Factors associated with concurrent wasting and stunting among children 6–59 months in Karamoja, Uganda

Children with concurrent wasting and stunting (WaSt) and children with severe wasting have a similar risk of death. Existing evidence shows that wasting and stunting share similar causal pathways, but evidence on correlates of WaSt remains limited. Research on correlates of WaSt is needed to inform prevention strategies. We investigated the factors associated with WaSt in children 6–59 months in Karamoja Region, Uganda. We examined data for 33,054 children aged 6–59 months using June 2015 to July 2018 Food Security and Nutrition Assessment in Karamoja. We defined WaSt as being concurrently wasted (weight-for-height z-scores <−2.0) and stunted (height-for-age z-score <−2.0). We conducted multivariate mixed-effect logistic regression to assess factors associated with WaSt. Statistical significance was set at p < 0.05. In multivariate analysis, being male (adjusted odds ratio [aOR] = 1.79; 95% confidence interval [CI] [1.60–2.00]), aged 12–23 months (aOR = 2.25; 95% CI [1.85–2.74]), 36–47 months (aOR = 0.65; 95% CI [0.50–0.84]) and 48–59 months (aOR = 0.71; 95% CI [0.54–0.93]) were associated with WaSt. In addition, acute respiratory infection (aOR = 1.30; 95% CI [1.15–1.48]), diarrhoea (aOR = 1.25; 95% CI [1.06–1.48]) and malaria/fever (aOR = 0.83; 95% CI [0.73–0.96]) episodes were associated with WaSt. WaSt was significantly associated with maternal underweight (body mass index <18.5 kg/m2), short stature (height <160 cm), low mid-upper arm circumference (MUAC <23 cm) and having ≥4 live-births. WaSt was prevalent in households without livestock (aOR = 1.30; 95% CI [1.13–1.59]). Preventing the occurrence of WaSt through pragmatic and joint approaches are recommended. Future prospective studies on risk factors of WaSt to inform effective prevention strategies are recommended. We conducted secondary data analysis of the June 2015 to July 2018 Food Security and Nutrition Assessment (FSNA) cross‐sectional survey datasets conducted in all the seven districts of the Karamoja Region in North‐Eastern Uganda. Data was pooled from seven FSNA datasets for the current study. These datasets have been described in detail elsewhere (Odei Obeng‐Amoako, Myatt, et al., 2020). The FSNA surveys were carried out by the Makerere University School of Public Health and the International Baby Food Action Network (IBFAN), Uganda, in collaboration with the United Nations (UN) Children's Fund (UNICEF), the UN Food and Agriculture Organization (FAO), the UN World Food Programme (WFP), the Department of Risk Reduction at the Office of the Prime Minister and the Ministry of Health, Uganda. The FSNA protocol adapted the Standardized Monitoring and Assessment of Relief and Transitions (SMART) survey methodology (Golden et al., 2006). The FSNA employed a two‐stage cluster sampling approach. First, a probability proportional to size sampling procedure was used to select clusters from a list of parishes in each district. Secondly, households in each of the clusters were selected by systematic random sampling technique (UNICEF, Department for International Development [DFID], WFP, FAO, & IBFAN, 2018). The estimated number of children and households were selected from each of the clusters per the FSNA protocol (UNICEF, DFID, WFP, FAO, & IBFAN, 2018). FSNA adapted a standardized semistructured questionnaire for data collection by trained research assistants. The FSNA questionnaire consists of household characteristics, food security and nutrition modules. The household and food security modules were administered to household heads or adults present at the time of the interview while the mothers or caregivers (hereinafter referred as caregivers) of children under 5 years were interviewed using the nutrition module. Weight and height measurements were collected from nonpregnant women with children aged 0–59 months in the household. Mid‐upper arm circumference (MUAC) was measured for both pregnant and nonpregnant women with children aged 0–59 months. Weight, height and MUAC measurements were collected from children aged 6–59 months found in the household. Recumbent length was measured for children <24 months old. Bilateral pitting oedema and haemoglobin levels among children aged 6–59 months were also examined. Child age in months was obtained from the child health cards or by maternal recall using a local event calendar. We defined wasted as weight for height z‐score (WHZ) <−2.0, stunted as height for age z‐score (HAZ) <−2.0, and underweight as weight for age z‐score (WAZ) <−2.0 based on z‐scores of the 2006 World Health Organization (WHO) growth standards (WHO, 2006). Degrees of anthropometric deficits were defined as no deficit as z‐scores ≥−2; moderate deficits as z‐scores <−2 and ≥−3 and severe deficits as <−3 z‐scores. We defined acute malnutrition as wasting (WHZ <−2) and/or MUAC <12.5 cm; severe acute malnutrition (SAM) as severe wasting (WHZ <−3) and/or MUAC <11.5 cm and moderate acute malnutrition (MAM) as moderate wasting (WHZ ≥−3 to −2) and/or MUAC ≥11.5 cm and ≤12.5 cm (United Nations High Commissioner for Refugees [UNHCR] & WFP, 2011). The outcome of our study was WaSt among children aged 6–59 months. We defined WaSt as concurrently having WHZ <−2 and HAZ <−2 (Myatt et al., 2018). We explored a range of factors likely to be associated with WaSt in children aged 6–59 months in this analysis based on previous literature and on the UNICEF conceptual framework for under‐nutrition (UNICEF, 1990). These explanatory variables were classified as child level, maternal level (i.e., caregivers) and household characteristics as well as markers of the food security seasons during which the surveys were conducted. We received approval to access the Karamoja FSNA datasets from the Office of the Prime Minister, Uganda. We used a de‐identified dataset for this analysis. We obtained a waiver of consent to use the FSNA datasets for this study from the Makerere University School of Medicine Higher Degrees Research and Ethics Committee. Ethical approval for the present analysis was granted by the Makerere University School of Medicine Higher Degrees Research