Consumption of iron rich foods and its associated factors among South Asian children aged 6–23 months using multilevel analysis Authors: Md. Jakaria Habib, Md. Saifullah Sakib, Abdul Baten, Muhammad Tareq Published: 2025 ABSTRACT The purpose of this study was to determine the prevalence of iron-rich food (IRF) consumption and its related characteristics among children aged 6–23 months in South Asian countries. This cross-sectional study utilized data from the most recent nationally representative demographic and health surveys (DHS) from six South Asian nations. The sample included 84,234 children aged 6–23 months who lived with their mothers. Eating iron-rich foods was the main outcome variable. India had the lowest IRF use at 17.4%, while the Maldives had the highest at 71.9%. Bangladesh followed with 69.6%, then Pakistan at 38%, Nepal at 35.3%, and Afghanistan at 30%. Overall, 21.5% of children from South Asian backgrounds consumed IRF. Additionally, male children, those who were still being breastfed, those who had experienced diarrhea in the past two weeks, mothers who were older and had less education, those with more than two children under five in the household, and those living in rural areas had lower odds of consuming the iron-rich food. This study demonstrates that a mother’s education and adequate prenatal care (ANC) visits have a significant impact on her child’s consumption of iron-rich foods. Overall, the consumption of IRF among children aged 6–23 months in South Asia remains low. Therefore, a focused communication program is needed to improve mothers’ awareness and practices regarding child nutrition, particularly to increase the intake of iron-rich foods during this critical growth stage. Introduction Most children under five living in low- and middle-income countries (LMICs) face the main types of childhood malnutrition. High levels of severe malnutrition often go hand in hand with elevated childhood death rates in certain areas. Malnutrition causes 3.1 million deaths among children under five worldwide each year, which accounts for 45% of all child deaths; the majority of these deaths occur in LMICs1,2. In recent decades, child malnutrition has declined worldwide, but it remains more prevalent in African and South Asian countries1. The United Nations launched the Sustainable Development Goals in 2015 to end malnutrition in all its forms by 2030. The World Health Organization and UNICEF jointly developed the Global Strategy for Infant and Young Child Feeding (IYCF) practices to reduce the impact of early malnutrition3. IYCF typically targets children between the ages of six and twenty-four months. Adequate complementary feeding practices throughout the early years of life ensure linear growth in later life and lower the mortality rate among children under five4–6. Therefore, the goal of complementary feeding is to help infants and young children transition gradually from exclusive breastfeeding to solid foods consumed as part of a family meal between the ages of 6 and 24 months7. To prevent malnutrition, especially iron deficiency (ID), which is the most common single-nutrient deficiency, the intake of iron-rich foods (IRF) is usually assessed as a separate IYCF indicator7,8. If left untreated, ID can progress to iron deficiency anemia (IDA), a persistent public health challenge, particularly in low- and middle-income countries. Importantly, ID often remains clinically silent until anemia becomes severe, highlighting the importance of early detection and preventive strategies8. IDA compromises oxygen transport in the body and has been strongly associated with impaired growth, elevated mortality risk, and delayed motor and cognitive development in infants and young children9,10. Consequently, these impairments contribute to reduced labor productivity later in life and collectively impose significant economic costs, estimated at a median of $2.32 per capita, or 0.57% of GDP11. Furthermore, the Global Burden of Disease Study (2017) identified iron deficiency anemia as one of the top five leading causes of years lived with disability (YLD), accounting for approximately 34.7 million YLD globally12. There is a special “window of opportunity” during the first two years of life to shape a healthier and more prosperous future. In terms of iron requirements, full-term, normal birth weight babies are nearly self-sufficient during the first six months. Due to rapid growth and brain development, the 6–11 month phase of life has a higher iron need than any other6,13,14. According to research by the American Academy of Pediatrics, a child’s daily iron requirement increases dramatically to 11 mg at 6 months of age15. Because of this, IDA is most common and severe in young children between 6 and 24 months. It was found that the prevalence of anemia in children was below 50% in some developing nations, such as 44% in Afghanistan, 36% in Sri Lanka, 35% in the Philippines, and 32% in Indonesia, but over 50% in some South Asian countries, such as 56% in Bangladesh, 59% in India, 58% in Myanmar, and 61% in Pakistan16. Inadequate dietary iron intake (quantity and bioavailability) is the most important factor causing anemia among vulnerable groups, like infants and small children. Therefore, it is recommended to introduce iron-rich complementary foods around 6 months of age to meet these increased daily requirements and prevent IDA in children3,15,17,18. Furthermore, dietary iron intake is generally low in the South Asian countries19–21. Although daily consumption of iron-rich foods could reduce the risk of IDA, the underlying factors contributing to such low intake remain insufficiently explored in South Asia. This study conducts a region-wide investigation of South Asia to identify factors associated with eating iron-rich foods to address this issue. Therefore, this research aimed to use multilevel logistic regression to examine the prevalence of iron-rich food consumption and related factors among children aged 6–23 months in South Asian countries. Methods Data source and sample size This study is based on the recent wave of nationally representative Demographic and Health Survey (DHS) datasets from six South Asian countries to determine the consumption of iron-rich foods among children aged 6–23 months. The DHS followed a stratified two-stage sample design to generate up-to-date data for a wide range of indicators in the areas of population, health, and nutrition. The stratification was achieved by dividing each region/province of the respective countries into urban and rural areas, except the Maldives, since no urban–rural classification exists for residential households22. In the first stage, enumeration areas (EAs) or enumeration blocks (EBs), considered as clusters, were selected with probability proportional to size. In the second stage, a systematic sample of households was drawn within each selected cluster. ICF’s Institutional Review Board (IRB) and the IRB in the host country approve the ethics of the survey. All ever-married women aged 15–49 who were usually members of the selected households and those who spent the night before the surveys in the selected households are eligible for interviews. More details about the survey methods are available at: https://dhsprogram.com/Methodology/index.cfm. After controlling for sample clusters, all analyses were conducted using the appropriate sampling weights from the country survey, following DHS guidelines for interpreting DHS data to obtain nationally representative estimates. The study population consisted of all sampled living children aged 6–23 months who were living with their mothers, while the target population encompassed all living children aged 6–23 months. A total of 84,234 children within this age range were included in the study based on the inclusion criteria. Table 1 below presents the sample sizes selected for this study from each of the six surveys, along with their respective survey years.Table 1List of surveys and corresponding samples considered in this study.NoSurvey nameSurvey yearSample size1Bangladesh Demographic and Health Survey2017–1824452Afghanistan Demographic and Health Survey201579843India National Family Health Survey2015–1668,9734Maldives Demographic and Health Survey2016–177705Nepal Demographic and Health Survey201614956Pakistan Demographic and Health Survey2017–182569Total84,234 Table 1List of surveys and corresponding samples considered in this study.NoSurvey nameSurvey yearSample size1Bangladesh Demographic and Health Survey2017–1824452Afghanistan Demographic and Health Survey201579843India National Family Health Survey2015–1668,9734Maldives Demographic and Health Survey2016–177705Nepal Demographic and Health Survey201614956Pakistan Demographic and Health Survey2017–182569Total84,234 List of surveys and corresponding samples considered in this study. Dependent variables The study’s dependent variable was whether children aged 6–23 months consumed iron-rich foods, categorized as “Yes” or “No.” The DHS considers consumption as having eaten at least one of these four foods: fish or shellfish, meat (beef, pork, lamb, chicken, etc.), eggs, organ meat (liver, heart, or other organs), or any of these foods at any time during the 24 h before the interview. Conversely, not having eaten any iron-rich foods in the 24 h before the interview was recorded as non-consumption. Explanatory variables In this study, fifteen explanatory variables were used. Explanatory variables were categorized into two levels: (i) individual levels and (ii) Community levels. Individual levels can be extended to further three categories (e.g., Child characteristics, Mother characteristics, and Household characteristics) includes—Child age, sex, Birth order, Currently breastfeeding status, Fever and Diarrhea in the last two weeks, Mothers’ age, education, exposure to Mass Media, Number of ANC visits, C-section delivery, Place of delivery, Wealth index Number of under 5 children and sex of household head while Place of residence were considered as community levels variable. These variables were selected based on prior literature identifying key determinants of IRF consumption among children aged 6–23 months. Statistical analysis Descriptive statistics were used to provide an overview of the study population’s background characteristics. The association between the predictor variables and the consumption of iron-rich foods was examined using chi-square tests. Before conducting additional analysis, case-wise deletion techniques were used to eliminate 1.3% of the missing observations. Multicollinearity, characterized by high interdependence among predictors, was also assessed, as it can distort regression estimates; tolerance values below 0.1 were used as the threshold for concern23,24. Multilevel models are particularly suitable for research designs in which participant data are hierarchically structured across multiple levels, requiring the dependent variable to be analyzed at the lowest level possible25,26. To account for the hierarchical structure of the data, this