Childhood Obesity and Overweight in Uganda: Evidence From the Uganda Demographic Health Survey 2016

Background: Childhood obesity is an emerging public health problem. Although previously a problem of high-income countries, low- and middle-income countries are now registering higher proportions of overweight and obese children. Studies in Africa have mainly focused on undernutrition among children. This paper explores the factors associated with childhood obesity and overweight in Uganda using data from the Uganda Demographic and Health Survey (UDHS) of 2016. Methods: We used Uganda Demographic and Health Survey (UDHS) 2016 data of 4,338 children less than ve years. Multistage stratied sampling was used to select study participants and data were collected using validated questionnaires. We used multivariable logistic regression to determine factors associated with obesity and overweight among children under the age of ve in Uganda. Results: The prevalence of overweight and obesity was 5.0% (217/4338) (95% CI: 4.3–5.6) with overweight at 3.9% (168/4338: 95% CI: 3.2–4.3) and obesity at 1.1% (49/4338: 95% CI: 0.8–1.5). Boys were more likely to be overweight or obese (adjusted odds ratio: aOR = 2.00; 95% CI 1.42–2.82) compared to girls. Furthermore, children from the Western region (aOR = 1.61; 95% CI 1.07–2.44) compared to those from the North, children below the age of 49 months and those with mothers who were overweight or obese (aOR = 3.36; 95% CI 1.53–7.34) were more likely to be obese or overweight compared to their counterparts who were above 48 months and those with underweight mothers respectively. Conclusion: The present study showed male sex, older age of the children, nutritional status of the mothers and region of residence were associated with overnutrition among under ve children in Uganda.

Tackling childhood obesity is critical because of its myriad health and socioeconomic consequences.
Childhood obesity is linked to premature mortality and adverse mental health in both the short and longterm. Furthermore, overweight and obese children suffer higher risks of early chronic diseases onset, for example, diabetes, dyslipidemias, cardiovascular illnesses and some cancers [1,14,15]. Likewise, they are reported to have low educational attainment because of poor psychosocial wellbeing [1,4]. Childhood obesity also results in higher costs to health systems, and greater nancial burden to households [1].
Studies in Africa have primarily focused on the study of undernutrition and not overnutrition among children. The few studies exploring childhood overweight and obesity have mostly come from Nigeria and South Africa, and only a few from other countries in sub-Sahara Africa [12]. In Uganda, demographic health surveys (DHS) reported a prevalence of 3-4% between 2011 and 2016 [16,17]. Nevertheless, the distribution and determinants of obesity were not articulated. Understanding the distribution and determinants of obesity is vital in designing public health messages and interventions [11,14]. This paper aimed at exploring the factors associated with childhood obesity and overweight in Uganda using data from the Uganda Demographic and Health Survey (UDHS) of 2016.

Methods
Study design and participants UDHS 2016 was a nationally representative cross-sectional study conducted using validated questionnaires. UDHS is a periodical survey that is carried out every ve years as part of the MEASURE DHS global survey and collects Information on demographic, health and nutrition indicators. The survey was conducted between June and December 2016 using strati ed two-stage cluster sampling design that resulted in the random selection of a representative sample of 20,880 households [18,19]. The households were randomly selected in two stages: clusters (or enumeration areas) were drawn in the rst stage and then a count within each cluster led to a list of households from which was conducted a systematic sampling with equal probability [18]. A detailed explanation of the sampling process is available in the UDHS 2016 report [18]. A systematic random draw was conducted amongst the selected households to choose households whose women/ mothers' and children's anthropometric measurements (weight and height) were taken. Anthropometric measurements were done on a subsample of about onethird of households [18]. Weight was taken with an electronic SECA 878 at scale while a Shorr Board® measuring board was used for height [18]. Children less than 24 months were measured lying down.
Our secondary analysis excluded children whose BMI z-score were missing or was recorded as "Flagged cases". In the children's dataset, a nal weighted sample of 4338 was analyzed after excluding agged cases and those with missing values. Written permission to access the whole UDHS database was obtained through DHS program website at the address https://dhsprogram.com/

Outcome variables
The BMI z-scores based on WHO 2006 reference population were used to assess over nutrition (obesity and overweight) [20]. The outcome over nutrition (coded as 1) combined both overweight and obesity. Children whose BMI z score was over two were considered as overweight and those with a BMI z score greater than three were considered as obese [20].

