Statistical Issues in the Investigation of Suspected Serial Killer Nurse:
Transcript of a lecture given by myself, Richard Gill
Slide 1.
I have given this talk to several different audiences so far, with different backgrounds and knowledge. All I ask is that you, the reader, have an open mind; an open mind to the possibility that the young lady on the right (Lucy Letby) might perhaps even be innocent, certainly I hope that you are open to the possibility that her trial was not completely fair and I hope you are open to the possibility that the preceding police investigation was not completely fair or unbiased either.
I am a scientist and a statistician. I certainly have opinions about Lucy Letby’s innocence. I see no reason to suppose she killed anybody actually and I believe that I have very good reasons for thinking that. My opinion is based on what I know about the science, and I know a lot of things that certainly the jury did not know, though which anybody can look up on the internet these days. Of course, we don’t know how reliable everything is. Obviously, you mustn’t believe everything you find on the internet. I do think there is powerful evidence for her innocence out there now which was not shown to the jurors and certainly not shown to the public during the trial. I hope there will be a retrial and that this evidence will be properly validated and used.
I am interested in Lucy Letby case because I was previously very much involved in the case of the young lady in the picture on the left, Lucia De Berk in the Netherlands. The picture was taken in the good old days before she became a suspect in the murder investigation and before she became convicted and widely known to be the most prolific serial killer nurse in the Netherlands, both of babies in a children’s hospital and of sick aged people in other hospitals where she had worked earlier. You see her here in around 1993, the photo taken by a friend around Christmas time (I know them both). I am not sure if she was already working as a nurse at that time. She was then about 30 years old, she already had a husband and a daughter. The photo was taken by her friend Carol Edrich, who Lucia met when Lucia was Carol’s cleaning lady. Carol had encouraged her to become a nurse, and so she did.
Lucia’s trial and tribulations began in 2001. A baby, Amber, died unexpectedly in September a few days after 9/11, and some people at the hospital thought the death was surprising and unexpected, though others at the hospital were not surprised at all. But this event certainly combined with suspicions which some doctors at the hospital previously had about Lucia, and there was gossip at the hospital because she certainly had a colourful life up to then. There were lots of weird things that were said, and she was one of those nurses who stands out in a crowd, says what she thinks, speaks up when things are not being done right and she did not let herself be walked over by anybody. Some people found her very odd, and some found her a bit scary; many on the other hand liked her very much.
In no time there was a police investigation and immediately Lucia was the key suspect. She was arrested in December 2001. There was a first trial in 2003, and at that trial the main evidence against her was statistical evidence. Statistics is of course my field, so that interests me, and it is statistical evidence that I see as linking these two cases even though in the Lucy Letby case there was no statistical evidence used apart from the spreadsheet which I will come to soon. At Lucy Letby’s trial, neither prosecution nor defence, as far as we know, employed a statistician or an epidemiologist, and there was no talk about statistical data and statistical analysis.
In the Lucia De Berk case, at her initial trial which resulted in her conviction for a number of murders and attacks, the main piece of evidence was a probability. calculation resulting in the chance of 1 in 342 million if I remember correctly, and it was calculated by a friend of mine who I respected very much, and at that time we the public but also professionals didn’t know exactly what he’d done but I knew he was at least a competent person and he got this number 1 in 342mil. Many people misunderstood it, it was not the chance that she was innocent, it was supposed to be the chance that were she innocent and were the coincidences between her shifts and bad things happening completely due to chance, what was then the chance of such an extreme coincidence or correlation or association. So it was that 1 in 342 million, was not the probability she was innocent, but it was the probability were she innocent and was everything just due to chance, then the probability that there would have been such extreme associations between her presence and the deaths. That same expert said obviously my hypothesis is wrong because we don’t believe a 1 in 342 mil event happened so there is some meaning behind it. This is a correlation and real and something causing it. He gave a list of possible causes and only one of them was that Lucia was murdering people which obviously would have been one explanation.
One of the big differences in these two cases is that he Netherlands case is done according to Netherlands criminal prosecution law and we do not have a jury, we do not trust a bunch of average people off the streets to decide guilt or innocence, we prefer to have a judges bench of three professional judges who have been to university, studied law and are very experienced. We have what you might think of as an inquiry by a board of judges, a chairman and two others, checking what the prosecution says is correct. We like it that way much better. This is due to differences to culture and history, but it’s a big difference between that and a jury trial that we have in the UK. Inquisitorial versus adversarial are the technical terms for these two kinds of trials. Inquisitorial trials the judges run an inquiry, they are free to call scientists and witnesses and experts, and they can change things on the way as they go along, but many things are agreed beforehand. There are pretrial meetings with judges and barristers (or the equivalent of) and scientific experts are very different category from ordinary witnesses and treated differently.
