An analysis of socio-demographic patterns in child malnutrition trends using Ghana demographic and health survey data in the period 1993–2008

Child malnutrition is a long-standing and continuing public health problem in many regions of the world, despite global-level progress in the reduction of this problem during the past two decades [1]. Globally, it is estimated that childhood stunting (short stature for age) reduced from 34% to 27% and underweight from 27% to 22%, between 1990 and 2000 [1] and stunting is predicted to reduce to 22% by the next decade [2]. However, evidence shows that global estimates cannot be used to monitor progress at the regional level [3]. In Africa, the prevalence of stunting declined marginally from 40.5% in 1980 to 35.2% in 2000 [3], and between 1990 and 2010 the prevalence of childhood stunting in Africa stagnated at about 40% (2). If this trend continues, we cannot expect significant improvement in the next decade. Global estimates are not good indicators of regional trends, and it can be misleading to use regional-level data to estimate the magnitude and trend in malnutrition at the sub-regional level. During the same period when stunting decreased in North Africa, it actually increased in Eastern Africa, while Western Africa showed very little change [3].

The lesson about the limits to generalizability of large area data to smaller areas is pertinent also to intra-national demographic divisions such as urban/rural, Regions, States and Districts. In 2008 World Health Organization (WHO) estimates of the prevalence of stunting in Ghana, the urban rate was 22% while the rural rate was 33%; the Region with the lowest rate was Greater Accra with 17% while the Eastern region had the highest rate of 39%, more than two-fold difference [4].

Aside from geographic variation, child malnutrition prevalence varies substantially by socio-demographic factors such as the child’s age and sex [5‐8]. Therefore, in order to study trends in malnutrition at a level useful to public health planning, analyses need to be stratified by defined socio-demographic segments; overall trends may mask important departures from the general trend.

It is also vital that malnutrition be defined consistently over the time that a trend analysis is undertaken, and this has been challenging in the area of child malnutrition. The National Centre for Health Statistics (NCHS) international growth reference, that had been in use by WHO since 1978, were changed in 2006. Therefore, the published estimates before and after 2006 are not comparable, except for studies that have used the old standards also in the analysis of data generated since 2006. This strategy results in consistency [9, 10].

Health ministries, as well as governmental and non-governmental health and development agencies, need access to new analyses of child malnutrition and long-term trends, that use the new WHO growth standards exclusively, and that provide disaggregated estimates for defined socio-demographic segments of the under-five population.

Turning to the socio-demographic factors that influence child under-nutrition, there is a vast theoretical and empirical literature advancing hypotheses and data on the link between socio-demographic factors and health generally and child health in particular [11‐13]; a review of that literature is well beyond the scope of this paper. However, below is a detail description of various factors that might influence the nutritional status of children under five years.

Child age and gender are of such fundamental importance to the study of malnutrition time trends that they should not be ignored. Normal child growth and malnutrition are defined by WHO growth standards separately for boys and girls and for age bands separating major phases of development such as pre- and post-weaning. Yet many, if not most published studies, tend to group boys and girls and/or fail to undertake age-specific analyses. When age- and gender-specific analyses are undertaken, they show consistent differences in malnutrition prevalence that would have been masked by less-differentiated analyses [5‐8].

Furthermore, a consistent finding in child malnutrition studies is that children who reside in rural areas have higher rates of stunting and underweight than those in urban areas [6‐8, 14, 15], while the pattern for wasting is somewhat inconsistent [16]. Even when controlling for poverty level, malnutrition prevalence is always lower in urban areas [17] than in rural area. Aside from urban/rural differences, significant intra-country regional variation in child malnutrition has been documented [6, 7]. In Ghana, the regions with the greatest burden of child malnutrition are those in the northern part of the country [5]. De Poel and colleagues [5] point out that the regional patterns observed in Ghana reflect ecological constraints, worse general living conditions, and poorer access to public facilities in the Northern regions. There is also an intergenerational aspect, with Northern women who suffered from malnutrition when they were children being more likely to give birth to children with low birth weight [5, 18], placing such children at risk of being malnourished [19]. As for urban/rural trends, new analyses are needed to track regional time trends in malnutrition.

