Lucy Letby - bad inference

[Gfamily]

I know there are some people who are deeply suspicious that the guilty verdicts for the Lucy Letby murder trials are based on bad evidence. I have no particular views on this - except to say that the jury were the ones who sat through all the evidence, and that if there were issues with the prosecution case, her defence lawyers didn't raise them at the time.

However, there was one charge on which the jury couldn't decide either way, and there is now a re-trial about to start. According to the prosecutor, because she was deemed guilty by the jury in respect of the other children, they should take that as significant in respect of the current case, saying:
"Those other cases that I’ve mentioned do have an importance in this case, because we suggest it’s not because you should convict her of this case because of what she’s done in other cases, but what we’re asserting is the relevance is – it gives you significant evidence of what her intention was at the time"
My view is that the very fact that the original trial jury were the ones that: having seen all the other evidence and being convinced of her guilt in "those other cases", if they were unable to come to a verdict in this last case, it suggests that they had good reason not to come to a verdict, so taking the other convictions as being meaningful is a gross overreach.

https://www.theguardian.com/uk-news/art ... cutors-say

[tenchboy]

Odd. The BBC are now reporting the same; whereas earlier today their coverage emphasized the judges words in telling the jury that this case must be decided on the evidence to be presented alone and that they must forget and not be swayed by anything that they may have heard about LL in the past.

[snoozeofreason]

If trials were conducted in a numerate manner then it would make sense to explain Letby's previous convictions to the jury so that they could avoid the defendant's fallacy.

However if trials were conducted in a numerate manner then the original jury would have been warned about the prosecutor's fallacy, in which case she might not have had those previous convictions.

Tom Chivers did a good article on the prosecutor's and defendant's fallacies, making passing references to Letby but without committing himself as to her guilt or innocence.

[noggins]

I dont see the point in this retrial. She’s already serving a whole life sentence. Given the jury found her unanimously guilty on some charges , majority guilty on others, im seeing the hung verdicts as sincere expressions of doubt.

[Gfamily]

I dont see the point in this retrial.

I suppose, for the bereaved parents it is an unresolved issue, so for them, a re-trial is an expression that the state is not willing to leave it.

[dyqik]

If trials were conducted in a numerate manner then it would make sense to explain Letby's previous convictions to the jury so that they could avoid the defendant's fallacy.

However if trials were conducted in a numerate manner then the original jury would have been warned about the prosecutor's fallacy, in which case she might not have had those previous convictions.

Tom Chivers did a good article on the prosecutor's and defendant's fallacies, making passing references to Letby but without committing himself as to her guilt or innocence.

The argument for prosecuting Letby at all was probably contaminated with the prosecutors fallacy. From what I've seen of evidence and analysis published in the US, and which it is currently illegal to publish in the UK, there is reasonable statistical doubt about whether any crime occurred at all.

[snoozeofreason]

If trials were conducted in a numerate manner then it would make sense to explain Letby's previous convictions to the jury so that they could avoid the defendant's fallacy.

However if trials were conducted in a numerate manner then the original jury would have been warned about the prosecutor's fallacy, in which case she might not have had those previous convictions.

Tom Chivers did a good article on the prosecutor's and defendant's fallacies, making passing references to Letby but without committing himself as to her guilt or innocence.

The argument for prosecuting Letby at all was probably contaminated with the prosecutors fallacy. From what I've seen of evidence and analysis published in the US, and which it is currently illegal to publish in the UK, there is reasonable statistical doubt about whether any crime occurred at all.

The prosecution seems to have learnt lessons from previous "prosecutor's fallacy" cases such as that of Sally Clarke. Unfortunately they were the wrong lessons. The lesson I would have wanted them to learn was that juries should be presented with relevant and correctly calculated probabilities. The lesson they actually learnt was that it is best not to present any numerical probabilities at all. Instead they should cover their backs by simply setting out a supposedly suspicious sequence of events and saying, effectively, "Coincidence? I think not."

In the Clarke case, numerical probabilities set out by supposed expert Roy Meadows may well have been pivotal in her original conviction. But they were also pivotal in her acquittal because genuine experts (such as that curly haired bloke) were able to point out that he had calculated his probability wrongly and - much more importantly - that it was the wrong probability to calculate.

