Longitudinal Diet Quality Trajectories Suggest Targets for Diet Improvement in Early Childhood

Abstract

Background:

There is little information about how diet quality evolves in early childhood, whether children exhibit varying diet quality trajectories, or which components of diet quality should be targeted for intervention.

Objective:

The goal of this study was to identify and evaluate trajectories of dietary quality in young children.

Design:

This was a secondary analysis of an observational, longitudinal cohort study of non-Hispanic African-American or white children and their parents from racially-concordant households with 4 years of follow-up (up to 13 study visits). Data on mother, infant feeding, and BMI were assessed at baseline. Diet was evaluated using 3-day diaries at each visit.

Participants/setting:

Of 372 children enrolled, 349 children had at least 3 study visits with dietary data for this analysis. Participants were enrolled at age 3 years between March 2001 and August 2002 in Cincinnati, Ohio. Final study visits were conducted between February 2005 and June 2006.

Main outcome measure:

The main outcome measure was the total Healthy Eating Index (HEI) 2005 score and HEI-2005 component scores.

Statistical analyses:

Diet quality trajectories were modeled using group-based modeling techniques.

Results:

The average total HEI-2005 score was low at age 3 years (55.1 ± 0.4 of maximum 100 points) and remained stable to age 7 years (54.0 ± 0.6, p=0.08 for trend). Five HEI-2005 trajectory groups were identified, of which one declined and one improved over time. HEI-2005 component scores, except milk intake and meat/beans scores, differed significantly (all p≤0.02) among trajectory groups at age 3 years, and most differences were maintained at age 7 years. Total vegetables, dark green and orange vegetables and legumes, and whole grains component scores were low for all trajectory groups. Whole fruit, total fruit, saturated fat and calories from Solid Fats, Alcoholic beverages and Added Sugars (SoFAAS) were highly variable among trajectory groups. Children in the lowest diet quality trajectory group were less likely to be breastfed and more likely to have been regular consumers of soft drinks (e.g., powdered drink mixes, sports drinks or soda pop) before age 3 years.

Conclusions:

Young childhood diet quality was low at age 3 years and remained stable to age 7 years. Improving intake of vegetables and whole grains is needed for all children. Focused attention regarding increasing fruit intake and reducing SoFAAS may be needed for families at increased risk for low overall diet quality.

Introduction

The quality of a child’s early diet, defined as an intake of adequate nutrients and avoiding energy-dense, nutrient-poor foods, may be important in establishing longer-term healthy dietary habits. Longitudinal studies have found moderate tracking of specific nutrients (e.g., sodium or percent of energy from fat) and food groups (e.g., fruit and vegetable intake) during childhood,^1–3 suggesting that early dietary habits can provide some information about later habits. For example, 50–65% of children in the highest quantile of energy or sugar intake are likely to remain in that quantile up to 3 years later, with various macronutrient intakes having correlations between 0.40 and 0.65 over time.³ Similarly, the Feeding Infants and Toddlers Study (FITS) has identified that the distribution of food groups, as a proportion of total energy, is similar across young age groups and approximates adult patterns by around age 18–24 months.^{4, 5} Data-driven classifications of dietary patterns in childhood show moderate tracking at young ages, although such analyses are complicated by changes in individual foods that are eaten at different developmental stages.^6–8 Thus, there is evidence that children’s dietary patterns are often carried forward into later life, making our understanding of childhood diet quality, or lack thereof, critical.

However, tracking, or the concept of remaining at a similar position relative to peers over time, does not address whether an individual’s diet is healthy, or whether the population distribution of diet quality is changing over time (e.g., improving or worsening). However, in young children, the healthiness of the diet and longitudinal changes in diet quality are not well studied. Use of a standardized diet quality metric that is applicable to any childhood stage is needed to fully understand trajectories of diet quality. In the United States, one such metric, the Healthy Eating Index (HEI), has been developed to assess how closely the diet of individuals 2 years of age or older meets key dietary intake recommendations in the Dietary Guidelines for America (DGA) published by the United States Department of Agriculture.^9–11 Because the DGA recommendations change over time, the HEI metric has been revised several times to remain concurrent with the DGA, with the 2005 HEI reflecting the 2005 DGA, and more recent HEI scoring (HEI 2010 and HEI 2015) reflecting the DGA updates in 2010 and 2015.^12–14 The 2005 HEI, used in this study, reflects an emphasis on intake of whole grains, various types of vegetables, and specific types of fat, while avoiding solid fats and added sugars.

