Conceived and designed the experiments: HN NI KS. Performed the experiments: HN NI. Analyzed the data: HN NI. Contributed reagents/materials/analysis tools: HN NI KS. Wrote the paper: HN. Interpreted data: NH NI AS RM KS. Revised the manuscript for intellectual content: NI AS RM KS. Revised the manuscript for language content: AS. Supervised the study: KS.
The authors have declared that no competing interests exist.
Recent analyses have suggested an accelerated decline in child mortality in Ghana since 2000. This study examines the long-term child mortality trends in the country, relates them to changes in the key drivers of mortality decline, and assesses the feasibility of the country's MDG 4 attainment.
Data from five Demographic and Health Surveys (DHS) between 1988 and 2008 and the Maternal Health Survey 2007 were used to generate two-year estimates of under-five mortality rates back to 1967. Lowess regression fitted past and future trends towards 2015. A modified Poisson approach was applied on the person-period data created from the DHS 2003 and 2008 to examine determinants of under-five mortality and their contributions to the change in mortality. A policy-modelling system assessed the feasibility of the country's MDG 4 attainment.
The under-five mortality rate has steadily declined over the past 40 years with acceleration since 2000, and is projected to reach between 45 and 69 per 1000 live births in 2015. Preceding birth interval (reference: 36+ months, relative risk [RR] increased as the interval shortened), bed net use (RR 0.71, 95% confidence interval [CI]: 0.52–0.95), maternal education (reference: secondary/higher, RR 1.71, 95% CI: 1.18–2.47 for primary), and maternal age at birth (reference: 17+ years, RR 2.13, 95% CI: 1.12–4.05) were primarily associated with under-five mortality. Increased bed-net use made a substantial contribution to the mortality decline. The scale-up of key interventions will allow the possibility of Ghana's MDG 4 attainment.
National and global efforts for scaling up key child survival interventions in Ghana are paying off ― these concerted efforts need to be sustained in order to achieve MDG 4.
Millennium Development Goal 4 (MDG 4) is targeted at reducing global under-five
mortality rates by two-thirds between 1990 and 2015.
In Ghana, the MDG 4 target of a two-thirds reduction in under-five mortality equates
to 40 deaths per 1,000 live births, relative to the figure in 1990 which was 120 per
1,000 live births.
In contrast, two recent surveys―the Maternal Health Survey (MHS) 2007 and the
Demographic and Health Survey (DHS) 2008―have both suggested a high rate of
decline since 2003 (26% and 28%, respectively).
This study, therefore, attempts to estimate the most up-to-date under-five mortality trends in Ghana, relates them to changes in the key drivers of child mortality decline, and assesses the feasibility of the country's MDG 4 attainment given the current child mortality reduction efforts in place.
We used data from the five DHS conducted in Ghana between 1988 and 2008, and the
MHS 2007. These were nationally representative cluster sample surveys that
covered 4,406, 5,822, 6,003, 6,251, 10,858, and 11,778 households in 1988, 1993,
1998, 2003, 2007, and 2008, respectively.
Mortality among children aged younger than five years was estimated for every
two-year period before each respective survey going back to 1967, using the
direct method based on the complete birth histories of women aged between 15 and
49 years. Periods of exposure and deaths were grouped into two calendar years,
and a separate life table was constructed for each period in the birth histories
to show the probability that children would die before their next birthday. We
then calculated the individual death probability by counting the number of
children born in a certain period of time and the number of children dying in
the same period. Then we used the synthetic cohort life table approach, in which
the probabilities of death for small age segments based on real cohort mortality
experience were combined into the more common age segments.
Mortality trends from 1967 to 2008 were estimated by fitting Lowess regression of
the natural log of mortality in children younger than five years to time with
bandwidths ranging from 0.3 (representing high sensitivity to recent data) to
2.0 (lower sensitivity), while the trend towards 2015 was forecast with the same
bandwidths. Mean annual reduction rates were calculated by using the results
from the Lowess regression analysis with the 0.3 bandwidth, which can precisely
detect changes that occur over time.
In order to examine potential determinants of under-five mortality in Ghana, we pooled data of children's records from DHS 2003 and 2008. We used only these two surveys as they allowed us to maximize the number of intervention variables that could be examined. The pooled data had 6,609 under-five children born to 4,259 ever-married women aged 15–49 years at the time of the survey. We then transformed individual records of children to survival-time data, in order to consider the time to event and censoring and reflect the changing rates of mortality during childhood. In this risk dataset, children first became at risk and came under observation at date of birth. A row represented each child for each age band in which he/she survived and was therefore at risk. The analysis-time variable was time since birth measured in months, and we split childhood into five age bands: the first month of life (age 0 months), months 1–5, months 6–11, months 12–23 and months 24–59. We obtained 27,011 episodes for 6,609 children. The total exposure time was 184,307 child-months and 499 children died. The median time of exiting the risk set due to experiencing the event or censoring was 26 months.
We employed Poisson regression with robust error variance to model a binary
outcome of death.
