Risk factors of stunting and wasting in Somali pre-school age children: results from the 2019 Somalia micronutrient survey

Background: Stunting and wasting in children less than 5 years of age are two key indicators of child malnutrition. Reducing their prevalence is a priority of the global public health community and for Somalia, a country suffering complex humanitarian emergencies such as drought, flooding, conflict and large-scale displacements. Methods: Data from the nationally representative cross-sectional Somalia Micronutrient Survey (SMS 2019) on 1947 children were analyzed to assess the prevalence and potential risk factors of stunting and wasting. Bivariate and multivariable analyses were conducted separately for children 0–5 months and 6–59 months, and population attributable fractions were calculated using adjusted risk ratios produced by Poisson regression models. Results: Among the 1947 children, the prevalence of stunting and wasting were 17.2% (95% CI: 15.0, 19.6) and 11.0% (95% CI: 9.3, 12.9), respectively. Among children 6–59 months of age, those residing in severely food insecure households had a higher risk of stunting (adjusted risk ratio [aRR] 1.47; CI: 1.12, 1.93) compared to those in food secure households. This risk of stunting was also higher in children with inflammation (aRR 1.75; CI: 1.35, 2.25) and iron deficiency (ID) (aRR 2.09; CI: 1.58, 2.80). For wasting, a dose-response relationship was found with household wealth, with the risk of wasting increasing significantly as the household wealth quintile decreased. On the other hand, the risk of wasting was lower in iron-deficient children (aRR 0.69; CI: 0.49, 0.98) than in iron-replete children. Among children 0–5 months of age no variables remained statistically significantly associated with stunting in the multivariable analysis. Wasting, however, was more common in children with recent diarrhea (aRR 3.51; CI: 1.68, 7.36). Conclusions: Nutritional status of children in Somalia may be improved by prevention of diarrhea and other infections and improvements in household food security. The 2019 SMS, a nationally-representative stratified cross-sectional household-based survey, was conducted between December 2018 and September 2019. Sampling was done from 6 strata: 1) Somaliland, 2) Puntland, 3) the Somalia states of Hirshabelle and Galmudug; 4) the Somalia states of Jubaland and South-West, 5) the Banaadir administrative region of Somalia, and 6) internally displaced persons (IDPs) settlements in all five aforementioned geographic strata combined. The survey used a two-stage cluster sampling procedure. Enumeration areas (EAs) and IDP camps were primary sampling units which were selected using probability proportional to size. Within each selected primary sampling unit, the required number of households were selected using simple random sampling. In the first 5 geographic strata, 25 EAs were selected in each stratum after excluding EAs that were not accessible due to insecurity; EAs were categorized as either urban or rural. In the 6th stratum, 25 IDP camps were selected. In each EA and IDP camp, 16 households were randomly selected resulting in a total sample of 2400 households. Further details of the selection procedure can be found elsewhere [12]. Within participating households, all children 0–59 months of age were eligible to be recruited into the survey, questionnaires in the child questionnaire were answered by the selected child’s mother or caretaker. Non-pregnant women were recruited from a randomly selected one-half subsample. Data on household characteristics were collected from the household head or knowledgeable adult household member. The household interview collected information on the household’s dwelling, durable goods ownership, water source, and sanitation facility using standard questions widely used in health and nutrition assessment surveys worldwide. Durable goods and dwelling characteristics were used to calculate a household wealth index using the standard methods [13, 14]. Household-level sanitation facilities and water source were respectively classified as adequate/inadequate or safe/unsafe based on WHO/UNICEF guidelines [15]. To estimate sanitation status at the community-level, the proportion of households in each cluster with inadequate sanitation facilities was calculated and categorized into sub-groups of 0–19%, 20–39%, 40–59%, 60–79%, and 80–100%. In addition, household food security was assessed using the Household Food Insecurity Access Scale (HFIAS) questionnaire module. The final scale was categorized into four categories: food secure, mild food insecurity, moderate food insecurity, and severe food insecurity [16]. The child questionnaire collected data on age, sex, recent morbidity, and consumption of vitamin and mineral supplements. Child age was calculated by subtracting the child’s date of birth from the date of the interview. As much of the population in Somalia was not calendar literate, a local event calendar for the past 5 years was developed to facilitate the approximation of the child’s birth date. The child questionnaire included standard infant and young child feeding questions [17] that were administered to children 0–23 months of age. Questionnaire data was also collected from women, however, as women were only recruited from ½ of all households, maternal-level data was not included in this analysis. The Open Data Kit (ODK) software was used for direct electronic questionnaire data entry. Anthropometrists were trained in standard anthropometric techniques [18]. Performance during a standardization exercise, the results of a post-test, and observations from trainers were used to select the best performing team