In addition to the overall diet quality, the 2005 version of the HEI¹³ includes 12 component-specific scores, which can provide information on the adherence to specific dietary recommendations. While a very high or very low score can arise only when most component scores are high or low, respectively, moderate HEI-2005 scores can result from very different patterns of component scores. Thus, in addition to understanding the changes in the overall diet quality over time, it is also important to understand how the underlying component scores contribute to changes in the overall score.

Early childhood is a critical age in the development of dietary habits, not only because diet tends to track over time, but also because children experience significant developmental shifts in diet selection during early childhood. Ages 2–6 are peak ages for the development of food neophobia, which can lead to a decreased diversity of food intake.¹⁵ However, a significant research gap thus remains regarding potential trajectories of dietary quality over early childhood, and whether and how specific components of diet quality change in these various trajectories. Understanding how children differ in their dietary development would enable a more targeted approach to nutrition messaging during early to mid-childhood.

The objective of this study, therefore, was to identify and evaluate the trajectories of overall HEI-2005 diet quality and HEI-2005 diet quality components between ages 3–7 years and identify key sociodemographic and early infant feeding factors associated with young children’s diet quality. The hypothesis was that overall diet quality would decline during these years, as children transition to school and make more of their own food choices.

Methods

This study leveraged a detailed, longitudinal data set originally obtained for the Epidemiology of BMI Rebound Study.^16–18 The study enrolled 372 healthy 3-year-old children between March 2001 and August 2002 from the metropolitan area surrounding Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, and evaluated them every 4 months for 4 years for a total of 13 face-to-face visits between 2001 and 2006. This study size was originally selected to assess longitudinal relationships between timing of BMI rebound and obesity risk in non-Hispanic African-American and white children, focusing on the two predominant racial/ethnic groups in the study area at the time of enrollment. Participants eligible for the study were healthy children with both biological parents available for study, free from known chronic disease that could affect growth and development, born at term (38–42 weeks gestation) and with a birthweight at least 4 pounds, 10 oz (2100 g). Recruitment was limited to children who were either non-Hispanic African-American or non-Hispanic white as reported by the parent, and with the child and both parents reporting the same race. Individuals of Hispanic ethnicity and of other races were not recruited, due to a low proportion (<1%) of residents of Hispanic descent and other racial groups in the study area at the time of recruitment. Participants were recruited from local physicians’ offices and Cincinnati Children’s Hospital Medical Center clinics serving populations spanning the socioeconomic spectrum. Children were not otherwise selected based on any dietary, nutritional or physiological characteristics. Parents were also enrolled for at least one assessment of their body mass index (BMI) and body composition. Informed consent was obtained from the parents for their own and their child’s participation, and the protocol was approved by the Cincinnati Children’s Hospital Medical Center Institutional Review Board.

Baseline data at age 3 years were collected by parental report regarding a number of sociodemographic, birth, and early infant feeding characteristics. Child race (African-American or white), mother’s highest level of education (categorical as: Grade school; Some high school; High school graduate or GED; Trade, technical or vocational school; Some college or junior college only (2 year degree); College graduate (4 year degree); Graduate school; Don’t know); and total pre-tax annual household income in the year 2000 (categorical as: <$20,000; $20,000–$49,999; $50,000–$74,999; and $75,000 or more) were collected once at study initiation by parent report. Current marital status was collected at baseline for both biologic mother and biologic father as a categorical variable (Married/living as married; Divorced; Separated; Widowed; Never married; Don’t know), with mother’s marital status used in analyses. Number of people in the household was reported by the parent at baseline. Maternal age at delivery was calculated from the mother’s and child’s date of birth, and the child’s weight at birth was reported by the mother during enrollment screening. Participation in the Special Supplemental Nutrition Program for Women, Infants and Children (WIC) program¹⁹ during pregnancy, after the birth of this child or at the baseline visit were assessed by maternal report; report of WIC participation at any of these times was classified as ever participating. Child health insurance coverage reported by parents at baseline was classified as public (e.g., Medicaid), private, or none.

Infant feeding characteristics (breastfeeding, age of regular consumption of specific foods, and infant feeding practices/concerns) were obtained at baseline by retrospective maternal report. Breastfeeding was defined as ever breastfed versus never for this analysis. Timing of regular consumption of foods were defined as the age at which each of the following foods was given at least once a day: infant cereal from a bottle; infant cereal from a spoon; fruit juice (from bottle or cup); baby food; chopped/mashed table food; regular table food (same as rest of family); cow’s milk (from bottle or cup); or “powdered drink mixes, sports drinks or soda pop” (this last group of drinks is hereafter referred to as “soft drinks”), with responses of never a regular consumer, 0–3 months, 4–6 months, 7–12 months, 12–24 months or 24–36 months of age. Infant feeding practices were assessed using the validated Infant Feeding Questionnaire, a 20-question assessment of 7 scored constructs, each with a range of 0 to 4: 1) concern about the infant undereating, 2) concern about the infant’s hunger, 3) awareness of the infant’s hunger and satiety cues, 4) concern about the infant overeating, 5) feeding infant on a schedule, 6) using food to calm the infant, and 7) social interaction during feedings.^{20, 21}

