BioMed Central Page 1 of 12 (page number not for citation purposes) BMC Public Health Open Access Research article National estimates for maternal mortality: an analysis based on the WHO systematic review of maternal mortality and morbidity Ana P Betr��n*1, Daniel Wojdyla2, Samuel F Posner3 and A Metin G��lmezoglu1 Address: 1UNDP/UNFPA/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction, Department of Reproductive Health and Research, World Health Organization, Geneva, Switzerland, 2Centro Rosarino de Estudios Perinatales, and Escuela de Estad��stica, Universidad Nacional de Rosario, Rosario, Argentina and 3Division of Reproductive Health, Centers for Disease Control and Prevention, Atlanta, GA, USA Email: Ana P Betr��n* - betrana@who.int Daniel Wojdyla - dwojdyla@arnet.com.ar Samuel F Posner - shp5@cdc.gov A Metin G��lmezoglu - gulmezoglum@who.int * Corresponding author Abstract Background: Despite the worldwide commitment to improving maternal health, measuring, monitoring and comparing maternal mortality estimates remain a challenge. Due to lack of data, international agencies have to rely on mathematical models to assess its global burden. In order to assist in mapping the burden of reproductive ill-health, we conducted a systematic review of incidence/prevalence of maternal mortality and morbidity. Methods: We followed the standard methodology for systematic reviews. This manuscript presents nationally representative estimates of maternal mortality derived from the systematic review. Using regression models, relationships between study-specific and country-specific variables with the maternal mortality estimates are explored in order to assist further modelling to predict maternal mortality. Results: Maternal mortality estimates included 141 countries and represent 78.1% of the live births worldwide. As expected, large variability between countries, and within regions and subregions, is identified. Analysis of variability according to study characteristics did not yield useful results given the high correlation with each other, with development status and region. A regression model including selected country-specific variables was able to explain 90% of the variability of the maternal mortality estimates. Among all country-specific variables selected for the analysis, three had the strongest relationships with maternal mortality: proportion of deliveries assisted by a skilled birth attendant, infant mortality rate and health expenditure per capita. Conclusion: With the exception of developed countries, variability of national maternal mortality estimates is large even within subregions. It seems more appropriate to study such variation through differentials in other national and subnational characteristics. Other than region, study of country-specific variables suggests infant mortality rate, skilled birth attendant at delivery and health expenditure per capita are key variables to predict maternal mortality at national level. Published: 12 December 2005 BMC Public Health 2005, 5:131 doi:10.1186/1471-2458-5-131 Received: 11 May 2005 Accepted: 12 December 2005 This article is available from: http://www.biomedcentral.com/1471-2458/5/131 �� 2005 Betr��n et al licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

BMC Public Health 2005, 5:131 http://www.biomedcentral.com/1471-2458/5/131 Page 2 of 12 (page number not for citation purposes) Background Since the launching of the Safe Motherhood Initiative in 1987 [1], there has been a worldwide effort to reduce maternal mortality and to identify its determinants. These efforts have been directed by the outputs of a number of international conferences over the past decade such as the International Conference on Population and Develop- ment in 1994, and the Fourth World Conference on Women in 1995 reinforced this commitment. The decla- ration of the Millennium Development Goals (MDGs) aiming at reducing by three-quarters the maternal mortal- ity ratio between 1990 and 2015 has also increased the demand for measuring maternal mortality at national and subnational levels [2]. Despite worldwide concern, an outstanding problem is how to monitor maternal mortality and to obtain reliable and comparable data. Measuring maternal mortality accu- rately is notoriously difficult except where there is com- prehensive registration of deaths and causes of death. Unfortunately, there are only a few countries where such registration could be characterized as complete [3] and even in these countries, poor attribution of cause of death results in significant underreporting of maternal deaths [4,5]. In addition, countries with complete death registra- tion are countries with low maternal mortality, and, con- sequently, countries where it is not a public health priority. It is in countries where a reliable vital registration system is not in place where maternal mortality represents a public health problem that cannot be accurately meas- ured. Several alternative techniques have been developed to fill the gap caused by poorly functioning vital registration sys- tems. Of these, the Reproductive Age Mortality Studies (RAMOS) are considered the gold standard for measuring maternal mortality because it involves identifying and investigating the causes of all deaths of women in repro- ductive age [6]. Another approach currently used in most developing countries derives estimates of maternal mor- tality from household surveys or surveys using the sister- hood method [7]. The sisterhood method is an indirect measurement technique that reduces sample size of the surveys by interviewing respondents about the survival of all their sisters [7]. Data on maternal deaths obtained through census has also been proposed as a means of esti- mating levels of maternal mortality [8]. Drawbacks include high costs in the case of RAMOS, large sample sizes required for household surveys and the use of esti- mates intrinsically referring to the past instead of the cur- rent situation in the case of sisterhood methods. Differentials in the definition of maternal death, varying efforts carried to capture maternal deaths, and the meth- ods used to confirm the deaths as 'maternal' are some of the inherent discrepancies in these methods that may affect estimates and impede comparisons. Unfortunately, a measure allowing for comparisons between these meth- ods is lacking. WHO, jointly with UNICEF and UNFPA, has made efforts to monitor maternal mortality by producing global, regional and national estimates for 1990, 1995 and 2000 [3,9,10]. Different methodologies used to calculate mater- nal mortality ratios as well as the lack of national data for many of the countries have been identified as major prob- lems in assessing the global situation as well as for moni- toring trends. Estimates for 2000 suggested 529,000 maternal deaths worldwide with an average maternal mortality ratio of 400 per 100,000 live births, and accounted for 173 countries with 99% of global births. However, 62 countries (27% of global live births) had no national data available, and maternal mortality estimates for those countries were developed using a regression model based on a set of explanatory country-specific vari- ables that are available for nearly all countries in the world [3]. An alternative model based also on country- specific variables was also proposed using the same data set [11]. The UNDP/UNFPA/WHO/World Bank Special Pro- gramme of Research, Development and Research Training in Human Reproduction (HRP), Department of Repro- ductive Health and Research at the World Health Organi- zation has conducted a systematic review of the prevalence/incidence of maternal mortality and morbidi- ties from 1997 to 2002 worldwide. The primary objective of this review is to assist in mapping the burden of repro- ductive ill-health by providing a comprehensive, stand- ardized and reliable tabulation of data on the incidence/ prevalence of maternal morbidity and mortality [12]. This article presents an analysis of the nationally representative maternal mortality data included in the review exploring the correlation between the maternal mortality estimates with study-specific and country-specific variables. Methods The methodology of the systematic review has been described elsewhere [13]. In brief, we searched for pub- lished and unpublished data on maternal mortality and morbidity reported between 1997 and 2002. The search strategy included 10 relevant electronic databases, hand searching, screening of reference lists of retrieved articles, congress abstract books, and contacting experts active in the field. Furthermore, we searched databases in develop- ing countries such as Index Medicus of the Eastern Medi- terranean Region (IMEMR) [14], African Index Medicus (AIM) [15] IndMED [16], a bibliographic database of Indian biomedical journals and HELLIS.ORG [17], a net- work of health science libraries across Asia. Results on the