Maternal mortality estimates are useful
Editor In the March 2001 issue of the Bulletin, Pierre Buekens asks "Is estimating maternal mortality useful?" (1). "Maternal mortality" can have at least three meanings: (i) total number of deaths of women from pregnancy- related causes in a given period; (ii) maternal mortality ratio: total number of deaths of women from pregnancy- related causes in a given period per 100000 live births; the ratio measures the risk of death a woman faces each time she becomes pregnant; and (iii) maternal mortality rate: total number of deaths of women from pregnancy- related causes in a given period per 100000 women of reproductive age. This measures both the obstetric risk and the frequency with which women are exposed to this risk.
In public health practice, how questions are posed is directly relevant to the definition of problems and why particular measures are selected. While the editorial by Buekens was written in relation to an article by Hill, AbouZahr & Wardlaw dealing with maternal mortality ratios (2), the question as actually posed has two broad interpretations. If concerned only with the ratio, it presupposes that we are dealing with women who are already pregnant. The focus then moves to major obstetric risk factors, namely: haemorrhage, sepsis, hypertensive disease of pregnancy or pre-eclampsia, prolonged or obstructed labour, and complications of unsafe abortion. These causes together are commonly said to account for up to 80% of all maternal deaths globally. However, if one is concerned with the rate, or simply the numerator alone, the frame of reference incorporates a more basic question: "Why do women get pregnant at the frequency they do?". If viewed as a function of fertility, in a country with a total fertility rate (TFR) in the vicinity of 7 (e.g. Uganda, Yemen), excess fertility accounts for >70% of all maternal deaths, taking replacement level as the criterion. (For illustrative purposes we are assuming 2.1 as the replacement level though a somewhat higher level could be justified, taking into account infant and child mortality rates depending on the setting.) With a TFR of 6, the attributable burden is approximately 65% (e.g. Oman, Rwanda), while at a TFR of 5 (e.g. Pakistan, Zambia), it is about 60%. The TFR is the number of children that would be born per woman if she were to live to the end of her childbearing years and bear children at each age in accordance with prevailing age-specific fertility rates (3). Clearly, efforts are required to deal with both excess fertility and pregnancy safety, and programming in both areas is actually taking place in most developing countries. Neither of these issues presents an easy challenge, and both have enormous sociocultural and political complexities that are beyond the present brief discussion.
The answer to Buekens's question must be "yes" at least at the level of policy, priority setting and resource allocation. Only by assessing maternal mortality (MM) can one place this alongside other causes of death and determine its relative magnitude and public health importance. While Hill et al. emphasize that no valid conclusions can be drawn from MM trend analyses because of major imprecisions in the data (2), even imprecise data give useful orders of magnitude supporting both lines of intervention mentioned above. Estimates of MM (at least both the numerator and the ratio) comprise important components, therefore, of a health situation analysis for any country. While one can also agree that MM is too difficult to measure to be programmatically useful and that process indicators are more applicable at this level, this observation is not unique to reproductive health: mortality is a useful, even if still imprecise, measure of disease burden in many other areas of public health where process indicators are also critical for programmatic purposes (e.g. HIV, malaria, hypertension, diabetes). Across the spectrum in public health there is a great need to improve existing data and the measures of disease burden derived from them (including mortality) and also to develop process indicators for the planning and evaluation of intervention programmes.n
Franklin White, Professor and Chair
Sarah Saleem, Senior Instructor
Department of Community Health Sciences
The Aga Khan University
Conflicts of interest: none declared.
1. Buekens P. Is estimating maternal mortality useful? Bulletin of the World Health Organization, 2001, 79: 179.
2. Hill K, AbouZahr C, Wardlaw T. Estimates of maternal mortality for 1995. Bulletin of the World Health Organization, 2001, 79: 182198.
3. The state of the world's children 2001. New York, UNICEF, 2001.