When selecting covariates for the model, we considered their relevance to child
mortality in Ghana and potential confounding factors. Covariates consequently
covered socioeconomic and demographic characteristics (urban/rural residence,
region, religion, economic status, maternal age at birth, maternal highest
education, maternal marital status, child's sex and multiplicity in birth)
and maternal and child health practices (birth order, preceding birth interval,
breastfeeding duration, use of a bed net, use of oral rehydration salt
[ORS] and the number of antenatal care [ANC] visits). As a
measure of economic status, we constructed wealth quintiles to make them
comparable between the two surveys. We appended household survey datasets to
estimate wealth index scores through a principal component analysis on the
ownership of household assets that were available from both surveys.
For the relative contribution of each covariate to the changes in child mortality from 2003 to 2008, first we used the estimated beta coefficients and means of the explanatory variables in 2003 to calculate the risk of mortality in that year. Then, for each explanatory variable, we changed its values to its mean in 2008 to compute the change in the risk of mortality, setting other variables to their means in 2003. We divided the difference by the risk of mortality in 2003 to obtain the relative contribution of each explanatory variable.
To estimate reductions in under-five mortality rates as a result of scaling up
child-survival interventions in Ghana, we used the Lives Saved Tool (LiST) in
the Spectrum Policy Modelling System.
LiST requires background information for projections of the future based on UN
population division estimates and under-five mortality trends from the UN child
mortality coordination group, child morbidity and nutritional status, deaths by
causes, and coverage of child health interventions as well as assumptions
concerning the efficacy of those interventions. The programme provides default
values for the required information based on a review of scientific studies
The tool projected the under-five mortality rates in 2009 and 2015 based on the
background information explained above, and made an alternative child survival
projection for 2015, in which the selected eight interventions were further
up-scaled in pursuit of the MDG 4 target. The interventions were selected based
on the causes of death among children aged under-five in Ghana (malaria,
32%; neonatal death, 29%; pneumonia, 15%; diarrhoea,
12%; HIV/AIDS, 6%; measles, 3%; and injuries,
3%)
The trend and determinant analyses were conducted using Stata/SE version 10.0 (StataCorp LP, College Station, TX, United States of America). LiST was used in the Spectrum version 3.45, downloaded from the website (Futures Institute, Glastonbury, CT, United States of America). The complex nature of the survey design, including such elements as stratification and clustering, was considered in the trend analysis and descriptive statistics of the study population.
Data are from Ghana Demographic and Health Surveys 1988–2008 and Ghana Maternal Health Survey 2007. Note: The MDG 4 target is 40 per 1000 live births, indicated as the red round-shaped point at the end of the dotted MDG track. MDG, Millennium Development Goal; DHS, Demographic and Health Survey; MHS, Maternal and Health Survey.
Data are from Ghana Demographic and Health Surveys 2003 and 2008. The change in mean predicted probabilities of death under five years of age between 2003 and 2008 was decomposed into contributory factors.
We suggested an alternative scenario with the further up-scaling of eight
interventions by 2015, which is indicated as “2015a” in the fourth
column of
We confirm a steady decline in under-five mortality between 1967 and 2008 with an
accelerated decline since 2000. The mean annual rate of reduction: 4.6% since
2000 exceeded the rate required for achieving MDG4 (i.e. 4. 4%).
This study analysed the trends and determininants of under-five mortality and the relative contribution of different factors to the decline in child mortality. This series of population-level analyses, together with the practical estimates of child survival intervention coverage as a supplementary tool, has suggested a way of directing policy-initiatives towards MDG 4 attainment, especially for other priority countries.
Specifically, we have identified bed net use for children aged under five and preceding birth interval as being important determinants of child mortality as well as major contributors to the decline in mortality. We also found that mothers' educational attainment, mother's age at birth and breatfeeding duration were important determinants of child mortality. The results of this study suggest that under-five mortality was successfully reduced by a combination of the recent scale up of child survival interventions and an overall improvement in the socioeconomic conditions in Ghana.
In addition to the factors identified in this study, there are several other possible
reasons for the decline in child mortality in Ghana. First, Ghana's economic
improvement must have been behind the success. In fact, Ghana has enjoyed constant
increases in gross domestic product (GDP) over the past 15 years
Ghana has also achieved equity in intervention coverage. A recent analysis revealed
that Ghana had the highest coverage in the richest quintile and the second poorest
quintile among the 16 Sub-Saharan African Countries with 2009 estimated GDP per
capita of less than International $1,000.
Several limitations of this study should be mentioned. First, the limited number of
variables in the DHS datasets did not allow us to explore a wider range of
intervention variables, particularly the direct impact of child survival
interventions on under-five mortality, including immunizations and the case
management of malaria, pneumonia and diarrhoea, as well as neonatal mortality, which
accounts for 29% of under-five mortality in Ghana.
In conclusion, Ghana has experienced a slow, but nevertheless, steady decline in its under-five mortality rate over the past forty years with an accelerated reduction since 2000. National and global efforts for scaling up key child survival interventions in Ghana are paying off and need to be sustained. Strategic scale-up of key interventions will lead the country even further towards the attainment of its MDG 4 target. Identification and enhancement of key child-survival interventions, following a thorough examination of each country's specific context, are recommended for other low-income, “No progress” countries, particularly in sub-Saharan Africa.
(DOCX)
(DOCX)
(DOCX)
We would like to express our sincere gratitude to Dr. Julie Rajaratnam for her technical support, and Dr. Ai Koyanagi and Dr. Moazzam Ali for their helpful comments.