Mothers’ BMI was calculated from height and weight measurements taken once but at varying times during the study period (median date in 2004). Child BMI and percent body fat at baseline (age 3 years) were also considered as potential covariates in this analysis. Maternal and child BMI were assessed by duplicate measurements of weight and height. Height was measured using a portable stadiometer, and weight was measured by a portable electronic scale. Each child was assessed for body fat composition using whole body dual energy x-ray absorptiometry (DXA) scans (Hologic Inc., Waltham MA) as has been previously described; only the baseline measure at age 3 was used in this analysis.^{18, 22}

Prior to each study visit, parents were asked to record their child’s dietary intake over 3 consecutive days (2 weekdays and 1 weekend day) in 3-day diet diaries, with dates of diet data collection assigned to the parent and with instructions to include portion quantities and preparation methods. Study dietitians reviewed the records with families at each visit, and analyzed the data using the Nutrition Data System for Research (NDS-R, Nutrition Coordinating Center (NCC), University of Minnesota, Minneapolis, MN). NCC Food and Nutrient Databases (versions 31–35 [released 2000–2004], 2005, and 2006)^23–25 and NDS-R software versions (v4.03–v4.06, v5.0, 2005 and 2006)^26–32 appropriate for the year of data collection were used. In order to obtain the specific files required for scoring the Healthy Eating Index, a newer version of the NDS-R program (NDS-R version 2015)³³ was then applied to the nutrient data to obtain food-specific files.

NDS-R files were used to calculate the 2005 Healthy Eating Index (HEI-2005), a derived measure designed by the United States Department of Agriculture and the National Cancer Institute to assess dietary quality relative to the 2005 Dietary Guidelines for Americans.^{13, 34, 35} The HEI-2005 includes 12 dietary component scores that address intake adequacy, insufficiency or excess; scores for each component range from 0–5, 0–10, or 0–20, and are summed for a maximum total HEI-2005 score of 100. All scores are calculated using a density standard (e.g., per 1000 kcal or as percent of energy) to control for differences in the quantity of intake. Nine of the 12 components of the HEI-2005 address nutrition adequacy. The remaining 3 scores assess dietary components that should be eaten in moderation, such as sodium, saturated fat, and calories from Solid Fats, Alcoholic beverages and Added Sugars (referred to as SoFAAS). Moderation components are reverse-scored with higher intakes receiving lower scores. Therefore, for all components, a higher score indicates a higher diet quality. Dietary intake was averaged for a given visit (median of 3 days of intake/visit), and this visit average was then used to calculate visit-specific HEI-2005 scores using the simple HEI scoring algorithm method.

Diet Quality Trajectories

Analysis was conducted using group-based trajectory modeling, as implemented in SAS PROC TRAJ,^{36, 37} to estimate HEI-2005 (diet quality) trajectory groups between 3 and 7 years of age. Between 3 and 13 HEI-2005 average total scores per person (one per study visit, median [Interquartile range, IQR] 13 [13, 13] visits/person) were used to estimate trajectory groups; participants with 1 or 2 study visits were excluded, as they could not provide adequate longitudinal information. Group-based trajectory modeling is a specialized application of finite mixture modeling and is designed to identify clusters of individuals following similar progressions of an outcome of interest over age or time.^{36, 37} The number of trajectory groups and the shape of each trajectory (e.g., linear, quadratic) were determined independently. In this analysis, the number of trajectories was determined first. Beginning with two trajectories, additional trajectories were added until Bayesian Information Criterion (BIC) values decreased by less than 4 units or the model failed to converge. Quadratic forms for each trajectory were then modified based on BIC improvement and significance of the terms in each trajectory model. Each individual in the analysis was then assigned to a specific trajectory group according to the highest posterior probability of group membership, based on their unique pattern of total HEI-2005 scores over time (median[IQR] posterior probability: 0.91 [0.77, 0.99]).

Analysis

Baseline characteristics were presented by diet quality trajectory group using means and standard errors, medians and IQRs, or numbers and percentages, as appropriate. Tests among groups were conducted using general linear models or Kruskal-Wallis tests for continuous variables, or chi-square tests for categorical variables. To identify baseline characteristics independently associated with diet quality trajectory membership, multinomial logistic regression was conducted using a cumulative logit link, modeling the probability of being in a higher diet quality trajectory. For all analyses, results are considered significant at p<0.05.

