1. Sample
The data for this cross-sectional study were sourced from the "Child Health Aspect: Anthropology, Development, Nutrition" survey, which was conducted in 2017 in the 31 Iranian provinces (11). The target population included all children under 5 years of age whose information was recorded using software packages that provide addresses and names of households; hence, the list of these households was used as a sampling frame for the random selection of households with children. A total of 18,600 households with children under the age of 5 were randomly selected using a two-stage sampling system (provinces and households). After inviting them to their health care center, the mother or tutor of the child was surveyed by a nutritionist using a face-to-face method. The questionnaire used was based on the DHS 2010 (Demographic and Health Survey, WHO) and MICS6 (Multiple Indicator Cluster Survey, UNICEF) questionnaires (12),(13). The participants were asked about their general household information (location, area of residence, marital status of the caregiver, sex of the caregiver, number of household members, number of children under 5 years old), child information (age, sex, birth weight, child diseases in the past 15 days, child birth order), type of child nutrition and food consumed in the last 24 hours, child growth rate, and demographic and socioeconomic information about the parents. In addition, the child's weight was measured with an electronic UNISCALE scale with an accuracy of 0.1 kg, and the child's height was measured with a stadiometer with an accuracy of 0.1 cm.
According to the surveys, 1 in 4 children under the age of 5 was below the standard child growth chart of the WHO (14). In the present study, the aim was to explore the effect of dietary diversity on stunting and overweight in children under 5 years of age in conjunction with socioeconomic indicators. Considering that children under 2 years of age primarily grow through breastfeeding and complementary feeding, measuring dietary diversity in this age group is not particularly pertinent. As a result, the focus was on children aged 24–59 months, resulting in a final sample of 11,147 observations, with no missing data.
2. Outcome
2. Outcome
The WHO mentions that malnutrition may take various forms, such as stunting, underweight, wasting, and overweight (14). Although stunting is one of the obvious consequences of growth failure in children (15), the prevalence of childhood obesity is increasing worldwide (16). Stunting and overweight were identified as the most challenging variables, prompting the decision to concentrate on these two aspects (17),(18). According to the Centers for Disease Control and Prevention (CDC), stunting is defined as a height-for-age z score (HAZ) less than − 2 standard deviations (SD) below the mean according to the WHO child growth standard (16). Stunting was coded as a binary variable. According to the CDC, overweight is defined as a weight-for-height z score (WHZ) above 2 (+ 2) SD above the mean according to the World Health Organization child growth standard. The overweight variable was also coded as binary.
3. Explanatory variables
The following socioeconomic variables were considered in the analysis, related to stunting and/or overweight in the literature: parental education (19),(20), parental occupation (7), and household assets as proxies for wealth (7),(21),(22).
The father’s and mother’s education were coded into five categories: illiterate, primary school, secondary school, high school or diploma, and university education.
The occupations of the fathers were classified into 9 groups according to employment status: employee, worker, farmer, shopkeeper, retired, student, unemployed, military and others. Some occupations were grouped in the study because of the expected income similarity and the small number of people within some categories; namely, retired, unemployed, and students were grouped into a single class.
Due to the large number of housewives and the small number of mothers in different jobs, mothers were classified into two groups, namely, working mothers and nonworking mothers. Employees, workers, farmers, and shopkeepers were classified as working mothers, and retirees, students, unemployed individuals, and housewives were considered nonworking mothers.
Regarding assets, the household's number of rooms, the house area and a list of available goods were used in the analysis. The number of rooms in each household was divided by its size, resulting in the categorization of the variable into four quartiles. The household's house area was divided by the household number of members, and the variable was divided into quintiles.
The household assets were coded as the sum of the number of assets, from zero to five. The following assets were considered: TV, freezer, microwave, computer, and car.
4. Mediation variable
Diet diversity score (DDS) is an index to measure variety of the dietary items consumed in the day (23). The DDS was used to assess nutrient adequacy (24). The DDS was calculated by counting the number of food groups that every child consumed on the last day. Mothers were asked qualitive questions about their children's dietary habits, namely, “Has your child consumed any of the following foods in the past 24 hours?”. Food items were classified into 7 groups: cereals, fruits and vegetables containing vitamin A, other fruits and vegetables, legumes, eggs, flesh foods, and dairy products. The minimum score was one, and the maximum score was seven. According to the children's ages, all the children had consumed 3 food groups— cereals, fruits, and legumes—in the last 24 hours, and the number of children who consumed less than that was rare. In this study, the quality of the diet was therefore calculated based on the consumption or lack of consumption of the remaining 4 food groups (eggs, flesh foods, dairy products, and fruits and vegetables containing vitamin A) by the child in the last 24 hours. If they consumed all 4 food groups, they were in the high-quality group; if they ate fewer than these 4 food groups, they were in the low-quality group.
Statistical analysis
Univariate logistic regressions were initially conducted for each explanatory variable independently, followed by a subsequent multivariate logistic regression that included all variables in the model. The results of multilevel logistic regression are presented as odds ratios (OR) with 95% confidence intervals (95% CI), and the significance level was defined as a p-value lower than 0.05. The first multivariate logistic regression allowed us to measure which socioeconomic variables were the most relevant for explaining stunting and overweight status, adjusting for age and sex. Logistic regression was subsequently performed including the DDS to measure the mediation role of the diet diversity score in the SES-nutrition relationship. The data analysis was performed with SPSS software version 21.