As someone or other said, if we express ourselves clearly, we risk being found out. At least Meadows set out his argument with sufficient clarity for it to be debunked. Vague handwaving about coincidence is harder to get at.

[IvanV]

Odd. The BBC are now reporting the same; whereas earlier today their coverage emphasized the judges words in telling the jury that this case must be decided on the evidence to be presented alone and that they must forget and not be swayed by anything that they may have heard about LL in the past.

I was a bit surprised by this, so I did a little search. You are allowed to mention someone's conviction as evidence to the extent it demonstrates the fact that they did the thing they were found guilty of.

So the argument would run, this child was murdered, this proven murderer was caring for her, what's the likelihood someone else murdered the child?

But, as Snoozey mentions, the real question is whether any of them were murdered at all. I have read articles suggesting that there are serious statistical questions over this.

Prosecutors prosecute when they have a good chance of a conviction. I find this a terrible criterion. They really ought to make an objective assessment of guilt, and at the very least refrain from prosecuting when an objective assessment is that there isn't a high probability of guilt. Often the likelihood in these cases is that someone else did it. Here, the potential likelihood is that no one was murdered.

The common idea, often seen as common sense, of "put it in front of a jury and let them decide", is completely misconceived. Because, as Snoozey says, judges and juries aren't very good at Bayesian probability. That ought to be done by the prosecutor in the prosecuting decision, and we ought to be able to trust them to do it properly. But their criteria militate against it. And the unfortunate lesson that prosecutors have learned from the prosecutor's fallacy is to present information in vague and tendentious ways that frustrates a good Bayesian assessment. Because the worst thing that can happen to a prosecutor is not to put the wrong person in prison, but to be detected having put the wrong person in prison.

[Sciolus]

Because the worst thing that can happen to a prosecutor is not to put the wrong person in prison, but to be detected having put the wrong person in prison.

Fortunately that takes at least 20 years in the UK, plenty of time to move on.

[raven]

But, as Snoozey mentions, the real question is whether any of them were murdered at all. I have read articles suggesting that there are serious statistical questions over this.

I haven't seen the statistics mentioned, but it did occur to me that even if a unit has a normal mortality rate, that doesn't exclude a savvy murderer picking terminal patients and shortening lives. Perhaps even with some dodgy justification about lessening suffering. Bit of a reach probably, but might just be possibile.

I read a little bit about the retrial. Prosecution case seems to be based on an eyewitness seeing something suspiscious, her being where she wasn't supposed to be while tubes got dislodged and alarms not going off. You'd need to know how often that stuff happens to know if that was odd or not. Maybe alarms failing and babies pulling their own tubes out is a routine occurence.

It's 'only' an attempted muder charge, though -- even though the baby did subsequently die, it was later and at another hospital, and they're not claiming that her actions caused the death.

[Woodchopper]

But, as Snoozey mentions, the real question is whether any of them were murdered at all. I have read articles suggesting that there are serious statistical questions over this.

As far as I remember there is forensic evidence that two babies were killed by being injected with insulin. Letby accepted in court that they had been murdered, but she asserted that she wasn't the murderer.

[dyqik]

But, as Snoozey mentions, the real question is whether any of them were murdered at all. I have read articles suggesting that there are serious statistical questions over this.

As far as I remember there is forensic evidence that two babies were killed by being injected with insulin. Letby accepted in court that they had been murdered, but she asserted that she wasn't the murderer.

The defense may have had to have accepted that, because statistical arguments aren't easily permissible, and they couldn't otherwise challenge it.

I believe that much of the forensic evidence is disputed by several other experts who the defense did not engage. I don't know about this specific evidence though.

[snoozeofreason]

But, as Snoozey mentions, the real question is whether any of them were murdered at all. I have read articles suggesting that there are serious statistical questions over this.

As far as I remember there is forensic evidence that two babies were killed by being injected with insulin. Letby accepted in court that they had been murdered, but she asserted that she wasn't the murderer.

The defense may have had to have accepted that, because statistical arguments aren't easily permissible, and they couldn't otherwise challenge it.