For the purposes of presentation in the figures, the visit-level HEI-2005 scores were averaged across an entire year (up to 3 visits [median of 9 days] of dietary intake spanning different seasons for years 1–4 [ages 3–6], and one visit [median of 3 days] in year 5 [age 7]) to obtain individuals’ average scores per year. For the purpose of comparing adequacy or excess across component categories, the HEI-2005 component scores are presented as percent of the maximum points within each component. Data and data collection forms from this study are available upon request from the authors.

Results

The study sample included 349 participants, 94% of the full cohort (n=23 excluded with only 1–2 visits with dietary assessment).^{18, 22} The cohort was predominantly non-Hispanic white (80%) with highly educated parents (50% with college education or greater), with the majority (72%) ever breastfed. Group-based trajectory analysis identified 5 diet quality trajectory groups (hereafter referred to as groups 1 [lowest total HEI-2005] through 5 [highest total HEI-2005]) based on linear trends of total HEI-2005 scores from ages 3 to 7 years (Figure 1). Baseline characteristics of the children in the five trajectory groups are presented in Table 1. Of the factors evaluated, higher diet trajectory groups were associated with higher mother’s education (p=0.002), higher likelihood the child was ever breastfed (p=0.0005) and later or no regular consumption of soft drinks (p<0.0001). Number of people in the household (p=0.03) and mother’s BMI (p=0.02) were both higher in groups 2 and 3 than group 5, while the child’s BMI percentile was higher in group 2 compared with group 4 (p=0.03). No significant differences were noted among diet trajectory groups with regard to regular consumption of any other foods or infant feeding style scores (data not shown; all p>0.05).

Figure 1: Total HEI-2005 score for 349 non-Hispanic African-American and white children from Cincinnati, OH, recruited at age 3 years in 2001–2002 and followed for 4 years to age 7 years, presented by child age and diet quality trajectory group.

Unadjusted mean ± standard error HEI-2005 score plotted by age of the cohort participants in a given year of the study for the overall cohort (black diamond/black solid line), and for the trajectory groups: Group 1 (filled circle/dash-double-dot gray line), Group 2 (open circle/long dash gray line), Group 3 (filled triangle/dash-dot gray line), Group 4 (open triangle/medium dash gray line) and Group 5 (filled square/solid gray line). All means differ from all other means in all years (p≤0.0001) except trajectories 3 and 4 at age 3 (p=0.62).

Table 1:

Baseline parental/household and child characteristics at age 3 years of 349 non-Hispanic white and African-American participants recruited in 2001–2002 in Cincinnati, Ohio, with dietary data from at least 3 study visits, presented overall and by Healthy Eating Index (HEI)-2005 diet quality trajectory group assigned using group-based modeling across 4 years of follow-up

