Newborn survival: A multi-country analysis of a decade of change

Neonatal deaths account for 40 of global under-five mortality and are ever more important if we are to achieve the Millennium Development Goal 4 (MDG 4) on child survival. We applied a results framework to evaluate global and national changes for neonatal mortality rates (NMR), healthy behaviours, intervention coverage, health system change, and inputs including funding, while considering contextual changes. The average annual rate of reduction of NMR globally accelerated between 2000 and 2010 (2.1 per year) compared with the 1990s, but was slower than the reduction in mortality of children aged 1-59 months (2.9 per year) and maternal mortality (4.2 per year). Regional variation of NMR change ranged from 3.0 per year in developed countries to 1.5 per year in sub-Saharan Africa. Some countries have made remarkable progress despite major challenges. Our statistical analysis identifies inter-country predictors of NMR reduction including high baseline NMR, and changes in income or fertility. Changes in intervention or package coverage did not appear to be important predictors in any region, but coverage data are lacking for several neonatal-specific interventions. Mortality due to neonatal infection deaths, notably tetanus, decreased, and deaths from complications of preterm birth are increasingly important. Official development assistance for maternal, newborn and child health doubled from 2003 to 2008, yet by 2008 only 6 of this aid mentioned newborns, and a mere 0.1 (US$4.56m) exclusively targeted newborn care. The amount of newborn survival data and the evidence based increased, as did recognition in donor funding. Over this decade, NMR reduction seems more related to change in context, such as socio-economic factors, than to increasing intervention coverage. High impact cost-effective interventions hold great potential to save newborn lives especially in the highest burden countries. Accelerating progress requires data-driven investments and addressing context-specific implementation realities.