Association between Sociodemographic Factors and Dietary Patterns in Children Under 24 Months of Age: A Systematic Review

Background: Understanding early-life complementary feeding dietary patterns and their determining factors could lead to better ways of improving nutrition in early childhood. The purpose of this review was to evaluate evidence of the association between sociodemographic factors and dietary patterns (DPs) in children under 24 months. Methods: Medline (PubMed), Cochrane Central, NICE guidelines, and Trip database were searched for observational studies that evaluated sociodemographic factors and their associations with DP. Results: Seven studies were selected for the present review. High education level among mothers was inversely associated with unhealthy DPs and positively associated with healthy DPs. Higher household income was negatively associated with unhealthy DPs. Four studies showed a positive association between low household income and unhealthy DPs and three studies showed a positive association between higher household income and healthy DPs. Additionally, in younger mothers, body mass index (BMI ≥ 30.0 kg/m2) and number of children were positively associated with unhealthy DPs. Conclusions: This review provides evidence of a positive association between mothers’ higher education level, higher household income, higher maternal age, and healthy dietary patterns as well as a negative association between these factors and unhealthy dietary patterns. Further studies from low- and middle-income countries are needed for comparison with associations showed in this review.


Introduction
Good nutrition through childhood, particularly during the first two years, is crucial to optimum growth, health, and behavioral development [1]. In this sense, during the first six months of life, breast milk fulfills infants' energy and nutrition needs. After six months, complementary feeding, defined as the inclusion of foods and liquids along with breast milk, must guarantee growth and development through the child´s remaining first two years [1][2][3]. Unfortunately, failure to meet these requirements through complementary feeding can often cause nutrient deficiencies, suboptimal growth, delayed development and disease, and may be due to a lack of variety and frequency in the types of foods provided to young children [4][5][6][7]. Sometimes infants cannot be feeding with maternal milk Nutrients 2019, 11,2006 2 of 17 due to maternal human immunodeficiency virus infection (HIV), maternal death or non-intention to breastfeed, thus formula milk would be considered in addition to complementary diet [2].
In general, complementary feeding has been assessed in terms of variety of foods and nutrient needs. For many years, this representation has been considered an important health indicator where each nutrient has a function in the health-disease process [8][9][10][11][12][13]. However, more recently, dietary pattern (DP) analysis, conceptually defined as representations of nutrient and food consumption groups, has emerged as an option to predict risk of disease better than individual foods or nutrients. Therefore, understanding early-life complementary feeding DPs, their determining factors, and their influence on later health could lead to better ways of improving nutrition in early childhood [14][15][16]. Since DPs are difficult to measure, different methodologies ("a priori" and "a posteriori") have been used to derive DPs. For the "a posteriori" approaches, multiple statistical techniques such as factor analysis, cluster analysis, and reduced rank regression analysis [17][18][19][20] have been employed to characterize DPs.
On the other hand, many complex social and demographic factors have been linked to the overall health of a population. The broad concept of socioeconomic status (SES) encompasses some of these factors, including parental education level, occupation, and family income, among others [21]. For this reason, SES should be understood as a compound measure of an individual's economic and sociological status that has great impact on the health of the population; however, we have to consider the influence of many other factors associated with DPs in early childhood. Previous studies have reported evidence of diet-socioeconomic status gradients across all levels of human development and that lower socioeconomic status is strongly linked to unhealthy DPs [22]. These studies have also described unhealthy DPs in children from low-income and poorly educated families [23]. Nonetheless, studies in low-and middle-income countries have contradicted the correlation between socioeconomic factors (income and education) and DP [24,25].
Additionally, maternal and family influences on complementary feeding practices are considerably more important than other factors. Women who follow healthy dietary recommendations are more likely to have infants whose diets are compatible with "infant guideline" pattern scores [26]. On the other hand, women whose diets are characterized by unhealthy food consumption are more likely to have infants with similar DPs, some of which are detectable even before birth [27,28].
In this context, links have been reported between younger mothers (<30 years old), multiparity, smoking during pregnancy, body mass index (BMI) ≥30.0 kg/m 2 ), and unhealthy DPs [29][30][31][32]. In general, "healthy diets" are found to go hand in hand with higher household income while "unhealthy diets" have been linked to lower household income [25,33]. The literature suggests that in families with low household income, children's diet is characterized by low consumption of fruit and vegetables and high intakes of ultra-processed foods (commonly energy-dense foods), which have been linked to increased risk of children becoming overweight or obese [32,34]. Other sociodemographic factors, such as place of residence (rural or urban), number of children in the family, employment status, and parents' diet have been suggested as important elements to complementary DP composition in children [35][36][37][38]. In addition, healthy DPs high in fruit, vegetables, and whole grains have been linked to high maternal income and education levels, and subsequently to good health in later stages of life [39,40]. While some studies reveal that parents and close relatives play an important role in early nutrition, the food industry is known for targeting mothers and encouraging the early introduction [41] of unhealthy food products and beverages that increase the risk of cardiovascular and metabolic diseases later in life [42].
As we noted above, the importance of nutrition for development in the first two years has been well-documented; however, there is little information on the impact of sociodemographic factors and the formation of DP in this period of life [43]. We now know that parents and children tend to share DPs [44], which is why establishing the sociodemographic factors that influence healthy or unhealthy DPs is more relevant than ever [45].
Identifying dietary patterns at an early stage would allow greater specificity in our recommendations and help us promote healthier complementary feeding DPs and other healthy lifestyles from the beginning of life [46]. Therefore, this review evaluates evidence of the association between sociodemographic factors and DP in children under 24 months.

