Sunday, December 02, 2007

Probability of being a serial killer: the case of Lucia de Berk

Here's a post on an extremely interesting story at the intersection of law, statistics and professional ethics in several disciplines. Inter alia, the issue has been covered in some of the best scientific journals. See for instance a). a report in the Science edition of the 16th of November (you need a subscription to access the PDF); b). a very good report on this case in Nature, January this year; you can read here that PDF (subscription free).

What is this about? One of the most interesting legal cases ever, involving multiple homicide, is now under review by the Supreme Court in the Netherlands. The case deals with the former trial of Lucia de Berk, a Dutch nurse, accused and sentenced to life imprisonment in spring 2003, for the alleged murder of seven hospital patients and the attempted murder on three others, in places where she had worked between 1999-2001. The issue is that there was never direct evidence to implicate De Berk (and she has always denied all accusations): she was condemned solely based on the fact that she happened to be always around when these patients died; basically the courts (an appeal court also upheld the initial verdict) decided that it was very unlikely, essentially only one chance in 342 million, according to the expert statistician who testified at the trial, that so many deaths could have occured accidentally while she was nearby.

What happens now is that the 2003 sentence is being challenged by several scientists who signed a petition to re-open the case (this is what the Supreme Court needs to decide on, after last months a justice department panel indeed recommended that the case be reopened). Richard Gill, a Leiden University mathematician and organizer of the petition, states that the previous conclusion, leading to the condemnation of De Berk, is based on "every statistical mistake in the book". Gill and others concluded that the previous statistical testimony was based on an incorrect analysis and that in fact the probability estimated earlier, of 1 in 342 million, is in fact as low as 1 in 48 or even 1 in 5, which are very unlikely to meet the criterion of "beyond reasonable doubt" needed for a criminal conviction. Here's the website of Gill dedicated to this case (with his detailed discussion of the statistical aspects in this case here and Gill's synopsis/reconstruction of the case + other interesting details here). There is even a whole book criticizing De Berk's conviction on scientific procedure, by Ton Derksen , a philosopher of science from the University of Nijmegen. Mark Buchanan, who wrote about the case in Nature (see the link above) summarizes the legal essence of the argument this way: "The court needs to weigh up two different explanations: murder or coincidence. The argument that the deaths were unlikely to have occurred by chance (whether 1 in 48 or 1 in 342 million) is not that meaningful on its own - for instance, the probability that ten murders would occur in the same hospital might be even more unlikely. What matters is the relative likelihood of the two explanations. However, the court was given an estimate for only the first scenario."

This is certainly not an easy case, despite the fact that it is not the first one that might involve wrong statistical evidence in a criminal sentence (the Nature report linked above also mentions another high profile case involving misuse of statistics, the case of Sally Clark, from 1999, in Britain). My opinions are the following. Firstly, I strongly believe that the Dutch Supreme Court has more than sufficient basis to re-open De Berk's case and carefully re-analyse all the previous evidence (I've obviously signed the petition as well). Further, one can only hope that a wrong verdict is overturned swiftly in De Berk's case, should the petitioners be right (and then I wouldn't want to be in the place of the 'expert statistician' who testified to start with, although it is true, as can be read on Gill's discussion, that this expert "always insisted that his analysis only showed that the observed coincidence could not be due to pure chance, not that Lucia caused the the deaths"; moreover, this expert himself wants the case to be reopen; but the scientific flaws would still be there, if the petitioners are right.). More generally, one can only hope that science will be used with the greatest care in any legal processes, especially criminal ones, given the extreme emotions and stakes typically involved (but not only: science has to be done properly, anytime, anywhere, anyway...).

No comments: