Experts warn Letby inquiry of misreading stats to explain 'suspicious' deaths
thejusticegap.com
thejusticegap.com
It is not unexpected of depressed people to blame themselves for bad things that happen around them, I think. …For multiple reasons, (1.) because you really do hate yourself that much, and might view those tragic events as also personal failures, and (2.) because for the suicidally depressed, it can be a way to manufacture a reason to believe why you deserve to be hated or physically injured.
IIRC the insulin levels or something were rather suspicious, but in general I sorta can't and shouldn't comment on the validity of rest of the case as a bystander. But when I read about that note in particular, my read on it was that somebody who actually did kill them versus somebody who just feels really bad about the situation and wants to vent in their own private space would write down the exact same thing in the exact same way with equivalent plausibility. She was a nurse, and babies died in her care; it is entirely plausible that, especially if she was already depressed, she would view that as being her fault and find some way to internalize the tragedy even if she didn't actually do anything to cause it.
And, IIRC, Letby did basically say that that was what she meant by the note— So, in sum, an innocent person who had been blaming themselves and a murderer trying to lie their way out of justice IMO would have written down the exact same thing with equal plausibility, and then tried to explain it the exact same way, which means that IMO that datapoint is basically useless for determining her guilt.
There was other evidence presented too, of course. I'm glad to hear that the conviction didn't rest on just that note, because given the situation, we don't actually know in isolation why she wrote that or what she meant by it— Most murderers try to avoid leaving confessions lying around to be used as evidence too.
I have no idea how much that affected the jury but it's just as important to remember that false confessions are a thing, and a common thing at that [2], and they have led to many false convictions. The wikipedia article I cite has a _very_ long list from many places in the world.
At the end of the day, innocent people can confess and guilty people can maintain their innocence. A confession is no more reason to accept guilt than protestations of innocence are reason to assume innocence.
Again, I don't know how much the jury were influenced by the note, but, anecdotally, the press has certainly held it up as some sort of evidence, and I've seen people on the internet take it as such, also. See for example a parent comment.
_____________
[1] For example, here:
https://www.liverpoolecho.co.uk/news/liverpool-news/lucy-let...
> the suspicious deaths all lined up perfectly with Lucy Letby's shifts
Once a nurse has been around the ward for long enough to have seen several deaths, you can produce a table showing a set of baby deaths correlating to any nurse's shifts [1]. The ward was for seriously ill newborn babies, so baby deaths were unfortunately common, and each nurse would regularly be on shifts where one would die.
As you say, these deaths were considered "suspicious", but some (e.g. [2]) claim otherwise, saying key tests that would rule the accused method of murder in or out were not done at the time of death.
> she was often caught interacting with those babies when she had been instructed not to
(Again, playing devil's advocate) the job selects for nurses who (apart from Letby, it seems) deeply care for young babies. Not a smoking gun.
> when she was taken off the ward the deaths instantly stopped
This has been contested. As far I can tell the hospital's death rate statistics disagree with the Office for National Statistics. The latter's data appears to show the deaths peaked after Letby was taken off the ward [3] (note, I haven't checked the underlying source mentioned in the tweet).
Replies to some comments below yours:
> And she took home documents from several of those patients, which she'd have no ordinary reason to have fished out of the trash.
Apparently this is quite common for nurses working on this and other intensive care wards. Nurses just off duty are often phoned by the nurses who took over the shift, asking for details on particular babies' treatment or behaviours that day. Having files on the babies to hand would let Letby give exact details without having to remember or misremember them.
> And the little old thing that she wrote "I killed them on purpose" on a post-it-note at her house.
This looks bad, yes, but again not a smoking gun. I think I read this happened after she was already taken off the ward or had somehow found out she was under investigation for causing baby deaths. I could believe a note like that could be written by someone on the verge of a breakdown or under extreme stress. The "I killed them" bit I could believe someone could write if they regret not having spotted something earlier that someone more experienced might have, especially if they have imposter syndrome.
As the blog makes clear, it's not really about whether Letby is innocent or guilty, it's about whether she got a fair trial. It seems many statistics were misused in her case, circumstantial evidence was used as smoking guns, and some evidence was not fully addressed [4]; basically enough to question the verdict.
[1] https://twitter.com/profnfenton/status/1692837686166368660
[2] https://gill1109.com/2023/05/24/the-lucy-letby-case/
[3] https://twitter.com/LawHealthTech/status/1692917506078376019
[4] Apparently during the trial it was mentioned that sewage pipes crossed the ceiling and backflow sometimes came into the sinks. Definitely not a good mix to have in a ward with very ill babies. Hopefully infection spread from bad plumbing was fully ruled out as causes of death during the trial. Some critics don't seem to think it was.