Types of Studies
We included observational studies (cohort) reported as full-text that were written in English and focused on the association between sociodemographic factors and DP in children under 24 months.

Types of Participants
The criteria of the studies included in this research focused on the following maternal sociodemographic factors: education, age, occupation, household income, marital status, ethnicity, parity, smoking during pregnancy, and BMI. We included studies carried out without time restrictions in countries with high, middle, and low incomes. We also included studies that used statistical dimensionality reduction techniques to identify DP, such as cluster analysis, factor analysis (principal component analysis), latent class analysis and reduced rank regression. Documents not available in English were excluded. Case-control studies, cross sectional, intervention studies, reviews, letters, and case reports were not considered.
Smoking status: never smoked, past smoker, and current smoker.

Search Strategy
Individual strategies were used to research the databases MEDLINE (PubMed), Cochrane Library, NICE guidelines, TRIP database, and CENETEC. A hand searching was made into the reference of the included studies. The final search was updated in April 2019. The articles were selected for full-text reading and were screened individually for potentially relevant studies that may have been missed. The database search strategy is shown in Figure 1. As a way of including infants under 24 in our Pub Med database search, we used an age filter from 0 to 23 months and discarded search results with children older than 24 months.
The strategic search was completed under the PICOS framework, which stands for: Population = children under 24 months; Intervention = sociodemographic factors; Comparison = different DP; Outcome = DP; Study = observational. The Database search strategy included different word combinations for each database and references were managed with Endnote ® Web software v. 3.1.1. (Thomson ResearchSoft, California, United States) Duplicated references were eliminated.

Selection of Studies
Selected studies were assessed by two methodological and clinical researchers (G.-C.C. and M.-S.L.) who independently reviewed titles, abstracts, and selection criteria (the selection process was done in Excel). Duplicated articles and those that did not meet inclusion criteria were not considered. Full-text study reports/publications were retrieved. It was not necessary to seek contact with the authors since all the information available in the articles was included. Relevant full-text study Full-text study reports/publications were retrieved. It was not necessary to seek contact with the authors since all the information available in the articles was included. Relevant full-text study reports/publications were retrieved ( Figure 1). We also assessed the risk of bias in every included study. A third reviewer (D.-G.E.) settled discrepancies.

Data Extraction and Management
Extracted information included country, study design, study duration, follow-up (cohort studies), study location and participants; age range, dietary assessment method, sample size, DP derivation method, economic indicators (household income), identified DP, direction of association between DP and interventions; and sociodemographic factors.
In order to reduce the number of DP described in our selected studies, we identified two main DPs: "Healthy/Prudent" and "Unhealthy/Western". The Healthy/Prudent DP was described as "adherence to guideline recommendations" according to age [1,46], and included fruits, vegetables, cereals, meat, eggs, olive oil, and whole grains. The "Traditional/Basic" DP was included in this group because it contained mainly local and home-cooked food which are commonly considered "healthy". The Unhealthy/Western DP included processed energy-dense food, low fiber "fast food", cookies, sweetened food, junk food, and puddings. For the present analysis, those DP including only beverages were not considered. The direction of association between sociodemographic factors and DP was described as: positive, converse or negative, exclusively when the associations were reported as statistically significant (P < 0.05) or if the 95% CI did not include a coefficient with a zero value or an odds ratio of 1.

Assessment of Potential Bias in Included Studies
The methodologies of the selected studies and potential bias in each study were evaluated using the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) method.
Two independent reviewers categorized the risk of bias as high, moderate, or low. Disagreements were resolved by a third reviewer. Agreement between reviewers was calculated using Cohen's kappa coefficient.

Data Extraction and Management
Extracted information included country, study design, study duration, follow-up (cohort studies), study location and participants; age range, dietary assessment method, sample size, DP derivation method, economic indicators (household income), identified DP, direction of association between DP and interventions; and sociodemographic factors.
In order to reduce the number of DP described in our selected studies, we identified two main DPs: "Healthy/Prudent" and "Unhealthy/Western". The Healthy/Prudent DP was described as "adherence to guideline recommendations" according to age [1,46], and included fruits, vegetables, cereals, meat, eggs, olive oil, and whole grains. The "Traditional/Basic" DP was included in this group because it contained mainly local and home-cooked food which are commonly considered "healthy". The Unhealthy/Western DP included processed energy-dense food, low fiber "fast food", cookies, sweetened food, junk food, and puddings. For the present analysis, those DP including only beverages were not considered. The direction of association between sociodemographic factors and DP was described as: positive, converse or negative, exclusively when the associations were reported as statistically significant (P < 0.05) or if the 95% CI did not include a coefficient with a zero value or an odds ratio of 1.

Assessment of Potential Bias in Included Studies
The methodologies of the selected studies and potential bias in each study were evaluated using the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) method.
Two independent reviewers categorized the risk of bias as high, moderate, or low. Disagreements were resolved by a third reviewer. Agreement between reviewers was calculated using Cohen's kappa coefficient.

Data Synthesis
Due to the variety of methods used to measure some variables of interest and due to a shortage of information in the included studies we decided not to carry out a meta-analysis in the present study.

Results of the Search and Study Selection
In total, 250 relevant articles were identified using electronic databases. During our screening phase, we reviewed abstracts and eliminated 215 articles that were either duplicated or had studies with descriptive, case-control designs, and cross-sectional designs. Of the remaining 35 articles, 28 did not meet the PICOS criteria: of these; nine included children older than 24 months of age or adults, 18 were excluded because they did not analyze DP, while one study did not correlate economic and sociodemographic factors with DP ( Figure 1). A concordance analysis assessed discrepancies between reviewers and revealed a "very strong" level of agreement. PubMed (0.55, P < 0.001), NICE (1.0, P < 0.001), Trip database (1.0, P < 0.001), Cochrane (1.0, P < 0.001), CENETEC (1.0, P < 0.001).

Included Studies
A total of seven studies based on "a posteriori" methods to derive dietary patterns were included in the final analysis. Of these, seven had cohort design. Updated searches were conducted in December 2018 and April 2019, and 10 new studies were identified. Of these, four articles met the phase one inclusion criteria but were eventually excluded due to their focus on maternal behavior (negative affectivity), or infant behavior (feeding difficulties), while others did not clearly describe DP associations with sociodemographic factors. No studies were included from this second review.
For the present review, we classified DP as "healthy/prudent" or "unhealthy/Western" pattern depending on which predominant foods in each study and in accordance with World Health Organization (WHO) international dietary recommendations for children's age. Healthy patterns were considered as: "core food" [49], "meat, vegetables and dessert" [48], "herbs, fruit, and raw vegetables" [48], "breastfeeding or milk formula with fruit and vegetables" [47,50], "conscious health pattern or Mediterranean diet" [51], "fruit, vegetables, and high protein food" [52], and "solid food according to international guidelines for children's age" [50] (Table 1).
Three of the included studies [48,50,52] positively associated multiparity with "unhealthy" DP too, while one also found the same association with having fewer children [48]. Maternal BMI ≥ 30 kg/m 2 during pregnancy was frequently associated with "unhealthy" DPs [48,51]. Other factors such as being married or unmarried were associated in converse directions with "healthy patterns" [48] and "unhealthy" patterns, respectively [47]. Being employed or unemployed carried the same association (Table 2) [52,53].

