Abstract
Background
Programs, such as Medicaid, the Supplemental Nutrition Assistance Program, and the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), provide access to vital medical and nutrition services. Few studies have investigated whether demographic, social, and economic determinants of health, including length of time spent on these programs, are associated with diet quality and weight status in early childhood.
Objectives
Classification and regression tree analysis, a machine learning method, was used to determine health predictors to identify patterns of children with higher compared with lower diet quality and higher compared with lower weight status.
Methods
Using the WIC infant and toddler feeding practices study-2 (unweighted N = 3051; weighted N = 413,211), classification and regression tree identified the sequence of binary splits that best differentiated the sample on Healthy Eating Index-2020 (HEI-2020; range 0–100) and HEI-2020 subscales (adequacy and moderation), and body mass index (BMI in kg/m2) z-score at 2–5 y. Predictors, including maternal BMI, child birthweight, sociodemographics, and length of time spent on safety net programs, were considered.
Results
Higher HEI-2020 scores were primarily predicted by race and ethnicity (e.g., Hispanic, Spanish, or non-Hispanic White), and longer WIC and shorter Supplemental Nutrition Assistance Program duration. In examining HEI-2020 subscales, higher HEI adequacy was primarily predicted by higher education, older maternal age, longer WIC duration, and race and ethnicity. Higher HEI moderation was primarily predicted by race and ethnicity and longer WIC duration. Higher BMI z-score was primarily predicted by higher birth weight.
Conclusions
Child diet quality and weight status were associated with different social determinants of health, which included maternal weight status, race and ethnicity, and food assistance program participation, particularly WIC.
Keywords: HEI-2020, classification and regression tree, BMIz, diet, children, WIC, SNAP, Medicaid, race, ethnicity
Introduction
Socioeconomic disparities in nutrition during childhood can contribute to poor diet quality and weight status [[1], [2], [3]]. Households with higher socioeconomic status (SES) are more likely to have lower food insecurity, healthier food habits, and more frequently meet national dietary guidelines compared to people with low SES; differences which can impact overall health [1,4,5]. Some families from low SES households have to choose between paying for housing, utilities, food, or medicine [6]. To help meet basic needs, social safety net programs, such as Medicaid, the Supplemental Nutrition Assistance Program (SNAP), and the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), provide children from low-income households access to vital medical and nutrition services for health and well-being. Historically, these safety net programs were created to provide health coverage and alleviate food insecurity and hunger for low-income people. Given the dramatic rise in childhood obesity over the past several decades [7], there has been an increased need for these programs to promote healthy dietary quality and weight status in low-income children [[8], [9], [10]].
Medicaid is the nation’s largest source of health coverage for low-income children [11]. Medicaid insures close to 39 million children [[11], [12], [13]]. Evidence shows that Medicaid coverage in early childhood puts individuals on a better health trajectory for long-term health [14,15]. Further, Medicaid coverage for adults has been associated with higher diet quality, lower obesity rates, improved access to care, and reduced mortality [16,17].
SNAP and WIC, 2 of the largest federal nutrition assistance programs, aim to improve nutrition and food security. SNAP provides households with an electronic benefit transfer card they can use to purchase eligible foods for approximately 42 million people, of which 36% are households with children [18]. WIC provides participating individuals a prescribed nutritious food package, nutrition education, and referrals to health and social services [19]. WIC serves >6 million pregnant and postpartum females and children aged ≤5 y per year [20]. Although WIC benefits are not intended to provide supplemental food for entire families, but only for pregnant females and children aged <5 y, when participants are also enrolled in SNAP, the effects of these 2 programs could be jointly beneficial. Prior research shows that participants who stay on SNAP tend to be more food insecure and have poorer diet quality than income-eligible nonparticipants [21]. In contrast, WIC participants tend to have better diet quality the longer they stay on the program, potentially due to the healthy food packages combined with nutrition education [22].
Many social determinants of health, including program participation, are related to health behaviors and health. Examining these predictors in relation to health behaviors and health cannot be modeled or understood easily using traditional regression methods. Classification and regression tree (CART) analysis is a powerful data-driven, machine learning method to identify prediction models without making parametric or distributional assumptions. CART builds trees through a sequence of binary splits that can best sort the sample on outcomes of interest. CART analysis has been used in studies related to child nutrition, such as examining childhood overweight and obesity [[23], [24], [25], [26]], childhood diet quality [27,28], risk factors for low and high birthweights [29,30], and breastmilk sodium concentration and weight loss in breastfeeding infants [31].
CART has also been utilized to investigate the effects of social safety net programs. For instance, it has been employed to identify predictors of WIC benefit usage behavior [32] and to determine factors associated with inpatient visits among Medicaid-insured patients [33]. CART analysis has not yet been used to investigate determinants of health; however, it includes the joint participation in WIC, SNAP, and Medicaid, and their association with child diet quality and weight status in early childhood. Thus, the objective of the current study was to apply CART analysis to identify the demographic, social, and economic determinants of health, including duration in specific safety net programs, that distinguish children from low-income households with higher compared with lower diet quality and higher compared with lower weight status.
Methods
Participants
The WIC Infant and Toddler Feeding Practices Study-2 (WIC ITFPS-2) is a nationally representative, longitudinal study of caregivers (>16 y) and their children followed from around birth up to age 9 y [34]. WIC ITFPS-2 is designed to examine feeding practices, the associations between WIC services and those practices, and the health and nutrition behaviors of children receiving WIC [34]. In 2013, study caregivers were recruited in person from 80 WIC sites across 27 states and territories. WIC sites were selected using a stratified 2-stage sampling approach [35]. Inclusion criteria included enrolling in WIC for the first time for that pregnancy or infant, and the ability to complete interviews in either English or Spanish. Exclusion criteria included: the child being aged >2.5 mo at the time of recruitment; being an adolescent mother aged <16 y; being a mother in foster care at the time of enrollment; and being a foster parent enrolling a foster infant.