I believe that much of the forensic evidence is disputed by several other experts who the defense did not engage. I don't know about this specific evidence though.

The children that Woodchopper is talking about are the ones referred to in court as Child F and Child L. Letby was accused of administering insulin to them, but neither child died. In fact, so far as I am aware no one even noticed anything out of the ordinary at the time of the alleged administration, despite the fact that the alleged dosage seems to have been very high. Letby was accused, and found guilty of, attempted murder in these two cases. The children involved are, one hopes, currently at primary school and blithely unaware that they are supposed to have had a miraculous escape from death.

There is an additional complication in that Letby was off shift at the time of the supposed administration of insulin to Child F. The prosecution case was that she had contaminated a bag with insulin in advance and knew, in some unspecified way, that it would be used on Child F despite the fact that bags were in fact used as needed.

The defence, bizarrely, accepted the fact that the bags had been contaminated despite the obviously inconvenient survival of the supposed victims, and the fact that the bags had been disposed of and were not available for further analysis.

[Woodchopper]

As far as I remember there is forensic evidence that two babies were killed by being injected with insulin. Letby accepted in court that they had been murdered, but she asserted that she wasn't the murderer.

The defense may have had to have accepted that, because statistical arguments aren't easily permissible, and they couldn't otherwise challenge it.

I believe that much of the forensic evidence is disputed by several other experts who the defense did not engage. I don't know about this specific evidence though.

The children that Woodchopper is talking about are the ones referred to in court as Child F and Child L. Letby was accused of administering insulin to them, but neither child died. In fact, so far as I am aware no one even noticed anything out of the ordinary at the time of the alleged administration, despite the fact that the alleged dosage seems to have been very high. Letby was accused, and found guilty of, attempted murder in these two cases. The children involved are, one hopes, currently at primary school and blithely unaware that they are supposed to have had a miraculous escape from death.

There is an additional complication in that Letby was off shift at the time of the supposed administration of insulin to Child F. The prosecution case was that she had contaminated a bag with insulin in advance and knew, in some unspecified way, that it would be used on Child F despite the fact that bags were in fact used as needed.

The defence, bizarrely, accepted the fact that the bags had been contaminated despite the obviously inconvenient survival of the supposed victims, and the fact that the bags had been disposed of and were not available for further analysis.

Thanks for the clarification. Looks like I mixed up being poisoned with being murdered.

I looked up the situation on Child F. It doesn't seem correct that no one noticed anything out of the ordinary at the time. Instead they were concerned about a very low blood sugar level and high heart rate. https://www.chesterstandard.co.uk/news/ ... g-insulin/

[Woodchopper]

But, as Snoozey mentions, the real question is whether any of them were murdered at all. I have read articles suggesting that there are serious statistical questions over this.

As far as I remember there is forensic evidence that two babies were killed by being injected with insulin. Letby accepted in court that they had been murdered, but she asserted that she wasn't the murderer.

The defense may have had to have accepted that, because statistical arguments aren't easily permissible, and they couldn't otherwise challenge it.

This is as reported in The Guardian

Nick Johnson KC, cross-examining Letby for a second day, asked her if she agreed that “someone” had “unlawfully” given Child F and Child L insulin. She agreed, saying that the feeding bags must have been tampered with by either someone on the unit or before the bags arrived on the ward.

“Insulin has been added by somebody – how or who I can’t comment on, only that it wasn’t me,” she said. “I don’t believe that any member of staff on the unit would make a mistake and give insulin.”

https://www.theguardian.com/uk-news/202 ... ourt-hears

I believe that much of the forensic evidence is disputed by several other experts who the defense did not engage. I don't know about this specific evidence though.

That's interesting, but in terms of the trial, if the accused agrees that the babies were poisoned then at least with those two babies the statistical evidence is irrelevant. It comes down to whether the prosecution can persuade the jury that Letby was the poisoner.

Of course its possible that Letby's defence was incompetent.

[snoozeofreason]

Thanks for the clarification. Looks like I mixed up being poisoned with being murdered.

