Procedures, Participants, and Measures
The University of Minnesota's Institutional Review Board approved this study, and all participants signed the appropriate parental consent and student assent forms. The study was conducted in three phases. First, brief field testing of the newly developed HFI was conducted with a small sample of adults (n = 5) to assess ease of completion and comprehension as the participants indicated which foods were difficult to inventory. The inventory food list was re-evaluated at this stage and foods were added if participants frequently indicated that there were foods to code without a place to do so on the inventory. Second, adults in the community were recruited to complete the home HFI in their homes while allowing trained staff members in their homes to complete the inventory independently (criterion validity testing with Sample 1). Third, parents and students were recruited for participation in the IDEA (Identifying Determinants of Eating and Activity) study [12] in which parents completed the HFI and Diet History Questionnaire (DHQ) and students participated in 24-hour recall dietary interviews (construct validity with Sample 2).
Sample 1
For the criterion validation phase, 51 adults were recruited from 19 area Minneapolis Park and Recreational Centers via posted flyers. Trained research staff traveled to the participants' homes to complete consent procedures and complete to the HFI. Although participants and staff completed inventories at the same time, they began their assessments in different parts of the home and were instructed not to communicate with each other as they completed the inventory. Participants were provided with a $30 gift card for their participation.
Sample 2
For the construct validation phase, 349 families (one student between the ages of 10 and 17 years and one parent/guardian or other adult caregiver) were recruited from the following sources: 1) an existing cohort of youth participating in the Minnesota Adolescent Community Cohort (MACC) Tobacco Study [13], 2) a Minnesota Department of Motor Vehicle (DMV) list restricted to the seven-county metro area, or 3) a convenience sample drawn from local communities. Of the 349 youth/adult pairs measured, 26% were recruited from the MACC cohort, 49% were recruited from the DMV sample and 25% were recruited from the convenience sample [12].
In the larger IDEA study, youth and adults pairs scheduled a visit to an IDEA clinic where anthropometric measures were taken and psychosocial surveys (that included demographic characteristics) were administered to both students and adults. Instructions for additional measures were given at this time and parents received a packet of instruments to take home to complete and return by mail. Included in this packet were the HFI and the DHQ. The DHQ is a food frequency that has been widely used with adults (NCI). Students were told to expect that three dietary recall interviews would be conducted with them by telephone within the next month, and were provided with a two-dimensional food model packet to help them estimate portion size. The final sample for construct validation includes data from the 342 families who completed the HFI, DHQ and dietary recalls (98% of sample).
Measures
Home Food Inventory
To develop the initial set of food items, the investigators examined existing instruments [8, 14] and reviewed the literature that identified major contributors to overall energy intake (e.g., [15]). This process allowed us to evaluate the foods listed in inventories that were developed for a specific limited purpose (e.g., a diabetic population, nonperishable foods, etc) as well as expand the items in our inventory to include foods known to be associated with energy intake in the population. In addition, based on literature that demonstrates a high correlation between readily accessible foods (i.e., foods in plain view) and their intake [3], two items were added to assess the accessibility of healthful foods within the main kitchen area and the refrigerator. Thus, we evaluated the literature and instruments to date, and added foods that provided a more comprehensive inventory of foods associated with dietary intake of adults in the US.
During the course of initial validity testing (Sample 1), changes in the number of items and instructions occurred. For example, participants were allowed to "write in" foods within given categories (e.g., dairy, fruits, vegetables), and if "write in" responses were found to be frequent, they were added as items on the inventory. During the validity testing with Sample 1, the HFI had 186 items while the final inventory administered to Sample 2 included 190 items.