Participants received a prenatal interview and ≤16 postnatal interviews during the first 5 y (1 mo, 3 mo, 5 mo, 7 mo, 9 mo, 11 mo, 13 mo, 15 mo, 18 mo, 24 mo, 30 mo, 36 mo, 42 mo, 48 mo, 54 mo, and 60 mo). Interviews were conducted by telephone by trained interviewers in English or Spanish. Interview questions included sociodemographic information, breastfeeding, and other feeding practices, as well as participation in WIC, SNAP, and Medicaid. Additionally, a 24-h dietary recall was collected at each time, except at 30 mo, 42 mo, and 54 mo. The study followed children at ages 6 y and 9 y, but the current analysis focuses on publicly available data of the core and supplemental samples through 5 y. The national study was approved by the Westat Institutional Review Board and the United States Office of Management and Budget. The national study is registered at clinicaltrials.gov as NCT02031978.
Dietary intake
The majority of interviews after the prenatal interview included a 24-h dietary recall, administered over the phone, using the USDA automated multiple-pass method [36]. During the automated multiple-pass method, a caregiver was asked to recall all their child’s dietary intake for the previous day, which could include the weekday or weekend. The caregiver was asked to report all foods, beverages, and dietary supplements for each eating event [37]. For daycare foods, the caregiver obtained any missing details about foods from a knowledgeable source (i.e., daycare provider). If they do not know the amounts, the amounts are estimated based on the USDA food and nutrient database for dietary studies, 5.0 [38].
Healthy eating index-2020
Healthy Eating Index-2020 (HEI-2020) total scores were computed using the population ratio method [39]. The HEI-2020 has been revised to reflect the 2020–2025 Dietary Guidelines for Americans and is considered a valid tool for assessing diet quality among individuals aged ≥2 y [40]. The total HEI-2020 score at 2–5 y was a primary outcome variable. In addition, the HEI-2020 adequacy subscale (sum of scores for total fruits, whole fruits, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood and plant proteins, and fatty acids) and the HEI-2020 moderation subscale (sum of scores for refined grains, sodium, added sugars, and saturated fats) were used as secondary outcomes.
BMI z-scores
As part of regular clinic visits, WIC sites use standardized protocols to directly measure outcomes for most enrolled children. For participants who have left WIC, the study attempted to collect weight and length information from the child’s health care provider. BMI z-scores (BMIz) were calculated at 2–5 y using the Centers for Disease Control and Prevention age- and sex-specific growth charts [41] and are used as a primary outcome.
Predictor variables
To capture the extent of program participation, a variable for the duration of Medicaid, SNAP, and WIC participation was calculated at each time point (24 mo, 36 mo, 48 mo, and 60 mo) using all prior survey rounds where the question was asked (Supplemental Table 1). The percentage of interviews where a caregiver reported that the child was receiving each program was used to define the duration groups, as has been done in previous studies [[42], [43], [44]]. Duration of program participation was then categorized using CART analyses, as described in the statistical analysis section.
Demographic characteristics of the study child, including sex, birth weight, and race and ethnicity, were collected from the caregiver at the first postnatal interview. Race and ethnicity were used in part as a marker of potential social and structural determinants of health that were otherwise unmeasured and could not be controlled for. Other sociodemographic variables collected from the study mother included maternal age at child’s birth, maternal race and ethnicity, language preference, maternal BMI, gestational diabetes, maternal depression score at the 3-mo visit, and household size. Variables collected annually included maternal BMI, marital status, household size, household income, household food security, and maternal employment status. Demographic characteristics were included using the closest in time measurement to the outcome to maximize their predictive relevance to the outcome and account for proximal confounding. Conditional multiple imputation was conducted to incorporate information from individuals with partial data using a conservative count of 10 imputations.
Statistical analysis
CART analysis was implemented using a minimum terminal node size of 10% of the sample with cross-validated complexity parameter selection that pruned trees by selecting the smallest tree within a 1 SE equivalence margin of the tree with the lowest coefficient of variation. This process identified a sequence of binary splits among the candidate variables that best partitioned the sample on differences in the outcomes (diet quality and BMIz) as measured by maximizing the between-nodes sum-of-squares. This process led to a tree with a root node (full sample) from which branches emerged, and derivative nodes at each point were developed, and subgroups split into terminal nodes [45].
For ages 2 y, 3 y, 4 y, and 5 y, a separate CART analysis was fit for each age and each outcome (diet quality and BMIz). CART analyses were conducted with all program participation variables (Medicaid, SNAP, and WIC) included at the same time to explore joint participation between variables. CART was preferred to linear regression because of the possibility of complex interactions even among the first predictors. Additionally, sensitivity analyses were done in which the program participation variables were included separately. Duration of program participation was categorized into:
-
•
Always: reporting participation in the program in the current and all previous timepoints
-
•
Sporadic: reporting participation and nonparticipation
-
•
None: reporting nonparticipation at each timepoint
Analyses were repeated using these duration categories for Medicaid (always/sporadic compared with none), SNAP (always, sporadic, and none), and WIC (always compared with sporadic), selected based on proportion of sample in each category and distribution of outcomes across the categories (Supplemental Table 2).
In all analyses, outcome-specific survey weights were used to have the sample represent the population and compensate for both the unequal sampling probabilities and nonresponse. In addition, estimation incorporated balanced replicated weights as recommended by the WIC ITFPS-2 study data use guidelines [35] to account for the complex survey design. The available sample size of 3051 participants provides 80% power to detect weak to strong associations between program duration and outcomes analogous to a strength of correlation of 0.06 or larger with the full sample, 0.08 or larger within a half-sample split, and 0.11 or larger within a quarter-sample split. The data analysis was conducted using the rpart statistical package in R version 4.1.24 [46].
Results
Most caregivers in the sample were aged between 20 y and 25 y, were Hispanic, spoke English, and had a high school education or more (Table 1). The average HEI-2020 score (out of 100 points) was 56.0–57.6 between 2 and 5 y. The mean child BMIz score at 2 y was 0.69, decreasing to 0.61 by 5 y. Most participants were on WIC at 2 y, and this decreased to largely sporadic participation at 4–5 y. SNAP participation was more intermittent, especially at older years (Supplemental Table 2). More participants were on Medicaid sporadically and always compared to never across all years.