	Overall	Group 1 (lowest total HEI-2005)	Group 2	Group 3	Group 4	Group 5 (highest total HEI- 2005)	p-value across groupsa
N (%)	349	25 (7%)	137 (39%)	106 (30%)	71 (20%)	10 (3%)
Parent/Household Characteristics
Mother’s age at delivery, years, median [IQR]b	30.0 [26.9, 33.5]	29.7 [27.5, 30.8]	28.9 [25.8, 33.5]	30.0 [27.2, 34.0]	31.0 [27.5, 34.2]	32.7 [31.3, 35.8]	0.054
Mother’s highest education, n (%)
≤High school graduate/GEDc	58 (17%)	8 (32%)	31 (23%)	13 (12%)	6 (8%)	0 (0%)	0.002
Some college/ technical	116 (33%)	6 (24%)	51 (24%)	39 (37%)	18 (25%)	2 (20%)
College graduate or higher	175 (50%)	11 (44%)	55 (40%)	54 (51%)	47 (66%)	8 (80%)
Mother works full time, n (%)	126 (36%)	8 (32%)	50 (37%)	46 (43%)	19 (27%)	3 (30%)	0.05
Number in the household, median [IQR]	1 [1, 2]	1 [1, 2]x,y	1 [1, 2]x	1 [1, 2]x	1 [1, 2]x,y	0 [0, 1]y	0.03
Ever had WICd, n (%)	90 (26%)	4 (16%)	44 (32%)	27 (25%)	14 (20%)	1 (10%)	0.15
Public health insurance, n (%)e	33 (12%)	3 (13%)	18 (17%)	9 (11%)	3 (5%)	0 (0%)	0.23
Household income <$50k in year 2000, n (%)	136 (39%)	6 (24%)	58 (42%)	45 (42%)	23 (32%)	4 (40%)	0.30
Mother’s BMIf, kg/m2, median [IQR]g	25.4 [22.1, 32.4]	23.8 [21.1, 33.3]x,y	26.5 [22.6, 33.6]x	25.8 [22.3, 31.8]x	25.0 [22.1, 31.4]x,y	21.2 [20.6, 22.2]y	0.02
Child Baseline Characteristics
Age, years, median [IQR]	3.29 [3.10, 3.57]	3.16 [3.05, 3.31]	3.31 [3.10, 3.57]	3.24 [3.12, 3.61]	3.33 [3.13, 3.56]	3.15 [3.06, 3.46]	0.10
Race (n, % African-American)	71 (20%)	0 (0%)	35 (26%)	20 (19%)	15 (21%)	1 (10%)	0.05
Sex (n, % Female)	165 (47%)	9 (36%)	70 (51%)	47 (44%)	34 (48%)	5 (50%)	0.65
BMI percentile, median [IQR]	64.6 [40.0, 84.0]	65.0 [41.5, 82.5]x,y	72.8 [45.8, 85.4]x	63.9 [44.0, 85.4]x,y	51.7 [17.0, 78.9]y	64.9 [51.5, 88.7]x,y	0.03
DXAh % body fat, median [IQR]	27.9 [24.0, 31.6]	27.3 [24.4, 33.2]	28.7 [24.2, 33.1]	27.6 [24.1, 30.5]	27.9 [23.0, 30.8]	26.3 [21.4, 28.7]	0.34
Ever Breastfed, n (%)	253 (72%)	14 (56%)	88 (64%)	79 (75%)	62 (87%)	10 (100%)	0.0005
Age at regular consumption of soft drinks, n (%)
7–12 months	15 (4%)	1 (4%)	11 (8%)	2 (2%)	1 (1%)	0 (0%)	<0.0001
12–24 months	80 (23%)	13 (52%)	32 (24%)	25 (24%)	10 (14%)	0 (0%)
24–36 months	99 (29%)	9 (36%)	48 (35%)	22 (21%)	19 (27%)	1 (10%)
Never a regular consumer	153 (44%)	2 (8%)	45 (33%)	56 (53%)	41 (58%)	9 (90%)

^aP-values from Chi-square or Kruskal-Wallis tests, for categorical and continuous variables, respectively. For significant overall Kruskal-Wallis tests, pairwise differences were evaluated with non-parametric DSCF test

^bIQR: Interquartile Range

^cGED, General Education Development Test

^dWIC: Special Supplemental Nutrition Program for Women, Infants and Children

^eN for health insurance=274 (24, 108, 79, 55, 8 by group)

^fBMI, Body Mass Index

^gN for mother’s BMI: 300 (22, 119, 85, 64, and 10 by group)

^hDXA: Dual Energy X-ray Absorptiometry

^x,yMedians not sharing the same superscript (x, y) are statistically different at p<0.05.

Multivariable multinomial logistic regression analysis identified that the child being breastfed (odds ratio [95% confidence interval]: 2.2 [1.4, 3.6], p<0.0001) and not regularly consuming soft drinks before age 3 years (2.7 [1.6, 4.3] versus regular consumption between 24–36 months, p<0.0001) were associated with a greater odds of being in higher diet quality trajectories, while a higher baseline BMI percentile (0.92 [0.86, 0.99] per 10 percentile points, p=0.03) and non-Hispanic white race (0.49 [0.29, 0.83], p=0.008) were associated with a lower odds of being in higher diet quality trajectories.

Diet Quality Trajectories

The mean (± SE) total HEI-2005 scores for the overall cohort by age were: 55.1 ± 0.4; 55.4 ± 0.5; 55.1 ± 0.4; 54.8 ± 0.5; and 54.0 ± 0.6 at ages 3, 4, 5, 6, and 7 years, respectively (Figure 1, p=0.08 for trend). All pairwise comparisons of total HEI-2005 among groups in each year were statistically significant (all p≤0.0001), with the exception of trajectory groups 3 and 4 at age 3 (p=0.62). Table 2 provides data on the linear trends over time of the HEI-2005 total and component scores, overall and by diet trajectory. Total HEI-2005 scores changed significantly only in two diet trajectory groups; group 3 declined significantly over time, while group 4 improved. However, the stability in total HEI-2005 scores masked significant changes in individual components over time. In all diet quality trajectory groups, scores for the total fruit component declined over time (decrease of 2% to 7% of maximum component score per year), while scores for total grains significantly improved (increase of 1% to 4%/year). The lower diet quality groups 1–4 also experienced improvements in meat/beans scores (increase of 1.5% to 4%/year) and deterioration of sodium scores (decrease of 1.5% to 4%/year) over time. Longitudinal changes in total vegetables, dark green and orange vegetables, whole grains and saturated fat scores were uncommon or not seen in any diet trajectories.

