Improving Toddlers’ Healthy Eating Habits and Self-regulation: A Randomized Controlled Trial

Participants

To be eligible for this trial, families had to meet the following requirements: (1) have a toddler 18 to 36 months old; (2) be enrolled in home-based Early Head Start with 1 of our 3 community partners during the 3-month recruitment period; and (3) be able to complete assessments in English. To qualify for Early Head Start, most families had to have incomes below the federal poverty threshold. As presented in Fig 2, 152 families were eligible for this trial; 73 families enrolled; and 66 were retained through posttreatment.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2

Consolidated Standards of Reporting Trials 2010 flow diagram.

Approximately 48% of families were non-Hispanic white; 29% were Black; and 23% were Hispanic or Latinx. Approximately 78% of families were enrolled in the Special Supplemental Nutrition Program for Women, Infants, and Children or the Supplemental Nutrition Assistance Program. Almost all parents were biological mothers; 61% were single; 29% had not graduated from high school; only 11% were employed full-time. On average, toddlers were 30.72 months old at trial entry (SD = 6.96); 44% were female.

Randomization and Masking

Home visitors asked parents if they were interested in learning about the trial. If families were interested, project interviewers explained everything in detail, including consent.

Once baseline assessments were complete, families were randomly assigned to trial conditions with an intended allocation of 1:1, blocked within Early Head Start centers, and based on a coin toss performed by the first author. Families received Recipe 4 Success, delivered as part of Early Head Start home visits, or usual practice Early Head Start home visits, which typically follow the evidence-based Parents as Teachers curriculum.¹⁸ Almost all home visitors had families in both conditions. Neither participating families nor home visitors could be blind to condition, but all project staff members who collected and coded data were.

Intervention Procedure

Recipe 4 Success consisted of 10 consecutive weekly lessons, implemented by families’ regular Early Head Start home visitors. The lessons required ∼45 minutes of the 90-minute home visits. Recipe 4 Success provided an alternative means of delivering Early Head Start–mandated content that promotes parents’ ability to support toddlers’ cognitive, social-emotional, language, pre-literacy, numeracy, and physical development. Families in Recipe 4 Success participated in different kinds of lessons for the intervention period but did not receive extra or longer home visits. There were no known harms or unintended effects of Recipe 4 Success; however, if families missed critical material from usual practice Early Head Start, they would have received that material once the trial ended.

Each Recipe 4 Success lesson included some didactic information, such as how toddlers often need to be exposed to a new food 10 to 20 times before they will eat it or how poor sleep can undermine emotional and/or behavioral control and lead to weight gain. However, most of each lesson was focused on active coaching of carefully structured and sequenced food preparation activities involving 3 to 6 ingredients we provided (adjusted as necessary to accommodate food allergies or limited cooking equipment and/or facilities). Toddlers are more likely to eat something they help make.¹⁹ In addition, food preparation provides multiple opportunities for toddlers to practice self-regulation, such as waiting patiently while ingredients cook, and for parents to practice sensitive scaffolding based on techniques used in parent-child interaction therapy,²⁰ such as behavior descriptions and specific praise. Lessons were tailored to toddlers’ skill levels. (Even the youngest toddlers could place a piece of cheese on a tortilla, stir dry ingredients in a bowl, or peel a banana.) As part of each lesson, home visitors helped parents identify opportunities for skill generalization in other contexts.

At the end of each lesson, home visitors rated 15 items assessing implementation fidelity, including “I was able to complete this lesson as planned,” with 1 = not true and 4 = very true and an average score of 3.53 (SD = 0.52). At the end of each lesson, parents rated 8 items assessing therapeutic engagement, including “My child and I made a good team during the cooking activity today,” with the same response options and an average score of 3.72 (SD = 0.43).

Before implementation, home visitors completed one day of training and received a detailed manual. Throughout the trial, home visitors continued to meet with their Early Head Start supervisors but were also encouraged to consult with trial investigators on regularly scheduled conference calls or by e-mail. The cost of providing Early Head Start is $8000 to $15 000 per year per child, depending on state of residence and the mix of center- versus home-based services.²¹ The additional one-time cost of Recipe 4 Success, provided within Early Head Start home visits, was <$300 per family: $100 for food, <$100 for manuals and intervention materials, and <$100 for group trainings and implementation support. All other costs associated with Recipe 4 Success, such as home visitor salaries, were fixed and incurred regardless of curriculum used.

Assessment Protocol

The same set of measures were collected at baseline, ideally one week before the first lesson, and posttreatment, ideally one week after the last lesson. There were no changes to primary outcomes after the trial began.