TABLE 1.
Characteristics for participants in the Special Supplemental Nutrition Program for Women, Infants and Children Infant Toddler Feeding Practices Study-2.
| Characteristic1 | Unweighted N = 3051 | Weighted N = 413,211 |
|---|---|---|
| Maternal age at childbirth (n, %)2 | ||
| 16–19 y | 355 (11.6) | 49,442 (12.0) |
| 20–25 y | 1256 (41.2) | 165,655 (40.1) |
| 26 y or older | 1440 (47.2) | 198,115 (47.9) |
| Maternal race (n, %) | ||
| Hispanic | 1217 (39.9) | 197,419 (47.8) |
| Non-Hispanic White | 895 (29.3) | 109,857 (26.6) |
| Non-Hispanic African American | 769 (25.2) | 82,432 (19.9) |
| Non-Hispanic other | 170 (5.6) | 23,504 (5.7) |
| Language preference (n, %) | ||
| Spanish | 569 (18.7) | 85,435 (22.2) |
| English | 2479 (81.3) | 299,263 (77.8) |
| Marital status at enrollment (n, %) | ||
| Married | 910 (29.8) | 137,991 (33.3) |
| Not married3 | 2141 (70.2) | 275,220 (66.6) |
| Maternal education level at enrollment (n, %) | ||
| None through grade 11 | 741 (24.4) | 100,996 (24.5) |
| High school | 1166 (38.3) | 156,343 (37.9) |
| More than high school | 1135 (37.3) | 155,458 (37.7) |
| Maternal employment status at 6 mo child age (n, %)4 | ||
| Full-time | 563 (21.3) | 74,390 (19.9) |
| Part-time | 530 (20.0) | 70,465 (18.9) |
| Not working for pay | 1551 (58.7) | 228,846 (61.2) |
| Maternal BMI at screening (n, %) | ||
| Normal or underweight | 1359 (44.5) | 188,500 (45.6) |
| Overweight | 833 (27.3) | 101,023 (24.4) |
| Obese | 859 (28.2) | 123,688 (29.9) |
| Household size at 6 mo child age (n, %)4 | ||
| 2 people | 241 (9.1) | 32,059 (8.6) |
| 3 people | 722 (27.1) | 100,618 (27.0) |
| 4 people | 712 (26.8) | 104,964 (28.1) |
| 5 people or more | 985 (37.0) | 135,287 (36.3) |
| Household poverty level at enrollment (n, %)4,5 | ||
| 75% of the poverty guideline or below | 1956 (64.1) | 257,597 (62.3) |
| Above 75% but <130% | 809 (26.5) | 111,536 (27.0) |
| Above 130% of the poverty guideline | 286 (9.4) | 44,079 (10.7) |
| Household food security at enrollment | ||
| Food secure | 1986 (65.5) | 272,571 (66.0) |
| Low food security | 677 (22.3) | 93,682 (22.7) |
| Very low food security | 368 (12.1) | 46,958 (11.4) |
| Child sex, female (n, %) | 1495 (49.0) | 198,129 (47.9) |
| Child weight at birth (g) | ||
| Low (<2.5 kg) | 222 (7.3) | 32,341 (7.8) |
| Normal (2.5 kg to <4.5 kg) | 2788 (91.4) | 375,350 (90.9) |
| High (>4.5 kg) | 41 (1.3) | 5520 (1.3) |
| BMIz score at 2 y (mean, SD) | 0.70 (1.27) | 0.69 (1.28) |
| BMIz score at 3 y (mean, SD) | 0.50 (1.32) | 0.52 (1.29) |
| BMIz score at 4 y (mean, SD) | 0.49 (1.39) | 0.53 (1.38) |
| BMIz score at 5 y (mean, SD) | 0.56 (1.30) | 0.61 (1.28) |
| HEI-2020 at 2 y (mean, SD) | 56.3 (11.1) | 56.7 (10.9) |
| HEI at 3 y (mean, SD) | 57.5 (12.1) | 57.6 (12.1) |
| HEI at 4 y (mean, SD) | 56.6 (12.5) | 57.0 (12.7) |
| HEI at 5 y (mean, SD) | 55.4 (12.7) | 56.0 (12.6) |
Abbreviations: BMI z-score, body mass index z-score; HEI-2020, Healthy Eating Index-2020; SD, standard deviation.
Values are means ± SDs or frequency (percentage)
Maternal includes other primary caregivers if not the mother (<1% of respondents are caregivers other than the infant’s biological mother at the time of enrollment).
Not married includes divorced, widowed, or separated.
Because of missing values, the total n is not the same for all variables.
Income at 100% of the federal poverty level was $23,550 for a family of 4 in 2013.
At 2 y, 3 y, and 5 y old, race and ethnicity were the first predictors to distinguish total HEI-2020 scores (Table 2). At 3 y, the second and third predictors fluctuated between WIC and SNAP duration, maternal education, race and ethnicity, and income. At 4 y old, WIC duration was the first predictor of HEI-2020, then SNAP duration and race and ethnicity.
TABLE 2.
Predictors of BMI z-score and Healthy Eating Index-2020 outcomes by classification and regression tree analysis for children in the Special Supplemental Nutrition Program for Women, Infants and Children Infant Toddler Feeding Practices Study-2.
In examining HEI-2020 subscales, at 2 y, HEI adequacy was primarily predicted by maternal education, race and ethnicity, then WIC duration (Table 2). At 3 y, maternal age was the first predictor to distinguish HEI adequacy, followed by WIC duration. At 4 y, WIC duration, then SNAP duration, and race and ethnicity defined HEI adequacy. After race and ethnicity, SNAP duration, and then maternal BMI were important predictors of HEI adequacy at 5 y. Race and ethnicity, then WIC duration, were the first 2 predictors to distinguish HEI moderation across 2–5 y. SNAP duration was the third most important predictor at ages 2 and 3 for HEI moderation.
