I wonder why the figure is so noisy in the 1800s. Admittedly, this is not related to the question.

The combined risk is really stunning: If one childbirth has a risk 4 % of the mother dying, and if we assume the risk in independent event, the combined risk for 5 deliveries is 1-0.96^5, which is almost 20 %!

So when people talk about "natural childbirth" they mean the kind with 4% mortality. In a society with very large families (10+) indeed this must have been a common way for a woman to die. Ouch.

@gerrit: From the paper: "From 1837, the Registrar General's office started to record maternal deaths; this was backed up by invitations from the Presidents of the Royal Colleges and the Master of the Society of Apothecaries to supply voluntarily copies of certificates of death, if possible with cause. It was not until about 1870, when the registration of cause of death was made mandatory, that rates became reasonably accurate." - but that doesn't explain the noise post-1870.

@RedSonja, not necessarily. There's plenty of room between unattended (by a professional with training and an understanding of hygiene) childbirth and heavily-medicalised childbirth. There are probably as many meanings of "natural childbirth" as people using the phrase, but most of those probably fall within this gap. For example the term is often used for anything other than a Caesarean, or delivery without the using of forceps/ventouse.

@RedSonja: Without taking sides on the "natural childbirth" debate, the British paper explains that after the 1930s: "Then a sudden precipitous reduction in maternal deaths occurred which could not be due to any natural factors involved in death. It was, in fact, due to the overcoming of maternal infection by chemotherapy and antibiotics." These are still available to women having modern homebirths. So it is unfair to categorise modern "natural" births as being at the 4% death rate.

@Suma: The pedant-inside-me-I-should-learn-to-keep-quiet is yelling in my ear that the assumption of independence is flawed, because you can't die twice giving birth.

It's quite astonishing to see how a dramatic reduction in the death rates of mothers giving birth correlates quite closely with the existence of the NHS (since 1948).

@Nobilis The NHS-claim in particular would need to be investigated by comparing reduction in maternal deaths to places of comparable wealth where no NHS-equivalent exists, such as the USA. I suspect it's not quite as easy to show the effect of NHS, because NHS emerged at the same time as medicine overall improved dramatically.

@Nobilis: Again the paper attributes the change to chemotherapy and antibiotics, not the NHS.

@Suma I have no specific knowledge of obstetrics, but I think it is conceivable that the risk associated with the first delivery is much larger than the mother's risk during subsequent births. Edit: I see the answer was updated just now to give information on this.

@Oddthinking Yes, but who would distribute these among the general populace if not the NHS? But I would second what gerrit states, it would be interesting to compare the results with a place where healthcare was not universally available (or indeed affordable).

@Oddthinking I think your pedant is wrong. You don't need indepence you need to use the conditional probability. Ie. you need the probability of dying during birth given the fact that you have not died in earlier births. So one could use the 4% figure as Suma does even though you know that very few dies twice(stating very few to prevent discussion about what constitutes death). Unfortunately as this answer points out you can't use the 4% figure because it is not constant, it's higher for the first birht, then follows a J curve.

@Taemyr - very true. I've done a bit of research and found an infographic on wikipedia which shows that the highest lifetime risk of maternal death currently in the world is between 1 in 7 and 1 in 8, i.e. about 13%. (Lifetime risk of maternal death being the chance of dying due to childbirth/related complications, assuming you're a female who made it to 15 years old).

"The first birth is the most dangerous, and then the next few are safer." - is there an actual mechanism for this, or is it an artifact of the fact that the risk depends more on the person's circumstances (access to healthcare, etc) and the people whose situations are bad already died before those "next few"?

At one extreme is "every childbirth has a 4% risk". At the other extreme is "4% of women will 100% die at the first childbirth, and the rest will never die from it at all" - the truth is likely somewhere in the middle.

@Random832: I have no time to search for references, but to give you a starting point: Getting the child through is a considerable strain for the birth canal and one of the major sources of injuries and complications. Another result however is that the birth canal is widened which makes subsequent births easier. Women are essentially training their birth canal with each birth.

All: use Skeptics Chat.

@Taemyr I interpreted the problem this way: Each event, n, has the outcomes "Died during child birth of nth child" and "Didn't die during child birth of nth child". Under that interpretation, the events are not independent, as the outcome of the first event affects the probabilities of the outcome of the second and subsequent events.

@Nobilis I think a simple thought experiment would suffice. If the NHS had been formed in 1900, but Howard Florey's work remained in the early 1940s, would maternal mortality have dropped 50 years earlier. I think the answer is clearly no.

@Oddthinking Think of a biased coin that comes up heads 96% of the time, and tails 4% of the time. You can compute the probability that your first 5 flips are heads exactly as suma has done.

@Oddthinking Switch to the pregnancy case. You can just as easily compute the probability that the mother survives the first 5 pregnancies.

@Oddthinking This isn't correct of course, but it's not because you can't die twice. It's because the events aren't independent for other reasons.

@nomenagentis Yes, I see your point.

@Oddthinking I never claimed that they where independent. But when working with non independent events you need the conditional probability. And that is what we have.

lol it's amusing to me that you told everyone off for having an extended discussion in the comments.. which you were part of :P