Participants and setting
Participants were parents (93 %) or other adult primary caregiving relatives (7 %; all called “parents” in this paper) of Medicaid-insured obese children. Participants were recruited from a large urban academic hospital-based pediatric primary care clinic and were enrolled with their obese child (body mass index (BMI) ≥95th percentile for age and gender) ages 2 to 11 years in a randomized clinical pediatric obesity trial targeting parents. The University of Pittsburgh Institutional Review Board approved the study that was carried out in 2010 in compliance with HIPAA guidelines. All participants provided written informed consent. One study participant, a recent immigrant, was excluded from the assessment because he did not recognize many foods queried in the assessment.
Study design and measures
Nutrition knowledge was assessed at baseline with a novel instrument, the Nutrition Knowledge Grid-Basic (NKG-Basic), developed by our research group, that assesses participants’ ability to accurately identify the food group and relative nutritional status of common foods. The NKG-Basic requires little reading ability and is presented visually as a grid formed by columns that display eight food groups and rows that represent three nutritional categories. Stickers portraying 23 common foods and beverages are available to be placed by participants in the grid cell that most accurately describes each food’s group and category (see Fig. ).
Fig 1 aColumn labels in brackets are for readability in the journal article only; actual NKG-Basic grid is 11×14 in., and labels are easily visible. b Sticker sizes relative to grid are enlarged for this figure
The instrument was developed using food groups from the Food Guide Pyramid [35
] and the National Heart Lung and Blood Institute’s WeCan
! categorization of foods into Go (high nutrition/low caloric content), Slow (good nutrition/moderate caloric content), and Whoa (low nutrition/high caloric content) (www.nhlbi.nih.gov/health/public/heart/obesity/wecan/
) nutritional categories. Columns display traditional food groups in six columns plus two additional columns, added because of our interest in assessing parents’ understanding of healthier vs. less healthy beverages and sweets/snacks. The additional snack and beverage groups are relevant because overconsumption of high calorie-low nutrition beverages is implicated in the development and maintenance of pediatric obesity, particularly in our study population (e.g., [36
]), and because parental understanding of energy-dense vs. nutrient-dense snacks is a key component in decreasing overall caloric intake in most pediatric obesity interventions (e.g., [37
]). Rows on the NKG-Basic grid display nutritional categories (Go, Slow, and Whoa). Food stickers were developed in consultation with a registered dietitian who has significant experience preparing nutrition education programs and materials for low-income groups in our region. The stickers represent a broad spectrum of healthy and not-so-healthy foods typically consumed by the residents of our geographical area who are primarily white (66.5 %) or black (25.8 %) and non-Hispanic (97.7 %) [38
The column and row format of the NKG-Basic was employed to allow for separate assessment of participants’ knowledge of each food’s group and of its nutritional category (e.g., participants can place a sticker in the correct column, but the incorrect row, indicating they know the food’s group, but do not understand its nutritional category) for each food presented. All food groups and nutritional categories are represented among the 23 foods.
The instrument is scored by assigning a binary accuracy score (1 = correct; 0 = incorrect) for the food group and, separately, for the nutritional category of each food sticker (excluding the broccoli and potato chip stickers that were used as examples). Six foods are scored as correct if placed in either of two food group columns (100 % orange juice is correct if identified as a fruit or beverage; skim, 2 % and whole milk are correct as both milk/milk like foods and beverages; doughnut as both a bread/cereal and sweets/snack; and tomato as both a fruit and vegetable). This scoring method resulted in a total of 27 possible correct food group placements (five vegetables, five fruits, two breads/cereals, three dairy, four meats and other proteins, one fat, two sweets/snacks, five beverages), and 21 correct nutrition category placements (eight Go, six Slow, seven Whoa foods).
Accuracy of food group and nutritional category identification was determined by dividing the number of correctly identified foods by the total number of foods in each group or category. For example, for food groups, because there were five vegetables, possible accuracy values were 0, 20, 40, 60, 80, and 100 % correct, whereas because there was only one fat, possible accuracy values were 0 and 100 %.
A trained research staff person began each assessment by orienting the participant to the instrument. Displaying the grid and stickers, the staff person said, “We want to understand how much families know about healthy and not-so-healthy foods.” The staff person explained the NKG-Basic column (food groups) and row (nutritional categories) headings and gave a brief description of what makes a food Go, Slow or Whoa (amount of calories compared with the amount of nutrition provided by the food; see row labels on Fig. ). The participant was then asked to “place each food sticker in the right box that tells its food group and whether it is a Go, Slow, or Whoa food.” Then, as an example, the researcher demonstrated how to place the broccoli sticker by talking the participant through the process of deciding the food group and nutritional category of broccoli and, finally, saying, “That’s right. You put the broccoli sticker under vegetables in the Go row,” and asking the participant to affix the sticker in the correct cell. The participant then completed an example by her/himself with the potato chips sticker and received correction and assistance if needed. If the participant had difficulty placing the chip sticker accurately, the procedure was completed a third time with the water sticker; this occurred only once in 135 assessments.
When the participant understood the use of the grid, the researcher instructed him/her to place the remaining stickers in the correct “boxes.” Foods that might be ambiguous from their pictures because of food preparation factors were clarified as the participant worked (e.g.: “That’s baked ham”; “Assume the corn-on-the-cob is plain with no butter on it”; “Ignore the bun and mustard; just focus on the hot dog”; “Those are hard-boiled eggs.”). No feedback about performance was provided to participants as they completed the task.
Participants completed a self-report, self-administered demographic questionnaire on which they checked off their employment status, household income, and race from multiple choice lists of categories; ethnicity was recorded separately as Hispanic or Non-Hispanic. Age was determined from reported date of birth. Weight and height of parent participants and their participating child were measured on a digital scale and research-quality portable stadiometer, respectively, by trained research staff in the research office that was housed within the pediatric clinic. BMI was calculated using the formula: weight in kilograms/(height in meters)2
Descriptive statistics appropriate to the distribution of the data were used to characterize the sample and their knowledge of food groups and nutritional status. Kruskal–Wallis one way analysis of variance tested whether categorical participant characteristics (i.e., race, employment status, homemaker status, and household income) were related to nutrition knowledge (i.e., percentage of correctly identified nutritional groups and categories). The Jonckheere–Terpstra trend test tested whether ordinal participant characteristics (i.e., education) were related to percentages of correct responses. Spearman correlation tested for correlations between continuous participant characteristics (i.e., age and BMI) and percentages of correct responses. Multivariable linear regression was used to determine whether age, race, education, employment status, and income were independently related to correct identification of nutritional categories, controlling for gender. Initial analyses with race, education, employment status and income led to collapsing categories when relationships did not differ significantly between all categories. Variables other than gender that were not significant in the model (i.e., p
0.05) were removed by using backward elimination. The characteristics of participants’ children who were enrolled with them in the intervention trial were not included in the multivariate analysis, as they were not hypothesized to be predictive
of parental nutrition knowledge.