At the end, the judges give their verdict, and they have to write up their motivation and reasoning for their verdict, whether they have used logical deduction to reach their conclusion, and this is published. This is an advantage for the defence of the Netherlands system because it means you can see what the reasoning was and if you can see a gap in the reasoning, you can attack it and use it in an appeal. If you see evidence being used and being crucial and if you disagree with that evidence for example scientific evidence that you think has been interpreted wrongly, you can point to that. I think such trials are fairer than the English ones in cases like this which are so complex and with so much evidence, as so much careful thinking has to be done to reach a reasonable decision.
The initial conviction of Lucia, at the lower court, where she got a life sentence, was almost entirely based on the statistical calculations. Immediately that was very controversial and lots of people wrote letters to newspapers saying you can’t do that, you can’t convict just on the basis of a probability calculation, it’s just not right. Also, statisticians started fighting each other because there are different kinds of statisticians, so they started quarrelling in public, as they tend to do, and arguing that the wrong statistical paradigm was used, and it should be Basian rather than frequentist. That was an interesting discussion and it certainly meant that there was an appeal. There was a new trial and that was quite rapid, a year later, and again Lucia was convicted and this time actually of more murders and more attempted murders ad the judges wrote explicitly at the start of their 150 page motivation (because it was a complex case) they wrote out exactly what they were so convinced of that she was such an evil person and how they came to that conclusion and the evidence which convinced them. It was the longest summary (called the ‘arrest’ in the Netherlands) in Dutch legal history. On the first page the judges are very explicit, the court does not make use of a statistical calculation; this is not statistics, no statistician has calculated the probability of this, they are not interested in that. It goes on to say Lucia is convicted by us on the basis of irrefutable scientific and medical evidence. That is very clear.
That stopped all the discussion in the press and amongst the public, on the nascent social media, about the statistics in this case because it was just medicine, and it was definitive and proved. Not many people read those 150 pages. That was in 2004. The defence made one more attempt by going to the supreme court because some evidence got forgotten and had stayed in a drawer at the forensic institute and it was said to be exculpatory but had arrived one day too late to be used in the trial. The supreme court turned down the application, but they did do one thing; Lucia had been handed life sentence, but the length of a life sentence is only 30 years and if you behave you can come out earlier. The judges were so convinced of her guilt and evil character that they tacked on to that life sentence, indefinite incarceration in a closed psychiatric hospital thus making her sentence a full life sentence.
Amusingly that was the 2004 verdict and sentence, and the supreme court removed that addition of the psychiatric bit because as they pointed out a psychiatric investigation had been done into Lucia’s state of mind and it had uncovered several disorders (everybody is weird in one way or another and all these have labels and names and pharmaceutical companies have made medicines to sell to you for them). But there was no psychiatric issue which she had which could be associated with her crimes. Therefore, she could not be sentenced to time in a psychiatric hospital because there was no treatment she could receive. That was 2005, so the sentence was reduced back down to the life sentence (not full life).
That was exactly a time when a couple of whistle-blowers started getting active and there was a philosopher of science professor (just retired) and a senior medical doctor, a brother and sister. They had personal connections to people at the hospital where Lucia had worked and had some inside knowledge and were careful not to use it. They started to talking to journalists and journalists starting stirring things up. An application was made to the Dutch equivalent of the CCRC (Criminal Case Review Commission in the UK). There is one in the Netherlands which was started at the same time as the UK, for the same reasons, as a result of enormous scandals of miscarriages of justice, involving either the police with tunnel vision or involving use of the wrong experts who weren’t actually experts and misinterpretation of scientific evidence.
So, exactly the same things as in the UK. And that team also published a book which was available in bookstores. Lucia’s case was completely closed, there was no contempt of court involved in this in any way (contempt of court rules are rather more reasonable I would say that the British ones). They started stirring things up, and a long process started which eventually only by 2010 (it took 5 years and we thought it was such a long time), it took 5 years to stir things up enough to get things working and to get the wheels in motion such that there was a new trial started in 2009. Lucia walked free in 2010 because there was a new trial, a fair trial, new evidence or at least new interpretations of scientific evidence which were much more plausible. And moreover, the public was also convinced by them and everybody (bar a few probably) now knows that Lucia was completely innocent. If you know a bit about the whole thing it is totally implausible.
It was 2005 when I got involved. It was a friend of mine who was very involved, a statistician. He got very upset by the fact that he discovered, that though the judges had said there was no statistics involved, the statistical thinking still reverberated through all those 150 pages of summary. Indeed there were doctors who said explicitly “normally cases like this would be seen as completely normal and natural, but because Lucia had been there and was there so often and I can’t explain it, I am inclined to believe that this particular death was unnatural”. So that ‘irrefutable scientific medical evidence’ was actually based, in part, on an opinion by medical doctors whose opinions were contaminated by the fact that they knew that a nurse was being investigated for a crime and she was always there when things happened. This was fact, she was surprised, it was a bit unusual, I have a probability calculation coming out of my 150, I would say that’s bad luck, maybe grounds for an investigation.