Table 2:

Healthy Eating Index (HEI)-2005 Total and Component Scores Unadjusted Linear Trends between ages 3 and 7 years for a Cohort of 349 non-Hispanic White and African-American Children enrolled in 2001–2002 in Cincinnati, Ohio, Overall and Separately by HEI-2005 Diet Quality Trajectory Group, presented as estimated change per year in the percent of the maximum component score

	Overall	Group 1 (lowest total HEI-2005)	Group 2	Group 3	Group 4	Group 5 (highest total HEI-2005)
	β ± SEa	β ± SE	β ± SE	β ± SE	β ± SE	β ± SE
Total HEI-2005	NCb	NC	NC	−0.9 ± 0.2 ***	0.6 ± 0.2 **	NC
Total Fruit	−5.9 ± 0.5 ***	−5.4 ± 1.5***	−6.5 ± 0.7***	−7.4 ± 0.8 ***	−2.4 ± 1.0 **	−5.8 ± 1.7 ***
Whole Fruit	−3.2 ± 0.5 ***	NC	−3.6 ± 0.7***	−4.8 ± 0.9 ***	NC	NC
Total Vegetables	0.5 ± 0.3 *	NC	0.8 ± 0.4 *	NC	NC	NC
Dark Green/Orange Vegetables and Legumes	NC	NC	NC	NC	1.7 ± 0.8 *	NC
Total Grains	1.9 ± 0.2 ***	3.9 ± 0.9***	2.1 ± 0.3 ***	1.8 ± 0.3 ***	1.1 ± 0.4 **	1.8 ± 0.8 *
Whole Grains	NC	NC	NC	NC	NC	NC
Milk	−1.4 ± 0.3 ***	−3.0 ± 1.2*	−1.6 ± 0.6 **	−1.7 ± 0.6 **	NC	NC
Meat/Beans	1.9 ± 0.3 ***	3.6 ± 1.4 **	1.5 ± 0.6 **	1.9 ± 0.6 **	1.8 ± 0.7 **	NC
Oils	2.2 ± 0.4 ***	NC	2.4 ± 0.6 ***	1.8 ± 0.6 **	3.0 ± 0.8 ***	NC
Saturated Fat	NC	NC	NC	NC	NC	NC
Sodium	−2.4 ± 0.3 ***	−4.3 ± 1.3 ***	−2.6 ± 0.5 ***	−2.3 ± 0.6 ***	−1.5 ± 0.7 *	NC
SoFAASc	NC	NC	0.8 ± 0.4 *	−1.5 ± 0.4 ***	1.3 ± 0.5 **	NC

^*p<0.05;

^**p≤0.01;

^***p≤0.001

^aAll values are unadjusted estimated mean (β ± SE) of change per year in the percent of the total component score (e.g., 100 is 100% of the maximum component score for all components). Positive changes indicate improving scores. Linear change across study years presented if significant.

^bNC, no significant linear change across years

^cSoFAAS: Calories from Solid Fat, Alcoholic beverages and Added Sugar

Almost all HEI dietary quality components differed significantly across trajectory groups at age 3 years (Figure 2A, all p≤0.0002 except sodium, p=0.02), with the exception of the milk and meat/beans intake scores. Groups 1–4 were all notably low (<40% of maximum score) for total vegetables, dark green and orange vegetables, and whole grains, while total grains and milk were reasonably high for all groups (>80% of maximum score). Striking component disparities across trajectories at age 3 years included total fruit (36% vs. 99% of maximum component score in the lowest and highest diet trajectory groups, respectively); whole fruit (22% vs. 96%, respectively); saturated fat (31% vs. 76%, respectively); and SoFAAS (26% vs. 81%, respectively).

Figure 2: HEI-2005 average component scores for 349 non-Hispanic African-American and white children from Cincinnati, OH, recruited at age 3 years in 2001–2002 and followed for 4 years to age 7 years, presented by diet quality trajectory group at ages 3 and 7 years.

Component scores, represented as mean percent of maximum component score, by trajectory group, A) at age 3 years (baseline) and B) age 7 years (final visit). Group 1 (filled circle/dash-double-dot line), Group 2 (open circle/long dash line), Group 3 (filled triangle/dash-dot line), Group 4 (open triangle/medium dash line) and Group 5 (filled square/solid line). P-values represent overall differences among trajectory groups from unadjusted general linear models.

