Child assessments of vegetable preferences and cooking self-efficacy show predictive validity with targeted diet quality measures

The diet quality of U.S. children remains an important public health concern. Even though solid fat and added sugar intakes have significantly decreased from 2003 to 2014 among 6–11 year olds [1], the vast majority of American children exceed the recommended intake for empty calories and fail to meet the recommendations for total vegetables, total fruit and dark green vegetables [2]. The 2015 Dietary Guidelines for Americans encourages increasing vegetable and fruit intake and reducing the consumption of solid fats and added sugars. Furthermore, higher intakes of vegetables and fruit were the only diet characteristics that were consistently identified in the Advisory Committee’s investigation of the associations between dietary patterns and key beneficial health outcomes in the 2015 Dietary Guidelines for Americans Advisory Report [3].

Cooking interventions have the potential to improve child diet quality because cooking involvement is associated with positive changes in dietary behavior [4–8]. Canadian children who cook at home consume more fruits and vegetables, on average, compared to children who are not involved with cooking at home [5]. Children who participated in the LA Sprouts cooking and gardening intervention increased their fiber intake by 22%, compared to a 12% reduction among controls (p = 0.04) [9]. In the UK, 9 to 11 year old children had increased vegetable consumption and cooking confidence after participating in a culinary education intervention [4].

Valid and reliable instruments are required to accurately assess the impact of cooking interventions on dietary behavior and diet quality. A variety of dietary assessment tools can be used to comprehensively and quantitatively evaluate food and beverage intake but many of these measures are resource intensive [10] and often too expensive especially for larger samples [11]. The 24-h recall is considered the least biased of the self-reported methods [12] and is generally considered the best method for assessing children’s diets. Regardless of the method, the challenges and cost associated with detailed quantitative measures of child diet to assess diet quality suggest a need for more feasible proxy measures that could be administered in studies with large samples. Any proxy measure of child diet quality would need to validated relative to gold-standard reference methods such as multiple pass 24-h recalls [4]. Criterion-related validity assesses how closely scores from a new measure relate to the scores from the reference measure. Predictive validity, one form of criterion validity, is used when scores from the new measure are used to predict the scores on a criterion measure that is administered at a later point in time, rather than concurrently [13].

Fruit and vegetable preference [6, 8, 14, 15], cooking attitudes [14, 15], and cooking self-efficacy [4, 6, 14, 15] are common outcome measures for child cooking interventions that can be feasibly collected using validated and reliable tools and would be appropriate to test as potential proxies for diet quality. According to research completed in 1996 by Domel et al and 1997 by Resnicow et al, fruit and vegetable preferences significantly predict fruit and vegetable consumption, [16, 17]. No studies have examined the predictive validity of fruit and vegetable preference for diet quality relative to 24 h dietary recalls, and the relationships between cooking attitudes and cooking self-efficacies and diet quality are not well understood.

The purpose of this cross-sectional study was to examine the validity of fruit and vegetable preferences (FP, VP, FVP), cooking attitudes (AT) and self-efficacy (SE) assessments to predict diet quality among Colorado 4th-graders participating in Fuel for Fun: Cooking with Kids Plus Parents and Play (FFF).

The VP and SE assessments utilized in FFF are consistent with dietary quality as assessed by HEI indicators generated via multiple pass 24-h recalls, demonstrating predictive validity of these Fuel for Fun outcome measures. These results also contribute to the evidence of the positive relationship between child dietary preferences and dietary behaviors. This link was previously demonstrated based upon estimates of intakes derived from 7-day food records [16, 17] and food frequency questionnaires [23]; the current study arrived at similar conclusions using a more robust diet assessment technique. However, previous studies established relationships between both fruit and vegetable preferences and consumption behaviors [16, 17, 23]. Yet, the current study found a significant relationship between vegetable preference and overall HEI score, but fruit preference did not significantly predict any of the targeted HEI outcomes. This may be because of children’s inherent tendency for high fruit preferences, [24–26] which may limit the ability of fruit preference to be a proxy for overall diet quality.

Our findings that cooking SE predicts Total HEI, Total Vegetable and Whole Fruit scores extends the authors’ conclusions from a previous study, that reported prior cooking involvement predicts changes in fruit and vegetable preferences after a cooking intervention [15]. Additionally, the relationship between cooking SE and diet quality is consistent with the results of Caraher and colleagues and Jarpe-Ratner and colleagues. Both studies concluded that cooking SE and vegetable consumption significantly increased after school-based cooking interventions, but both of these studies used a screener rather than dietary recalls to measure vegetable intake [4, 6]. In the current study, we estimate that for every 3 point increase in cooking self-efficacy we expect a 82.5% increase in total vegetable score (OR = 1.222, 95% CI: 1.021–1.464), suggesting a positive relationship between child cooking self-efficacy and meeting the dietary guidelines for total vegetable consumption. However, the study design does not establish the temporality of this relationship.

Additional limitations include a relatively small sample size and challenges with dietary assessments. Domel et al’s 1996 assessment and Resnicow et al’s 1997 assessment of the relationship between FVP and fruit and vegetable consumption featured 392 and 1398 participants, respectively. However, the dietary intake of the 108 students included in our study was similar to national averages [21], suggesting generalizability of our results. The challenges of child dietary assessment have been previously discussed [10, 27, 28]. In addition, child diet assessment methods, including 24-h recalls, have been infrequently validated [27]. Having a parent present during the 24-h recall telephone interviews may have imparted some social desirability bias, but parent influence was helpful in reminding their children of forgotten foods and portion sizes. The 24-h, multiple pass recall method is the best diet assessment available, but the method has inherent limitations, such as systematic and random error [12]. However, all students in the sample had two or more dietary recalls, reducing the influence of the random error associated with day-to-day variation in dietary intake [12]. Dietary recall invitations were planned so that diet assessments included weekday as well as weekend day intake. Both weekday and weekend day recalls occurred for 73% of participants. However, for any of the targeted HEI outcomes or FFF survey measure predictor variables the differences were not significant between children who had heterogeneity in day type (i.e. recalls included both weekend and weekdays) vs. those with only weekday or weekend day recalls. Lastly, the study sample had limited racial or ethnic diversity. Future studies are needed to assess the predictive validity of these measures in more diverse populations.

This study has several strengths. The majority of results were robust across two different analyses methods. Our sample had a slightly higher mean HEI (62.5 + 10.7) and Green Vegetables and Beans HEI component score (1.3 + 1.7) compared to the national average (55.1 + 0.72 and 07.0 + 0.09, respectively), but total vegetable and whole fruit scores were very similar to the national average [21]. In addition, we used the multiple pass interviewer-administered 24-h dietary recall method conducted by a skilled dietary assessment center, and the parents were present for most of the dietary assessments. The FVP, AT, and SE assessment tools are reliable and have been previously validated.

The study findings have important implications. The predictive validity of the FFF outcome measurements suggests that using these VP and SE assessments may be particularly useful to researchers and practitioners when indicators of diet quality are needed but constraints of time or cost abrogate the use of dietary assessment methods. The positive, predictive relationship between cooking self-efficacy and diet quality also supports the use of cooking interventions to improve overall diet quality, vegetable consumption, and whole fruit consumption. Similarly, the significant relationship between vegetable preference and overall diet quality suggests that interventions to improve child vegetable preferences are warranted.

This study provides evidence that low-cost measures of vegetable preferences and cooking self-efficacy predict diet quality in 4th grade children. These easy to administer, valid and reliable instruments encompass constructs common to many cooking interventions, increasing their value as evaluation tools. These results also reinforce the relationship between cooking and healthful dietary behavior.