Menu
Menu
Menu
Menu
Background Understanding the burden and contextual risk factors is critical for developing appropriate interventions to control undernutrition. Methods This study used data from the 2014 Ghana Demographic and Health Survey to estimate the prevalence of underweight, stunting, and wasting. Single multiple logistic regressions were used to identify the factors associated with underweight, wasting and stunting. The study involved 2720 children aged 0–59 months old and mother pairs. All analyses were done in STATA/IC version 15.0. Statistical significance was set at p<0.05. Results The prevalence of underweight, wasting and stunting were 10.4%, 5.3%, and 18.4% respectively. The age of the child was associated with underweight, wasting and stunting, whereas the sex was associated with wasting and stunting. Normal or overweight/obese maternal body mass index category, high woman’s autonomy and middle-class wealth index were associated with a lower odds of undernutrition. The factors that were associated with a higher odds of child undernutrition included: low birth weight (<2.5 kg), minimum dietary diversity score (MDDS), a higher (4th) birth order number of child, primary educational level of husband/partner and domicile in the northern region of Ghana. Conclusion There is still a high burden of child undernutrition in Ghana. The age, sex, birth weight, birth order and the MDDS of the child were the immediate factors associated with child undernutrition. The intermediate factors that were associated with child undernutrition were mainly maternal related factors and included maternal nutritional status and autonomy. Distal level factors which were associated with a higher odds of child undernutrition were the wealth index of the household, paternal educational status and region of residence. We recommend that interventions and policies for undernutrition should address socioeconomic inequalities at the community level while factoring in women empowerment programmes.
The conceptual framework for this study was founded on previous studies that have identified and described risk factors of malnutrition in children [5,9,23,26]. The framework used is based on the premise that distal factors may determine the nutritional status of children by acting directly or indirectly through some interrelated mediating factors except for age and gender of the child. Briefly, according to our framework, the immediate causes of childhood undernutrition include food, birth weight, the birth order number of the child and diseases. Infections and diarrhoea can decrease food intake and nutrient utilization resulting in poor nutrition, growth, and development of the child [27,28]. Also, the immediate causes of childhood undernutrition are rooted in problems at the household level. Maternal undernutrition during pregnancy can result in low birth weight at birth, which is associated with an increased risk of undernutrition in early childhood [29]. Large family sizes may lead to inadequate food intake, as do poor access to safe water and sanitation facilities lead to an increase in diseases, which in turn affects food intake and utilization [30]. Caregivers may seek care for their sick children when health care services are accessible and affordable. Furthermore, each household level problem, in turn, has its correlated factors at the distal level. As a fact, some household behaviours are modelled by cultural and religious norms prevalent in the community [5]. Education, employment, household wealth and place of residence are indices of socioeconomic status and may reflect access to resources by the household [28]. A higher maternal educational level and household wealth index are associated with increased access to household dietary needs, health care services and better living conditions, which are inhibitors of childhood undernutrition [31,32]. Our framework lays out the hierarchical relationship between the risk factors for childhood undernutrition that were examined in this study (Fig 1). We used the child recode dataset of the 2014 GDHS. Approval for the use of the dataset was obtained from ICF international. The Demographic and Health Survey (DHS) is a nationally representative survey that provides coverage data at the population level on key health indicators including reproductive health, fertility, child health, and nutrition from which differences can be assessed by bio-demographic, socioeconomic and geographic characteristics after disaggregation. Details about the survey can be found in the DHS Methodology report [33]. The 2014 GDHS was carried out by the Ghana Statistical Service (GSS), Ghana Health Services (GHS), and the National Public Health Reference Laboratory (NPHRL) of the GHS. The survey employed a multistage and multi-sampling technique. Sampling units (clusters) were selected in the first stage. The second stage involved the systematic selection of 12, 831 households. Three different questionnaires were used to collect information on household characteristics, fertility, morbidity, mortality and child health. Eligible women for interview were all women aged 15–49 years who were either permanent residents or visitors who stayed in a selected household the night preceding the survey. Weight and height measurements were collected from eligible women and children 0–59 months. Children from selected households were measured irrespective of whether their mothers were interviewed. The sampling frame used was updated from the 2010 population and housing census (PHC). The response rate was 97% for the women’s questionnaire. Height and weight measurements were taken for 3,118 children 0–59 months. However, anthropometric information was available for 2,895 children in the dataset. We excluded children who were flagged for z-scores of nutritional status indices (n = 175), which led to the final sample of 2720 (weighted n = 2636) children under five years of age for analysis. Further details on the survey design and data collection process have been explained elsewhere [20]. Some variables were recategorized to produce enough sample for data analysis. The three dependent variables in this study were underweight, wasting and stunting. Weight-for-age (WAZ), weight-for-height (WHZ) and height-for-age (HAZ) z-scores of less than -2 standard deviations (SD) from the median according to the 2006 child growth standards of the World Health Organization (WHO) were used to define underweight, wasting and stunting respectively [34]. The z-scores cut-off point was used to construct binary measures of underweight (WAZ < -2SD), wasting (WHZ < -2SD) and stunting (HAZ < -2SD). A dummy variable with a value of “1” was used in each case to identify children who were underweight, wasted or stunted and “0” for children who are not underweight, wasted or stunted. Control variables: the age (in months) and