Lucy Letby and the statisticians

They haven't got a clue

I hope you’ll forgive one more Letby post. I promise it will be the last unless there are further developments of any significance. Further developments have been thin on the ground since her appeal was rejected, which is why the Telegraph’s Sarah Knapton has had to resort to repeating the same discredited talking points to keep the clickbait flowing. Her most recent effort focuses on a small gathering of statisticians at which doubts were raised about Letby’s conviction.

Speaking about the evidence after the meeting, Dr Jane Hutton, a professor of statistics at the University of Warwick, said: “It’s a large pile of crockery, much of which is broken. Such a pile does not hold water however big it is. [Sorry, what?! - CJS]

“We’re not saying the conviction is unsafe, but we consider that if the concerns we are raising are essential to the decisions of the court, then the convictions are unsafe.”

Alas for Dr Hutton, the concerns she is raising have always been irrelevant and she is attacking a straw man. As I mentioned in my previous post, it is worth listening to this interview in which she reveals her profound ignorance of how the investigation proceeded and what evidence was presented at the trial.

The same seems to be true of the other prominent statistician who has been banging the drum for Letby, Professor Richard Gill. Gill is quoted extensively in this recent article from the newspaper of Leiden University where he used to teach. I mention it because it makes the same false claims about how the investigation and trial proceeded that have become widely accepted by ignorant Letby supporters. As usual, it centres on a misunderstanding of the spreadsheet that showed that Letby was the only nurse on duty when all of the suspicious events took place.

According to Gill, the doctors’ bias is apparent in several ways: ‘How did they arrive at these 25 cases? They argue that Lucy is present every time something suspicious happens, but the deaths are only considered suspicious because Lucy was present.’

Gill is adamant: ‘They concluded that Letby was guilty based on their prejudices. The police expert then made up a story about how those babies could have been killed, ignoring all other evidence.’

This is astonishingly ignorant and conspiratorial. Gill has been commenting on this case for a long time. If he is so willfully ignorant to believe this theory then it is no wonder that relative newcomers have talked themselves into the same delusion. I have addressed the Texas sharpshooter nonsense in previous posts, but I think it would be useful to put down in print what two of the key players in the investigation have said in interviews.

The first is Dr Dewi Evans who reviewed the medical records for the police in the summer of 2017. He identified 15 incidents which he believed to have involved deliberate harm. It turned out that Letby was on duty for all of them, but he did not know what when he reviewed the evidence.

On the Tortoise podcast recently, Evans said:

“This trial, this case, had nothing at all to do with statistics. Absolutely nothing. The statisticians are making a hue or cry, and their hue and cry is based on information that is simply wrong. These deaths were unexplained and unexpected. The diagnosis made by me and supported by other clinicians was based on clinical evidence. In other words, the deaths or collapses were either suspicious, that is, they could not be explained on the basis of some natural cause, or, to put it more strongly, they were consistent with inflicted harm.

In other words, there was sufficient evidence in, these cases, to confirm that the death was the result of an injection of air into the bloodstream. And if the cases - if there was only one case of this nature, there would be sufficient there, in my opinion, to make a diagnosis of inflicted injury.

The fact that all of this occurred when one nurse was on duty was not known to me when I was reviewing these notes. And it is quite important, I think, that I did not know that there was a suspect. I did not know that at the time I was reviewing these notes, she’d been taken off clinical duties, sometime earlier. I knew nothing about that. So the trial was based on clinical evidence and the statisticians can make as much of a fuss as possible. Their opinion is totally irrelevant in relation to how we reached our diagnosis.”

Evans’ account of this has been consistent throughout. In a podcast with Raj Persaud last year, he said:

“So right from the beginning [after seeing Baby’s O’s records in July 2017], I said, look, this is inflicted injury. We can - we need to investigate this. And I also told them, and I think this was crucial as far as the trial is concerned, I said, if you suspect somebody or you suspect some people, I don’t want to know.”

And this is what he said on the Trial of Lucy Letby podcast last year:

Evans: “At the time [July 2017] there was no mention of any suspect. There was certainly no mention of Lucy Letby. And there was certainly no mention even of someone saying ‘look, I am concerned, I have seen somebody doing something that is concerning. So both the police and I were working on a blank sheet.”

