Comments and additional information
Inter-Agency Group (MMEIG) has been updated, revised and improved the previous methods for estimating maternal mortality, with special emphasis on the development of methods for estimating maternal mortality trends from 1990 to 2010. Estimates of MMR were launched to bring to the United Nations General Assembly in monitoring progress towards achieving the MDGs.
This report presents estimates of maternal mortality by 2010 at global, regional, and country, and the conclusions of the assessment of trends in maternal mortality levels since 1990.
Following the development of estimates of maternal mortality, including consultations with countries which were conducted during July and August 2010: In order to give countries the opportunity to review the estimates of country, the sources data and methods, to obtain primary data sources that might not have been previously reported or used, and to promote mutual understanding of the strengths and weaknesses of available data and to ensure wide performance monitoring. Following consultations, the statistical model was run again to incorporate the new data provided by countries.
We found substantial differences in the methods used to estimate maternal mortality in 2008 compared with the 2005
• In the current round of estimation, we developed a multilevel regression model to obtain estimates of maternal deaths in multiple time points from 1990 to 2010. The multilevel approach (with random intersections on the axis "Y" for countries and regions) provides a means of statistically sound data representing the country on levels and trends of maternal mortality in a global model can also be used for predictions outside the sample.
• Unlike the 2005 estimate, this round of estimates took into account the issue of incidental and indirect maternal deaths among women with HIV-positive in the estimation of maternal mortality in countries with high prevalence of HIV / AIDS ( details described in Appendix 5).
• In the analysis of the estimates for 2010, an effort was made to exclude accidental deaths during pregnancy in order to better capture the true maternal deaths.
• For estimates of 2010, countries were classified into three groups based on available data, instead of the eight groups used for the estimates of 2005 (the latter was based on the data source used to derive an estimate for the more recent time period). One reason for this change is that for the estimation of 2010 some countries had multiplied the sources of data over time so could not be assigned to a single data source. For example, El Salvador had vital registration data (which was not complete), reproductive health surveys and RAMOS.
Therefore, the results of this exercise should not be compared with estimates previously released by the group. This exercise generated MMR updated figures for 1990, 1995, 2000, 2005 and 2010 with new and improved methodology. The findings presented in Annex 3, allow analysis of trends in maternal mortality between 1990 and 2010.
First, it should be noted that the data and methods have improved over time. Estimates of 2010 should not be compared with those of previous years to assess changes over time. Trends in maternal mortality calculated using the same methodology improved and presented for the years 1990, 1995, 2000, 2005 and 2010 in this report show changes over time.
Second, depending on the type of data source used, raw data from different countries had to adjust to the specifics. These features included the degree of potential underreporting of maternal deaths, which is even a problem in countries with highly developed civil registration, to obtain estimates of maternal mortality that are comparable across study designs. This setting allows regional and global aggregation. For this reason, the point estimates presented often differ from the figures reported by countries.
Third, despite the improvement of the methodology, the global database on maternal mortality remains weak. Only just over a third of countries or territories have complete civil registration systems and good definition of the cause of death (Group A). The ability to generate estimates at national, regional and global with greater precision and accuracy would be greatly enhanced if the national civil registration systems will improve even more. This improvement would avoid the need for special studies of maternal mortality - time consuming, costly and of limited utility in monitoring trends - or by using statistical models that have their own weaknesses.
A technical report of the methodology, the data set and statistical analysis code used to prepare these estimates is available in a technical report available in Trends in Maternal Mortality: 1990 to 2010 (only English version).