Risk of Bias in the Studies Included
Each article included was critically evaluated under the scope of STROBE criteria. Articles were classified as "high risk", "moderate risk", "low risk or unclear risk" and "very low risk" according to GRADE nomenclature (Table 3). 14 months: "core food", "basic combination" 24 months: "core food", "non-core food" Maternal: education level, age (during delivery), smoking during pregnancy, breastfeeding duration, marital status, weight, parity, economic status (decile), Australian nationality.
14 months Higher education levels (university) were associated with high consumption of "core food". Younger mothers at time of delivery and smokers were conversely associated with the "basic combination pattern". 24 months Australian nationality and younger mothers were associated with "non-core food". Hispanic vs. non-Hispanic white race/ethnicity mothers were associated with "fruits and vegetables "and "formula feeding and low variety". Multiparity was associated with "formula feeding and low variety patterns". High school education or less was associated with both "formula feeding and low variety" and "mixed patterns". Low household income was associated with the "formula feeding and fruit and vegetables pattern". Older, high-income, married, and high education-level mothers were less associated with milk formula, low variety and milk formula, high energetic density food. High pre-pregnancy body mass index was associated with milk formula, low variety and milk formula, and high energetic density food. Milk formula, low variety was associated with mothers who returned to work at 3 months. No differences were found between weight gain during pregnancy, child sex, body weight at birth (Z score), and use of childcare. Low paternal education, low household income, parental smoking, high maternal body mass index during pregnancy, high intake of carbohydrates, and multiparity were associated with the "Western-like pattern". High fiber intake during pregnancy and older parents were conversely associated with the "Western-like pattern". Folic acid intake during pregnancy, high fiber maternal diet, and single parenthood were positively associated with the "healthy pattern". Mothers who consumed alcohol during pregnancy and had a history of comorbidity, and those with daughters, were less associated to the "health conscious" pattern. Unemployed mothers, daily smokers during pregnancy, lower education levels, lower household income, higher number of children, were associated with "confectionary and sweetened beverages". Non-smokers during pregnancy, high education levels, longer duration of breastfeeding, full-time employment, and higher household income with a rice, fish and vegetable intake pattern were less associated with "confectionary and sweetened beverages". Mothers with more than 13 years of education and a rice, fish, and vegetable dietary pattern, were less associated with "confectionary and sweetened beverages". "High sugar, fat and protein pattern", "infant guideline solids", "formula milk", "high dairy and regular cereal".
Low household income, maternal non-Hispanic Afro-American ethnicity, low education level, high body mass index were associated with the "high sugar, fat and protein dietary pattern. High household income, maternal non-Hispanic, white ethnicity, and high education were associated with "Infant guideline solids". Emergency cesarean-section, higher maternal age, Asian-pacific ethnicity, and low gestational weight gain were associated with the "formula, baby cereal dietary pattern". Vaginal non-induced, maternal non-Hispanic, white race/ethnicity, low education, being employed, and obesity were associated with "high dairy and regular cereal dietary pattern".