Interviewer: “We’ve interviewed the SIO [Detective Superintendent] Paul Hughes and he said that at that early stage they were all looking to see whether a crime had been committed, not investigating a crime per se and were hoping that the experts - so you, essentially - would review the notes and go “oh no - this explains why this baby died.”

Evans: “That’s right. What I said to the police is ‘if you have a suspect, I don’t want to know’. And I think what was crucial, I think, in our investigation was that I told the police: ‘Send me the notes of every baby who has died. Send me the notes of every baby that has collapsed or deteriorated, between early 2015 and July 2016, emphasising not just the suspicious ones or the ones that could not be explained. I wanted to see all of them.

“So I looked at 32 cases and in a number of cases it was quite easy to rule out or explain why the baby had died or why the baby had collapsed. It was usually due to complications of prematurity or issues involving significant congenital problems or infection or haemorrhage. And these are the common factors causing death or disability in a baby on a neonatal unit. So having ruled out a number of cases I ended up with 15 cases where the collapse and/or death could not be explained on the basis of some natural cause.”

You could, I suppose, accuse Evans of lying in all these interviews (and in court), but if you accept that he is telling the truth, we have to reject the notion that incidents were only deemed suspicious because Letby was on duty. The association with Letby had certainly occurred to some of the doctors at the hospital - which is why she was moved from night shifts to day shifts and later moved to a desk job - but it was not known to Evans.

Nor was it known to the detectives who initially reviewed the cases. As Detective Superintendent Paul Hughes has explained, he allocated each case to a different detective precisely because he wanted to ensure a ‘sterile corridor of evidence’

“I decided that we’d allocate one case per detective. And the reason I decided to do that was - two reasons. One, it gives a unique ownership to a detective that they don’t usually get the way we work major crime in this country, but it also separates the potential sharing of the information initially so that other people’s mindset would be altered by what they’ve heard in another investigation. So an issue was to keep it separate, to create a sterile corridor of evidence. So allow people to come to a determination of what they were finding on their own and then at the right time introducing them weekly team meetings of all the detectives together with the analysts where they would start sharing information.

It was chilling, really, at times to see it drop into effect. So a detective would go on to give the update of their investigation and the fact that: ‘Well, what happened in my case was, according to the medical evidence, the collapse took place at this time. At this time, Lucy Letby’s [sic] designated nurse went on a break handing over care to Lucy Letby, the parents left and the child collapsed’. To hear another detective go ‘Oh my God, that’s exactly what happened in my case’ and then, and then, and then. And you start building that realism of ‘hang on, independently this has been investigated in isolation’ - and now people are saying, ‘my God, that is exactly what I have seen’.”

Once the police noticed that Letby was the only person present on every occasion, they interviewed her because she was obviously better placed to provide information on all the incidents than anyone else. She was, Hughes said, “our biggest source of evidence”. The investigation went from there, with more evidence being gathered, leading eventually to Letby’s prosecution.

Whatever you think about Evans’ interpretation of the medical records and the strength of the evidence generally, it is a fanciful conspiracy theory to claim that Evans or the doctors or the police cherry-picked the collapses and deaths that Letby was present for and ignored the rest. As far as Evans and the detectives were concerned, the suspicions came first and Letby’s association came afterwards.

Now let us return to what Gill has to say in the university newspaper.

According to Gill, the fact that those 25 specific cases were singled out and listed on the schedule is a sign that doctors started looking at incidents related to Letby with a huge confirmation bias: they only found what they wanted to find.

The doctors did not assemble the spreadsheet.

‘Any statistician will immediately wonder why, when and by whom it was decided that specifically those 25 incidents were considered suspicious.’

Evans initially identified 15 of them and more were identified as the investigation proceeded, including two cases of attempted murder by insulin poisoning. Note that Letby was charged with 22 offences, not 25, so there were incidents when Letby was present for which she was not charged, presumably because of insufficient evidence. And the jury did not convict her of all charges, which suggests that the supposedly ‘statistical’ evidence of the spreadsheet did not impress them too much.