HFI items are listed in a checklist type format with yes/no (1/0) response options. Higher scores represent greater availability. In addition, participants were instructed to check whether the vegetable, fruit, and bread items were fresh, frozen, dried or canned, as appropriate. The category order was set up to facilitate ease of completion, beginning with the refrigerated items, followed by frozen items, and non-perishable items. Participants are instructed to look for these foods in all areas of the home where food is stored, including the refrigerator, freezer, pantry, cupboard, and other areas (e.g., basement). Participants were informed that lower fat products may be labeled as "reduced-fat," "low-fat," "light," "nonfat," or "skim." Foods in the dairy, added fats, frozen desserts, prepared desserts, and savory snacks were categorized into regular-fat or reduced-fat groupings; beverages were categorized into regular sugar and low sugar categories; and foods in the two ready-access categories were further subgrouped into healthful and less healthful categories. Although the categorization of foods into healthful and less healthful categories may not be entirely straightforward, we assessed each food by its typical fat and sugar content when determining its category. To assess the overall obesogenic home food availability, a summative score was created that includes regular-fat versions of cheese, milk, yogurt, other dairy, frozen desserts, prepared desserts, savory snacks, added fats; regular-sugar beverages; processed meat; high-fat quick, microwavable foods; candy; access to unhealthy foods in refrigerator and kitchen. The obesogenic home food availability score potential range was from 0–71 (present sample: range = 9–53, M = 29.4, SD = 7.6). The HFI can be requested from the primary author. A table reflecting which foods are included in each food group/subgroup is provided in Additional file 1. The inventory took approximately 30–45 minutes to complete depending upon the amount of food stored in the home.
24-Hour Recall Interviews
Students in Sample 2 completed three telephone-administered 24-hour dietary recalls following their clinic visit (response rate for students completing three recalls was 86%). Dietary recalls were conducted for two weekdays and one weekend day, with the aim of having each of the three recalls completed within a 2-week period. In general, multiple dietary recalls are widely accepted as a valid and reliable method for dietary assessment, and have yielded acceptable validity in children as young as 10 years [16]. Trained and certified staff from the Nutrition Coordination Center (NCC) at the University of Minnesota administered the recalls, using the Nutrition Data System Research (NDS-R) software [2006, Nutrition Coordinating Center, University of Minnesota, Minneapolis, MN] with an interactive, interview format with direct data entry linked to a nutrient database [17]. The NDS-R data set allows for the examination of both nutrient intakes and food group information (e.g., servings of fruits and vegetables consumed).
Diet History Questionnaire
Dietary assessment for parents was conducted using the Diet History Questionnaire (DHQ) food frequency instrument developed by the National Cancer Institute (NCI). Parents received the DHQ at the clinic visit and were asked to mail it in when completed. This instrument consists of 144 food items and includes both portion size and dietary supplement questions. Requiring approximately one hour to complete, it has been widely used to characterize usual food and nutrient intakes in numerous adult populations.
Several studies have been conducted to assess the validity and calibration of the DHQ. Findings indicate that the DHQ provides reasonably valid estimates for usual intake of most nutrients and that it performs as well or better than other well-known food frequency instruments available in the field [18–20]. The food list and nutrient database used for standardized analysis of the DHQ are derived using national dietary data from the US Department of Agriculture's Continuing Survey of Food Intakes by Individuals (1994–96) [20].
Data Analysis
To assess criterion validity using data from Sample 1, participants' and research staffs' assessment of home food availability were compared. Consistent in research of criterion validity, the staff report was considered the gold standard as they were trained on how to use the inventory. Validity was evaluated by calculating kappa, sensitivity, and specificity between participant and staff reports on the presence of individual foods. To summarize these results, we calculated the average of these individual kappas across both major and minor food groupings. In addition, to test the performance of the instrument's assessment of broad food categories (rather than individual foods), we assessed the extent of agreement between food group summary scores between participant and staff reports by creating additive summary scores for major and minor food groupings (e.g., overall dairy score, cheese score, respectively). Using data from Sample 2, construct validity was assessed by examining Spearman correlations of five major food category scores on the HFI (i.e., dairy, vegetables with and without potatoes, fruit, and meats & other nondairy protein) with number of servings of the same foods as well as nutrients that should be correlated with these foods (e.g., calcium with dairy, Vitamin C with fruit) with foods and nutrients from the DHQ and 24-hour recall interviews. These five categories were chosen since it was possible to create similar categories across the three measures. In addition, we assessed construct validity of the obesogenic home food availability score by comparing it to both parental and adolescent reports of energy intake. All analyses were conducted in SAS (v9.1, SAS Institute, Inc., Cary, NC, 2003).