Participants and FCCHs
The present study utilizes baseline data from Healthy Start/Comienzos Sanos study, an 8-month cluster randomized controlled trial examining the efficacy of a multicomponent intervention to improve nutrition and physical activity environments in English and Spanish-speaking FCCH.(21) Details about study recruitment, intervention, and evaluation as well as study results have been described in full elsewhere,(21,22) and the methods relevant to the current analyses are described below. The Institutional Review Boards of the University of Connecticut, Brown University, and University of Rhode Island approved all study procedures and materials.
To meet study eligibility requirements, FCCHs had to be within 60 miles of Providence, Rhode Island, and had to have been in operation for at least 6 months. FCCH providers had to read and speak Spanish or English, provide meals and snacks for children, and care for at least two 2-to-5-year-old children for at least 10 hours per week. Data was collected from November 2015 to July 2018. Eligible providers completed a baseline telephone survey and in-person survey at the FCCH. A two-day observation was scheduled when at least one parent of an eligible child consented for that child to participate. All measures were conducted or administered by trained project staff. Providers received $25 for completing the baseline in-person survey and $50 for the two-day observation. Children received a reusable water bottle as a thank you gift and parents received a $20 gift card.
Measures
Demographics and other provider characteristics. Providers were asked to provide information about their sex (male, female, or refuse to answer), race (White, Black or African American, American Indian or Alaska Native, Asian, Native Hawaiian or Other Pacific Islander, other races not mentioned above, unknown),(23) ethnicity (Hispanic/non-Hispanic)(23) in a telephone survey and the following variables on an in-person survey: age, household income (less than 25k, 25k – 50k, 50k -75k, 75k-100k, more than 100k), marital status (single, married or living with a partner, divorced, separated, widowed), education (less than high school, high school or GED, associate’s degree, Bachelor’s degree, Master’s degree or higher), years in the U.S., country of origin (U.S./non-U.S.), years as a childcare professional, number of children currently in their care (and how many are their own children or grandchildren) and whether the FCCH was enrolled in the CACFP.
Weight Status. Children’s body weight and height were measured using an established research protocol.(24) The research staff member conducting measurements set up equipment in a space visible to, but located away from the main childcare activities. Children with parental consent who assented to the measurement came to the area to be measured one at a time. Height was measured using a SECA portable stadiometer to the nearest 8th of an inch. Weight was measured using a Tanita digital scale to one decimal place. Measurements were repeated three times and averaged for each child. Body mass index (BMI) was calculated as weight(kg)/height(m²) and BMI z-scores were calculated based on each child’s sex, age and BMI according to the Centers for Disease Control (CDC) growth charts. Per CDC guidance, overweight was defined as BMI > 85th percentile and obesity as BMI > 95th percentile.(25)
Food and Beverages Environment in FCCH. To assess the food and beverage environment, we used the validated Environment and Policy Assessment and Observation (EPAO) instrument.(26) The EPAO, originally developed for use in childcare centers, was modified to assess the FCCH environment.(27,28) One or two field observers (two observers if FCCHs with more than three children) conducted the EPAO observation in each FCCH for two full childcare days, which included at least two eating occasions (breakfast, morning snack, lunch, afternoon snack, and/or dinner). Observations began before children ate breakfast and ended when children left for the day. The nutrition-related sections of the EPAO assess compliance with 38 nutrition best practices. Each practice is rated on a scale of 0–3, where higher scores indicate better compliance. The best practices are grouped and averaged into 7 sub-scale scores, each reflecting an aspect of the overall nutrition environment within FCCHs. In the current study, we used the first two subscales: foods provided (amount, type, and quality of foods provided to children during meals and snacks), and beverages provided (amount, type, and quality of beverages provided to children during meals and snacks). The food provided subscale included serving of 12 types of food (i.e., whole fruit; fruit with syrup; total vegetables; dark green, orange, yellow vegetables; vegetables with added fat; fried potatoes; fried meat; high-fat meat; low-fat meat; high-fiber whole grain foods; high-sugar high-fat foods; and high-salt high-fat snacks). Beverages provided subscale included serving 5 beverages (i.e., water; fruit juice; sugary drinks; milk; flavored milk). Observations from the two days were combined to create a single, continuous set of subscores and an overall score. Detailed notes about the FCCH environment and providers’ nutrition and physical activity practices were recorded by the observer during the home visit. Forms were reviewed for accuracy and completeness by field staff. Additional review was conducted by data staff.(21)