Birth weight was consistently the most important predictor distinguishing BMIz across 2–5 y (Table 2). After birth weight, maternal BMI, gender of the child, and SNAP duration were important predictors of BMIz at 4 y old. At 5 y, maternal BMI was the second and race and ethnicity were the third predictors to distinguish BMIz.
Although there was variation in selected predictors across child ages, the directions of the associations were generally consistent within an outcome, as illustrated in Figure 1 and Supplemental Figures 1–3. Specifically, for BMIz at 4 y (Figure 1A), higher birthweight (≥3.4 kg) was associated with a higher BMIz, which was then split by sex, with higher BMIz among female children. Among those with lower birthweight, having a mother who was normal weight or underweight was associated with lower BMI and among those with a mother who was overweight or obese, sporadically participating in SNAP corresponded to having a lower BMIz, whereas those always or never participating in SNAP having higher BMIz (0.32 BMIz score compared with 0.73 BMIz score). For HEI at 4 y (Figure 1B), those with sporadic participation in WIC had lower HEI as compared to those always participating in WIC (55 compared with 60). With no participation in SNAP, participants had even lower HEI as compared to those sporadically or always participating in SNAP (54 compared with 58). Among those who were always participating in WIC, being of Hispanic descent and predominantly Spanish speaking was associated with higher HEI compared with those who were not. The pattern of association was the same for the HEI adequacy subscale (Figure 1C), but not for the HEI moderation subscale (Figure 1D), which split first on race and ethnicity (with higher HEI moderation among Hispanic-Spanish speaking children) and then on WIC duration, with higher scores among those always on WIC.
FIGURE 1.
Classification trees for 4-y-old outcomes in the Special Supplemental Nutrition Program for Women, Infants and Children (WIC) Infant Toddler Feeding Practices Study-2: (A) BMI z-scores (BMIz); (B) Healthy Eating Index-2020 (HEI) scores; (C) HEI adequacy (HEIad) scores; and (D) HEI moderation (HEImod) scores. BMI, body mass index; SNAP, Supplemental Nutrition Assistance Program.
Discussion
Child diet quality was associated with several social determinants of health predictors, which included measures of health, race and ethnicity, and food assistance program participation, particularly WIC. The key predictor distinguishing weight status was birth weight or maternal BMI, whereas race and ethnicity, and WIC and SNAP duration were the dominant predictors of total HEI-2020 scores. Race and ethnicity and longer WIC duration were the primary predictors of higher adherence scores on the HEI subscale for moderation, which reflects how much a person adheres to the foods to limit in the United States Dietary Guidelines for Americans, such as refined grains, sodium, added sugars, and saturated fats [40].
Diet quality was primarily predicted by race and ethnicity and, to a lesser extent, by the duration of participation in food assistance programs, particularly consistent participation in WIC. These findings align with previous research showing an association between race and ethnicity and diet quality [47], with higher diet quality among Spanish-speaking Hispanics compared to English-speaking Hispanics [48], and diets of Hispanic children having better nutrient densities and lower caloric densities than those of non-Hispanic (NH) Whites, whereas NH Black children had poorer nutrient densities [49]. Additionally, within the normal weight category, Mexican American children have been shown to have better diet quality than NH Black children [50]. The differences in diet quality observed within and across racial and ethnic groups can be partially attributed to whether the primary caregiver was born in the United States [47], the degree of acculturation [48,51], and the child’s BMIz categorization [50]. Additionally, there has been a documented association between lower SES and poorer diet quality, with a growing trend observed among Black children [52] – a pattern that was also evident in this study population.
Although the mean HEI scores for children aged 2 through 5 in this study population were consistently lower than the HEI-2020 and HEI-2015 United States population average of 61 out of 100 for ages 2 through 4 [53,54], which is consistent with previous research [50,51], this study identified subpopulations of low-income children who may have better diet quality, specifically children whose caregivers were Hispanic-Spanish speaking and also participated in WIC longer. Notably, prior research has shown that being Hispanic-Spanish speaking is associated with higher odds of longer WIC participation, which in turn is associated with better diet quality in children [43]. It remains unclear why this subgroup of participants consistently has better diet quality, but it may be due to a greater appreciation of the WIC food package and/or the nutrition education provided for the full period of WIC eligibility to age 5. Given these findings, the possible importance of long-term WIC participation in improving diet quality cannot be downplayed, particularly for some racial and ethnic groups. By providing a nutritious food package, WIC may potentially reduce the intake of nutrients of concern, such as added sugar and sodium, while promoting the consumption of healthy food groups, such as fruits and whole grains, as highlighted in the Dietary Guidelines for Americans [53].
Findings in previous literature [55,56] have shown that children from households participating in SNAP had lower HEI scores at ages 2–5 y compared with income-eligible nonparticipants, suggesting poorer diet quality. This study observed a similar but more nuanced pattern. Specifically, sporadic or continuous participation in SNAP was associated with lower diet quality, but only among children with intermittent WIC participation or those who were NH White. In contrast, children who were never on SNAP but consistently participated in WIC or belonged to racial and ethnic groups other than NH White had higher diet quality. Notably, SNAP participants have been reported to experience higher food insecurity rates compared to nonparticipants, although this may not be causally attributed to SNAP use [[57], [58], [59], [60]]. Although individuals participating in 2 or more safety net programs are more likely to be food insecure [61], have lower incomes, and experience economic hardships [62], consistent joint participation in WIC may improve the diet quality of SNAP participants, as shown by our study findings. Regarding the sporadic participation trend that was observed for SNAP participants, previous research shows that half of all participants stop receiving benefits within 12 mo, and two-thirds leave within 2 y [63]. In addition, SNAP participants often experience income volatility and job turnover [63], affecting program eligibility, and the SNAP program has requirements for frequent recertification of eligibility, leading to sporadic participation, which potentially affects long-term impacts on child diet quality in these households and the assessment of those impacts itself. Finally, although previous studies of adults [16,17] have indicated an association between Medicaid coverage and better diet quality, this study found that the duration of Medicaid did not significantly impact child diet quality. This may be partly because this study asked about household Medicaid and household SNAP participation rather than direct participation by the child.