members. All survey procedures were practiced in a field test under close supervision prior to the start of data collection. During the survey field work, the weight and height or length of each child was measured according to standard procedures [19]. Child weight was measured using an automatically taring bathroom scale (SECA®, Hamburg, Germany); all scales were calibrated each morning before the start of data collection. The standard wooden height board (UNICEF item number S0114540, Copenhagen, Denmark) was used to measure the child height or length. Length/height and weight measurements were taken twice by the same measurer and both values recorded. The second length/height measurement was taken after the child was removed and replaced on the height board to account for incorrect positioning during the first height measurement. For weight measurements, children that could not stand were directly weighed on the scale, and the scale’s tare function was used for children that could not stand. For the second measurement, children were either re-weighed directly or the tare process was repeated. Capillary blood was collected from all children 6–59 months; no blood was taken from children younger than 6 months of age to prevent potential injury. Blood was collected from the heel in children 6–11 months of age and from the finger in children 12–59 months of age. After sterilization with an alcohol pad, the puncture site was wiped dry and punctured with a “contact activated” high-flow blade lancet (Becton Dickinson, Franklin Lakes, NJ, USA). After wiping the first drop of blood away, the second and third drop of blood were used to measure hemoglobin concentration and recent or current malaria infection, followed by the collection of 300–400 μl of capillary blood into a silica-coated blood collection tube (Sarstedt, Microvette® 300 Z, Nümbrecht, Germany). Following on-site measurements of haemoglobin and malaria parasitemia, the labeled microtainers were placed in a cool box at 2–8 °C in the dark for transport in the evening of the same day to one of four state laboratories. Samples were centrifuged at 3000 rpm for 7 min to separate the serum, which was then aliquoted into labeled cryovials. Aliquots were stored at − 20 °C and later shipped on dry ice for analyses. Serum samples were analyzed for retinol-binding protein (RBP), ferritin, C-reactive protein (CRP), and alpha 1-acid glycoprotein (AGP) at the VitMin-Lab (Wilstaett, Germany) using an ELISA method. Anthropometry, anemia and malaria data were collected on paper forms, and subsequently entered into ODK on the same day. Data analysis was done using Stata/IC version 14.2. All analyses of questionnaire data were conducted using sampling weights to account for the unequal probability of selection in the six strata. Z-scores were calculated based on WHO’s 2006 Growths Standards [20], and children with height-for-age z-scores (HAZ) and weight-for-height z-scores (WHZ) < − 2.0 standard deviations were classified as stunted and wasted, respectively [20]. Prior to data analysis, the quality of anthropometric measurements was examined for all teams. As a result of this examination, child height measurements taken by all teams working in Galmudug State were excluded, and length, height, and weight measurements taken by one team in South-West State and one team in Jubaland State were excluded [12]. Further details about anthropometric data quality are presented elsewhere [12]. Any inflammation was defined as concentrations of CRP > 5 mg/L and/or concentrations of AGP > 1 g/L [21]. After adjusting serum ferritin and RBP concentrations for inflammation using the Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia (BRINDA) approach, iron deficiency (ID) was defined as serum ferritin concentrations < 12 μg/L [22] and vitamin A deficiency was defined as RBP concentrations < 0.7 μmol/L [23]. For categorical variables, proportions were calculated to derive the prevalence. All measures of precision, including 95% confidence limits and chi square p values for differences in subgroup prevalence, were calculated accounting for the complex cluster and stratified sampling used by the SMS 2019. For this analysis, we identified potential risk factors from the recent WHO conceptual framework for stunting [24] and other potential risk factors of wasting identified from a review of the literature. Bivariate and multivariable analyses were conducted separately for children 0–5 months and 6–59 months to identify potential risk factors of stunting and wasting. Separate analyses for these age groups were done to account for the fact that blood biomarkers were only available in children 6–59 months of age. For all analyses, significance was accepted at P < 0.05. Variables with chi-square p-values < 0.1 during categorical bivariate analyses were included in the four subsequent multivariable models. When applicable during the bivariate analysis, a nonparametric test for trend was conducted using Stata’s nptrend command [25] to identify dose-response relationships. Following tests of collinearity (i.e., variance inflation factor), multivariable Poisson regression models were run using backwards elimination until only variables with statistically significant associations were remaining. To account for age differences, age in months was included in all regression models. Following the multivariable analyses, the population attributable fraction (PAF) was calculated for all statistically significant potential risk factors. PAF was calculated using the equation pdaRR-1aRR using the adjusted risk ratios (aRRs) produced by the Poisson regressions and the proportion of cases with the potential risk factor of interest (pd) [26].