Changes in Diet Quality of American Preschoolers Between 1977 and 1998

Abstract

Objectives. We determined diet quality trends among nationally representative samples of preschoolers between 1977 and 1998.

Methods. Adjusted diet quality index scores, overall intake, and tertiles of total score were compared for combined samples and 2 age groups using t tests with Bonferroni correction; surveys used were the US Department of Agriculture’s National Food Consumption Survey 1977–1979 (n = 2342), Continuing Survey of Food Intake by Individuals (CSFII) 1989–1991 (n = 858), and CSFII 1994–1996 and 1998 (n = 5355).

Results. Total scores increased slightly. Consumption of grains, fruits, and vegetables improved while added sugar and juice intake worsened.

Conclusions. Diet quality improved marginally since 1977. Consumption of fruits and vegetables needs to be increased and that of total and saturated fat, juice, and added sugar decreased.

Obesity has become an epidemic in the United States. In 1998, the prevalence of overweight children aged younger than 5 years was estimated to be 22%.1 Childhood obesity tracks into adult years2–4 (especially among females5), and comorbidities of high body weight during childhood include hypertension, dyslipidemia, insulin resistance, sleep apnea, and psychosocial consequences.6–8 Thus, the prevention of childhood overweight has become an issue of utmost importance.

It has been recognized that a safe and effective approach to decreasing the prevalence of overweight among children could be the improvement of overall diet quality and thus the establishment of healthier eating patterns.9–12 A high-quality diet is one that provides the nutrients and foods needed to maintain optimal health and promote growth and development. Assessment of diet quality aims to capture dietary intake patterns rather than intakes of individual nutrients or foods and is based on established dietary intake recommendations.13

Since dietary behavior tracks well from early childhood,14–17 the establishment of healthy eating patterns at a very young age might support better diet during childhood and thus in adults. As previous studies showed, preschoolers’ diets typically do not provide sufficient servings of fruit and vegetables but exceed the recommended intake in fat, saturated fat, and juice.18 In this population, the 2 main food sources for fat, saturated fat, added sugar, and grains are high-fat desserts and high-fat grain-based mixed dishes (data available from the authors upon request). Thus, targeting diet quality to improve childhood obesity might entail messages to limit intake of certain foods and food groups rather than focusing solely on increasing consumption of certain nutrients.

This study aimed to assess trends in overall diet quality in a study using nationally representative samples of the American preschool population from the US Department of Agriculture (USDA) National Food Consumption Survey (NFCS) for 1977 to 1979 (referred to here as NFCS77), the USDA Continuing Survey of Food Intake by Individuals (CSFII) for 1989 to 1991 (CSFII89), and the 1994–1996 CSFII combined with the supplemental survey of children for 1998 (CSFII94). We developed an overall Diet Quality Index for Children (C-DQI) and examined component scores in each survey to establish whether there are trends over the 3 survey time periods (scoring criteria of the index are available from the authors upon request). To observe differences by age group, we conducted an analysis using the full sample in each survey year, as well as 2 mutually exclusive age groups (children aged 2 years and 3 years, compared with children aged 4 years and 5 years).

METHODS

Data Sets

The NFCS77 and CSFII89 used stratified area probability samples of noninstitutionalized persons in US households in the 48 contiguous states. Dietary intake was collected by 1 day of interviewer-administered 24-hour recall and 2 additional days of self-administered 1-day food records. The CSFII94 survey data were also collected through a stratified, multistage, area probability sample to obtain a nationally representative sample of noninstitutionalized persons living in households. However, data for day 1 of dietary intake were collected via a household interview, whereas data for day 2, which were collected 3 to 10 days later but not on the same day of the week, could be reported during a phone interview. In 1998, the same methods were used for a supplemental survey of children aged younger than 9 years. The supplement was designed to be merged with the data set of 1994 to 1996; hence, we combined the samples of both surveys to represent the children of the 1994–1996 and 1998 CSFII (called CSFII94).