Discussion
This review suggests the presence of an association between sociodemographic factors (such as maternal education level, household income, maternal BMI, and maternal age) and DP in children under 24 months.
Five studies included in the present review showed that higher maternal education levels were associated with "healthy" DPs [47][48][49]52,53], whereas lower maternal education were associated with "unhealthy" DPs [48,[51][52][53]. Our results are similar to previous studies where maternal education has been considered a determining factor for DP in mothers as much as in children [54][55][56][57]. For example, a study conducted by Hidaka et al. [58] on 190 American mother-child dyads with children between 2 to 4.5 years old evaluated and correlated DP with maternal SES factors such as education and household income. They found that mothers with healthier diets had on average higher SES scores (difference 1.05, CI 95%: 0.80 to 1.30) than mothers with "unhealthy" DPs. Furthermore, this study concluded that education was directly linked to favorable nutrient intake which was strongly associated with healthier nutritional behaviors in participating mothers. Financial support is another important factor frequently associated with DP in children. A systematic review conducted by Bazzano et al. [59] showed available qualitative research on DPs and practices in infants and young children from the perspective of parents and families in low-income settings in the USA. This review reported four categories of barriers obstructing recommended breastfeeding practices such as lack of support for breastfeeding from families, health workers and due to time poverty, and three categories of barriers to recommended complementary feeding practices showing how higher quality foods were replaced with lower quality foods for financial reasons, lack of diversity in available foods, lack of water for cooking, and inability to buy food. These last findings highlight the importance of economic factors when defining early DPs, especially those considered a risk factor for nutritional outcomes such as obesity, [60,61] malnutrition, and double burden malnutrition. However, further studies are required to corroborate these associations.
In this review we found an association between higher household income and healthy dietary patterns; however, all studies included in this review were carried out in developed countries and could not be extrapolated to populations with different characteristics. However, a previous systematic review developed in children and adolescents from countries with high, middle, and low incomes also reported an association between higher economic income and "healthy" DP [62].
Likewise, the results from this systematic review suggest that maternal age is associated with children's DP. We found that children whose mothers were ≥30 years old had healthier DPs [48] mainly represented by high intake of fruits, vegetables, whole grains, seeds, cheese, and eggs. In contrast, children whose mothers were <30 years had unhealthier DPs characterized by sugar sweetened beverages, high-fat, highly processed foods, and candies [49]. It is possible that younger mothers may know less about nutrition than older mothers, however, further studies would be required before drawing any conclusions. An example of the importance of maternal age can be found in a cohort study performed on women in the United Kingdom, where women under 19 often consume unhealthy foods such as ultra-processed (energy-dense food) paired with a low intake of fruit and vegetables, while women aged 35 and older showed high consumption of fruits, vegetables, whole grains and starch [63]. Additionally, a study conducted in New Zealand [28] also found a positive association between increasing maternal age and "health conscious" and "fusion/protein" DPs, while also reporting the opposite "junk" and "traditional/white bread" pattern associations in younger mothers.
This review also found an association between maternal BMI ≥ 30.0 kg/m 2 and unhealthy DPs in children [47,53]. Some studies have described an association between overweight or obese mothers and increased obesity risk in their children [64]. However, current literature on the association between the quality of parents' diets and children's diets is limited. There is little research on the relationship between parental and child DPs, particularly empirically derived DPs. Despite being limited, there is evidence in current literature supporting an association between the quality of parents' diet and children's diet, although theoretically derived patterns based on national nutrition guidelines differ between countries. Consequently, any significant relationships found using a particular country-specific index may not apply to other populations. While this review found three studies that included an association between having more children and unhealthy dietary patterns [47,49,51] we did not find enough evidence to establish number of children in a family as a factor associated with DP. However, having more than two children may contribute to a less restrictive feeding environment due to elevated maternal stress levels, which may lead to unhealthy diets [65]. Mothers are influential in their children's eating behaviors [66] and time-restrictions as a result of having multiple children and pressuring children to eat may favor unhealthy feeding practices.
A strength of this review is that it looks at studies conducted in countries across North America, Europe, Asia, and Oceania. As a result, we were able to describe similarities in dietary patterns around the world, as well as associations between sociodemographic factors and DP in children under 24 months of age [2, 46,67,68]. Another strength is that information from those studies was collected through tools such as multiple-pass 24-h recall and FFQs or designed by professional staff [69]. Dietary pattern derivation methods (PCA and latent class analysis) used in the included studies are also validated [70]. The PCA method used for factor analysis added specific food items or food groups on the basis of how food items in the dataset correlated with each other [14].
Albeit, this review has limitations. The studies included in this review were carried out in developed countries and their results are unlikely to apply to developing or underdeveloped populations. Also, measurements such as household income were based on a regional index which makes them difficult to interpret.

Conclusions
The results of our systematic review suggest an association between sociodemographic factors (such as maternal education, maternal age, and household income) and dietary patterns in children under 24 months of age.
We recommend studies in developing and less developed countries to find out which sociodemographic factors are associated with dietary patterns in childhood in different contexts, as well as with different health outcomes such as malnutrition, obesity, and double burden malnutrition. Most of these are preventable with adequate complementary feeding. Meanwhile, we recommend the ongoing promotion of healthy eating habits that include fruits, vegetables, whole grains, seeds, cheese, eggs, and others, while reducing consumption of unhealthy sugary, high-fat, highly-processed foods and candies in the early stages of life [64].