Independent variables
Independent variables were categorized into children, parents' and household characteristics that were chosen basing on previous studies [20][21][22] and availability in the UDHS data base.
Household characteristics. Wealth index of household (categorized into quintiles: richest, richer, middle poorer and poorest), type of residence (urban and rural), number of household members (less than 5 and 5 and above), sex of household head (female and male) and region (North, East, West and Central).

Statistical analysis
We used the SPSS analytic software version 25.0 Complex Samples package for this analysis. Weighted data was used to account for the unequal probability sampling in different strata. Frequency distributions were used to describe the background characteristics of the children. Pearson's chi-squared tests were used to investigate the signi cant differences between childhood obesity and overweight and the explanatory variables. Bivariable logistic regression was also conducted and we present crude odds ratio (COR), 95% con dence interval (CI) and p-values. Independent variables found signi cant at p-value less than 0.25 in the bivariable analysis were included in the multivariable model. Adjusted odds ratios (AOR), 95% Con dence Intervals (CI) and p-values were calculated with statistical signi cance level set at pvalue < 0.05. All variables in the model were assessed for collinearity, which was considered present if the variables had a variance in ation factor (VIF) greater than 10. However, none of the factors had a VIF above 3. Sensitivity analysis was done comparing children who were obese or overweight with those who were normal after excluding those who were thin.

Results
The mean age of children was 28.34 with a standard deviation (sd) of 17.20 months while that of the mothers was 28.80 (sd 6.83). Boys made up 52% of the study participants, and the majority of children were between 13-24 months (28%), resided in rural areas (77.8%) and belonged to households with size of 5 and above members (67.2%). The mean weight, height, number of household members and BMI z-score were 11.5 kilograms (sd 3.57), 83.5 centimeters (sd 14.0), 6 members (sd 2.7) and 0.20 (sd 1. 19) respectively. Of the 4,338 children, 5.0% (217/4338) (95% CI: 4.3-5.6) were overweight or obese (overweight 3.9% (95% CI: 3.2-4.3) and obesity 1.1% (95% CI: 0.8-1.5). More detailed characteristics of study participants are shown in Table 1. the Western region (aOR = 1.61; 95% CI 1.07 to 2.44) compared to those that are from the North. Children who were older (above 12 months) and those with mothers who were overweight or obese were more likely to have obesity or overweight compared to those with underweight mothers and their counterparts who were less than 12 months, respectively.

Sensitivity Analysis
After excluding children who were thin, and remaining with only children who were obese, overweight and had normal BMI z-score, the prevalence of over-nutrition was 5.2% (95% CI: 4.4-5.7) with overweight at 4.0% (95% CI: 3.3-4.5) and obesity 1.2% (95% CI: 0.9-1.5). Similar to the original analysis, age and sex of the child, region and nutritional status of the mother remained positively associated with childhood obesity and overweight.