The C-peptide level in Child F's discharge letter have been called "the golden thread" of the case by the leading Detective Sergeant of the investigation Paul Hughes.
The deputy senior investigating officer DCI Nicola Evans called it a "real milestone" in the investigation.
The Guardian called it "a smoking gun in this medically complex case".
Dr Stephen Breary, the hospital's neonatal unit head consultant, said about its discovery: "If there was a iota of doubt, it was removed then."
I think it is fair to say that if this letter would not have been found, after already a year of rather fruitless investigation, there would quite likely not have been a case.
Yet, the very same Dr Breary, who was also under pressure because of the apparent rise of deaths in his department, came up with this document more than two and a half years after the child was released and just a couple of days before Letby was about to return to his department against his will.
I'm not saying Dr Breary is under suspicion or Letby is innocent, but think the discharge letter should have seen more scrutiny.
My point is that I don't think there is anything in this case that really holds water if you seriously probe it and if you pile up flawed arguments - no matter how many - it will not bring you any closer to the truth. If I can wagger a prediction, and I already said this when Letby was convicted: I see a successful appeal and this verdict be shredded. Not because I think Letby is innocent, but because a working legal system should not put anyone behind bars for life with such a weak case.
The "smoking gun" turned up inconveniently late, in February of the following year, when he'd been asked to go through the records of the kids who survived.
That's what you say. I neither said nor insinuated any of that.
The case is supported by two main pillars. One is the idea that the stats support suspicions clusters of deaths. I expected that to be challenged back in August and we see it happen now.
The second pillar is the "smoking gun" of Child F's raised peptide levels. In my opinion this pillar is far from being strong enough to support this case alone. If we don't see any additional props appear, I expect it to crumble in the appeal.
"your timelines are all wrong."
- August 2015: Child F blood sample [2]
- February 2017: Police investigation launched
"On the night before Valentine’s Day in February 2018, nearly a year after police launched their investigation" [1]
- July 2016: Letby removed from frontline care [1]
- February 2018: Smoking gun found [1]
- May 2017: Letby scheduled to return [1]
[1] https://www.theguardian.com/uk-news/2023/aug/19/trust-me-im-...
[2] https://www.theguardian.com/uk-news/ng-interactive/2023/aug/...
So unless you're going to insist on a special calendar in which February 2018 occured two days before March 2017, you're basically confirming that I was correct to point out that your original statement that "Dr Breary, who was also under pressure because of the apparent rise of deaths in his department, came up with this document more than two and a half years after the child was released and just a couple of days before Letby was about to return to his department against his will." got the timeline all wrong.
What I initially misread was "Letby had been just six days from returning to work on the neonatal unit at this point." from [1]. This sentence refers to the 27 April 2017 two paragraphs before and not to the smoking gun discovery in the following paragraph. When I compiled the timeline I totally focussed on the months. 27 April plus six days would be May. But 2017, not 2018.
I stand corrected and thank you for pointing it out.
From https://www.google.com/search?q=c-peptide+false+positive
https://labs.selfdecode.com/blog/c-peptide/
> However, while healthy kidneys are efficient at breaking down C-peptide, in people with impaired kidney function and kidney disease, blood C-peptide levels will falsely increase and urine levels will falsely decrease.
I'm going to go out on a limb and say that the medical professionals who determined the evidence was conclusive had sufficient expertise not be relying on Google searches to interpret blood peptide readings..
The hospital's expert witness supposedly said the line had been cut. Dig deeper: No, the witness was asked if the damage to the line was consistent with being cut. Yes, it was, but that doesn't mean it wasn't also consistent with simply breaking.
In the end it's pretty clear the hospital tried to throw the nurse under the bus to cover up the fact that it was their shoddy supplies that killed the baby. No idea on what sort of settlement the parents got, no details on what the nurse got but it was substantial.
1) understand that there are two types of ambulance (generally) - basic life support and advanced life support - the latter carries invasive care options, and narcotics, the former does not. They called us for an interfaculty transport approximately 200 miles. Patient was given pain meds before transport. Late in transport, EMTs on a BLS ambulance (no narcotics physically on board the ambulance) noted the patient was in profound respiratory depression and treated that - arriving at the hospital, it was discovered that the patient had been given vastly too much narcotics. Originating hospital insisted that it must have been given in transport. Kept insisting all the way to a DOH hearing. The EMT and the ambulance company were cleared and the hospital and nurse faced repercussions for documentation, administration, cover-up issues, but they still had to have their day in "court" to get to that point.