study employs a multilevel model approach to assess the likelihood of IRF consumption across the individual, community, and country levels, where community and country were considered as random effects. Models I and II are unconditional models, containing no independent variables at level two (community) and three (country), respectively. Model III incorporates individual-level variables and is a two-level structure. Model IV includes individual- and community-level factors at level two, whereas Model V extends the model to a three-level framework, incorporating variables at the individual and community levels, with country-level clustering accounted for as a third level. The intraclass correlation coefficient (ICC) and median odds ratio (MOR) have been calculated to quantify the effect of clustering. The ICC is a key concept in multilevel analysis as it measures the similarity between responses within a specific level of the hierarchy. The median odds ratio (MOR) indicates substantial heterogeneity between clusters27. Furthermore, the purpose of the MOR is to translate the area-level variance into the widely used odds ratio (OR) scale, which offers a consistent and intuitive interpretation. Additionally, model selection criteria such as the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) were used to assess how well a model fits the data. The analysis in this study was conducted using the Statistical Package for Social Sciences (SPSS) version 20 and R programming (version 4.0.0). The R package “lme4” was used for multilevel logistic regression analysis. Geographical maps were created using the “ggplot2” package. Results Background characteristics and prevalence of iron-rich foods consumption The frequency distribution of child-, mother-, and household-level characteristics for the total of 84,234 children aged 6–23 months is shown in Table 2. More than two-thirds of the children were 6–17 months old. Most of the children were reported to be second or third children of their parents and had been breastfed at the time of the survey. About three-fourths of the mothers were aged 20–29. Slightly less than one-third of the mothers had no education. Around one-fifth had received adequate ANC visits. The majority of the children belonged to households with a low wealth index. Of the 84,234 children aged 6–23 months sampled from all six South Asian countries, 0.013 percent of observations were missing ANC information, 0.012 percent for C-section delivery, and 0.004 percent for place of delivery.Table 2Socio-demographic characteristics and prevalence of iron-rich food consumption of the children aged 6–23 months and respondents.VariablesCategoryn (%)Iron-rich food consumption n (%)p-valueChild characteristicsChild age (in months)6–1129,295 (34.8)3316 (11.3)< 0.00112–1729,389 (34.9)7365 (25.1)18–2325,551 (30.3)7470 (29.2)Child sexMale44,218 (52.5)9218 (20.8)< 0.01Female40,016 (47.5)8934 (22.3)Birth orderFirst29,777 (35.4)6304 (21.2)< 0.01Second or third39,845 (47.3)8946 (22.5)Fourth and up14,613 (17.4)2901 (19.9)Currently breastfeedingNo18,020 (21.4)5259 (29.2)< 0.001Yes66,215 (78.6)12,892 (19.5)FeverNo66,541 (79.0)13,780 (20.7)< 0.001Yes17,693 (21.0)4371 (24.7)DiarrheaNo69,979 (83.1)15,160 (21.7)0.071Yes14,255 (16.9)2991 (21.0)Mothers’ characteristics of the childrenMother age (in years)15–194466 (5.3)1164 (26.1)< 0.0520–2432,674 (38.8)6671 (20.4)25–2928,945 (34.4)6220 (21.5)30–3412,090 (14.4)2761 (22.5)35–394590 (5.4)1004 (21.9)40–441136 (1.3)263 (23.2)45–49332 (0.4)67 (20.1)Mother’s educationNo education27,025 (32.1)4931 (18.2)< 0.001Primary11,461 (13.6)2542 (22.2)Secondary36,245 (43.0)8311 (22.9)Higher9503 (11.3)2368 (24.9)Number of ANCNo ANC15,142 (18.0)2819 (18.6)< 0.001Less than 853,849 (64.0)10,674 (19.8)8 +15,232 (18.0)4654 (30.6)Missing11 (0.0)–C-section deliveryNo68,592 (81.4)13,764 (20.1)< 0.001Yes15,632 (18.6)4383 (28.0)Missing10 (0.0)–Place of deliveryHealth Facility66,137 (78.5)13,999 (21.2)< 0.001Home18,094 (21.5)4152 (22.9)Missing3 (0.0)–Household characteristics of the childrenWealth indexPoor37,681 (44.7)7179 (19.1)< 0.001Middle17,255 (20.5)3874 (22.5)Rich29,298 (34.8)7098 (24.2)Number of under 5 children2 or less68,792 (81.7)15,199 (22.1)< 0.0013 or more15,443 (18.3)2953 (19.1)Sex of HH headMale74,354 (88.3)15,925 (21.4)< 0.05Female9880 (11.7)2226 (22.5)Community level factorsPlace of residenceUrban23,977 (28.5)6193 (25.8)< 0.001Rural60,257 (71.5)11,959 (19.8)Total84,23418,151 (21.5)ap < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001.n Indicates frequency. Table 2Socio-demographic characteristics and prevalence of iron-rich food consumption of the children aged 6–23 months and respondents.VariablesCategoryn (%)Iron-rich food consumption n (%)p-valueChild characteristicsChild age (in months)6–1129,295 (34.8)3316 (11.3)< 0.00112–1729,389 (34.9)7365 (25.1)18–2325,551 (30.3)7470 (29.2)Child sexMale44,218 (52.5)9218 (20.8)< 0.01Female40,016 (47.5)8934 (22.3)Birth orderFirst29,777 (35.4)6304 (21.2)< 0.01Second or third39,845 (47.3)8946 (22.5)Fourth and up14,613 (17.4)2901 (19.9)Currently breastfeedingNo18,020 (21.4)5259 (29.2)< 0.001Yes66,215 (78.6)12,892 (19.5)FeverNo66,541 (79.0)13,780 (20.7)< 0.001Yes17,693 (21.0)4371 (24.7)DiarrheaNo69,979 (83.1)15,160 (21.7)0.071Yes14,255 (16.9)2991 (21.0)Mothers’ characteristics of the childrenMother age (in years)15–194466 (5.3)1164 (26.1)< 0.0520–2432,674 (38.8)6671 (20.4)25–2928,945 (34.4)6220 (21.5)30–3412,090 (14.4)2761 (22.5)35–394590 (5.4)1004 (21.9)40–441136 (1.3)263 (23.2)45–49332 (0.4)67 (20.1)Mother’s