CART analysis identified birth weight as the primary predictor for early childhood weight status, which aligns with previous studies that observed higher BMIz scores in children with increased birth weights [64,65] and accelerated weight gain patterns [66]. Additionally, in a similar cohort that was predominantly from low-income and multiethnic households, increased birth weight was independently predictive of higher childhood obesity rates [67]. Other studies that included birth weight in their models also identified it as a strong predictor of childhood obesity [68,69]. In addition to birth weight, increased maternal BMI was associated with higher BMIz scores [65], doubling the risk of obesity at 2–4 y among low-income children enrolled in WIC [70], consistent with this study’s findings.
Strengths of this study include the use of a national, longitudinal cohort with information about Medicaid, SNAP, and WIC self-reported participation collected at 8 different time points through the first 5 y of life. In addition, BMIz and HEI-2020 scores, which require time-intensive data collection, were available at 2–5 y of age, which are not available in many longitudinal data sets that include safety net program duration. Because participants in this study were predominantly WIC participants recruited from WIC sites, their SNAP and Medicaid participation may be systematically different and influenced by self-selection bias, potentially limiting the representativeness of households participating in Medicaid and SNAP within this population of low-income households. Because Medicaid, SNAP, and WIC participation were not assessed between study visits, it is possible that the duration groups do not accurately reflect sporadic program participation or churning, a topic worth further research. Further, because self-reports of program participation may be underreported, there may have been a misclassification of some people who participate less extensively as nonparticipants [71]. Further, there might be a selection as to who stays on the programs or leaves them. Such selection may lead to systematic differences in the estimates of the effects of some programs. Despite this limitation, the extensive longitudinal data collection of WIC ITFPS-2 makes it well-suited to use CART analysis to investigate the social determinants of health, including duration in specific safety net programs, that distinguish children from low-income households with higher compared with lower diet quality and higher compared with lower weight status. Additionally, a single day of 24-h dietary recall was used to calculate HEI-2020 scores, which may limit the ability to assess usual intake and may contribute to potential energy underreporting [72].
In conclusion, child diet quality and weight status were associated with several social determinants of health and often occurred at the intersection of multiple predictors, which included maternal BMI, race and ethnicity, and food assistance program duration of participation, particularly WIC. The CART categorization of social safety net programs into ordinal categories provides a framework for examining needs specific to the duration of participation. These findings can help inform tailored recommendations, such as staying on the WIC program for longer, for improving diet quality and weight status of young children.
Author contributions
The authors’ responsibilities were as follows – LEA: conceptualized, designed the study, drafted the manuscript, and revised the manuscript; CDA: designed the study, carried out the analyses, critically reviewed, and revised the manuscript; CD: assisted in the investigation, helped to draft the manuscript, and revised the manuscript; LDR, SEW, MB: designed the study, critically reviewed, and revised the manuscript; EAF: designed the study, critically reviewed, and revised the manuscript; and all authors: read and approved the final manuscript.
Data availability
Data described in the manuscript and a code book are publicly and freely available without restriction at: https://data.nal.usda.gov/dataset/wic-infant-and-toddler-feeding-practices-study-2-wic-itfps-2-prenatal-infant-year-5-year-datasets and https://osf.io/suahz/.
Funding
This publication was supported by grant #81359 from Healthy Eating Research, a national program of the Robert Wood Johnson Foundation and the USDA/National Institute of Food and Agriculture Hatch Project# CA-D-NTR-2689-H. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Robert Wood Johnson Foundation or the USDA.
Conflict of interest
The authors report no conflicts of interest.
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.cdnut.2025.107505.
Appendix A. Supplementary data
The following is the Supplementary data to this article:
References
- 1.Darmon N., Drewnowski A. Does social class predict diet quality? Am. J Clin. Nutr. 2008;87(5):1107–1117. doi: 10.1093/ajcn/87.5.1107. [DOI] [PubMed] [Google Scholar]
- 2.Pechey R., Monsivais P. Socioeconomic inequalities in the healthiness of food choices: exploring the contributions of food expenditures. Prev. Med. 2016;88:203–209. doi: 10.1016/j.ypmed.2016.04.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Vazquez C.E., Cubbin C. Socioeconomic status and childhood obesity: a review of literature from the past decade to inform intervention research. Curr. Obes. Rep. 2020;9(4):562–570. doi: 10.1007/s13679-020-00400-2. [DOI] [PubMed] [Google Scholar]
- 4.Chaparro M.P., Cruthirds S., Bell C.N., Wallace M.E. State-level socioeconomic racial inequity and food insecurity in the U.S. Am. J Prev. Med. 2022;63(6):971–978. doi: 10.1016/j.amepre.2022.06.019. [DOI] [PubMed] [Google Scholar]