One of the most robust findings in the public health literature is that health is associated with socio-economic status in a graded way. Increments in household material standards (as measured for example by the Wealth Index) are associated with increments in health [20], including child nutritional status [8, 21, 22]. Children from wealthy households are better off compared to children from poor households. In Ghana for instance, children in the poorest households are at more than twice the risk of being stunted compared to those from the richest households [23]. Hence, prevalence and trend studies of child malnutrition should differentiate among socio-economic status groups.

Additionally, the level of maternal education is consistently, although sometimes only weakly, associated with child malnutrition [24‐26].

As the time when an accounting of progress on development goals for the Global South is coming due – especially the Millennium Development Goals in 2015—planners and policy makers need reliable and valid information on child malnutrition trends that are sufficiently detailed in their socio-demographic breakdown to illuminate where goals are being met, and where challenges remain. The needed data exist, gathered in four Ghana Health and Demographic Surveys spanning 15 years from 1993 to 2008. However, these data require reanalysis, using current WHO Child Growth Standards and providing breakdowns by key socio-demographic segments.

The aims of this paper are: [1] to present findings on national child malnutrition prevalence trends in Ghana in the period 1993 to 2008; [2] to disaggregate the national level data to socio-demographic and geographic sub-groups, and [3] to compare and contrast the national level findings on malnutrition with analysis based on the sub-group.

This study examined long-term trends in child malnutrition among children less than three years of age in Ghana, by region, by rural–urban setting and by socio-economic and demographic characteristics. The findings at the national level show that overall child malnutrition is significantly decreasing. Stunting, the effect of chronic under-nutrition, declined from 33.5% in 1993 to 25.7% (23% reduction) by 2008. This is in sharp contrast with the complete stagnation of stunting trends in the West African Sub-region as a whole [2]. Thus, Sub-regional trends cannot be used to estimate trends in countries within a region, at least not in the case of Ghana. Yet, given the current trend, the rate of decline in stunting is not likely to move Ghana to the level where it will meet the WHO target of a 40% reduction by 2025 [35].

Child underweight has also declined significantly in Ghana, from 25.1% in 1993 to 15.1% in 2008, a 60% reduction. Thus, by 2008 Ghana had already achieved the Millennium Development Goal (MDGs) target of halving 1990 levels of underweight by 2015. This achievement could be attributed to significant reduction in extreme poverty in Ghana over the past two decades. For instance, overall poverty declined from 51.7% in 1991 to 28.5% in 2006 [36, 37]. Similarly, the number of people living below the extreme poverty line in the country decreased by more than half in the last two decades. Total food production also increased significantly in Ghana during this period [36], essential for an increase in the availability of food to the population. Another possible explanation of the achievement made by Ghana with regard to underweight reduction is the introduction of the National Plan of Action on Nutrition (NPAN) by the Government of Ghana in 1995 [38], which aims to combat malnutrition in the country. The NPAN is multimodal, including the Baby–Friendly Health Facility Initiative, the Community Based Nutrition Behavior Change Communication (BCC) strategy, mother-to-mother support groups for promoting optimal breastfeeding and complementary feeding practices, and community-based growth promotion projects, are a few of the NPAN’s key elements.

The situation for wasting is not as positive as it is for stunting and underweight. Even though Ghana has made some progress in the reduction of wasting between 1993 and 2008, the national prevalence remains high at 12.4% for children less than 36 months, a level the WHO classifies as requiring urgent response [39]. Wasting, which is usually the result of acute significant food shortage and/or disease, is a strong predictor of mortality among children under-five years [18]. Hence, for Ghana to be on track to achieve MDG 4, there is a need for re-doubled effort to address childhood wasting.