‘Because there were more deaths than those listed on that schedule, but no one seems to know exactly how many.’

Keep up, Richard. There were nine other deaths, none of which were remotely unexpected or suspicious. Four were due to “congenital problem or birth defect” one baby was “asphyxiated or deprived of oxygen at birth, and the other four “died of infection and their deaths were precipitated with a period of time consistent with infection”.

‘If a child dies within the first six days after birth, English hospitals classify this as a stillbirth in other cases. That’s better for the hospital’s statistics, because these stillbirths don’t count towards mortality rates, so they don’t appear on this schedule either.’

This is simply not true. No live births are counted as stillbirths in England. A neonatal death is defined as “a baby born at any time during the pregnancy who lives, even briefly, but dies within four weeks of being born”. Indeed, some of babies Letby killed did not survive for six days and were not counted as stillbirths. I don’t know where Gill gets this idea from.

Several newspapers are currently speculating that there were between six and nine other deaths that were not considered during the trial.

There were nine and they were not considered at the trial because they were not suspicious, even though Letby was present for at least two of them.

According to Gill, the doctors’ bias is apparent in several ways: ‘How did they arrive at these 25 cases? They argue that Lucy is present every time something suspicious happens, but the deaths are only considered suspicious because Lucy was present.’

Utter rubbish.

‘Moreover, it’s bizarre that only nurses are included in this schedule. Why weren’t doctors or cleaning staff considered in the investigation? They assumed from the start that it was a nurse because they already had their eye on Letby.’

Reading the court of appeal judgement is enough to show that different doctors were on duty when the babies died and they were generally called in from other parts of the hospital to help resuscitate them. In other words, they were not in the nursery when the collapses occurred and therefore did not have the opportunity to cause them.

The idea that a cleaner had the opportunity to attack the babies without being observed is too ridiculous to comment on.

‘Some of the nurses didn’t work night shifts, which are riskier hours because no doctors are present. But Letby did. In addition, not all of them have the same level of training, while Letby was one of the more experienced nurses on the ward. Therefore, it’s plausible that she would tend to look after the more vulnerable patients; patients with a higher risk of death.’

Doctors are present during night shifts and Letby was moved onto day shifts in April 2016 after which the deaths and collapses started happening in the day time (with the exception of Baby N when Letby was given a night shift). The prosecution argued (and I agree) that Letby preferred working night shifts because the parents were less likely to be around.

Furthermore, not all of the deaths were independent incidents, Gill notes. Several children on the list are twins, and there is one set of triplets. If one of them dies, it is more likely that the others are also at a higher risk.

Two of the triplets in this case died (Baby O and Baby P). The other survived by being moved to a different hospital at the insistence of the parents. The consultant agreed with their request because, by this point, she had concluded that Letby was ‘a mortal danger’ and that transferring the last triplet was ‘the only way he was going to live’.

Baby A and Baby B were twins. Baby A died and Baby B survived.

Baby E and Baby F were twins. Baby E died and Baby F survived.

Baby L and Baby M were twins. Both of them survived.

Letby seems to have had a penchant for killing twins and triplets but she never fully succeeded. So much for the idea that if one dies, they all die.

‘If you happen to be working when a seriously ill pair of twins comes in, you’re more likely to have two deaths occur during your shift.’

That never happened.

Based on the figures, he concludes that no murders were committed at all. ‘During the years in which Letby allegedly murdered the children, mortality rates spiked in all hospitals in the region. The Countess of Chester Hospital wasn’t even an outlier.’

The case was never about ‘spikes’. It was never about statistics. You cannot conclude that no murders were committed from looking at ‘the figures’. As I said to Peter Hitchens during our debate about this recently, the babies are not data points for statisticians to ponder over. The suspicious circumstances of their collapses and deaths were discussed in meticulous detail during the trial and if statisticians can’t be bothered to engage with the totality of the evidence then they should STFU.

Previous posts on this subject:

Is Lucy Letby innocent?

Further thoughts on Lucy Letby

Lucy Letby and the Texas sharp shooter