- 5.Torres R., Santos E., Orraca L., Elias A., Palacios C. Diet quality, social determinants, and weight status in Puerto Rican children aged 12 years. J Acad. Nutr. Diet. 2014;114(8):1230–1235. doi: 10.1016/j.jand.2014.01.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Fiese B.H., Koester B.D., Waxman E. Balancing household needs: the non-food needs of food pantry clients and their implications for program planning. J Fam. Econ. Issues. 2014;35(3):423–431. [Google Scholar]
- 7.Ogden C.L., Carroll M.D., Fakhouri T.H., Hales C.M., Fryar C.D., Li X., et al. Prevalence of obesity among youths by household income and education level of head of household — United States 2011-2014. MMWR Morb. Mortal. Wkly. Rep. 2018;67(6):186–189. doi: 10.15585/mmwr.mm6706a3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Tester J.M., Leung C.W., Crawford P.B. Revised WIC food package and children’s diet quality. Pediatrics. 2016;137(5) doi: 10.1542/peds.2015-3557. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Verghese A., Raber M., Sharma S. Interventions targeting diet quality of Supplemental Nutrition Assistance Program (SNAP) participants: A scoping review. Prev. Med. 2019;119:77–86. doi: 10.1016/j.ypmed.2018.12.006. [DOI] [PubMed] [Google Scholar]
- 10.Chaparro M.P., Anderson C.E., Crespi C.M., Whaley S.E., Wang M.C. The effect of the 2009 WIC food package change on childhood obesity varies by gender and initial weight status in Los Angeles County, Pediatr. Obes. 2019;14(9) doi: 10.1111/ijpo.12526. [DOI] [PubMed] [Google Scholar]
- 11.Donohue J.M., Cole E.S., James C.V., Jarlenski M., Michener J.D., Roberts E.T. The US Medicaid program: Coverage, financing, reforms, and implications for health equity. JAMA. 2022;328(11):1085–1099. doi: 10.1001/jama.2022.14791. [DOI] [PubMed] [Google Scholar]
- 12.Sommers B.D., Grabowski D.C. What is Medicaid? More than meets the eye. JAMA. 2017;318(8):695–696. doi: 10.1001/jama.2017.10304. [DOI] [PubMed] [Google Scholar]
- 13.US Department of Health and Human Services . 2023. Who enrolls in Medicaid & CHIP?https://www.medicaid.gov/state-overviews/scorecard/who-enrolls-medicaid-chip/index.html Available from. [Google Scholar]
- 14.Currie J. Child health as human capital. Health Econ. 2020;29(4):452–463. doi: 10.1002/hec.3995. [DOI] [PubMed] [Google Scholar]
- 15.Currie J., Decker S., Lin W. Has public health insurance for older children reduced disparities in access to care and health outcomes? J Health Econ. 2008;27(6):1567–1581. doi: 10.1016/j.jhealeco.2008.07.002. [DOI] [PubMed] [Google Scholar]
- 16.Nguyen B.T., Han X., Jemal A., Drope J. Diet quality, risk factors and access to care among low-income uninsured American adults in states expanding Medicaid vs. states not expanding under the affordable care act. Prev. Med. 2016;91:169–171. doi: 10.1016/j.ypmed.2016.08.015. [DOI] [PubMed] [Google Scholar]
- 17.Sommers B.D., Baicker K., Epstein A.M. Mortality and access to care among adults after state Medicaid expansions, N Engl. J Med. 2012;367(11):1025–1034. doi: 10.1056/NEJMsa1202099. [DOI] [PubMed] [Google Scholar]
- 18.US Department of Agriculture . 2023. Supplemental Nutrition Assistance Program (SNAP)https://www.fns.usda.gov/data-research/data-visualization/snap/household-characteristics Available from: [Google Scholar]
- 19.US Department of Agriculture, WIC Program. 2023. https://www.fns.usda.gov/wic Available from: [Google Scholar]
- 20.US Department of Agriculture . 2024. Data Tables WI.https://www.fns.usda.gov/pd/wic-program Available from. [Google Scholar]
- 21.Andreyeva T., Tripp A.S., Schwartz M.B. Dietary quality of Americans by Supplemental Nutrition Assistance Program participation status: A systematic review. Am. J Prev. Med. 2015;49(4):594–604. doi: 10.1016/j.amepre.2015.04.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Weinfield N.S., Borger C., Au L.E., Whaley S.E., Berman D., Ritchie L.D. Longer participation in WIC is associated with better diet quality in 24-month-old children. J Acad. Nutr. Diet. 2020;120(6):963–971. doi: 10.1016/j.jand.2019.12.012. [DOI] [PubMed] [Google Scholar]
- 23.Riedel C., von Kries R., Buyken A.E., Diethelm K., Keil T., Grabenhenrich L., et al. Overweight in adolescence can be predicted at age 6 years: a CART analysis in German cohorts. PLOS One. 2014;9(3) doi: 10.1371/journal.pone.0093581. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kitsantas P., Gaffney K.F. Risk profiles for overweight/obesity among preschoolers. Early Hum. Dev. 2010;86(9):563–568. doi: 10.1016/j.earlhumdev.2010.07.006. [DOI] [PubMed] [Google Scholar]
- 25.Toschke A.M., Beyerlein A., von Kries R. Children at high risk for overweight: a classification and regression trees analysis approach. Obes Res. 2005;13(7):1270–1274. doi: 10.1038/oby.2005.151. [DOI] [PubMed] [Google Scholar]
- 26.Gamliel A., Ziv-Baran T., Siegel R.M., Fogelman Y., Dubnov-Raz G. Using weight-for-age percentiles to screen for overweight and obese children and adolescents. Prev. Med. 2015;81:174–179. doi: 10.1016/j.ypmed.2015.08.017. [DOI] [PubMed] [Google Scholar]
- 27.Miller T.L., Jacobson D.L., Somarriba G., Neri D., Kurtz-Vraney J., Graham P., et al. A multicenter study of diet quality on birth weight and gestational age in infants of HIV-infected women. Matern. Child Nutr. 2017;13(4) doi: 10.1111/mcn.12378. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Yannakoulia M., Lykou A., Kastorini C.M., Saranti Papasaranti E., Petralias A., Veloudaki A., et al. Socio-economic and lifestyle parameters associated with diet quality of children and adolescents using classification and regression tree analysis: the DIATROFI study. Public Health Nutr. 2016;19(2):339–347. doi: 10.1017/S136898001500110X. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Zhang C., Garrard L., Keighley J., Carlson S., Gajewski B. Subgroup identification of early preterm birth (ePTB): informing a future prospective enrichment clinical trial design. BMC Pregnancy Childbirth. 2017;17(1):18. doi: 10.1186/s12884-016-1189-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Garmendia M.L., Mondschein S., Montiel B., Kusanovic J.P. Trends and predictors of birth weight in Chilean children. Public Health. 2021;193:61–68. doi: 10.1016/j.puhe.2021.01.019. [DOI] [PubMed] [Google Scholar]