educationNo education27,025 (32.1)4931 (18.2)< 0.001Primary11,461 (13.6)2542 (22.2)Secondary36,245 (43.0)8311 (22.9)Higher9503 (11.3)2368 (24.9)Number of ANCNo ANC15,142 (18.0)2819 (18.6)< 0.001Less than 853,849 (64.0)10,674 (19.8)8 +15,232 (18.0)4654 (30.6)Missing11 (0.0)–C-section deliveryNo68,592 (81.4)13,764 (20.1)< 0.001Yes15,632 (18.6)4383 (28.0)Missing10 (0.0)–Place of deliveryHealth Facility66,137 (78.5)13,999 (21.2)< 0.001Home18,094 (21.5)4152 (22.9)Missing3 (0.0)–Household characteristics of the childrenWealth indexPoor37,681 (44.7)7179 (19.1)< 0.001Middle17,255 (20.5)3874 (22.5)Rich29,298 (34.8)7098 (24.2)Number of under 5 children2 or less68,792 (81.7)15,199 (22.1)< 0.0013 or more15,443 (18.3)2953 (19.1)Sex of HH headMale74,354 (88.3)15,925 (21.4)< 0.05Female9880 (11.7)2226 (22.5)Community level factorsPlace of residenceUrban23,977 (28.5)6193 (25.8)< 0.001Rural60,257 (71.5)11,959 (19.8)Total84,23418,151 (21.5)ap < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001.n Indicates frequency. Socio-demographic characteristics and prevalence of iron-rich food consumption of the children aged 6–23 months and respondents. ap < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001. n Indicates frequency. Table 2 also demonstrates the prevalence of iron-rich food consumption among South Asian children and crude odds ratio (cOR). Among children between the ages of 12–17 months, only 25.1% consumed iron-rich food [cOR = 2.62], and 29.2% of children aged 18–23 months consumed iron-rich food [cOR = 3.24]. The prevalence of consumption was found to be higher (22.5%) among the children with second or third birth order [cOR = 1.08]. Non-breastfeeding children were more likely to consume iron-rich food than breastfeeding children (29.2% and 19.5% respectively). The table shows that 26.1% of the children whose mothers were aged 15–19 years consumed IRF. The proportion of children consuming IRF increases with increasing mother’s education, from 18.2% among children whose mother had no education to about 25% among children whose mother had higher education. The proportion of consumption was 30.6% among the children whose mother had adequate ANC [cOR = 1.92]. The prevalence of consumption was lower among children living in households with low wealth status. Urban children were more likely to consume IRF than rural children. Children’s feeding practices on iron-rich foods indicators Figure 1a,b show the status of iron-rich food consumption among children aged 6–23 months across South Asian countries. Among these countries, children from the Maldives consumed the highest amount of iron-rich foods (71.9%; 95% CI 68.8–75.2), followed by Bangladesh (69.6%; 95% CI 67.8–71.4). Over one-third of children from Pakistan (38%; 95% CI 36.2–39.9) and Nepal (35.3%; 95% CI 32.8–37.7) consumed IRF. About 30% [95% CI 29.4–31.4] of Afghan children consumed IRF, while only 17.4% [95% CI 17.1–17.5] of Indian children did. Overall, 21.5% [95% CI 21.3–21.8] of South Asian children consumed iron-rich food in the 24 h before the survey (Table 2).Fig. 1Prevalence of iron-rich food consumption by (A) Country, and (B) administrative boundaries. Prevalence of iron-rich food consumption by (A) Country, and (B) administrative boundaries. Random effect analysis and model comparison Table 3 displays the results of random-effect analysis and model comparisons. The ICC value in Model I (unconditional model at level two) suggests that the correlation between two randomly selected children within the same cluster was 0.48. The unconditional model (Model II) at the level three for IRF consumption shows that the ICC values at the community and country level were 0.51 and 0.23, respectively. Thus, the correlation between children from the same country was 0.23, while the correlation between children from the same country and the same community was much higher at 0.51. Therefore, IRF consumption was much more similar among children residing in the same community and country. In the full model (Model V), the decrease in ICC values from 0.51 to 0.47 and from 0.23 to 0.17 indicates the reduction in the clustering effect compared to Model II.Table 3Results from random effect analysis and model comparison.ParametersModel IModel IIModel IIIModel IVModel VRandom effects Random effect variance (SD) Community level3.03 (1.74)1.86 (1.36)2.89 (1.70)2.85 (1.69)1.84 (1.35) Country level–1.49 (1.28)––1.08 (1.04) ICC Community level0.480.510.470.460.47 Country level–0.23––0.17 MOR Community level5.263.675.085.083.63 Country level–3.22––3.06Model fit statistic & comparison Likelihood ratio test (LRT)–p-value < 0.001p-value < 0.001p-value < 0.001p-value < 0.001 Deviance (-2LL)68,86466,97864,92264,87962,837 AIC68,86866,98464,97764,93562,895 BIC68,88767,01265,23065,19763,166ICC, intra class correlation coefficient; MOR, median odds ratio; AIC, Akaike’s information criteria; BIC, Bayesian information criteria.Model I—unconditional two-level hierarchical model; Model II—unconditional three-level hierarchical model; Model III—Model with individual level factors at two-level (cluster as random effect); Model IV—Model with individual, and community level factors at two-level (cluster as random effect); Model V—Model with individual, and community level factors at three-level (cluster and country as random effect). Table 3Results from random effect analysis and model comparison.ParametersModel IModel IIModel IIIModel IVModel VRandom effects Random effect variance (SD) Community level3.03 (1.74)1.86 (1.36)2.89 (1.70)2.85 (1.69)1.84 (1.35) Country level–1.49 (1.28)––1.08 (1.04) ICC Community level0.480.510.470.460.47 Country level–0.23––0.17 MOR Community level5.263.675.085.083.63 Country level–3.22––3.06Model fit statistic & comparison Likelihood ratio test (LRT)–p-value < 0.001p-value < 0.001p-value < 0.001p-value < 0.001 Deviance (-2LL)68,86466,97864,92264,87962,837 AIC68,86866,98464,97764,93562,895 BIC68,88767,01265,23065,19763,166ICC, intra class correlation coefficient; MOR, median odds ratio; AIC, Akaike’s information criteria; BIC, Bayesian information criteria.Model I—unconditional two-level hierarchical model; Model II—unconditional three-level hierarchical model; Model III—Model with individual level factors at two-level (cluster as random effect); Model IV—Model with individual, and community level factors at two-level (cluster as random effect); Model V—Model with individual, and community level factors at three-level (cluster and country as random effect). Results from random effect analysis and model comparison. ICC, intra class correlation coefficient; MOR, median odds ratio; AIC, Akaike’s information criteria; BIC, Bayesian information criteria. Model I—unconditional two-level hierarchical model; Model II—unconditional three-level hierarchical model; Model III—Model with individual level factors at two-level (cluster as random effect); Model IV—Model with individual, and community level factors at two-level (cluster as random effect); Model V—Model with individual, and community level factors at three-level (cluster and country as random effect). In the null model at two levels (Model I), heterogeneity in IRF consumption between communities was indicated by the Median Odds Ratio (MOR) of 5.26. When interpreting the MOR, we can say that if children moved to another cluster with a higher likelihood of consumption, their median increase in odds would be 5.26. The Likelihood Ratio Test (LRT) shows that all five models had likelihood ratios significantly different from Model I. Model V had the lowest deviance, AIC, and BIC values as compared to the other models, and was therefore selected as the best-fit model. Identify the contributing factors of iron-rich food consumption The results of the random intercept multilevel logistic regression, with two unconditional or empty models, are presented in Table 4. This table displays the adjusted odds ratio (aOR) along with its 95% confidence interval and the corresponding p-value. From the full model (Model V), we found that the likelihood of IRF consumption increases significantly with the child’s age and female children were more likely to consume than male (aOR = 1.11; 95% CI 1.06–1.17). Compared to first children of their parents, children of second or third birth order were 24% more likely (aOR = 1.24, 95% CI 1.17–1.32) to consume IRF. The odds of consumption of iron-rich food were 1.39 times higher (aOR = 1.39, 95% CI 1.29–1.49) among the children of mothers with secondary education as compared to children of mothers with no education. Children whose mothers had adequate ANC visits had almost twice the odds (aOR = 1.86, 95% CI 1.69–2.04) of consuming iron-rich food compared to children from mothers who had no ANC visits. However, the results show that the odds of consumption were significantly lower among children from a family with a rich wealth index (aOR = 0.88, 95% CI 0.82–0.95) compared to children from a poor household. Children from a household that had 3 or more under-5 children were significantly 10% less likely (aOR = 0.90, 95% CI 0.84–0.96) to consume iron-rich food than the children whose household had 2 or fewer under 5 children. Children from rural areas had 25% less (aOR = 0.75, 95% CI 0.69–0.81) odds of consuming iron-rich food than children from urban areas.Table 4Multilevel Logistic regression analysis of iron-rich food consumption in South Asia.VariablescORModelsModel III aOR (95% CI)Model IV aOR (95% CI)Model V aOR (95% CI)Child age (in months) 6–111.0111 12–172.62***3.36*** (3.16–3.58)3.35*** (3.15–3.57)3.65*** (3.42–3.89) 18–233.24***4.52*** (4.23–4.83)4.51*** (4.51–4.81)5.21*** (4.87–5.56)Child sex Male1.0111 Female1.09***1.13*** (1.08–1.18)1.14*** (1.08–1.19)1.11*** (1.06–1.17)Birth order First1.0111 Second or third1.08***1.31*** (1.23–1.39)1.32*** (1.24–1.40)1.24*** (1.17–1.32) Fourth and up0.92**1.22*** (1.10–1.34)1.20*** (1.09–1.32)1.03 (0.94–1.14)Currently breastfeeding No1.0111 Yes0.59***0.65*** (0.61–0.69)0.63*** (0.59–0.67)0.70*** (0.66–0.75)Fever No1.0111 Yes1.26***1.27*** (1.20–1.35)1.27*** (1.20–1.35)1.07* (1.00–1.13)Diarrhea No1.0111 Yes0.96a0.91* (0.85–0.98)0.88*** (0.82–0.94)0.92* (0.86–0.99)Mother age (in years) 15–191.0111 20–240.73***0.68*** (0.61–0.76)0.67*** (0.60–0.75)0.82** (0.74–0.92) 25–290.78***0.66*** (0.59–0.75)0.65*** (0.58–0.73)0.83** (0.73–0.93) 30–340.84***0.69*** (0.61–0.79)0.68*** (0.59–0.77)0.83** (0.72–0.95) 35–390.79***0.60*** (0.51–0.71)0.62*** (0.52–0.72)0.71*** (0.60–0.83) 40–440.86*0.70** (0.55–0.89)0.73* (0.58–0.93)0.80 (0.63–1.01) 45–490.71*0.49*** (0.33–0.72)0.54** (0.37–0.78)0.54** (0.37–0.79)Mother education No education1.0111 Primary1.28***1.14** (1.05–1.24)1.20*** (1.10–1.30)1.10* (1.01–1.20) Secondary1.33***1.32*** (1.23–1.42)1.36*** (1.26–1.46)1.39*** (1.29–1.49) Higher1.49***1.48*** (1.33–1.64)1.52*** (1.37–1.69)1.43*** (1.29–1.59)Number of ANC No ANC1.0111 Less than 81.08**1.13** (1.05–1.21)1.14** (1.06–1.22)1.06 (0.98–1.14) 8 +1.92***1.87*** (1.70–2.06)1.87*** (1.70–2.06)1.86*** (1.69–2.04)C-section delivery No1.0111 