I looked up the situation on Child F. It doesn't seem correct that no one noticed anything out of the ordinary at the time. Instead they were concerned about a very low blood sugar level and high heart rate. https://www.chesterstandard.co.uk/news/ ... g-insulin/

Thanks for that. I was taking my information from the BBC report at the time that didn't mention the low blood sugar. However it did say this

Mr Myers [defence barrister] said the defence cannot dispute the blood test results as the samples have long since been disposed of, but there was nothing in fact to say Ms Letby injected the boy with insulin.

"The evidence has demonstrated that there have been bag changes when Ms Letby wasn't present," he said, before reminding the jury that Child F's bag, which was meant to run for 48 hours, was changed due to cannula issues after Ms Letby had finished her shift.

He said the evidence showed that a stock maintenance bag was subsequently hung by two other nurses, which the prosecution has claimed was also contaminated with insulin.

Ms Letby would need a "Nostradamus-like" ability to foresee Child F's first bag would need changing and then know which "random" stock bag would be chosen by her colleagues, he said.

Prof Hindmarsh, the expert quoted in your article, accepted that, if insulin was the cause of the low blood sugar, Letby would have had to have contaminated both bags to “more or less to the same degree," despite having no way of knowing that a second bag would be needed, or which bag would be used, and not being on shift at the time of the bag change.

[dyqik]

As far as I remember there is forensic evidence that two babies were killed by being injected with insulin. Letby accepted in court that they had been murdered, but she asserted that she wasn't the murderer.

The defense may have had to have accepted that, because statistical arguments aren't easily permissible, and they couldn't otherwise challenge it.

This is as reported in The Guardian

Nick Johnson KC, cross-examining Letby for a second day, asked her if she agreed that “someone” had “unlawfully” given Child F and Child L insulin. She agreed, saying that the feeding bags must have been tampered with by either someone on the unit or before the bags arrived on the ward.

“Insulin has been added by somebody – how or who I can’t comment on, only that it wasn’t me,” she said. “I don’t believe that any member of staff on the unit would make a mistake and give insulin.”

https://www.theguardian.com/uk-news/202 ... ourt-hears

I believe that much of the forensic evidence is disputed by several other experts who the defense did not engage. I don't know about this specific evidence though.

That's interesting, but in terms of the trial, if the accused agrees that the babies were poisoned then at least with those two babies the statistical evidence is irrelevant. It comes down to whether the prosecution can persuade the jury that Letby was the poisoner.

Of course its possible that Letby's defence was incompetent.

I think they were prevented from bringing some evidence as a matter of law, chose not to dispute some claims to avoid complicating things, and they really should have called an expert witness for the defence to dispute many of the assertions of the prosecution expert.

[IvanV]

I think they were prevented from bringing some evidence as a matter of law, chose not to dispute some claims to avoid complicating things, and they really should have called an expert witness for the defence to dispute many of the assertions of the prosecution expert.

What is the practical reality of trying to finance, first sufficient expertise to realise that such would be helpful, and then executing such a defence, as a criminal defendant of modest means in England these days?

[snoozeofreason]

I think they were prevented from bringing some evidence as a matter of law, chose not to dispute some claims to avoid complicating things, and they really should have called an expert witness for the defence to dispute many of the assertions of the prosecution expert.

I think that, at the original trial, it was mainly a matter of the defence choosing not to bring evidence. But once that choice has been made, the evidence becomes inadmissible at appeal.

[IvanV]

I think they were prevented from bringing some evidence as a matter of law, chose not to dispute some claims to avoid complicating things, and they really should have called an expert witness for the defence to dispute many of the assertions of the prosecution expert.

I think that, at the original trial, it was mainly a matter of the defence choosing not to bring evidence. But once that choice has been made, the evidence becomes inadmissible at appeal.

Is there any sense those might have been good choices? Or the lawyer just couldn't be bothered to go to the effort?

[snoozeofreason]

I think they were prevented from bringing some evidence as a matter of law, chose not to dispute some claims to avoid complicating things, and they really should have called an expert witness for the defence to dispute many of the assertions of the prosecution expert.

I think that, at the original trial, it was mainly a matter of the defence choosing not to bring evidence. But once that choice has been made, the evidence becomes inadmissible at appeal.

Is there any sense those might have been good choices? Or the lawyer just couldn't be bothered to go to the effort?