- 31.Manganaro R., Marseglia L., Mamì C., Palmara A., Paolata A., Loddo S., et al. Breast milk sodium concentration, sodium intake and weight loss in breast-feeding newborn infants, Br. J Nutr. 2007;97(2):344–348. doi: 10.1017/S0007114507280572. [DOI] [PubMed] [Google Scholar]
- 32.Woelfel M.L., Abusabha R., Pruzek R., Stratton H., Chen S.G., Edmunds L.S. Barriers to the use of WIC services. J Am. Diet Assoc. 2004;104(5):736–743. doi: 10.1016/j.jada.2004.02.028. [DOI] [PubMed] [Google Scholar]
- 33.Wu S.S., Perry A., Zimmerman N.M., Bryant G. Predictors of flare-related inpatient or emergency department stay in systemic lupus erythematosus: A real-world analysis of Medicaid claims in the United States, J Manag. Care Spec. Pharm. 2024;30(1):61–70. doi: 10.18553/jmcp.2024.30.1.61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.US Department of Agriculture . 2017. WIC Infant and Toddler Feeding Practices Study 2: Infant year report.https://www.fns.usda.gov/data-research?keywords=itfps&sort_bef_combine=created_1_DESC [Google Scholar]
- 35.US Department of Agriculture . 2022. WIC Infant and toddler feeding practices Study 2: fifth year report.https://www.fns.usda.gov/data-research?keywords=itfps&sort_bef_combine=created_1_DESC [Google Scholar]
- 36.Moshfegh A.J., Rhodes D.G., Baer D.J., Murayi T., Clemens J.C., Rumpler W.V., et al. The US Department of Agriculture Automated Multiple-Pass Method reduces bias in the collection of energy intakes. Am. J Clin. Nutr. 2008;88(2):324–332. doi: 10.1093/ajcn/88.2.324. [DOI] [PubMed] [Google Scholar]
- 37.US Department of Agriculture. DoAFaN. WIC Infant and toddler feeding practices Study 2: second year report. Available from: https://www.fns.usda.gov/data-research?keywords=itfps&sort_bef_combine=created_1_DESC.
- 38.Montville J.B., Ahuja J.K., Martin C.L., Heendeniya K.Y., Omolewa-Tomobi G., Steinfeldt L.C., et al. USDA food and nutrient database for dietary studies (FNDDS), 5.0. Procedia. Food Sci. 2013;2:99–112. [Google Scholar]
- 39.National Cancer Institute . 2024. Healthy Eating Index SAS Code.https://epi.grants.cancer.gov/hei/sas-code.html Available from. [Google Scholar]
- 40.Shams-White M.M., Pannucci T.E., Lerman J.L., Herrick K.A., Zimmer M., Meyers Mathieu K., et al. Healthy eating Index-2020: review and update process to reflect the Dietary Guidelines for Americans, 2020-2025. J Acad. Nutr. Diet. 2023;123(9):1280–1288. doi: 10.1016/j.jand.2023.05.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Control CfD. About Child and Teen BMI. Available from: https://www.cdc.gov/growth-chart-training/hcp/computer-programs/.
- 42.Insolera N., Cohen A., Wolfson J.A. SNAP and WIC participation during childhood and food security in adulthood, 1984-2019. Am. J Public Health. 2022;112(10):1498–1506. doi: 10.2105/AJPH.2022.306967. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Anderson C.E., Martinez C.E., Ritchie L.D., Paolicelli C., Reat A., Borger C., et al. Longer special Supplemental Nutrition Program for Women, Infants, and Children (WIC) participation duration is associated with higher diet quality at age 5 years. J Nutr. 2022;152(8):1974–1982. doi: 10.1093/jn/nxac134. [DOI] [PubMed] [Google Scholar]
- 44.Borger C., Paolicelli C.P., Sun B. Duration of Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) participation is associated with children’s diet quality at age 3 years, Am. J Prev. Med. 2022;62(6):e343–e350. doi: 10.1016/j.amepre.2021.12.019. [DOI] [PubMed] [Google Scholar]
- 45.Choi S.K., Fram M.S., Frongillo E.A. Very low food security in US households is predicted by complex patterns of health, economics, and service participation. J Nutr. 2017;147(10):1992–2000. doi: 10.3945/jn.117.253179. [DOI] [PubMed] [Google Scholar]
- 46.Therneau T., Atkinson B., Ripley B. 2025. rpart: recursive partitioning and regression trees.https://CRAN.R-project.org/package=rpart Available from: [Google Scholar]
- 47.Othman S.I., Fertig A., Trofholz A., Berge J.M. How time in the US and race/ethnicity shape food parenting practices and child diet quality. Appetite. 2022;171 doi: 10.1016/j.appet.2021.105870. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Au L.E., Arnold C.D., Ritchie L.D., Lin S.K., Frongillo E.A. Differences in Infant diet quality index by race and ethnicity predict differences in later diet quality. J Nutr. 2023;153(12):3498–3505. doi: 10.1016/j.tjnut.2023.10.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Zimmer M.C., Rubio V., Kintziger K.W., Barroso C. Racial/ethnic disparities in dietary intake of U.S. children participating in WIC. Nutrients. 2019;11(11):2607. doi: 10.3390/nu11112607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Thomson J.L., Landry A.S., Tussing-Humphreys L.M., Goodman M.H. Diet quality of children in the United States by body mass index and sociodemographic characteristics. Obes. Sci. Pract. 2020;6(1):84–98. doi: 10.1002/osp4.388. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Molitor F., Doerr C. Diet quality differs by race/ethnicity among mothers and their Children from Supplemental Nutrition Assistance Program-Education households. Health Equity. 2021;5(1):633–636. doi: 10.1089/heq.2021.0007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Ding G., Wen C., Chen Y., Vinturache A., Zhang Y. Diet quality among children. JAMA Pediatr. 2024;178(9):944–946. doi: 10.1001/jamapediatrics.2024.1880. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.US Department of Agriculture . 2020. Dietary Guidelines for Americans; pp. 2020–2025. [Google Scholar]