and Ethics Committee and the Uganda National Council for Science and Technology. Child level factors included in the study were age in months, sex and morbidity. We categorized the age of children into 6–11, 12–23, 24–35, 36–47 and 48–59 months (Garenne et al., 2019). Childhood morbidity was reported by the caregivers. The presence of infectious disease was defined as ARI, malaria/fever, and diarrhoea episodes in the last 2 weeks before the survey. We considered age, the number of live‐births and nutritional status of caregivers. We categorized maternal age as 15–19, 20–29, 30–39 and >40 years (UBOS & ICF, 2018). Maternal educational level was described using the cumulative number of years of formal education attained. Educational level was categorized as no education, 1–7 years (primary) and ≥8 years (secondary and above). We also classified the number of live‐births of caregivers into 1–3, 4–6 and >7 number of live‐births. We categorized maternal body mass index (BMI) as normal (BMI = 18.5–24.9 kg/m2), underweight (BMI < 18.5 kg/m2), and overweight/obese (BMI ≥ 25 kg/m2) (UNHCR & WFP, 2011). Maternal height was categorized as tall (height ≥170 cm), moderately tall (height = 160–170 cm) and short (height <160 cm). Maternal MUAC signifying acute malnutrition was classified as low MUAC (MUAC <23 cm) and normal MUAC (MUAC ≥23 cm) (MoH Uganda & UNICEF, 2016; Ververs, Antierens, Sackl, Staderini, & Captier, 2013). The socioeconomic factors included in the analysis were household heads' sex, maternal and household heads' educational level, wealth index, livestock ownership, access to land for cultivation and season of survey. Similar to maternal educational level, we described household head's educational level as no education, 1–7 years (primary) and ≥8 years (secondary and above). We used principal component analysis to generate a wealth index based on household assets ownership (Rutstein & Staveteig, 2014). These household assets were bed, table, chair, mattress, radio/tape, cell phone, sewing machine, bicycle, automobile, motorcycle, television, axe, panga/machete, hoe, ox‐plough, water tank, seed and food stores (UNICEF, DFID, WFP, FAO, & IBFAN, 2018). We categorized the wealth index from the lowest to the highest as poorest, poor, middle, rich and the richest households using quintiles. We used household livestock ownership and access to land for cultivation as an independent factor for wealth given that Karamoja is an agro‐pastoral setting. Household food consumption scores reflect the diversity and frequency of food items from the different food groups reportedly consumed over the previous 7 days. The food items were grouped into main staples, pulses, vegetables, fruits, meat/fish, milk, sugar and oil. The food consumption scores were weighted according to the relative nutritional value of the consumed food groups (International Dietary Data Expansion [INDDEX] Project, 2018). The thresholds of food consumption scores that were used to determine food consumption status of households were 0–21 (poor), 21.5–35 (borderline), >35 (acceptable) (INDDEX Project, 2018). We evaluated the presence of a latrine in a household as a probable factor associated with WaSt. Sources of drinking water were defined as improved water sources (piped/tap, protected well/spring and borehole fitted with a hand‐pump) and unimproved water sources (surface water, river, dam, runoff, rainwater collected in a tank and water from open well/spring) (WHO & UNICEF, 2006). The association between food security seasons and the occurrence of WaSt was also examined. FSNA surveys were conducted in May/June for the lean (hunger or preharvest) and in November/December for the nonlean (postharvest) seasons. Children aged 6–59 months were included in the analysis. We used WHO flags for outliers; HAZ <−6 and HAZ >6, WAZ <−6 and WAZ >5, and WHZ <−5 and WHZ >5 were excluded as biologically implausible z‐scores values (WHO, 2009). Given that WaSt was the outcome variable, children with incomplete data on WHZ or HAZ or both were excluded from the study. Children with bilateral pitting oedema, height <45 and >120 cm and MUAC >20 cm were also excluded from the dataset (MoH Uganda & UNICEF, 2016). The maternal MUAC measurements which were out of range of upper quartile +3.0× interquartile range (IQR) and lower quartile −3.0× IQR were censored as outliers (Healy, Chambers, Cleveland, Kleiner, & Tukey, 1984). All statistical analyses were conducted using STATA 13.0. We also analysed anthropometric z‐scores per WHO growth standards in STATA (Leroy, 2011). We used descriptive statistics to summarize continuous variables using medians and IQRs and categorical variables using percentages. We used multivariate mixed‐effect logistic regression to assess child level, maternal and household level, socioeconomic and seasons factors associated with WaSt. Mixed‐effect logistic regression method adjusted for the clustering effect of the multistage sampling design used in the FSNA survey. In the multivariate analysis we only included explanatory variables with p < 0.2 from the bivariate regression analysis. Multicollinearity of the variables was assessed by using the variance inflation factor (VIF > 10). In the multivariate analysis, we used a backward stepwise method to remove the nonsignificant variables (p > 0.05). To identify variables with interaction effects in the model, we assessed the significance (p > 0.05) of each interaction term one at a time in the basic model, but none of the interaction terms was conceptually meaningful. We found no confounding factors in our model after testing for a ≥10% change in the effect measure in the presence of another variable. We retained maternal stature and season of the survey in the final model although they were dropped during the backward stepwise process based on literature (Black et al., 2013; Harding et al., 2018). Akaike information criterion (AIC) and Bayesian information criterion (BIC) methods were used to assess for the goodness of fit of our model. Factors associated with WaSt were reported by adjusted odds ratio (aOR) at 95% confidence interval (CI). We set statistical significance at p < 0.05 in this analysis.