The HEI-2005 component differences at age 7 years were similar to those seen at age 3 years (Figure 2B), with a few exceptions. Overall, the diet quality advantages seen in group 5 relative to the other groups became less pronounced, especially with regard to total fruits, whole fruits, saturated fat, SoFAAS and total HEI-2005, although these components remained significantly different across trajectory groups (all p≤0.003). In addition, trajectory groups 3 and 4 differed from each other by age 7 years for most components, while the two groups were indistinguishable at age 3 years.

Discussion

Among children in this study, overall diet quality in young childhood, as assessed by the HEI-2005, is already relatively poor by age 3 years and remains relatively stable through age 7 years. Five diet quality trajectory groups were identified during early childhood, which differed by both demographic factors and early feeding practices. Most individual components of the HEI-2005 differed significantly among the diet trajectory groups, and these differences were largely maintained between ages 3 and 7 years. Thus, the establishment of diet quality likely primarily occurs prior to age 3 years. Efforts to improve children’s diet quality should focus on early food exposures and establishing healthy eating habits, especially by encouraging breastfeeding and avoiding regular intake of soft drinks before age 3 years.

The relatively low quality of preschoolers’ diets seen in this cohort at baseline is consistent with results from other studies,^{1, 38, 39} but appears to conflict with micronutrient-based analyses that find most children have adequate intakes of many micronutrients.^40–42 However, micronutrient-based studies are not sufficient for assessing overall dietary composition or quality. Food-based guidelines and indices have been developed to emphasize the nutritional benefits of eating whole foods, such as phytochemicals, antioxidants, and other beneficial compounds such as fatty acids or trace vitamins and minerals.^9–11

Contrary to expectation, increased child age and transitioning to a school environment did not have a major impact on overall diet quality in the current study. This finding is, however, consistent with a previous longitudinal study reporting relative stability in HEI-1995 scores between ages 3–7 years in Brazil.³⁹ Given the already-poor diet quality identified at age 3 years, prevention of poor diet quality may be needed prior to age 3 years. Indeed, the transition from a milk-based infant diet to a fully diverse toddler diet may be an ideal time to ensure children initiate healthy diets. Data from our previous studies have shown that introduction of food groups⁴³ and dietary diversity in infants develops rapidly in the second six months of life.⁴⁴ In the Avon Longitudinal Study of Parents and Children (ALSPAC) study, the quality of food-based intake declined significantly between ages 1.5 and 3.5 years old, primarily due to an increase in sugary foods, processed meats and other nutrient-poor “non-core” foods, with a corresponding decrease in nutrient-rich “core” vegetables, fruit, dairy and lean meats.⁴⁵ Lower diet quality was found at age 24 months compared with 13 months of age in the Infant and Toddler Feeding Practices Study-2.⁴⁶ Furthermore, dietary quality, assessed by a modified Diet Quality Index Score, decreases steadily from age 1 to age 4 years.⁴⁷ Cross-sectional dietary data from the Feeding Infants and Toddlers Study (FITS) suggests that intake of specific food groups as a proportion of total energy stabilizes by 18–24 months of age, resembling adult patterns.^{4, 5} Older children and adolescents are even more prone to having poor-quality diets,^{38, 48} although it is not clear when later declines occur. This suggests not only that prevention of poor diet quality prior to age 3 years is important, but that continuing focus on preventing further deterioration in diet quality later in childhood is also needed.

Component-level analyses also suggest differing strategies may be needed to improve overall diet quality in young children. Some components were low in all diet trajectory groups, such as total vegetables, dark green and orange vegetables, and whole grains. Longitudinal analysis of diet quality components also identified these components as unlikely to change spontaneously during ages 3–7 years. Furthermore, diet trajectory groups demonstrated significant declines in total fruit and sodium scores during this time. Improving such widespread deficiencies in meeting diet quality goals during early childhood thus appears to require focused attention, and may require population-level education campaigns about why incorporation or avoidance of these dietary components is important and strategies for doing so. However, education is often not sufficient to result in behavior change,⁴⁹ and it is unclear whether or in what format a population-level intervention would be effective. Further research is needed to identify barriers to improving diet quality in children and effective strategies to overcome these barriers. Evidence that these dietary components are amenable to intervention comes from the WIC program, which changed its food package to emphasize fruits, vegetables, whole grains and low-fat milk, resulting in a 3-fold increase in the Greens/Beans HEI-2010 component, and the overall HEI-2010 score in WIC participants compared with non-participants.⁵⁰ By contrast, most children, regardless of overall diet quality, had adequate or high scores for milk and total grains components, so interventions focused on these components would have minimal impact on overall diet quality.