sex of the child were considered as control variables. Age was categorized as 0–5; 6–11; 12–23; 24–35; and 36+ months of age. The immediate factors (child level factors) included in the study were child’s birth weight; child’s birth order number among other living children; dietary diversity score (DDS); fever and cough episode in the last two weeks before the survey; and diarrhoea episode in last two weeks before the survey. Fever, cough and diarrhoea were considered measures for child’s health status. A DDS comprising 7 food groups was created for the children based on the available data. The food groups included grains, roots and tubers; legumes and nuts; dairy products (cheese, milk, and yoghurt); flesh foods (meat, fish, poultry); eggs; vitamin A rich fruits and vegetables; and other fruits and vegetables [35]. In the DDS, a score of ‘1’ else ‘0’ was assigned if the child consumed at least one food item from each of the food groups. The aggregated scores of the 7 food groups comprised the DDS which ranged from 0–7. The acceptable minimum DDS was the consumption of foods from at least four food groups. The intermediate factors (household and maternal factors) included in the study were mother’s age; mother’s parity; mother’s Body-Mass-Index (BMI) categorized as thin (BMI< 18.5kg/m2), normal (BMI 18.5–24.9 kg/m2) and overweight/obese (BMI ≥25 kg/m2); the timing of the first ANC visit; the place used for delivery by the mother; health insurance coverage; woman’s autonomy; household size; type of toilet facility; and source of drinking water. Antenatal care use in Ghana is almost universal [20]; hence, the timing of the first ANC visit and the place of delivery were used as measures of mother’s health-seeking behaviour during pregnancy and for childbirth. The woman’s autonomy was measured by her involvement in household decision making, attitude towards wife beating and property ownership. A cumulative autonomy index score was created from the summation of individual scores (see supporting material S1 Table). Tertiles of woman’s autonomy was constructed from the final autonomy index score to provide a measure for woman’s autonomy. This method has been used in other studies [36–38]. The categorization of the type of toilet facility and source of drinking water was guided by the World Health Organization & United Nations Children’s Fund definitions [39]. The definition of improved household toilet facility was adapted to take into consideration the housing system in some parts of Ghana [40–42]. Therefore, the use of ‘Improved household toilet’ in this study defined all households with access to improved toilet facilities, including those shared with other household members. The distal factors (socioeconomic and cultural factors) considered were the administrative region, place of residence (rural or urban), mother’s educational level, husband/partner’s educational level, mother’s employment status, household wealth index and religion of the mother. Mother and husband/partner’s educational level was measured by three dummy variables; no formal education, primary, and secondary or higher. The wealth index is a composite measure of a household's cumulative living standard and was calculated in the GDHS by using easy-to-collect data on a household’s ownership of selected assets, such as televisions and bicycles; materials used for housing construction; and types of water access and sanitation facilities. Household wealth quintiles ranging from the poorest to the richest were used as a measure of the wealth index. It was presumed that the independent variables would exhibit different patterns of relationship across and within hierarchical levels of each of the dependent variables. Therefore, multiple single logistic regressions were preferred over multivariate regression models in identifying the determinants of underweight, wasting and stunting [43]. Besides, the “svy” command prefix used to adjust for the survey design used by the DHS program can be used with single logistic regression models and not multivariate regression models to produce robust coefficients, standard errors and confidence intervals that are representative. The model fitting process in this study involved three stages. Firstly, the independent association of each of the distal factors with each of the forms of undernutrition in the absence of the intermediate and immediate factors was assessed (model 1). Secondly, distal factors were fitted with the intermediate factors to assess the association between the intermediate factors and undernutrition adjusting for the confounding effects of distal factors (model 2). Finally, the distal factors and intermediate factors were fitted with immediate factors; this produced the “best fit” independent association between the immediate factors and undernutrition while adjusting for the confounding effects of the distal and intermediate factors and the independent relationship between distal factors and undernutrition (model 3). The age and sex of the child were considered as control variables and maintained in each of the models. The model fitting process was guided by Victora, Huttly, Fuchs, & Olinto [44]. To avoid an excessive number of parameters and unstable estimates in subsequent models, only variables with a p-value <0.1 were retained in subsequent models [45]. We entered pairwise interaction terms in order to explore potential nonlinearities, but none of these interactions was statistically significant in the final models. Prevalence estimates with their corresponding confidence intervals (CI) were calculated for the dependent variables, and the Chi-square (χ2) test was used to assess significant differences between the groups. Adjusted odds ratios (AOR) with their corresponding 95% CIs were reported for risk factors. All data analyses were done using STATA/IC version 15.0 for Windows (StataCorp LLC, College Station, Texas USA). The ‘svyset’ and ‘svy’ command prefix, as well as weights, were used to adjust for the complex study design used by the DHS program. This study did not need any ethical clearance because it is a secondary analysis of data from the 2014 GDHS. The dataset used for the analyses did not contain personal identifiers to respondents or households; the DHS Program protects the privacy of respondents and household members in the surveys. The DHS survey procedures were approved by the Institutional Review Board of ICF Macro International (Calverton, Maryland USA) and the Ethics Review Committee of the Ghana Health Services. Information on the ethical considerations of the DHS survey can be obtained online (www.dhsprogram.com). Nonetheless, permission was obtained from ICF to use the dataset. Moreover, the dataset was used for the sole purpose of this study and the sources from which relevant ideas were obtained for this study have been duly referenced.