Samples

This analysis includes only children aged between 2 and 5 years who were not breastfed and for whom sociodemographic information was provided and data were available on at least 2 days of intake: 2342 children in the NFCS77, 849 in the CSFII89, and 5437 in the CSFII94. In all 3 surveys, interview respondents or another adult in the household reported diets of children aged younger than 6 years. The data sets provide sociodemographic information, dietary intakes, and a food servings database file with the number of Food Guide Pyramid19 serving sizes per 100g of a food code. Food codes were linked to intake data similarly in all 3 survey years to accommodate new food codes in the latest survey wave and to make data comparable across all survey waves.

Variables

Sociodemographic information, such as age, race, gender, total household income, years of education, and employment, was ascertained during the survey. In this study, education and employment data of the female head of household were used as a proxy for mothers. In an effort to capture cultural differences, the variables race (Black, White, and other) and Hispanic origin (Mexican, Puerto Rican, Cuban, other Spanish subgroup, or non-Hispanic) were recoded to define 4 distinct ethnic groups: non-Hispanic White, non-Hispanic Black, non-Hispanic other, and Hispanic. Similar coding has been employed by Odgen et al.20 Urbanicity was defined as living in a nonmetropolitan statistical area (non-MSA), MSA central city, or MSA non–central city.

The C-DQI was employed to determine overall diet quality. This composite assessment tool was specifically designed to assess over- and underconsumption of nutrients, foods, or food groups important for American children of preschool age. On the basis of the definition by Ibrahim,21 only areas of public health importance were included, resulting in 8 components: added sugar, total fat, saturated fat, grains, fruits and vegetables, dairy products, excessive juice, and iron. Added sugar, fat, and saturated fat were assessed as percentage of total energy. Grains, fruits, vegetables, and dairy products were assessed by number of servings, excess juice by number of 6-ounce servings of 100% fruit juices in excess of 1 serving per day, and iron by milligrams per day. To accommodate the consumption of smaller portions by younger children, serving sizes for children aged 2 and 3 years were two thirds of the serving size for older children.19 Component scores were allocated according to dietary intake, with increasing points reflecting improved diet quality. The measured component scores were compiled into a composite overall dietary intake score.

Analysis

Descriptive statistics (means and standard errors) were employed to describe the study sample populations by age groups. Standard errors were employed to allow estimation of the stability of the mean after weighted analysis was conducted. To account for differences in the distribution of sociodemographic characteristics in the American population over time, estimates were adjusted for ethnicity, urbanicity, maternal employment, and education status. To ascertain whether there was a significant trend in diet quality status between surveys, Student t tests were performed on the adjusted mean total C-DQI scores, as well as component scores, between individuals in the NFCS77 and CSFII89, CSFII89 and CSFII94, and NFCS77 and CSFII94. A P value of .05 that was corrected for multiple comparisons by Bonferroni correction was assumed to detect significance between mean scores (P ≤ .0019).

To further examine the changes of dietary intake over time, we subsequently compared average intakes of the nutrient, foods, and food groups that comprise the C-DQI of each of the 3 survey waves in the full sample as well as by age group (P ≤ .0006). Additionally, we assessed whether there was a significant difference in intakes between individuals in the 2 age groups (P ≤ .0019). To assess where areas of improvement were, we compared the average intakes of the C-DQI components in the lowest and highest tertile of total C-DQI points from the sample in the NFCS77 and CSFII94.

Changes occurring in dietary intake or dietary intake patterns can be assessed by several methods.22,23 If one assumes that the population investigated is the same (such as in a longitudinal study) or comparable and that the methods are similar, one can determine differences over time to establish trends. There is no longitudinal nationally representative data set available at this time. Thus, the nationally representative character of the 3 surveys was used as bases for comparability of the data sets to estimate trends over time. Although the dietary data collection slightly varied between the 2 older surveys (NFCS77 and CSFII89) and the most recent one (CSFII94), the methods were comparable and satisfactory for assessing dietary intake in large populations. Similar assumptions have been used by others to establish dietary trends over time in the US population.22,24–29 The statistical package used for all analysis was Stata (version 7.0; Stata Corp, College Station, Tex), which allowed adjustment for sample design effect and weighting to maintain the nationally representative character of the data.