It's not clear why the defence didn't introduce more expert evidence. The Royal Statistical Society had produced a report on medical “murder” cases before the trial started and I get the impression that Richard Gill, one of the authors and also a participant in the strikingly similar case of Lucia De Berk, was keen to testify. It could be that the defence thought that the jury would not understand statistical analysis of the probability of "coincidences", or it could have been that they didn't understand it themselves. It might also be that they were put off by Gill, who seems to have a somewhat prickly personality, and felt that he, in particular, would not have made a good witness. There were also experts willing to question the medical evidence, but who were not called for some reason or other (in the end the only expert witness called by the defence was a plumber!).

Whatever the defence's strategic thinking was, it obviously didn't pan out well because Letby is now serving a whole life sentence. Although whether any other strategy would have worked better is unknowable.

It seems that, at her retrial, she is being represented by the same barrister as for the original trial. That may not be a wise option but, as you point out, she may not have that many options, and may not have the expertise to choose between them.

[discovolante]

Pure speculation and I could be entirely wrong, but assuming her case is funded by legal aid, it might be tricky to get funding for a new barrister because of the costs involved in them getting up to speed with the case. But I dunno, maybe that kind of thing is less of a big issue in cases as large and already costly as this one.

ETA it's also the solicitor who instructs the barrister, not the client. A private paying client may be able to exert some influence but again she may have less under legal aid funding.

[bob sterman]

Some other odd choices were made by the defence. I understand that they agreed (as an "agreed fact") that exogenous insulin must have been administered to one of the babies via parental nutrition and IV bags (as the only explanation for very high insulin levels in a particular blood sample). So the question just became who put it there (and why?).

Now maybe that is the most compelling explanation - but I would have thought a defence team would have challenged this? There are other plausible explanations they could have put forward (e.g. lab test error). Not sure why they didn't? Given the threshold here is "beyond reasonable doubt" there may have been a chance to argue reasonable doubt on this issue.

[snoozeofreason]

Some other odd choices were made by the defence. I understand that they agreed (as an "agreed fact") that exogenous insulin must have been administered to one of the babies via parental nutrition and IV bags (as the only explanation for very high insulin levels in a particular blood sample). So the question just became who put it there (and why?).

Now maybe that is the most compelling explanation - but I would have thought a defence team would have challenged this? There are other plausible explanations they could have put forward (e.g. lab test error). Not sure why they didn't? Given the threshold here is "beyond reasonable doubt" there may have been a chance to argue reasonable doubt on this issue.

So far as I can make out (from the New Yorker article by Rachel Aviv, which dyqik mentioned, and also from Richard Gill's blog), the insulin assay was performed at Royal Liverpool University Hospital using a test that was not appropriate for the detection of factitious (i.e. intentionally induced) hypoglycaemia. Gill claims that the description of the test can be found at http://pathlabs.rlbuht.nhs.uk/insulin.pdf. I can't be certain that this document does indeed describe the particular test used, but it is from RLUH and does contain the following warning (printed in red).

Please note that the insulin assay performed at RLUH is not suitable for the investigation of factitious hypoglycaemia. If exogenous insulin administration is suspected as the cause of hypoglycaemia, please inform the laboratory so that the sample can be referred externally for analysis.

Aviv claims that a biochemist at RLUH had called the hospital where Letby worked and explicitly recommended that the sample be verified by a more specialised lab.

Of course, both Gill and Aviv have taken positions on this case, so you have to allow for a certain amount of confirmation bias. But I don't think either can be dismissed as conspiracy theorists. Gill is an eminent statistician and Aviv is a serious journalist who seems to have a decent track record, and their arguments seem to make sense (at least to me, although I am sure I have my own biases).

As I said, it is unclear why the defence did not push back on the medical evidence a bit harder. Aviv speculates that it might be because, although the defence would have been able to produce experts who would testify that the evidence did not prove Letby's guilt, the prosecution's first question to such an expert would have been "Well does it prove her innocence?" to which the answer would be "No". Of course that shouldn't be an obstacle if the jury understands the standard of proof that they are supposed to apply, but the defence might not have been confident that they would.

[dyqik]

Any question on those lines would likely be objected to and struck off the record by a competent defense and judge.