- 54.Thomson J.L., Tussing-Humphreys L.M., Goodman M.H., Landry A.S. Diet quality in a nationally representative sample of American children by sociodemographic characteristics. Am. J Clin. Nutr. 2019;109(1):127–138. doi: 10.1093/ajcn/nqy284. [DOI] [PubMed] [Google Scholar]
- 55.Singleton C.R., Young S.K., Kessee N., Springfield S.E., Sen B.P. Examining disparities in diet quality between SNAP participants and non-participants using Oaxaca-Blinder decomposition analysis. Prev. Med. Rep. 2020;19 doi: 10.1016/j.pmedr.2020.101134. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Litvak J., Parekh N., Juul F., Deierlein A. Food assistance programs and income are associated with the diet quality of grocery purchases for households consisting of women of reproductive age or young children. Prev. Med. 2020;138 doi: 10.1016/j.ypmed.2020.106149. [DOI] [PubMed] [Google Scholar]
- 57.DeWitt E., Gillespie R., Norman-Burgdolf H., Cardarelli K.M., Slone S., Gustafson A. Rural SNAP participants and food insecurity: how can communities leverage resources to meet the growing food insecurity status of rural and low-income residents? Int. J Environ. Res. Public Health. 2020;17(17):6037. doi: 10.3390/ijerph17176037. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58.Singleton C.R., Fabusoro O., Teran-Garcia M., Lara-Cinisomo S. Change in employment status due to the COVID-19 pandemic, SNAP participation, and household food insecurity among Black and Latino adults in Illinois. Nutrients. 2022;14(8):1581. doi: 10.3390/nu14081581. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59.Harper K., Belarmino E.H., Acciai F., Bertmann F., Ohri-Vachaspati P. Patterns of food assistance program participation, food insecurity, and pantry use among U.S. Households with children during the COVID-19 pandemic. Nutrients. 2022;14(5):988. doi: 10.3390/nu14050988. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60.U.S. Department of Agriculture Economic Research Service . 2024. Supplemental Nutrition Assistance Program (SNAP)https://www.ers.usda.gov/topics/food-nutrition-assistance/supplemental-nutrition-assistance-program-snap/key-statistics-and-research/ Available from: [Google Scholar]
- 61.Chaparro M.P., Lopez M.A. Trends in food insecurity among households with children participating in WIC and SNAP in California, 2005-2017, Matern. Child Health J. 2024;28(2):315–323. doi: 10.1007/s10995-023-03829-8. [DOI] [PubMed] [Google Scholar]
- 62.Urban Institute . 2014. Characteristics of families receiving multiple public benefits.https://www.urban.org/research/publication/characteristics-families-receiving-multiple-public-benefits Available from: [Google Scholar]
- 63.Center on Budget and Policy Priorities . 2018. Most working-age SNAP participants work, but often in unstable jobs.https://www.cbpp.org/research/most-working-age-snap-participants-work-but-often-in-unstable-jobs Available from. [Google Scholar]
- 64.Taveras E.M., Rifas-Shiman S.L., Belfort M.B., Kleinman K.P., Oken E., Gillman M.W. Weight status in the first 6 months of life and obesity at 3 years of age. Pediatrics. 2009;123(4):1177–1183. doi: 10.1542/peds.2008-1149. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 65.Vehapoglu A., Goknar N., Turel O., Torun E., Ozgurhan G. Risk factors for childhood obesity: do the birth weight, type of delivery, and mother’s overweight have an implication on current weight status? World J Pediatr. 2017;13(5):457–464. doi: 10.1007/s12519-017-0030-9. [DOI] [PubMed] [Google Scholar]
- 66.Danner F.W., Toland M.D. The interactive role of socioeconomic status, race/ethnicity, and birth weight on trajectories of body mass index growth in children and adolescents. J Early Adolesc. 2013;33(3):293–314. [Google Scholar]
- 67.Gómez-López L., Van Hulst A., Barnett T.A., Roy-Gagnon M.H., Tremblay A., O’Loughlin J., et al. Does parental body mass index status modify the associations among birth weight, early growth and childhood adiposity? Paediatr. Child Health. 2013;18(2):e2–e9. doi: 10.1093/pch/18.2.e2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 68.Qiao Y., Ma J., Wang Y., Li W., Katzmarzyk P.T., Chaput J.P., et al. Birth weight and childhood obesity: a 12-country study. Int. J Obes. Suppl. 2015;5(2):S74–S79. doi: 10.1038/ijosup.2015.23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 69.Rooney B.L., Mathiason M.A., Schauberger C.W. Predictors of obesity in childhood, adolescence, and adulthood in a birth cohort. Matern. Child Health J. 2011;15(8):1166–1175. doi: 10.1007/s10995-010-0689-1. [DOI] [PubMed] [Google Scholar]
- 70.Whitaker R.C. Predicting preschooler obesity at birth: the role of maternal obesity in early pregnancy. Pediatrics. 2004;114(1):e29–e36. doi: 10.1542/peds.114.1.e29. [DOI] [PubMed] [Google Scholar]
- 71.Meyer B.D., Mok W.K., Sullivan J.X. Household surveys in crisis. J Econ. Perspect. 2015;29(4):199–226. [Google Scholar]
- 72.Thompson F.E., Subar A.F. In: Nutrition in the prevention and treatment of disease. 4th ed. Coulston A.M., Boushey C.J., Ferruzzi M.G., Delahanty L.M., editors. Academic Press; 2017. Chapter 1. Dietary assessment methodology; pp. 5–48. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data described in the manuscript and a code book are publicly and freely available without restriction at: https://data.nal.usda.gov/dataset/wic-infant-and-toddler-feeding-practices-study-2-wic-itfps-2-prenatal-infant-year-5-year-datasets and https://osf.io/suahz/.