RESULTS

A description of the sample populations used in this study can be found in Table 1 ▶. While the number of subjects aged 2 to 5 years varied greatly among the surveys, the distribution of sociodemographic characteristics within each group remained constant for most of the characteristics, indicating comparability of the data sets. However, more female heads of households had a higher educational level in CSFII89 and CSFII94 and almost twice as many were employed as in NFCS77. Approximately half of the children were male, two thirds or more were non-Hispanic White, and nearly half lived in households with a total income of less than 185% of the poverty level. Average energy intakes of the children significantly increased, from 1389 kcal per day in NFCS77 to 1558 kcal per day in CSFII94.

TABLE 1—

Selected Sociodemographic Characteristics (%) of Sample of Preschoolers From the National Food Consumption Survey for 1977–1979, the Continuing Survey of Food Intake by Individuals for 1989–1991, 1994–1996, and 1998

	NFCS77		CSFII89		CSFII94
Characteristics	2–3 y (n = 1127)	4–5 y (n = 1215)	2–3 y (n = 460)	4–5 y (n = 389)	2–3 y (n = 2805)	4–5 y (n = 2632)
Gender: male	51.2	53.9	47.7	55.7	51.9	50.7
Ethnicity
Non-Hispanic White	72.7	72.6	71.8	77.0	61.1	61.7
Non-Hispanic Black	16.2	16.9	13.5	10.9	16.7	16.7
Non-Hispanic other	1.4	1.8	4.6	2.6	16.0	16.7
Hispanic	9.7	8.7	10.1	9.5	6.2	4.9
Income
< 185% of poverty level	41.9	42.3	41.4	40.0	44.6	45.8
185%–350% of poverty level	42.2	42.0	30.2	38.0	29.3	30.7
> 350% of poverty level	15.9	15.7	28.4	22.0	26.1	23.5
Urbanicity
MSA central	30.0	31.2	31.5	25.5	32.4	32.5
MSA noncentral	37.2	37.2	45.8	52.8	48.2	47.5
Non-MSA	32.8	31.6	22.7	21.7	19.4	20.0
Region
Northeast	23.4	23.9	15.9	21.0	19.4	19.2
Midwest	28.1	26.3	24.1	22.6	23.9	23.9
South	29.6	33.5	32.6	35.0	33.1	33.9
West	18.9	16.3	27.4	21.4	23.6	23.0
Household size
1–4 members	61.1	54.2	62.1	57.0	62.2	58.4
> 5 members	38.9	45.8	37.9	43.0	37.8	41.6
Educationa
<High school	28.1	27.7	19.6	17.6	16.7	18.0
High school	39.7	41.7	38.7	30.6	31.5	33.5
> High school	32.2	30.6	41.7	51.8	51.8	48.5
Employment statusa: employed	36.8	34.8	45.4	48.8	59.4	61.8

Note. NFCS77 = National Food Consumption Survey for 1977–1979; CSFII89 = Continuing Survey of Food Intake by Individuals for 1989–1991; CSFII94 = Continuing Survey of Food Intake by Individuals for 1994–1996 and 1998; MSA = metropolitan statistical area central = inner city; non central = urban; non-MSA = rural.

^aOf female head of household.

Adjusted mean total and component C-DQI scores by USDA survey years for children aged 2 and 3 years compared with those aged 4 and 5 years, as well as total samples, are shown in Table 2 ▶. Average total scores were 43.7 in the NFCS77, 45.0 in the CSFII89, and 45.7 in the CSFII94, indicating a small increase between the first and second, second and third, and first and third USDA survey wave. Higher component or total points indicated improved diet quality. However, the significance of changes in the full sample were similar to those in the 2 age groups, with the exception of the decrease in scores for excess juice between the NFCS77 and the CSFII89 and between the CSFII89 and the CSFII94 (P = .0493 and .0085). Scores for added sugar, dairy, and iron decreased, while total and saturated fat, grains, and fruits and vegetables as well as total scores increased over time.