In this study, behavioral and feeding practices during infancy were key contributors to diet quality among preschool-aged children, more than demographic characteristics of the mother or family. In particular, breastfeeding was positively associated with diet quality, and earlier regular consumption of soft drinks were negatively associated with diet quality at age 3 years. This is consistent with previous work identifying breastfeeding as a contributor to improved diet quality.^{51, 52} Breastfeeding may impact diet quality in a number of ways, including the fact that mothers who decide to breastfeed may also make healthier choices generally.⁵³ However, breastfeeding also enables early exposure to a variety of tastes through the mother’s milk,⁵⁴ which may enable greater acceptance of new foods when introduced,⁵⁵ and increase children’s later preferences for bitter or sour foods.⁵⁶ Clearly, early regular consumption of sugar-sweetened beverages (SSB) is directly related to the SoFAAS component of the HEI-2005, and SSB consumption has been linked with lower overall and component-specific HEI scores in children, even after removing SSB intake from the HEI calculation.⁵⁷

While not independently associated, mothers with lower education were also at high risk for establishing poor diet quality in their young children, and this finding is also consistent with previous studies.^{8, 58, 59} In this study, racial differences in diet trajectory group assignment were not significant in unadjusted analyses, in contrast to previous studies where African-American children often have lower diet quality.^{60, 61} However, African-American children in our sample were also significantly less likely to have been breastfed and more likely to have been early regular consumers of soft drinks than their white counterparts. Confounding by these and other factors may have resulted in the counterintuitive finding that African-American children were more likely to be in higher diet quality groups in multivariable analysis.

Related to this, scores for whole fruit, total fruit, saturated fat and SoFAAS intake were highly differentiated between the lowest and highest trajectory groups. Targeted messages or interventions regarding these specific food groups among families at higher risk for poor diet quality should also be a priority, beginning in infancy and toddlerhood. Identifying these families would require dietitian or primary care provider assessment of early eating patterns and parental feeding styles and concerns, along with effective early intervention. There is some indication that early interventions can impact later diet quality in children, particularly regarding specific components of the HEI-2005. For example, increased availability of fruits and vegetables in the home was associated with higher whole fruit, total fruit and total vegetable component HEI scores among fourth graders, although it had little impact on overall HEI scores.⁶² In particular, because of the heavy weighting of the SoFAAS component on the HEI-2005 scoring (20% of total score), and the significant differences in the intake of SoFAAS between those with the highest and lowest diet quality trajectories, reducing added sugars (including SSB) and solid fats may have the largest impact on total HEI-2005 diet quality in this age group. Indeed, a focused intervention to reduce sugar-sweetened beverage (SSB) intake, albeit among low-income adults, had wide-ranging impacts on improving other components of HEI-2005 diet quality.⁶³

The particular strengths of this work included analysis of a large longitudinal cohort of children with up to 9 days of diet records per year for 4 years, allowing for a more detailed analysis of diet quality and diet quality trajectories among young children than many previous studies. In addition, the breadth of the study measures allowed for evaluation of several potential covariates, including infant feeding practices and body composition. However, some limitations of this work should be taken into consideration. First, the data included in this study were collected in the early 2000’s, so may not reflect contemporary dietary patterns among young children. The study was conducted in a single geographical location, and the racial distribution of the study was limited to non-Hispanic white or African-American families with predominantly higher incomes, and given the sample size, racial distribution, and income of this sample, these findings are not generalizable to all U.S. children.

Conclusions

In this longitudinal study of non-Hispanic white and African-American children, children at 3 years of age had relatively poor diet quality that remained consistent through early childhood to age 7 years. These findings point to earlier ages in infancy and toddlerhood as key periods for the development of healthy dietary patterns. Most children, regardless of overall diet quality, appeared to have adequate intake of milk and total grains components. Conversely, the intake of some dietary components is inadequate for all children, regardless of total HEI-2005 trajectory group, such as total vegetables, dark green and orange vegetables and legumes, and whole grains. Such dietary components, which remain deficient in many older children and adults,⁶¹ point to a need for research to identify effective strategies to improve the diet quality of all children. In addition, future research could explore whether targeted approaches may be better suited to dietary components with wide variability among the overall diet trajectory groups, such as whole and total fruit, as well as added sugars and solid fats.

Research Snapshot.

Research Question:

Do all young children experience the same diet quality trajectory over time? If not, what components of diet quality differentiate dietary trajectories in this age group?

Key Findings:

In this longitudinal cohort study of non-Hispanic African-American and white children aged 3 through 7 years, diet quality was low at age 3 years and remained stable to age 7 years. Intake of vegetables and whole grains was low for all children. In contrast, fruit intake and calories from solid fats and added sugars were highly variable among trajectory groups.