2) A deputy cracked his neck, and unbeknownst to him, partially severed a blood vessel. He contacted EMS complaining of nausea and vomiting. Among other things, a stroke test was done (negative) and he was taken to the hospital. CTs were done, and the physician at the hospital was insistent that all was clear, and was about to discharge him with vertigo, while other deputies insisted something was seriously wrong, and getting worse with him. Another provider sent scans to a major trauma center who noted that he had indeed had a stroke. He was also now failing a stroke test. He was transferred and treated but irreversible damage had occurred needing 24/7 nursing care, and lawsuits were launched. He died a month or so later. The hospital tried to throw EMTs under the bus, saying that they failed to diagnose or tell the hospital that the patient was having a stroke. Even though at the hospital another initial stroke test was performed by a physician with a negative result. The hospital then said there were subtle signs that in hindsight the EMTs "should have recognized" (but apparently a physician should not have been expected to...). That case is still ongoing, but the ambulance company managed to extricate itself after settling for policy limits (which was farcical in itself).
Before trial the experts will go through all the discovery and try to come up with theories that refute the prosecutions idea. You essentially have infinite time to go through the discovery. But, you do not know in advance what the witnesses and opposing experts are going to testify to on the stand.
During trial things move very, very quickly and you don't have any time to really take into account everything the witnesses are saying, and if there are things that say needed some time to run a statistical analysis to see if they were true, well you'd be screwed because you don't have that time. So, essentially witnesses can say things that seem really bad to a lay jurist, but if you had time to run the math they would easily be explained as common and not statistically significant.
In a complex trial like this I think it would be fairly easy for it to go the wrong way.
The justice system is definitely designed to produce convictions. Acquittals are a rare bird.
This is only a piece of evidence if:
- it's not a common behavior among nurses in her unit.
- she did that specifically to the victims.
Otherwise it could be that this rule isn't applied in practice, or that she is just the kind of people who don't really listen to instructions if they think they should not, which albeit a questionable attitude in an hospital isn't the same thing as “she killed them”.
She should be in prison if guilty of these horrific crimes. However, sending an innocent person to prison for them is substantially worse than no one going to prison for them.
https://news.ycombinator.com/item?id=37814339
Those arguments, as presented to the jury, relied entirely on a spreadsheet of the nurses' duty roster. There's an image here (on the Daily Mail website):
https://i.dailymail.co.uk/1s/2023/08/18/22/74487747-12286051...
Looking at this spreadsheet the one thing that immediately jumps out to me is that this spreadsheet is not the entire data.
The spreadsheet shows only the days where Lucy Letby was on shift and there was a suspicious incident (i.e. when a baby died or collapsed).
It is very difficult to believe that there were no shifts were incidents occurred and Lucy Letby was not on shift, or times Lucy Letby was on shift but no incident occurred.
Obviously, without that data that is not included in the spreadsheet it is impossible to draw any conclusions about Lucy Letby's correlation with the incidents she was convicted of causing. That's pretty much what the RSS letter points out, that:
it is far from straightforward to draw conclusions from suspicious clusters of deaths in a hospital setting – it is a statistical challenge to distinguish event clusters that arise from criminal acts from those that arise coincidentally from other factors, even if the data in question was collected with rigour.
And the data in that spreadsheet was certainly not presented with any rigour. Rather, this spreadsheet is a textbook example of cherry-picking. It only shows the data that justifies the prosecutor's claim.
I have no way to know whether this was done consciously, in order to mislead, or it was merely the result of poor understand of statistics by police officers and prosecutors. I'm inclined to believe the latter was the case.
In any case, _if_ the jurors were convinced by that spreadsheet of Lucy Letby's guilt, to whatever extent (weighing it in view of other evidence), then they did so without having enough information.
I agree. It's very strange that there was not at least one case when she was not there. Death are very clear, but it's easy to use different criteria to classify an weird case as suspicious.
I'd expect more cluster of nurses. There are a few that work (almost?) in all the cases during the day, but the nights are more random. Some nurses appear only one. Which one had a full time job there and which where hired only a few times?
