Friday, 23 November 2012

More on Irish Maternal Mortality

[Update 25th November 2012: corrected a couple of errors.]

It's always good to follow the example of the masters, so I'll start this post in the style of Cosma Shalizi, who regularly leads off with an Attention Conservation Notice, telling his readers what return they can expect for the time invested in reading further. That's especially appropriate in this case, because Ireland's low maternal mortality figures are being cited frequently in the course of arguments against abortion liberalisation. But that's not what this post is about. To my mind abortion law has to do with rights, not numbers. I want to know more about the numbers, but even if I knew them with certainty I don't believe that would settle any moral questions.

Another issue that I'm not proposing to tackle is the underlying causes of variations in maternal mortality from one country to another. There is quite a bit of literature on that but I haven't read it. When I do, I will share whatever I learn here, so please drop by again. That's what blogging is all about: sharing one's scraps of newfound knowledge (and frequently getting told one is full of shit, usually by people who know even less).

If you're still reading I take it that you want a few numbers. So far, the best primary source I have found is the clinical reports of the Coombe Hospital, which are available online. The table below combines two tables from page 34 of the report for 2010 (click to enlarge).

We could add one more column, for 2003, by looking at the corresponding tables from the 2009 report, which tells us that 8,288 mothers attended the hospital in 2003 and there were no maternal deaths in that year. So over an 8-year period, 71,229 mothers attended the hospital and there were just three maternal deaths. That's a reasonably good outcome, with mortality averaging 4.2 per 100,000 mothers (or slightly better worse if the number is expressed as a proportion of live births). From press reports I know that there were two deaths in 2001-2002. If we include those, assuming that the number of mothers wasn't much different from 2003, we get a mortality figure of about 5.6.

Does what goes for the Coombe Hospital, go for the country as a whole? Well, maybe. The pessimist's reading is that the Coombe is an unusually well-equipped hospital so others are unlikely to do as well. The optimist's reading is that the Coombe probably handles more difficult cases -- see the footnotes to the table. The truth is that we can't safely make inferences for the country as a whole from the figures for a major hospital or even a few of them.

So we're back to the need for better data. When I wrote my previous post, I hadn't read this recent article by Niall Hunter which mentions recent research highlighting the inadequacies of CSO data. I haven't seen that research but I doubt that it would surprise me greatly.

Wednesday, 21 November 2012

Good Midwives or Bad Statisticians?

In the controversy surrounding the death of Savita Halappanavar many commentators have been quick to point out that Ireland's maternal mortality rate is really very low by international standards. I'm not disputing that, but a word of caution is in order.

According to the CSO's Report on Vital Statistics 2010, there was just one maternal death in Ireland in that year. This figure is certainly incorrect. The Rotunda Hospital recorded three maternal deaths in 2010, while the Coombe Hospital recorded one. (See this Irish Times report). I don't know what the true figure for the entire country is but obviously it can't be less than four. It is unlikely to be in double figures: maternal mortality figures are typically quoted as deaths per 100,000 live births and Ireland has about 75,000 births per year.

So the CSO's numbers are a bit dodgy. What else is new? It won't come as a huge surprise to anyone who has wrestled with CSO figures for migration and the balance of payments. As Paul Krugman remarked in the latter context, Ireland is a major importer of errors and omissions. Even so, an error of at least 300% should make us hesitate before making brash claims about the quality of medical care.

Efforts are being made to produce more meaningful figures. Until these are available, treat with scepticism anyone who refers you to numbers published by outfits like the World Health Organization, Unicef or the CIA Factbook. Their models rely in one way or another on figures supplied by national sources; garbage in, garbage out.

Once again, because I really don't want to be supporting scaremongers: I'm not suggesting that Irish maternal care is poor, even by the standards of other highly developed countries. I'm simply saying that we don't have data that entitles us to make strong claims about our relative performance. Maybe the English really are a bit worse at maternity care. Or maybe they are just a bit better at counting.