A number of important departures from the national trends raise cause for concern. The situation at the regional level is a complete departure from the national level. Our results show that the Middle region is the only region that exhibited significantly declining trends in stunting and underweight both for both males and females, revealing that not all regions enjoyed the declining trends seen in the national-level analyses. In terms of prevalence, the Northern and Upper regions have the highest levels of malnutrition, and this has also been documented in previous studies [5, 14]. The disproportionate burden of malnutrition on children in the Northern part of Ghana may be attributable the North’s high rate of poverty [36, 37]. While the reduction in overall poverty at the national level between 1992 and 2006 was remarkable, the three Northern regions did not record significant poverty reduction [36, 37]. Additionally, over 70% of people whose incomes are below the poverty line can be found in the Savannah areas [36, 37]. The observed improvement in poverty statistics at the national level has not been equitably distributed. As poverty is a strong driver of malnutrition [40], sustained high poverty rates in the North translate into sustained malnutrition in the children of the North. Poverty and malnutrition are mediated by inadequate food consumption and infectious disease. Insufficient availability of food at household level due to poverty and/or improper feeding practices places children at an elevated risk of malnutrition, both directly and by malnutrition’s influence on a child’s susceptibility to disease and infection, and immune system dysfunction [41].

It is clear from the above discussion that the overall positive national trends mask important variation at the sub-national level. This demonstrates that sub-national level analyses are important for identifying regions and social groups that need better support and interventions to improve the malnutrition situation. Further, there is evidence that even the regional level prevalence data may mask differences in other important segments of the country [42].

Our data suggest that differentials in child malnutrition by place of residence have substantially narrowed over time in Ghana, consistent with other studies [15]. This trend in urban–rural differentials is primarily due to static trends in urban malnutrition coupled with rural improvement. A number of factors might have contributed to the improving trends in rural areas. One such factor is the significant increase in food production in Ghana between 1992 and 2008 [36, 37]. Since much of the food consumed in Ghana is produced in the rural areas, it is plausible that food supplies became increasingly available to rural households during this period, and when food is available in the household, it is assumed that children will be well fed. Another possible explanation for the urban–rural differentials is that poverty is viewed in Ghana as an overwhelming rural phenomenon [43]. Therefore, most antipoverty initiatives are directed at the rural population. However, the emphasis on the alleviation of rural poverty has led to a degree of neglect of the problems of urban poverty and urban food insecurity. In fact, urban poverty and associated health problems are growing in Ghana [44, 45]. The introduction of nutrition rehabilitation services and the Supplementary Feeding and Health and Nutrition Education Programmes [38], could also explain the narrowing gap between rural and urban settings. These programmes are targeted at deprived communities, mostly rural communities. These explanations are sensible, but remain merely speculative, because the possible contributory factors identified above were not measured in the surveys that provided the data for this study.

The gap in malnutrition by maternal education level has significantly narrowed in Ghana, to the extent that malnutrition rates for children of mothers who have no education are indistinguishable from children of educated ones in 2008. These differences in trends are partly due to increasing malnutrition in children of mothers with more than primary education. Further analysis stratified by place of residence was undertaken as part of this study, to examine the possibility that the effect of education is confounded by place of residence. This analysis did not provide evidence of such confounding.

The findings above are unexpected because it is assumed that mothers who have high education will be more empowered to be able to take decisions on the type of nutrition and care the child should receive. Educated mothers could also be in a better position to make informed nutritional decisions that buck unhealthy cultural norms about child feeding [46]. The puzzling failure of maternal education to have the expected effect on child nutritional status in Ghana could be due to structural factors. One such factor is intervention targeting. Most nutrition and antipoverty interventions in Ghana are targeted at the population considered disadvantaged, in this case, the non-educated and rural population [38]. This places the so-called advantaged groups, who also have pressing needs, at a disadvantage. The re-coding of the education variable could also be a plausible explanation. Our data have few people who have higher education and as a result, secondary education was collapsed with higher education. This could mask the importance of higher maternal education as a protective factor. Additionally, it is important to note that Ghana has very high unemployment rates [36, 37], and as a result most mothers with secondary education are unlikely to find a decent job to earn a living, and consequently may find it difficult to provide for the basic needs of their children.

