From the issue of The Lancet published today:
There is little hope of obtaining precise estimates of maternal mortality rates, as we do for under-5 mortality, for instance. The sources of data are heterogeneous, data quality varies substantially, and the issue of death after induced abortion remains important in countries where it is illegal. It seems a better strategy to separate estimates of obstetric deaths for countries with vital registration, and pregnancy-related deaths for countries that rely on surveys, to increase internal consistency and produce more reliable trends.
Maternal deaths and pregnancy-related deaths are not necessarily the same thing, the article states. A maternal death is one that "could have been prevented by proper antenatal and obstetric care," while a pregnancy-related death "can include infectious, non-communicable, and external causes."
The article's authors, Michel Garenne and Robert McCaa, also say that "one could note a decline in maternal deaths despite an increase in pregnancy-related deaths when confounding with other causes is very strong, as is the case in countries with increasing death rates from HIV, tuberculosis, accidents, and violence."
I would say two things about this article's thesis. First, in a country ravaged by HIV/AIDS and war, some women who are stricken with illness or murdered will be pregnant, but how those deaths are pregnancy-related is a mystery to me. A pregnancy-related death to me would be a woman murdered by her husband for being pregnant, or a woman whose pregnancy contributed to her death from swine flu, for example.
Second, just as a note, it's pretty grandiose to say that "proper antenatal and obstetric care" can head off every true pregnancy-related disaster, like amniotic fluid embolism, for example. Sometimes, in spite of the best efforts, women die.
Oh, well, the authors' point is a good one: Precise estimates of maternal mortality are hard to come by.