).  Data in the Questionnaire File are not edited.  The FFQ Questionnaire file is rarely used; rather, analysts of FFQ data almost always use values from the FFQ Daily Frequency Covariates File.  On the rare occasion that one might wish to combine data from these files, they can be linked by the individual ID number, SEQN. 

The FFQ Daily Frequency Covariates File contains multiple records per person; there is a separate record for each food subtype/season combination.  NCI’s DietCalc software applies algorithms to the raw frequency responses from the FFQ Questionnaire File to convert them into average daily frequencies over the past year (for more details on these algorithms, please see the FFQ documentation on the NHANES Web site).  The DietCalc software also imputes data in cases where there are inconsistent results for the stem and follow-up questions (e.g. when a participant reports that they never ate a certain food, but provided answers to the follow-up questions). 

Average daily frequencies are indexed by two variables; these two variables are:

• FFQ_VAR: indexes food by season (i.e. in summer/winter OR “in season” versus rest of the year).
• FFQ_FOOD:  indexes subtypes of food, such as different types of beverages. 

The FFQ Daily Frequency Covariates file contains a separate record for each subtype and season combination of each food asked on the FFQ.  Each record designates the season (using FFQ_VAR), the subtype (using FFQ_FOOD), and the contribution of that subtype/season combination to the daily frequency over the year(using FFQ_FREQ).  For foods that are not queried by season, FFQ_FREQ represents the average daily frequency of consumption of that food.  When foods are queried by season and/or subtype, frequencies are presented in such a way that they can be summed to get the average frequency of consumption.  These frequencies were weighted using algorithms to take into account nonequivalent lengths of time (e.g. “summer” is considered to be 4 months whereas “rest of year” is 8 months) so that summing the frequencies provides the overall daily frequency over the past year.  Therefore, frequencies for a food or beverage consumed in a particular season do not accurately represent daily frequencies for that season.  Instead, when foods are queried by season, it is necessary to sum by FFQ_FOOD (i.e. foods with the same value for FFQ_FOOD) to capture the contributions of the different seasons. 

 The last two files listed above, FFQ Variable Look-up File (VARLOOK) and FFQ Food Look-up File (FOODLOOK), are used to cross-reference numerical codes in the FFQ Daily Frequency Covariates File to corresponding food descriptions.  VARLOOK provides descriptions for FFQ_VAR codes and FOODLOOK provides descriptions for FFQ_FOOD codes.  These files can be merged with the FFQ Daily Frequency Covariates File using the sample SAS program (FFQmerge.SAS) available on the Downloads page.

A complicated example of using the FFQ data occurs with soda/soft drinks, for which participants are asked to report intake by season, and by subtypes of regular/diet or with/without caffeine.  As shown below in the snippet of output for Participant Number 21007, in order to determine the average daily frequency of intake of diet soft drinks, you would need to sum all of the FFQ_FREQ values designated by FFQ_FOOD values of 182 or 183 (shown in yellow).  In this case, you would be summing by values of FFQ_FOOD irrespective of what time of year the diet soft drinks were consumed.  Thus, participant 21007 consumed diet soft drinks an average of 0.35 times per day, or about once every three days.  Similarly, if you were to add the FFQ_FREQ values designated by FFQ_FOOD values of 180 or 181 (shaded in orange), you would get the average daily frequency of regular soft drink consumption.  Participant 21007 consumed regular soft drinks an average of 0.35 times per day, as well. 

It would also be possible to determine how many times per day this participant consumed regular caffeinated soft drinks or regular decaffeinated soft drinks by summing rows where FFQ_FOOD is 180 or 181, respectively.  Furthermore, you could determine how many times per day participant 21007 consumed all types of caffeinated soda or all types of decaffeinated soda by summing rows where FFQ_FOOD is 180 & 182 OR 181&183, respectively.      

 

FFQ_FREQ Output for Respondent Sequence Number 21007

FFQ_FREQ