TABLE 2—

Total and Component Children’s Diet Quality Index Point Score for 2 Age Groups of Preschoolers From the National Food Consumption Survey for 1977–1979, the Continuing Survey of Food Intake by Individuals for 1989–1991, 1994–1996, and 1998

	1977–1979		1989–1991		1994–1996 and 1998
Component (maximum score)	2–3 y (n = 1127), Mean ±SE	4–5 y (n = 1215), Mean ±SE	2–3 y (n = 460), Mean ±SE	4–5 y (n = 389), Mean ±SE	2–3 y (n = 2805), Mean ±SE	4–5 y (n = 2632), Mean ±SE
Added sugar (10)	2.7 ±0.1297	2.1 ±0.1146	3.4 ± 0.3501a	2.7 ± 0.3308a	2.5 ± 0.0922b,c	1.7 ± 0.0853b,c
Total fat (5)	3.4 ± 0.0241	2.0 ± 0.0462	3.6 ± 0.0590a	2.5 ± 0.1208a	3.9 ± 0.0208b,c	2.9 ± 0.0374b,c
Saturated fat (5)	3.6 ± 0.0127	1.0 ± 0.0476	3.7 ± 0.0352a	1.5 ± 0.1333a	3.8 ± 0.0156b,c	2.1 ± 0.0458b,c
Grains (10)	6.1 ± 0.0683	7.0 ± 0.0649	6.7 ± 0.1637a	7.8 ± 0.1421a	8.8 ± 0.0447b,c	8.2 ± 0.0494b,c
Fruits and vegetables (10)	4.7 ± 0.0707	5.1 ± 0.0714	5.4 ± 0.1763a	5.3 ± 0.1855a	7.2 ± 0.0521b,c	6.2 ± 0.0573b,c
Excess juice (10)	9.6 ± 0.0449	9.5 ± 0.0506	8.1 ± 0.2734	8.6 ± 0.1850	6.4 ± 0.1004b,c	8.0 ± 0.0853c
Dairy (10)	6.3 ± 0.1116	6.4 ± 0.1124	6.4 ± 0.2616a	6.1 ± 0.3083a	5.9 ± 0.0868b,c	6.2 ± 0.0900b,c
Iron (10)	9.4 ± 0.0400	8.7 ± 0.0493	9.3 ± 0.0941a	8.8 ± 0.1170a	9.2 ± 0.0465b,c	8.4 ± 0.0574b,c
Total score (70)	45.7 ± 0.2049	41.9 ± 0.2132	46.6 ± 0.5456a	43.4 ± 0.5185a	47.7 ± 0.1714b,c	43.8 ± 0.1826c

Note. Estimates are sample design corrected and weighted; Student t test (P ≤ .0019) is used to note statistical significance for each comparison.

^aP ≤ .0019 for comparison between 1977 and 1989.

^bP ≤ .0019 for comparison between 1989 and 1998.

^cP ≤ .0019 for comparison between 1977 and 1998.

Table 3 ▶ shows the comparison of average dietary intakes of the nutrients, foods, and food groups that comprise the components of the C-DQI between the 2 age groups and the total sample for each USDA survey. Over the last 21 years, the main changes in intake were that percentage of total energy from fat and saturated fat decreased, whereas consumption of added sugar as percentage of total energy and servings of grains, fruits, vegetables, dairy products, juice, and iron increased.

TABLE 3—

Average Intakes of Children’s Diet Quality Index Components by Preschoolers in 2 Age Groups and the Full Sample From the National Food Consumption Survey for 1977–1979, the Continuing Survey of Food Intake by Individuals (CSFII) for 1989–1991, and the CSFII for 1994–1996 and 1998