Project interviewers conducted in-home assessments, which lasted ∼90 minutes. All interviewers had experience from previous research projects, completed training on administering measures for this trial, and were closely supervised by the Pennsylvania State University Survey Research Center.

Undergraduate research assistants coded all video recordings of parent–child interactions. Research assistants completed extensive training on the rating scales until they were reliable with a master coder (intraclass correlation coefficient > 0.80). They also double-coded 30% of all videos to ensure they were reliable with each other (intraclass correlation coefficient = 0.79).

Outcome Measures

Toddlers’ healthy eating habits were assessed by parent recall of what food and drinks toddlers consumed in the previous 24 hours. Interviewers collected these recalls on 3 separate occasions, once in person and twice over the telephone, usually within the same week and never including intervention days. These recalls were structured so parents were separately asked about breakfast, lunch, and dinner, and morning, afternoon, and evening snacks, as well as anything else toddlers ate or drank. To summarize overall healthy eating habits, we computed the percentage of all meals and snacks that included: (1) a fruit and/or vegetable, (2) a source of protein, and (3) no sweets and/or junk food, including ice cream, potato chips, and soda.

Toddlers’ self-regulation was assessed with 3 measures tapping different dimensions of the umbrella construct: delay of gratification, task orientation, and emotional and/or behavioral control. The first measure relied on the well-validated snack delay task,²² in which toddlers are presented with an M&M in 4 separate trials and asked to wait 5, 15, 30, and 60 seconds before eating it, with scores ranging from 0 = ate the M&M before time was up to 2 = waited the entire time (Cronbach’s α [consistency of scores across trials] = 0.76). The second measure was the task orientation and/or regulation subscale²³ of the Infant Behavior Record.²⁴ After completing the entire assessment battery, interviewers rated toddlers’ attention span with scores ranging from 1 = fleeting attention to 9 = long continued absorption; interviewers also rated toddlers’ object orientation with scores ranging from 1 = does not indicate interest in objects to 9 = reluctantly relinquishes test materials (r = 0.50, P < .001). The third measure was from the well-validated Infant-Toddler Social and Emotional Assessment.²⁵ Parents rated 8 items on toddlers’ abilities to stop doing what they wanted and comply with parents’ requests, with scores ranging from 1 = not true or rarely and 3 = very true or often (Cronbach’s α [consistency of ratings across items] = 0.71). For a summary score, these 3 measures of self-regulation were each standardized, with mean = 0 and SD = 1.00, and averaged.

Parents’ responsive feeding practices²⁶ were assessed with video recordings of how parents introduced toddlers to 3 novel healthy foods, such as dried seaweed, which were not part of any intervention lesson. Research assistants rated whether parents praised their toddlers and whether they exerted negative pressure with 0 = behavior absent and 1 = behavior present; the individual ratings were then added together across the 3 foods. Research assistants also rated parents’ sensitivity and/or responsiveness and their demandingness across the entire snack-sharing interaction on a 5-point Likert scale with 1 = not at all and 5 = very much. For a summary score, negative pressure and demandingness were reverse scored, and all 4 ratings were standardized and averaged.

Parents’ sensitive scaffolding²⁷ was also assessed with video recordings. After viewing each of 3 3-minute interaction tasks (bowling, building a block tower, and completing a shape sorter), research assistants rated parents’ tendencies to structure the activity in a developmentally appropriate manner that promotes self-regulation, with scores ranging from 1 = almost never to 5 = almost always. For a summary score, ratings were averaged.

Plan for Analysis

To test the effectiveness of Recipe 4 Success, linear regression equations were estimated for each of the 4 primary outcomes. The following were independent variables in each equation: treatment status, site, child age, days between baseline and posttreatment assessments (range = 90–215), and baseline score of the outcome. All results are reported as Cohen’s d, which represents the difference in means between families in the treatment and control conditions divided by the pooled SD and adjusted for model covariates. By convention, a Cohen’s d of 0.20 is considered small, 0.50 is medium, and 0.80 is large.²⁸

Previous preventive interventions most similar to Recipe 4 Success had medium effect sizes.¹⁰ Optimal Design²⁹ software indicates that a trial needs a sample of 94 families to have 0.80 statistical power to detect a medium effect size of 0.50 with a probability value of 0.05 and covariates, including the baseline score, accounting for 25% of the variance in the outcome. Our recruitment goal was 100 to 105 families. With a sample of 73 families, this trial had 0.80 statistical power to detect an effect size of 0.57.