Children’s Diet. Children’s dietary quality was measured by calculating the 2015 Healthy Eating Index (HEI) score(29) with two days of dietary data collected using the Dietary Observation in Child Care (DOCC). The DOCC is a reliable, valid visual observation technique for measuring children’s dietary intake developed by Ward and her team.(30,31) During the DOCC certification process, field staff need to accurately estimate at least 80% of 20 measured portions of food that a child would typically eat. Field staff must achieve 80% inter-rater reliability with a “gold standard” observer in the field at a FCCH to pass the certification. The quality of observations was continually assessed throughout dietary data collection such that observers needed to pass the certification process annually, and participate in structured monthly practice, quarterly validity checks, and semi-annual inter-rater reliability checks. Trained and certified data collectors observed all meals/snacks over two days and estimated the amount of food and beverages served and consumed for each child. These data were entered into Nutrition Data System for Research (NDSR)(32) to calculate an average daily HEI score. The total HEI score is a sum of 13 dietary components sub-scores based on two-day averaged score, ranging from 0 to 100, with higher scores indicating better diet quality.(29) A score of 80 or higher reflects a high-quality diet among preschool aged children.(33) HEI component scores are calculated as intake per 1000 calories (except for fatty acids which is scored as a ratio of unsaturated to saturated fatty acids) including total vegetables (5), greens/beans (5), total fruit (5), whole fruit (5), whole grains (10), dairy (10), total proteins (5), seafood plant protein (5), fatty acids (10), sodium (10), refined grains (10), added sugars (10), and saturated fats (10).(29)
Analysis
A multilevel mediation analysis was conducted to examine two questions. First, are there any direct effects of the food and beverages provided in FCCHs (measured by the EPAO) on children’s BMI z scores? Second, are these effects mediated by children’s diet quality, as measured by HEI scores? Building on the classic mediation model that assumes independent observations,(34) many mediation analyses have been extended to the multilevel context.(35–37) Many intervention programs are conducted in group settings such as schools and community groups.(38,39) As individuals within a cluster tend to be more similar than those selected from different clusters, the statistical assumption of independence of units is violated. Thus, statistical multilevel analysis are needed to account for clustering(40) or it may lead to invalid results.(41) The data collection of the current study occurs at two levels. Because multiple children were sampled within each FCCH, all models account for children clustered within FCCHs. In each model, data collected at the child level are called level-1 data, while data collected at the FCCH level are called level-2 data. In the current study, a 2-1-1 model was employed, which corresponds to measurement levels of the independent (i.e., FCCH food provided/beverage provided – level 2 data), mediator (i.e., child diet quality (HEI) – level 1 data) and outcome (i.e., weight status – level 1 data) variables, respectively (Fig. 1). This two-level model, which allows for grouping of child outcome within FCCHs included residuals at the child and FCCH level.(41)
Parameters for the models, including fixed effects of independent variables and mediators as well as random effects of mediators (only random intercepts between mediators and the outcome were estimated), were estimated using restricted maximum likelihood estimation.(42) The FCCH level independent variables did not vary at the individual level; therefore, the within-group effects of the independent variables were omitted and the mediation effects are presented as between-group indirect effects (i.e., the effect of the group differences in independent variables on the outcome through the mediator).(42) The significance level was set at p < .05. A bootstrapping approach was used to determine if this mediation effect is statistically significant.(43) All analyses were conducted in Stata SE 16.