Yes1.55***1.36*** (1.27–1.46)1.34*** (1.25–1.43)1.28*** (1.20–1.37)Place of delivery Health Facility1.0111 Home1.11***1.29*** (1.20–1.38)1.30*** (1.21–1.39)1.02 (0.95–1.09)Wealth index Poor1.0111 Middle1.23***1.04 (0.97–1.12)1.06 (0.98–1.14)1.02 (0.95–1.09) Rich1.36***0.99 (0.92–1.07)0.98 (0.91–1.06)0.88** (0.82–0.95)Number of under 5 children 2 or less1.0111 3 or more0.83***0.91** (0.85–0.98)0.91** (0.85–0.97)0.90** (0.84–0.96)Sex of HH head Male1.0111 Female1.07*1.04 (0.97–1.13)1.10* (1.02–1.19)1.02 (0.94–1.10)Place of residence Urban1.011 Rural0.71***0.76*** (0.69–0.83)0.75*** (0.69–0.81)cOR, Crude odds ratio; aOR, Adjusted odds ratio; Model III—Model with individual level factors at two-level (cluster as random effect); Model IV—Model with individual, and community level factors at two-level (cluster as random effect); Model V—Model with individual, and community level factors at three-level (cluster and country as random effect).*p < 0.05; **p < 0.01; ***p < 0.001. Table 4Multilevel Logistic regression analysis of iron-rich food consumption in South Asia.VariablescORModelsModel III aOR (95% CI)Model IV aOR (95% CI)Model V aOR (95% CI)Child age (in months) 6–111.0111 12–172.62***3.36*** (3.16–3.58)3.35*** (3.15–3.57)3.65*** (3.42–3.89) 18–233.24***4.52*** (4.23–4.83)4.51*** (4.51–4.81)5.21*** (4.87–5.56)Child sex Male1.0111 Female1.09***1.13*** (1.08–1.18)1.14*** (1.08–1.19)1.11*** (1.06–1.17)Birth order First1.0111 Second or third1.08***1.31*** (1.23–1.39)1.32*** (1.24–1.40)1.24*** (1.17–1.32) Fourth and up0.92**1.22*** (1.10–1.34)1.20*** (1.09–1.32)1.03 (0.94–1.14)Currently breastfeeding No1.0111 Yes0.59***0.65*** (0.61–0.69)0.63*** (0.59–0.67)0.70*** (0.66–0.75)Fever No1.0111 Yes1.26***1.27*** (1.20–1.35)1.27*** (1.20–1.35)1.07* (1.00–1.13)Diarrhea No1.0111 Yes0.96a0.91* (0.85–0.98)0.88*** (0.82–0.94)0.92* (0.86–0.99)Mother age (in years) 15–191.0111 20–240.73***0.68*** (0.61–0.76)0.67*** (0.60–0.75)0.82** (0.74–0.92) 25–290.78***0.66*** (0.59–0.75)0.65*** (0.58–0.73)0.83** (0.73–0.93) 30–340.84***0.69*** (0.61–0.79)0.68*** (0.59–0.77)0.83** (0.72–0.95) 35–390.79***0.60*** (0.51–0.71)0.62*** (0.52–0.72)0.71*** (0.60–0.83) 40–440.86*0.70** (0.55–0.89)0.73* (0.58–0.93)0.80 (0.63–1.01) 45–490.71*0.49*** (0.33–0.72)0.54** (0.37–0.78)0.54** (0.37–0.79)Mother education No education1.0111 Primary1.28***1.14** (1.05–1.24)1.20*** (1.10–1.30)1.10* (1.01–1.20) Secondary1.33***1.32*** (1.23–1.42)1.36*** (1.26–1.46)1.39*** (1.29–1.49) Higher1.49***1.48*** (1.33–1.64)1.52*** (1.37–1.69)1.43*** (1.29–1.59)Number of ANC No ANC1.0111 Less than 81.08**1.13** (1.05–1.21)1.14** (1.06–1.22)1.06 (0.98–1.14) 8 +1.92***1.87*** (1.70–2.06)1.87*** (1.70–2.06)1.86*** (1.69–2.04)C-section delivery No1.0111 Yes1.55***1.36*** (1.27–1.46)1.34*** (1.25–1.43)1.28*** (1.20–1.37)Place of delivery Health Facility1.0111 Home1.11***1.29*** (1.20–1.38)1.30*** (1.21–1.39)1.02 (0.95–1.09)Wealth index Poor1.0111 Middle1.23***1.04 (0.97–1.12)1.06 (0.98–1.14)1.02 (0.95–1.09) Rich1.36***0.99 (0.92–1.07)0.98 (0.91–1.06)0.88** (0.82–0.95)Number of under 5 children 2 or less1.0111 3 or more0.83***0.91** (0.85–0.98)0.91** (0.85–0.97)0.90** (0.84–0.96)Sex of HH head Male1.0111 Female1.07*1.04 (0.97–1.13)1.10* (1.02–1.19)1.02 (0.94–1.10)Place of residence Urban1.011 Rural0.71***0.76*** (0.69–0.83)0.75*** (0.69–0.81)cOR, Crude odds ratio; aOR, Adjusted odds ratio; Model III—Model with individual level factors at two-level (cluster as random effect); Model IV—Model with individual, and community level factors at two-level (cluster as random effect); Model V—Model with individual, and community level factors at three-level (cluster and country as random effect).*p < 0.05; **p < 0.01; ***p < 0.001. Multilevel Logistic regression analysis of iron-rich food consumption in South Asia. cOR, Crude odds ratio; aOR, Adjusted odds ratio; Model III—Model with individual level factors at two-level (cluster as random effect); Model IV—Model with individual, and community level factors at two-level (cluster as random effect); Model V—Model with individual, and community level factors at three-level (cluster and country as random effect). *p < 0.05; **p < 0.01; ***p < 0.001. Discussion The study analyzed data from 84,234 South Asian children aged 6–23 months to assess the prevalence and factors associated with iron-rich food consumption and found that only 21.5% had consumed iron-rich food (IRF) in the previous 24 h. IRF consumption was highest in the Maldives (71.9%) and lowest in India (17.4%). Older children, non-breastfed children, those born second or third, and children whose mothers had higher education, and more ANC visits were more likely to consume IRF. Urban children had higher IRF intake than rural children, while children from wealthier households surprisingly had lower odds of consuming IRF. The study employed multilevel analysis to account for layered sources of variability, considering the data’s structure, including individuals within communities and countries. This method is well-suited for complex hierarchical data, where individuals within the same cluster tend to behave more similarly than those in different clusters28. In this study, multilevel analysis revealed strong clustering of IRF consumption at the community and country levels, and the best-fitting model explained much of this variation both between and within clusters. The study found that 21.5% of children in South Asia, aged 6–23 months, consumed foods high in iron. This is lower than the findings from research in Australia (82.6%), Mexico (63.1%), Ireland (90%), East Asia and