	1977–1979			1989–1991			1994–1996 and 1998
Component	2–3 y (n = 1127)	4–5 y (n = 1215)	Total (n = 2342)	2–3 y (n = 460)	4–5 y (n = 389)	Total (n = 849)	2–3 y (n = 2805)	4–5 y (n = 2632)	Total (n = 5437)
Added sugar, % total energy	13.9 ± 0.2146	14.8 ± 0.2035	14.3 ± 0.1589	13.1 ± 0.5093a	14.2 ± 0.5215a	13.6 ± 0.4005a	14.9 ± 0.1782b,c	16.5 ± 0.1844b,c	15.7 ± 0.1373b,c
Total fat, % total energy	36.2 ± 0.1848	36.2 ± 0.1648	36.2 ± 0.1329	34.0 ± 0.4248a	34.0 ± 0.4431a	34.0 ± 0.3278a	32.0 ± 0.1453b,c	32.5 ± 0.1383b,c	32.3 ± 0.1075b,c
Saturated fat, % total energy	14.4 ± 0.0965	14.2 ± 0.0873	14.3 ± 0.0685	13.4 ± 0.2106a	13.4 ± 0.2686a	13.4 ± 0.1785a	12.3 ± 0.0739b,c	12.2 ± 0.0688b,c	12.2 ± 0.0532b
Grains, servings	3.8 ± 0.0510	4.4 ± 0.0499	4.1 ± 0.0396	4.1 ± 0.1039a	5.0 ± 0.1098a	4.5 ± 0.0856a	6.9 ± 0.0668b,c	5.7 ± 0.0599b,c	6.3 ± 0.0480b,c
Vegetables, servings	1.5 ± 0.03115	1.8 ± 0.0333	1.7 ± 0.0254	1.6 ± 0.0628a	1.7 ± 0.0859a	1.6 ± 0.0553a	2.5 ± 0.0391c	1.9 ± 0.0303c	2.2 ± 0.0263b,c
Fruit, servings	1.1 ± 0.0348	1.1 ± 0.0336	1.1 ± 0.0264	1.4 ± 0.0758a	1.3 ± 0.0775a	1.3 ± 0.0581a	2.9 ± 0.0556b,c	1.8 ± 0.00360b,c	2.4 ± 0.0360b,c
Excess juice, 6-oz servings	0.2 ± 0.0160	0.2 ± 0.0169	0.2 ± 0.0126	0.6 ± 0.0647a	0.5 ± 0.0564	0.5 ± 0.0463a	1.2 ± 0.00443b,c	0.6 ± 0.00236b,c	0.9 ± 0.0265b,c
Dairy, servings	1.8 ± 0.0309	1.9 ± 0.0307	1.9 ± 0.0238	1.9 ± 0.0739a	2.1 ± 0.0832a	2.0 ± 0.0620a	1.9 ± 0.0261b,c	2.0 ± 0.263b,c	2.0 ± 0.0195c
Iron, mg	9.4 ± 0.1305	10.5 ± 0.1291	9.9 ± 0.1016	9.6 ± 0.2910	11.1 ± 0.3724a	10.4 ± 0.2692a	11.5 ± 0.1153b,c	13.1 ± 0.1387b,c	12.3 ± 0.0953b,c

Note. Estimates are sample design corrected and weighted; Student t test (P ≤ .0006) is used to note statistical significance for each comparison.

^aP ≤ .0007 for comparison between 1977 and 1989.

^bP ≤ .0007 for comparison between 1989 and 1998.

^cP ≤ .0007 for comparison between 1977 and 1998.

Owing to space constraints, 95% confidence intervals for the mean estimates were not presented in the tables but are available from the authors upon request. The changes displayed in the table are not trivial. For example, for children aged 2 and 3 years, consumption of total fat as percentage of total energy decreased over time from 36.20% (95% confidence interval [CI] = 35.84, 36.56) in 1977 to 34.00% (95% CI = 33.17, 34.83) in 1989 and to 32.00% (95% CI = 31.72, 32.28) in 1998, and total dairy consumption in servings per day in the total sample changed from 1.90 (95% CI = 1.85, 1.95) in 1977 to 2.00 (95% CI = 1.96, 2.04) in 1998.

Examination of diets between age groups within each survey year showed that the diets of children aged 2 and 3 years were significantly different from those of children 4 and 5 years old in most of the components in the NFCS77 and CSFII94 but not in the CSFII89.

Comparison of intakes in the lowest and highest tertile of C-DQI total points in the sample from the NFCS77 and CSFII94 (Table 4 ▶) indicated that intake of servings of grains and fruit increased over time across tertiles, whereas intake of fat and saturated fat decreased; thus, in the CSFII94, children in the lowest tertile consumed less fat or saturated fat as percentage of total energy than children in the highest tertile in 1977. Added sugar increased in both survey years in the lowest as well as highest tertiles, while dairy intake remained fairly stable. Excess juice consumption was equal in both tertiles in the NFCS77 and was much lower than in either tertile in the CSFII94 (0.2 serving, compared with 1.0 serving in the lowest tertile and 0.7 serving in the highest tertile in the CSFII94). Iron consumption also increased over time; consumption in the lowest tertile in the CSFII94 was not much lower than in the highest tertile in the NFCS77 (10.4 mg/day vs 11.3 mg/day). Likewise, total energy consumption increased over time; children in the lowest tertile of the C-DQI score in 1998 only consumed 21.5 kcal/day less than children in the highest tertile in 1977.