Our data show that child malnutrition trends in the poorest wealth quintiles are decreasing while trends in the richest quintiles remain static. Even though this suggests that children in the poorest quintiles are getting better over time, malnutrition remains the bane of the poor in Ghana. The results show that malnutrition prevalence among children in the poorest wealth quintile is as much as twice that of the richest quintile. These findings corroborate earlier studies in Ghana using DHS data, which found that children in the poorest households are more than twice at risk of being malnourished compared to their counterparts in the richest households [5, 23]. However, the declining trends among the poorest quintiles implies that the halving of people living below the extreme poverty line, and the significant increase in food production in Ghana between 1991 and 2008 [36, 37] have helped to close the malnutrition equity gap.

The child’s age is an important factor in level of risk for malnutrition. Malnutrition is more prevalent in older than in younger children [5‐7]. In our data, the oldest children (24–35 months) have the highest levels of malnutrition, and the youngest age group (0–5 months) has the lowest levels. With regard to time trends, the older children exhibited significant improvement compared to the younger ones in the case of stunting. However, both younger and older age groups exhibited significant declining trends in the case of underweight. The higher proportion of malnutrition among older children could be due to inappropriate child feeding practices and/or increased morbidity, while the declining malnutrition trends may be explained by systematic improvement in the availability and quality of food [36, 37] and other care practices during this period. This is an empirical issue that could be addressed using the GDHS data on child care and feeding; however such analyses are beyond the scope of this paper.

There are limitations in the analysis reported in this paper. The core data of this study come from the anthropometry measurements and the birth date data used to calculate the growth variables. As reported in our results, data of sufficient quality was obtained from between 87 and 90 percent of eligible children depending on survey year. There are a myriad reasons why useable anthropometry data might be missing, discussed in detail by Pullum [47]. These include poor technical work by data collectors, faulty equipment, sick or uncooperative children, refusal by the mother or another household member, the child being away from the household during the data collection window, and data entry errors at the time of data collection and/or in the transfer of data to analyzable data files, among other reasons. While missing data is of inevitable concern in survey research, what would be of greater concern would be a systematic pattern over the survey years wherein various reasons for missing data increased or decreased in prevalence from survey to survey. We do not have detailed enough missing data analyses from the four surveys to evaluate the seriousness of this potential source of bias. There is evidence that poor birth data was a minor cause of missing growth data in all the survey.

Besides missing data, the validity of the malnutrition trends reported in this paper may be compromised due to method variation in determining which children were eligible for measurement. In 1993 and 1998 surveys, anthropometric measurements were restricted to children born to the women who were interviewed. Children were excluded if their mothers were not in the household, if their mothers were not eligible for the individual interview, or the mother did not complete an interview. The methodology changed in the 2003 and 2008 surveys, and children who slept in the household the night before data collection were eligible regardless of the interview status of their mother. As a result, orphans and children whose mothers were away were excluded in 1993 and 1998 and included in 2003 and 2008. This would pose a validity issue if orphans, for example, are more likely to suffer malnourishment than non-orphans in the same household. At least one study has examined this issue, comparing South African orphans and non-orphans in the same households [48]. There were no significant differences between the two groups on health outcomes. Nevertheless, the change in the sampling protocol in 2003 is a source of concern for trend analyses such as this paper reports.

The analysis at the national level shows that child malnutrition is significantly declining in Ghana. However, the aggregate national trends mask the fact that not all segments of the population benefit from improvement of the same magnitude, as our findings by various geographic and demographic characteristics reveal. There is a need for policies that address the specific constraints of households left out of progress so that children from all segments of the country benefit. Additionally, the study also found that the widened gap between rural and urban settings in Ghana is closing. While child malnutrition is progressively decreasing in the rural areas, it remains static in the urban settings. This contributed greatly to narrowing the rural/urban child malnutrition gap. In addition, the malnutrition trends by maternal education in Ghana have narrowed to the extent that the differences between the educated and non-educated are not easily distinguishable. A perplexing finding of this study is the increasing trends of childhood stunting among children of mothers who have higher than primary education, in sharp contrast with existing literature. This anomaly needs further investigation.