the Pacific (62.5%), China (51%), Bangladesh (60%), Ireland29, and sub-Saharan Africa (42.1%)30. However, the results align with those of previous studies in Madagascar (19.6%) and Ethiopia (21.41%)31,32. Economic and cultural factors may contribute to the lower prevalence of iron-rich food consumption in South Asia. Religions can also significantly influence dietary habits—for example, Muslims and Jews are not permitted to eat pork, Hindus avoid beef, and many schools of Buddhism discourage eating meat altogether33. On the other hand, compared to children aged 6–11 months, those aged 12–17 and 18–23 months were more likely to consume foods high in iron. The findings align with previous studies14,30,34–36. However, it is important to interpret this result in light of the developmental stage of the children studied. The period between 6 and 23 months is marked by rapid physical and developmental growth, during which children gradually acquire the ability to chew, swallow, and tolerate a wider variety of foods. Therefore, increased consumption of complex or iron-rich foods among older children may partly reflect their greater physiological readiness for such foods. Additionally, maternal knowledge and attitudes are critical determinants of infant feeding practices, particularly for children aged 6–11 months. Mothers occasionally face pressure from female elders to adhere to traditional beliefs that discourage feeding animal-source foods to infants under one year34,37,38. The delayed initiation of complementary feeding at 12 months is also shown by another study39. Children aged 6–11 months were less likely to receive the recommended meal frequency40, and children in the 6–11 age group were four times less likely than older children to receive adequate complementary feeding14. Furthermore, older children often eat iron-rich meals shared with the family41 Furthermore, a child’s birth order indicates their position within the family relative to their siblings. Children in second or third birth positions are 24% more likely to consume IRF than their parents’ firstborns. Birth order appears to significantly influence parental care practice, especially when comparing first- and second-born children. In terms of care, firstborns receive special attention42. However, the higher likelihood of IRF consumption among children of second or third birth order compared to firstborns might be explained by the fact that, as a mother’s parity increases, she gains more experience in preparing and feeding a varied diet to her child43. In the current study, we found that children who were still being breastfed had a lower likelihood of consuming IRF than those who were no longer breastfeeding. This aligns with recent research5,44, which indicates that children who were not breastfed scored slightly higher on dietary diversity than those who were breastfed. Additionally, a child’s illness in the week before the survey influenced how much they consumed. Compared to children who did not experience diarrhea in the previous two weeks, those with diarrhea were much less likely to consume IRF. The findings from earlier studies45,46 can be linked to the reduced intake observed during diarrheal episodes. Misperceptions, lack of knowledge about proper feeding techniques and caregiver supervision, or decreased appetite during diarrhea can all significantly affect intake. Conversely, children who had a fever in the previous two weeks were more likely to consume IRF. Caregivers might increase feeding during illness, aiming to provide additional nutrients to support recovery as suggested by the Pan American Health Organization and the World Health Organization47. This variability underscores the complexity of caregiver feeding behaviors during illness and highlights the need for evidence-based, context-specific interventions to ensure adequate nutrition during and after illness. According to this study, children whose mothers were over 35 had a 29% lower likelihood of consuming iron-rich foods. This finding is consistent with previous research suggesting that the older mother is likely to delay introducing complementary foods and provide a healthy diet48,49. Additionally, a Pakistani study revealed that young mothers had a higher likelihood of enhanced dietary diversification50. This pattern is best explained by cohort effects as younger generations of mothers benefit from distinct societal advantages, primarily higher educational attainment and greater exposure to modern health information51,52. Compared to children born to uneducated mothers, children of educated mothers ate more foods high in iron. Other research findings30,31,53 also agree with this. Furthermore, children of educated parents are more likely to eat iron-rich meals, which may be due to their parents’ better nutritional knowledge, awareness of and access to information about children’s feeding habits, and their consumption of iron-rich foods54. Additionally, mothers with higher education levels are more likely to understand and act upon nutrition-related messages, which can promote healthier feeding behaviors for their children. According to this study, children whose mothers received sufficient ANC visits (8 +) were nearly twice as likely to consume iron-rich foods as children whose mothers received no ANC visits. This finding aligns with previous studies30,55,56. The reason for the higher odds of IRF consumption among children whose mothers had adequate ANC visits could be that mothers are motivated to feed their children iron-rich