TABLE 4—

Average Intakes of Children’s Diet Quality Index (C-DQI) Components and Total Energy by Preschoolers in the Lowest and the Highest Tertile of C-DQI Total Scores in the National Food Consumption Survey for 1977–1979 (NCFS77) and the Continuing Survey of Food Intake in Individuals for 1994–1996 and 1998 (CSFII94)

	NFCS77		CSFII94
C-DQI Component	Lowest Tertile of Total C-DQI Points (n = 788)	Highest Tertile of Total C-DQI Points (n = 778)	Lowest Tertile of Total C-DQI Points (n = 1791)	Highest Tertile of Total C-DQI Points (n = 1764)
Added sugar, % total energy	16.6 ± 0.2601	11.5 ± 0.2487	17.6 ± 0.20520	13.2 ± 0.3.61
Total fat, % total energy	37.3 ± 0.2115	35.5 ± 0.2437	33.2 ± 0.1846	31.3 ± 0.2033
Saturated fat, % energy	14.8 ± 0.1147	14.0 ± 0.1223	12.8 ± 0.0963	11.7 ± 0.1013
Grains, servings	3.4 ± 0.0559	4.7 ± 0.0704	5.1 ± 0.0720	7.3 ± 0.0816
Fruit, servings	0.8 ± 0.0354	1.4 ± 0.0478	1.9 ± 0.0633	2.9 ± 0.0774
Vegetables, servings	1.3 ± 0.0322	2.0 ± 0.0459	1.8 ± 0.0345	2.7 ± 0.0588
Excess juice, 6-oz servings	0.2 ± 0.0224	0.2 ± 0.0197	1.0 ± 0.0437	0.7 ± 0.0649
Dairy products, servings	1.7 ± 0.0468	2.0 ± 0.0294	1.9 ± 0.0485	2.0 ± 0.0253
Iron, mg	8.3 ± 0.1379	11.3 ± 0.1849	10.4 ± 0.1363	13.9 ± 0.1805
Energy, kcal/day	1265.2 ± 2.7679	1497.6 ± 2.0594	1476.1 ± 1.1823	1631.7 ± 2.2301

Note. Estimates are sample design corrected and weighted.

DISCUSSION

We examined nationally representative samples of American preschoolers to investigate the change in overall diet quality in the past 21 years. Our results show a small yet significant improvement in overall diet quality since 1977. A closer look at the components of the diet quality index revealed that, while most scores after the CSFII89 increased (indicating improvement of diet), scores for added sugar, excess juice, dairy products, and iron decreased. Despite the drop in average scores in the iron component to 8.8 of 10 possible points, mean intake still exceeds the current estimated average recommendations.30 This finding shows that American children consume on average much higher iron levels than indicated in the current recommendations. Intake trends also indicated that, although percentage of total energy from fat and saturated fat decreased, consumption of both in absolute amounts was virtually the same in 1977 and 1998. This indicates that the energy increase in the diets of children in 1998 was from sources other than fat, such as carbohydrates or sugars. Interpretations of children’s diets based on percentage of total energy from fat might therefore be misleading: despite the decrease in this percentage, consumption of total fat has remained stable and thus children might not be at lower risk for cardiovascular diseases. Intake of added sugar (as percentage of total energy as well as teaspoons per day) and fruit juice increased significantly. Similar results for other age groups have been found by others.31–33

Total energy consumption increased significantly between 1977 and 1998, which might have contributed to the increase in the prevalence of childhood obesity. However, one has to consider that the methods of dietary intake assessment have changed over time and that some of the increased energy might have owed to improved measurement methods. Our results illustrate that although overall diet quality has improved over time, overall energy intake has risen significantly, and there has been a shift from consumption of fats as a high proportion of total energy intake to one of nonfat macronutrients (for instance, added sugars). As comparison of the lowest and highest tertiles of the total C-DQI scores indicates, over a 21-year period, consumption of certain items, such as grains and fruit, increased, while others (fat and saturated fat) decreased to levels below those of 1977. Hence, the change in dietary intake has affected the total pre–school-age population, not just the proportion of children who had higher or lower diet quality in 1977.

