Friday, February 17, 2006

Sudden Infant Death - Chance or Design?

Professor Sir Roy Meadows has won his appeal to overturn the decision that led to him being struck off the medical register.

Professor Meadows, one of the UK's leading paediatric doctors, acted as an expert witness in several cases. In the case that led to him being struck off, he said, according to the BBC report, that the chances of two natural unexplained cot deaths in a family was one in 73,000,000. My understanding is that on the basis of this testimony, a mother was found guilty of murdering two children. She was later freed on appeal.

The BBC adds:
The figure was later disputed by the Royal Statistical Society and other experts said that once genetic and environmental factors were taken into consideration, the odds were closer to 200 to one.
There are several issues here I wanted to comment on.

The first is the obvious "design inference" that was made by the jury when they found the accused guilty to start with. How probable is this specified event (that a family should have two natural unexplained cot deaths? If the answer is, as Prof. Meadows said, one in 73,000,000, then since there are probably only about 20,000,000 mothers in the country at any one time, this explanation doesn't seem likely. Of course, if the odds are one in 200 (by which I suspect they mean, "if a family has one cot death, the likelihood of their having a second one is about one in 200"), then the "chance" explanation is reasonable - even plausible.

The next is what he said at the original trial. Some justifications for reinstating him on the medical register were because expert witnesses need to be able to testify without fear of retribution, and (in effect) because he made an honest mistake with statistics. Now I agree that expert witnesses need to give their honest opinion, and I don't have a fundamental problem with him being reinstated - this case says little about his ability to practice as a doctor (and he has now retired in any case - his reinstatement was about restoring his credibility). What I disagree with is that he ought to be called an expert witness. On the basis that Prof. Meadows was a doctor, a paediatrician, a scientist and a knight, his testimony effectively represented the key plank of the prosecution case. As somebody with some scientific training, it took me about two minutes reading the account of the trial to raise questions - the key one being: "If there is one cot death in a family, can we exclude the possibility that further cot deaths won't have a causal link - genetic or environmental?" The answer to that is no. Which for all Meadows' years of teaching and training is effectively what discredits his "expert opinion". But if somebody is introduced as an "expert" to the jury, even if they have the training to understand the statistics and science involved, are they really likely to think that the expert could possibly be wrong?

And yet, the implication of the reinstatement judgement, as reported by the BBC, is that Meadows made an "honest mistake" over statistics. But that was the whole point of him being there. He was there to be an expert witness about the likelihood of this having a chance explanation - and it sounds as though this was the area in which he was not competent. Sure - no reason to be struck off the medical register - but how on earth did he end up as a credible expert witness in the first place?!

Finally, there's the issue of payment for expert witnesses. This typically ranges from £20 to £200 per hour. On the back of Meadows' testimony, someone was wrongfully imprisoned, and lost months or years of their life, in addition to having their reputation destroyed. In the circumstances, I'd have thought it would be a minimal matter of courtesy if Meadows paid the acquitted woman his fee.

Update - for more commentary and report from the Daily Telegraph, see here and here. These reports supply a lot more detail than I had been able to glean from the BBC reports I had heard. They do little to change my mind.