Making Sense of New Maternal Mortality Numbers: Four Take-Aways for Policy and Research Action

This is a joint post with Katherine Douglas.A new Lancet paper by a team of researchers at the University of Washington’s Institute for Health Metrics and Evaluation has caused quite a stir about the progress we are making towards Millennium Development Goal (MDG) 5: To reduce maternal mortality ratios by three quarters from 1990 to 2015! I have long wondered why no one was making an effort to question the ostensibly stagnant “500,000 maternal deaths per year” estimate, so this team’s effort to provide us with the new numbers (and reinvigorated focus!) is very welcome.  With a few exceptions (Karen Grepin’s excellent blog being one), I have yet to see much of a response from the global health community, although there has been quite a lively discussion in the development community at large (see Bill Easterly’s blog for example) about differences in modeling  outcomes, whether these new maternal mortality numbers are better than the older ones and whether these new figures indicate the efficacy of safe motherhood programs, among other topics.   However, two important issues are missing from these conversations: The extreme limitations of existing maternal mortality data, and what we can take away from these new estimates.Maternal Mortality Data and Measures 101We only have ESTIMATES—yes, that’s what they are—because we don’t have the actual  maternal death data from vital registration systems in most countries. Because many developing countries lack the capacity to accurately gather, analyze, disseminate and report data on a regular basis, we can’t express maternal mortality accurately either as absolute numbers or as rates (i.e. the number of maternal deaths in a specific time interval /total number of women of reproductive age in a specific time interval).  So, the next best option is estimating the number of maternal deaths from different data sources—vital registration, household surveys, census, health service records and specific studies on reproductive-age mortality (RAMOS).  These estimated numbers of maternal deaths are also expressed as a ratio of 100,000 live births to calculate a different measure of mortality: Maternal Mortality Ratio (MMR).  MMR is the MDG Goal 5 indicator, and measures the risk associated with each pregnancy (i.e. obstetric risk).  While monitoring MMRs is particularly useful for policy making and decisions regarding the accessibility to and the quality of prenatal and obstetric care, it does not allow us  to determine whether these deaths are due to direct or indirect causes (see below).The numbers in the Hogan et al. paper are also estimates—of both total numbers of maternal deaths and MMR.  The authors were able to gather more data than in previous years, although the quality of the data remains unavoidably poor (see Karen Grepin’s excellent analysis of the new estimates and why they might in fact be low because of data and assumptions used in modeling these numbers).My Take-Aways for Policy and Research1) Improve the quality and coverage of vital registration dataIt’s good news that we have new estimates. It’s bad news that these estimates continue to be derived from poor data and that some countries have no data at all! (Note that 21 countries--mostly in North Africa and the Middle East--have no data on maternal mortality for the entire period from 1980-2008!) Health Metrics Network (HMN) at WHO is  charged with strengthening health information systems, including vital registration systems in the developing world, but we still have a situation where we are debating estimates because of poor quality and coverage of vital registration data. WHY?  Richard Horton, Chair of the Board at HMN and Editor of the Lancet has widely touted the Hogan et al. paper.  Perhaps Horton and HMN Executive Secretary, Sally Stansfield, can offer us some insights on how this debate about the maternal mortality estimates will spur efforts to strengthen information systems.  What will it take besides resources to establish and maintain functioning vital registration systems?2) Learn what has worked and why Unfortunately, these new estimates do not shed any light on the interventions and programs that can successfully reduce maternal mortality. We are just beginning to learn more about the effects of contraceptive use on the number of maternal deaths (fertility decline has a lot to do with reducing a woman’s risk of dying during pregnancy because it reduces the frequency of her “exposure” to pregnancy). We also know that while proven technologies needed to prevent most of the [350,000-500,000] maternal deaths that occur every year already exist (a fact which has prompted the WHO to designate such deaths as "avoidable”) we have almost NO evidence (see here for a succinct summary of the empirical evidence or lack thereof of interventions including traditional birth attendant training, increased skilled attendance at birth, antenatal care, community mobilization, and dissemination of clean delivery kits) from developing countries to support the claim that any single intervention can  effectively reduce maternal mortality.At another level, as former CGD Vice President Ruth Levine explores in one of her case studies in Millions Saved:  Proven Successes in Public Health, evidence from Sri Lanka shows how long-term government commitment to broad, systematic improvements of health services for pregnant women can have a dramatic affect on maternal mortality rates.  The Hogan paper shows that in 18 years (1990-2008), some countries like Egypt, Romania, Bangladesh, India and China have had substantial success in reducing MMR, but only a few policy case studies have documented the actual program let alone evaluated the outcomes of these programs.3) Distinguish between direct and indirect causes of maternal mortality “Direct obstetric deaths” refer to deaths that  result from obstetric complications of the pregnant state (i.e. pregnancy, labor and up to 6 weeks after the delivery--also known as puerperium), from interventions, omissions or incorrect treatment, or from a chain of events resulting from any of the above. “Indirect obstetric deaths” refer to deaths that result from a previously existing disease or a disease that developed during pregnancy and which was not due to direct obstetric causes but which was aggravated by the physiological effects of pregnancy (eg. HIV/AIDS, Malaria, cardio-vascular disease, diabetes). The new estimates from the Hogan et al. team highlight the negative effect of HIV/AIDS on maternal mortality, particularly in east and southern Africa.A paper from Cross, Bell, and Graham(2010) shows that distinguishing between direct and indirect causes of maternal mortality is even more important now as maternal death due to indirect causes such as HIV/AIDS and Malaria begin to outnumber deaths due to direct causes in some regions of the world.  Highlighting the difference between direct and indirect causes of maternal deaths will be necessary for effective  monitoring and evaluation of interventions aimed at reducing mortality.  For example, evaluating a safe motherhood program with interventions that address obstetric risk might include all causes of death, including those (like HIV/AIDS) that are not targeted by the safe motherhood program, perhaps underestimating its effect. In addition, if the outcome being evaluated is only all-cause maternal death, it will hide the need to provide additional or alternative interventions to combat indirect maternal death such as due to HIV/AIDS.4) Let’s try something newWe need a new idea to catalyze progress towards the maternal mortality MDG, re-energized by these new numbers and strong policy opportunities in developing countries and with donors (see here, and here).  Okay, so the numbers may have budged, (although we aren’t sure because the data are unavailable and/or of poor quality, particularly in countries where ratios are highest), but even if these were actual counts of maternal deaths, 330,000 in a year is still unacceptably high.  Health system strengthening continues to be a slow, grinding activity, and we don’t even know whether or not safe motherhood programs are working, and why. Personally, I’m tired of reading the same old rants about policy and planning and what governments and donors should do to reduce maternal deaths. Other than more and better evaluation to find out how and why packages of interventions work, why not try something new like the COD Aid idea from my colleagues at CGD?  As I think through (with several colleagues in and outside of CGD) the possibility of applying this fresh thinking to maternal and neonatal mortality, the many challenges about addressing maternal mortality—weak health systems for service delivery  and measurement—loom large.  But I do think that providing incentives for countries to accelerate their progress towards their MDG 5 targets by paying them a bonus directly proportional to the number of maternal lives saved is one way for countries to take on these challenges. Of course, the devil is in the detail, and each country will work this out differently.  But if they can, with all the help they need and ask for, there is some hope of a way forward for what has remained an intractable global health challenge.   What do you think?  Thoughts on a COD Aid application for maternal mortality or any other ideas?