Menu
Menu
Menu
Menu
Background To examine changes in under-5 mortality, coverage of child survival interventions and nutritional status of children in Ethiopia between 2000 and 2011. Using the Lives Saved Tool, the impact of changes in coverage of child survival interventions on under-5 lives saved was estimated. Methods Estimates of child mortality were generated using three Ethiopia Demographic and Health Surveys undertaken between 2000 and 2011. Coverage indicators for high impact child health interventions were calculated and the Lives Saved Tool (LiST) was used to estimate child lives saved in 2011. Results The mortality rate in children younger than 5 years decreased rapidly from 218 child deaths per 1000 live births (95% confidence interval 183 to 252) in the period 1987-1991 to 88 child deaths per 1000 live births in the period 2007-2011 (78 to 98). The prevalence of moderate or severe stunting in children aged 6-35 months also declined significantly. Improvements in the coverage of interventions relevant to child survival in rural areas of Ethiopia between 2000 and 2011 were found for tetanus toxoid, DPT3 and measles vaccination, oral rehydration solution (ORS) and care-seeking for suspected pneumonia. The LiST analysis estimates that there were 60 700 child deaths averted in 2011, primarily attributable to decreases in wasting rates (18%), stunting rates (13%) and water, sanitation and hygiene (WASH) interventions (13%). Conclusions Improvements in the nutritional status of children and increases in coverage of high impact interventions most notably WASH and ORS have contributed to the decline in under-5 mortality in Ethiopia. These proximal determinants however do not fully explain the mortality reduction which is plausibly also due to the synergistic effect of major child health and nutrition policies and delivery strategies.
We used full birth and death history data collected from women aged 15 to 49 years in nationally representative surveys: namely the 2000 Demographic and Health Survey (DHS) the first DHS to be undertaken in Ethiopia, 2005 DHS and the 2011 DHS to calculate under–5 mortality. The surveys covered 14 072, 13 721, and 16 702 households respectively. To assess trends in coverage of child survival interventions and nutritional status we used the same three Ethiopian DHS surveys. The surveys provide detailed information about the health and nutritional status of women and children and coverage of health care services. The analysis included all survey data sets available with full data, including sampling weights, to allow for re–analysis (see Table S1 in the Online Supplementary Document(Online Supplementary Document) for further details on the surveys). To assess coverage of malaria interventions two separate Malaria Indicator Surveys (MIS) were used since these surveys sample specifically from malaria endemic areas. Malaria is seasonal in most parts of Ethiopia, with variable transmission and prevalence patterns affected by the large diversity in altitude, rainfall, and population movement. The MIS from 2007 [6] and 2011 [7] focus on malarious areas defined as <2000m in altitude mapped by global positioning system (GPS); hence these provide a more appropriate estimate of coverage of malaria interventions than the DHS surveys [7]. All of the surveys provided cross–sectional data on intervention coverage in their respective years; however for the MIS, primary data are not available and only point estimates are presented. Definitions and data sources for all indicators can be found in Table S2 in Online Supplementary Document(Online Supplementary Document). We used a direct method for estimating under–5 mortality based on the synthetic cohort approach [8,9]. Under this concept, age–specific mortality probabilities for narrow age ranges and defined periods are calculated using death events and exposures. These probabilities are combined to compute the probability that a child has not died before reaching age 5 years [9]. Under–five mortality rates were computed for successive five year periods preceding the 2011 DHS. For the purposes of this analysis, mortality rates were calculated for 5–year periods starting from 1987–1991 up until 2007–2011 (the 5–year period immediately prior to the 2011 DHS). Survival probabilities were calculated over age ranges; 0, 1–2, 3–5, 6–11, 12–23, 24–35, 36–47, 48–59 months as recommended by DHS (Section B in Online Supplementary Document(Online Supplementary Document)) [9]. The standard errors for the computed mortality estimates were obtained using the Jackknife variance estimation, a repeated sampling method [8]. A series of mortality estimates were obtained by deleting and replacing each primary sampling unit; this produced a sample of under–5 estimates, from which the variance was computed in turn. We also estimated the average annual change (AAC) in mortality using mortality estimates for the periods 1987–1991 and 2007–2011 (Section B in the Online Supplementary Document(Online Supplementary Document)). We analyzed primary data from three Ethiopia DHS surveys to assess coverage trends for 10 indicators which represent high impact maternal and child health interventions; three additional malaria intervention indicators are presented as point estimates. We re–calculated all coverage indicators using standard indicator definitions [10] for tracking progress toward MDG 4. The sampling design of these DHS surveys, such as clustering at enumeration areas and sampling weights (due to non–proportional sampling), were taken into account. Except for the malaria indicators, coverage estimates for rural areas are presented to reflect the focus of the HEP on universal access. We considered malaria indicators for endemic areas only. The 95% confidence intervals were used to assess whether the changes were significantly different across the three time periods. We computed anthropometric indicators for stunting (height–for–age) and underweight (weight–for–age) in children younger than three years of age from information on age, height and weight in the surveys applying the WHO child growth standards [11]. Moderate or severe (below minus two standard deviations (SD) from the median) and severe (below minus three standard deviations (SD) from the median) were calculated for both nutritional measures. Infant feeding indicators such as exclusive breastfeeding and micronutrient intake (vitamin A supplementation) were calculated by age of the child. We used Stata (version 13) (Stata Corporation, College Station, Texas, USA) for all mortality and coverage analyses. We used the Lives Saved Tool (LiST) to estimate the number of deaths averted in 2011 due to changes in coverage since 2000. We compared the changes in mortality produced in LiST with single year estimates from IGME [12] as well as the five–year estimates produced in this analysis using DHS data. LiST uses country–specific or region–specific baseline information on mortality rates and causes of death as well as background variables (fertility, exposure to Plasmodium falciparum, stunting rates) and current coverage of more than 60 interventions and their associated effectiveness values [13–16] relative to specific causes of death and risk factors to estimate the deaths averted, overall and by specific interventions. The modeling methods have been widely published including discussion of the limitations [16–18]. We used 2000 as the baseline year and projected forward to 2011 using all available national data on changes in intervention coverage and nutritional status (Section C and Table S5 in the Online Supplementary Document(Online Supplementary Document)). Specific input values used in this LiST application are available in Table S6 in Online Supplementary Data(Online Supplementary Document). The analysis was done with the program Spectrum/Lives Saved Tool, version 5.04 (Johns Hopkins University, Baltimore Maryland, USA).
