Menu
Menu
Menu
Menu
Background: A small but growing body of research indicates that progress in reducing child malnutrition is substantially uneven from place to place, even down to the district level within countries. Yet child malnutrition prevalence and trend estimates available for public health planning are mostly available only at the level of global regions and/or at country level. To support carefully targeted intervention to reduce child malnutrition, public health planners and policy-makers require access to more refined prevalence data and trend analyses than are presently available. Responding to this need in Ghana, this report presents trends in child malnutrition prevalence in socio-demographic groups within the country’s geographic regions. Methods. The study uses the Ghana Demographic and Health Surveys (GDHS) data. The GDHS are nationally representative cross-sectional surveys that have been carried out in many developing countries. These surveys constitute one of the richest sources of information currently available to examine time trends in child malnutrition. Data from four surveys were used for the analysis: 1993, 1998, 2003 and 2008. Results: The results show statistically significant declining trends at the national level for stunting (F (1, 7204) = 7.89, p ≤.005), underweight (F (1, 7441) = 44.87, p ≤.001) and wasting (F (1, 7130) = 6.19, p ≤.013). However, analyses of the sex-specific trends revealed that the declining trends in stunting and wasting were significant among males but not among females. In contrast to the national trend, there were significantly increasing trends in stunting for males (F (1, 2004) = 3.92, p ≤.048) and females (F (1, 2004) = 4.34, p ≤.037) whose mothers had higher than primary education, while the trends decreased significantly for males and females whose mothers had no education. Conclusions: At the national level in Ghana, child malnutrition is significantly declining. However, the aggregate national trend masks important deviations in certain socio-demographic segments, including worsening levels of malnutrition. This paper shows the importance of disaggregated analyses of national child malnutrition data, to unmask underlying geographic and socio-demographic differences. © 2013 Amugsi et al.; licensee BioMed Central Ltd.
This study uses data from the Ghana Demographic and Health Surveys [27]. The surveys were conducted in 1993, 1998, 2003 and 2008 by the Ghana Statistical Service and Ghana Health Service, with technical and financial support from ICF Macro through the USAID-funded MEASURE DHS programme. The surveys were designed to be representative at the national, regional and rural–urban levels. A two-stage probabilistic sampling design was used to select clusters (census districts) at the first stage. The second stage involved the selection of households from these clusters. All women and men aged 15–49 in the selected households were eligible to participate in the surveys. The household response rates were 98.4% in 1993, 99.1% in 1998, 98.7% 2003, and 98.9% in 2008. The data were collected at two levels—the household and individual levels. At the household level, information was collected on household characteristics such as source of drinking water, toilet facilities, cooking fuel, and assets of the household. At the individual level, questionnaires were administered to women aged 15–49 and men aged 15–59 to gather information on individual characteristics and health behaviors, and information on children in the household. Child nutritional status was assessed by height-for-age z-scores, weight-for-height z-scores and weight-for-age z-scores using the new WHO Child Growth Standards [28]. A child was considered stunted, wasted or underweight if their height-for-age, weight-for-height or weight-for-age z-scores were further than −2 standard deviations from the median of the reference sample used to construct the WHO 2006 growth standards. The DHS 2008 survey used the new WHO Child Growth Standards [29], while the earlier DHS surveys used the NCHS growth reference [30-32]. For the purposes of cross-survey comparability, we calculated z-scores using the new WHO Child Growth Standards, using a syntax file provided by the WHO [33]. This syntax file automatically flagged all biological implausible values. Thus, height-for-age z-scores less than −6.0 and greater than +6.0, weight-for-height z-scores less than −5.0 and greater +5.0 and weight-for-age z-scores less than −6.0 and greater than +5.0 are excluded from our analysis. The socio-demographic variables included child sex and age, mother’s education, urban/rural residence, region of residence and Wealth Index (composed using factor analysis to combine household-level information about housing quality and ownership/access to material goods). Some of the variables were re-coded in order to attain reasonable sample sizes, and also based on suggestions in the literature. For maternal education, incomplete and complete primary were recoded as “Primary”, and incomplete secondary, complete secondary and tertiary as “Some high school or higher”. The region variable was recoded into five categories—Upper East and West regions as “Upper”, Ashanti and Brong Ahafo regions as “Middle”, Western, Central, Volta and Eastern regions as “South” while Greater Accra and Northern regions remained “Accra” and “Northern” respectively [34]. We used SPSS for windows version 19.0 to perform the data analysis. Using the definitions described above, children were classified as stunted/not stunted, wasted/not wasted and underweight/not underweight. All analyses were stratified by sex. We used the Chi Square test for homogeneity to calculate the confidence intervals for prevalence estimates, which are reported in the Tables. We used logistic regression to test the significant of trends over time. The results of these tests are given in the text only (and not in the Tables). A trend was considered statistically significant if the p-value was less than 0.05. Since the DHS sampling design includes both under- and over-sampling, all analyses were conducted with sample-weighted data. The weights also accounted for non-response. It is possible to use multi-level methods to adjust for cluster-level design effect. This should be done in analyses that are sensitive to within-census district social commonalities. We have not adjusted for the possible design effect of cluster, due to the implausibility that census district is an important source of dependency in the child growth data. This strategy avoids over-adjustment of the analyses. The DHS project sought and obtained the necessary ethical approvals from ethics committees in Ghana before the surveys were carried out. Informed consent was obtained from study participants before they were allowed to participate in the surveys. The survey data sets used in this report were completely anonymous with regard to participant identity.
N/A
