Maternal Mortality: The Correct Assessment Is Everything
In the past 25 years, maternal mortality has been tabulated, evaluated, and processed from death certificates and International Classification of Diseases, 9th Revision (ICD-9) codes. The World Health Organization has added on the bookend of death less than 42 days after birth. Statistics had seemed relatively good with this method of counting. During the mid-1980s though, maternal mortality rates leveled off and stopped declining. From 1982, the United States rate has stayed at or above 7.5 deaths per 100,000 live births.2 We are 20th in the industrial world in maternal mortality rate; that is 19 countries have lower rates of maternal death than does the United States. Multiple investigators began to take closer looks at mortality, presumably to evaluate the etiologies and further reduce this rate. Curiously, studies from Australia, Austria, Canada, Finland, France, Illinois, Japan, Maryland, Netherlands, New York, North Carolina, the United Kingdom, and other sites noted a dramatic underreporting of maternal death. In some places, the maternal mortality rate was underreported by 90%, and when maternal deaths occurring more than 6 weeks after delivery were included, the underreporting increased further.
We cannot reduce maternal mortality until we know how large the scope of the problem is, and more importantly, the causes. That is obvious.
In this month’s issue of Obstetrics & Gynecology, Deneux-Tharaux and her coworkers have reported a study that compares 2 methods of assessing maternal mortality: the traditional death certificate/ICD-9 technique with what they have termed a “standardized enhanced method.”3 Databases from Finland, France, Massachusetts, and North Carolina were examined for the period 1999–2000. The results are important for every public health official and health care provider who works with pregnant women. The differences are very large. Using the standardized enhanced method, which links death certificates with birth certificates and then abstracts the records with evaluation by trained specialists, the authors found differences of approximately 93% underreporting in Massachusetts, 22% in France, 67% in Finland, and 27% in North Carolina. The corrected maternal mortality varied from 9.6 deaths per 100,000 live births in Finland to 16.8 deaths per 100,000 live births in North Carolina. Additionally, the various causes of mortality differed greatly between regions: hemorrhage was more of a problem in France, whereas cardiomyopathy was more of a problem in North Carolina.
There are many valuable lessons from this report. The first is that abstractions from large databases, such as death certificate registries, may provide misleading results when not accompanied by follow-up as well as cross-checked validation. There is an old expression about large databases: “nonsense in then nonsense out.” The second lesson is that, if we are to reduce maternal mortality, one of the worst tragedies in health care, we need to accurately assess the problem.