Discussion
This study provides evidence on the factors associated with childhood obesity and overweight in Uganda from the recent UDHS of 2016. We revealed that, 5% (217/4338, 95% CI: 4.3-5.6) of the children were either overweight or obese.
In this study, childhood overnutrition was associated with maternal overnutrition. This is in agreement with ndings, a similar study in Nepal which showed that the risk of childhood obesity is increased in cases of maternal overweight and obesity both pre and during pregnancy which predisposes them to obesity in later life [23]. Various plausible mechanism can explain this relationship. Firstly, genetic factors and obesity-promoting environment enhances the effect of genes related to body fatness. For example, changing environmental conditions that would allow a pregnant woman to have easier access to foods that are high energy dense, would make changes in the expression of genes related to body fatness [24].
Therefore, consumption of these foods is associated with maternal overnutrition which explains its role in predisposing newborns to high birthweight and obesity [24]. Furthermore, sedentary behaviors coupled with low levels of physical activity and poor food choices which are generally energy dense increases their odds being obese [25,26].
It is implicit that the prevalence of childhood overweight and obesity is relatively high in urban areas and households with poor socio economic status compared to those in rural areas [25,27]. A review that compared overweight prevalence in low and middle income countries did not show signi cant differences between the richest and the poorest quintiles [23], agreeing with our ndings. Overweight and obesity was not associated with wealth index and place of residence. This may be partly explained by the economic growth resulting from rapid nutrition transition ensuing in rural and urban areas across difference socioeconomic groups often characterized by a dramatic shift towards consuming inexpensive foods that are high in fat and sugar [28]. Further, the wealth index as a measure of socioeconomic status expresses some methodological weakness that partly explain a weak association [29].
This study demonstrated that the male sex was signi cantly associated with childhood overweight or obesity. A study by Beatrice J et al. [30] and another by Dubois et al. [31] were consistent with our nding. Biological in uences may partly explain the observed association as body composition between sexes start early in the fetal and postnatal periods [32,33]. Girls have been shown to have less third trimester fetal growth compared to boys [33]. After birth, girls tend to have greater fat mass and less fat-free mass, which is linked with less energy intake and less calorie needs for girls compared to boys [32,34,35].
Furthermore, girls have more circulating leptin, a hormone that suppresses appetite and promotes energy utilization [32,36]. Gender-based stereotypes that are practiced by some parents such as a feminine identity being characterized by eating less portions may partly explain the observed nding [32]. Evidence has shown that some parents tend to be more concerned about girls' weight status than boys' hence the boys are usually given more food [32,37]. Gender behavioral differences in sleeping, physical activity and television-watching have been observed with girls being shown to sleep less and engage in less physical activity and watch television more than boys [32,38,39]. These behaviors have been signi cantly associated with overweight and obesity [39,40].
In this study the age of a child was associated with overweight and obesity. A study conducted in Cameroon found a similar association [20]. Children aged 49 months and below had increased odds of having over-nutrition compared to the older counterparts aged between 49 and 59 months. This association was stronger among the younger children less than 37 months and weakened in older children between 37 and 48 months. Toddlers and preschoolers' total diet and activity level play an important role in determining a child's weight [41]. As such, older children are more active compared to the younger ones, hence the more energy they expend [41]. Other plausible mechanism that explains this association include environmental and behavioral factors such as socio-economic conditions, consumption of high calorie and fast foods and lifestyle changes [20,41].
Children from Western Uganda were more likely to have over nutrition compared to those from the Northern region. Evidence shows that the Western region has the second highest GDP per capita, with the Northern region, having some of the poorest districts in Uganda [42,43]. Additionally, the Western region receives higher amounts of rainfall and produce more crop yields which have led to a higher level of food security [43]. All these factors lead to increased food availability and hence over-nutrition. Northern Uganda has some of the poorest and most food insecure sub-regions which could partly be attributed to the fact that the region experienced a long civil war which greatly affected the economy and agricultural production [19]. Furthermore, most people in the Northern part of Uganda are pastoral communities with some being nomadic which may negatively affect the production and consumption of the foods from agricultural origin (crops) as they mostly focus on mainly pastoral activities [19]. The increased poverty and decreased agricultural production led to decreased food availability and access to food in Northern Uganda.

Strengths
We used a nationally representative sample and weighed the data for analysis and therefore our results are generalized to all Ugandan children below ve years. Secondly, we used data with a large sample size which was collected, entered and cleaned by a team of trained and highly experienced scientists hence limiting mistakes in the data set.

Limitations
Lack of other measures of overweight and obesity e.g., skin fold measurements, data on chronic disease biomarkers e.g. lipid pro le, measure of physical activity, analysis of nutritional characteristics such as dietary habits which would allow assess the development of overweight and obesity. The cross-sectional design is limited by lack of temporality hence causality inferences cannot be made.

Conclusion
This study established the determinants of overweight and obesity as maternal nutritional status (BMI), region of residence, sex, and age of the children. children from western Uganda were more likely to be overweight and obese compared to those from northern Uganda.
Strategies to improve the nutritional status in children should be focused across all socioeconomic groups. Advocating for a sustained political commitment and the collaboration of many private and public stakeholders is paramount to curb childhood obesity epidemic. Preventive interventions need to be strengthened especially in Western Uganda and among boys, children aged below 49 months and those whose mothers have a high BMI. Initiatives like lifestyle modi cations and proper nutrition should be encouraged to reduce overweight and obesity in mothers and promoting measures such as surveillance of weight gain during antenatal consultation and nutritional follow-up of boys. Further studies including nutritional characteristics are needed to understand the association with child age and sex and will help in re ning preventive strategies against childhood overnutrition in Uganda.
There is need to address socio-economic (contextual) factors mainly poverty and regional inequalities. There is also need for further studies to explain why stunting is highest among women with the highest wealth index.