foods during their ANC appointments and have the chance to learn from healthcare professionals about healthy child nutrition and proper feeding practices57,58. This also aligns with the WHO recommendation for a minimum of eight ANC visits, which provides greater opportunities for repeated counseling, nutrition education, and reinforcement of optimal infant feeding practices59. A noteworthy difference in odds ratios between sufficient (8 +) and fewer ANC visits (less than 8) also suggests that substantial benefits are realized only when mothers complete the recommended ANC schedule. However, since only 18% of mothers in this study attained eight or more ANC visits, the finding also reflects low coverage of optimal ANC services, indicating gaps in access, continuity, and utilization of comprehensive maternal care. Children delivered via cesarean section had 1.28 times higher chances of eating iron-rich meals compared to their peers. This might be because mothers who had cesarean births were more likely to visit prenatal clinics, where they could build strong relationships with medical staff who could give advice and help with proper child-feeding techniques. However, our results differ from those in the earlier study60. Surprisingly, the outcome reveals that children from wealthy households had a far lower probability of IRF consumption. However, our findings contrast with conventional evidence suggesting that children from wealthier households are more likely to consume foods high in iron30–32. This unexpected pattern may reflect a dietary transition among affluent and urban families, where greater access to processed and ultra-processed foods, combined with aggressive marketing, has already shifted perceptions of these products as convenient substitutes, displacing traditional nutrient-rich diets61. Such transitions have been observed in several low- and middle-income countries, where rising income levels are associated with reduced intake of traditional foods and increased consumption of ultra-processed items. For instance, one in three children in India was fed confectionery daily, one in four children in Nigeria were fed instant noodles daily, and two in five children in Ethiopia were fed processed bread daily62. Furthermore, children from households with more than two under-five children were significantly 10% less likely to consume iron-rich food compared to children in families with two or fewer under-five children. Another study found a similar result63. It has also been suggested that housing areas affect IRF usage. Compared to children from urban regions, children from rural areas had a 25% lower likelihood of consuming an iron-rich diet. Additionally, research using national survey data from Sri Lanka50 and Madagascar32 reached the same conclusion. Moreover, urban living may offer improved household socioeconomic status, better access in terms of distance and food diversity, universal access to education, and adequate use of maternity healthcare, all of which can enhance dietary practices and knowledge64–66. The use of a large dataset from the six Demographic and Health Surveys (2015–2018), which represent the population of South Asian countries, is a key strength of the study. Therefore, we believe that the findings of this study apply to all South Asian children aged 6–23. However, the following minor limitations should be considered when interpreting the results: The study’s cross-sectional design prevents it from establishing a cause-and-effect relationship. Additionally, our measure of iron-rich food consumption only reflects intake on a single day (within the past 24 h) of such foods; it does not account for usual intake, amounts consumed, or whether the intake meets the WHO’s recommended daily iron intake. Some important factors-such as mothers’ employment status and indicators of child stunting, wasting, and undernutrition-could not be analyzed due to missing data. Given the multidimensional nature of the factors influencing IRF consumption, addressing this issue requires a comprehensive, multi-sectoral approach. Along with nutrition-specific interventions, broader strategies are vital, such as improving access to health services, strengthening educational opportunities for women, and expanding social protection programs to support vulnerable households. Tailored efforts should focus on empowering mothers with limited access to ANC, those with lower education levels, and those from disadvantaged socioeconomic backgrounds. To boost IRF consumption during early childhood, integrated programs that combine behavior change communication, health system strengthening, and community-based support mechanisms are essential. Conclusion The study highlights substantial variation in the consumption of iron-rich foods among children across the region, with overall intake remaining relatively low. Current feeding practices do not sufficiently address the increased iron requirements that occur during the critical developmental period from six to twenty-three months. Key factors associated with lower IRF consumption included male gender, ongoing breastfeeding, older maternal age, lower maternal education, fewer ANC visits, having multiple young children in the household, and residing in rural areas. The findings underscore the important role of maternal education and adequate antenatal care in promoting the intake of iron-rich foods among children, suggesting that interventions targeting these factors could improve child nutrition outcomes across South Asia.