It is noteworthy that the C-DQI scores of children aged 2 and 3 years were higher for most components of the C-DQI, indicating that younger children have a healthier diet than older children. This might be because of the higher level of control in young children’s environment by parents or other caretakers. Older children are able to make more food choices independent of an adult and appear to select less healthy items.

The scoring scheme of the C-DQI uses different cutoff points for children aged 2 and 3 years compared with those aged 4 and 5 years in the total fat and saturated fat component, to better accommodate the most recent dietary intake recommendations of the American Academy of Pediatrics.34 In addition, the recommended dietary intakes for younger children are lower than for children aged older than 4 years, and thus younger children scored differently in the iron component than older children. However, this did not distort the relationship between actual dietary intake and C-DQI scores. As shown in Table 3 ▶, the intakes of older children were significantly different from the diets of the younger ones, and as previous analysis has shown, diet quality deteriorated within each age group with increasing age (data available from the authors upon request).

The apparent contradiction between higher total diet quality scores and the increasing prevalence of childhood obesity can be explained by the shift in the C-DQI component scores over time. As Table 4 ▶ indicates, while consumption of grains, fruit, vegetables, and dairy products increased even in individuals in the lowest tertile of C-DQI total points to levels that are closer to the recommended intakes, percentage of total energy from added sugar strongly increased; it is even higher for the children in the highest C-DQI tertile in the CSFII94 than in the same tertile in the NFCS77. While consumption of fat in terms of percentage of total energy has decreased, data indicated that consumption in absolute amounts increased and contributed to higher energy intakes. Thus, although our index indicates more healthy food choices, overall energy consumption has increased, which might be a contributor to the increasing prevalence of childhood obesity.

One of the limitations of this study was that dietary intakes for children were collected by proxy in all survey waves. Children’s diets were reported by an adult who may have lacked complete knowledge of everything the child ate. Dietary assessment using 24-hour recalls is prone to incomplete data owing to memory lapses and may not be typical of usual intake. However, in large population studies, such as those conducted for nutrition monitoring purposes, more complicated dietary assessment methods are not feasible, and 2 days of data provide reasonable estimates at the group level. Additionally, bias might have been introduced owing to increased public awareness of the importance of limiting fat and saturated fat, especially in the later surveys; items high in fat and saturated fat might therefore have been underreported.

Another potential limitation to this study is the possibility of a bias in unknown direction owing to the change in dietary intake assessment methods between surveys. In the most recent survey, dietary intake data were collected in the form of 2 24-hour recalls, whereas in 1989 and 1977 there was only 1 such recall, but 2 consecutive days of food records were collected. Also, increases and changes in the food supply have occurred during the time span investigated. Although we used a system to link the food codes between survey years, additional alterations in consumption might have owed to the change in food supply. The direction and magnitude of this bias has not yet been specified.

It may be necessary to revise our C-DQI owing to the most recent release of new dietary reference intakes for macronutrients by the National Academy of Sciences.35 Once the appropriateness of the new recommendations for dietary intake for preschool age children has been established by authorities in pediatric nutrition, such as the American Academy of Pediatrics, a revised C-DQI might be developed.

Our research found that overall diet quality among preschool children has marginally improved over the past 21 years; however, total energy intake and consumption of fruit juices and added sugar have significantly increased. Nutritional concerns for American children have shifted from problems of deficient intakes to overconsumption of energy-contributing food groups, which this diet quality tool assesses. Overall diet in children needs to be improved to aid prevention of chronic diseases in adulthood.36,37 Studies using the Healthy Eating Index drew similar conclusions to the ones reported here.38,39 However, results of those studies were limited to the assessment of items provided by the Food Guide Pyramid. Despite their limited diet variety compared with children aged 4 and 5 years, younger children had better diets39; efforts to support high-quality diets from an early age might therefore improve health outcomes in the future.15,16,33,40