There is a strange diagonal that starts at Child D - Nurse H. How did they sort the nurses? Alphabetically? By some internal HR number? [It may be an artifact, because people is too good detecting patterns like diagonals. I was generating a similar random grid a few years ago, and I saw too many patterns that I suspected the random generator was wrong. But it was just me overfitting the data.]
So, there should be far fewer than one death per thousand of the newborns in the hospital. Since her activities took course over a year, it's quite possible that no unexplained deaths happened apart from the ones that she was involved with.
It's also possible there might be one death not caused by her that just happened to occur on her shift.
Further, as I understand it, those were babies that were being cared for in the Neonatal Intensive Care Unit and may have been vulnerable and less likely to survive to begin with.
For example, I remember the press noting that a baby who died was no bigger than a hand. The press reported this to tug at readers' heartstrings, but it also has the effect of reminding that a baby born so early is in a precarious situation and may well not make it.
In any case, the spreadsheet above doesn't give any of that information, either, so there is not enough information to know whether the deaths were suspicious in the first place. Looking at the spreadsheet, all we can say is that the only reason those incidents are associated with Lucy Letby is because Lucy Letby was on shift. Which is a tautology and not useful to make any decisions.
> The numbers of unexplained collapses were particularly abnormal: there had previously been only two or three deaths a year in the neonatal unit.
She was convicted for cases between June 2015 and June 2016, so I'd expect to see like 2 or 3 cases of natural deaths. Assuming she works there for 40 hours of the 168 hours of the week, it's like 1 coincidence and 1 or 2 when she was not there. Been there when 7 babies died is a LOT of bad luck.
Assuming a Poisson distribution with lamba=2.5/4, P(7)=7E-13 ...
There was corroborating evidence which lead to her conviction.
At least that is exactly what happened to Lucia de B [1] (which they mention in the article), which at first glance looks eerily similar.
Though unlike the Lucia de B case it doesn't seem likely these deaths could all have been natural. The evidence pointing towards Letby seems mostly probalistic in nature though, which should be treated carefully.
Chief among these was the duty roster spreadsheet showing staff shifts, through which the Crown Prosecution Service (CPS) was "able to show the jury that Letby was the one common denominator in the series of deaths and sudden collapses on the neonatal unit".
https://theweek.com/law/lucy-letby-and-the-importance-of-und...
That is totally statistical information - even though it's, shall we say, lay statistics and not what you'd expect to see from a statistician.
However if Lucia de Berk herself wishes people to use her full name then that is a very noble decision. The edit window has passed unfortunately, but I will use her full name in future.
However, the public perception around the time of the trial was substantially different. Especially early on, the media focused strongly on the fact that it "could not possibly be a coincidence". Reading the trial coverage I got a very strong feeling of déjà vu, seeing the coverage pretty much mirroring the Lucia de Berk case.
Considering the mountain of other evidence in the Letby case, I highly doubt it was a miscarriage of justice. But I definitely agree with the Statistical Society that just because something looks like a duck, it doesn't have to be a duck.
When people have studied these types of investigations, and has the evidence re-evaluated by unbiased investigators, the results are usually that nothing untoward occurred, or that deaths are better explained by systemic failures rather than the actions of specific individuals.
There’s a good reason why we don’t let police go on fishing expeditions. Medical investigations, frequently initiated by untrained administrators, and performed by individuals within a hospital with no formal training in performing unbiased investigations, nor given enough to perform proper investigations, simply don’t hold up to the standards we would expect from a criminal investigation.
None of this is to say that the Letby verdict is unsafe, but it wouldn’t be the first time a medical investigation with lots of evidence resulted in an awful miscarriage of justice.
The police have said that the duty roster spreadsheet shown to the jurors was an important piece of evidence used to convince the jurors [2].
The spreadsheet in question has been circulated in the media (I'm assuming, by the police) and it can be found in various places [3].
Looking at that spreadsheet there is something that immediately jumps out to me and that makes me think that at least that piece of "statistical" evidence was downright dodgy.
I am curious to know if anyone else on HN sees the same thing. Here's a direct link to the image on the Mail website:
https://i.dailymail.co.uk/1s/2023/08/18/22/74487747-12286051...
To be clear: I'm not trying to trap anyone, I'm just really curious to see what everyone else thinks _about this particular piece of evidence_.
____________
[1] https://www.bbc.co.uk/news/uk-england-merseyside-66823777
Summing up at the end of her trial, judge Mr Justice Goss told the jury it was a case in which the prosecution "substantially, but not wholly" relied on circumstantial evidence.
[2] https://theweek.com/law/lucy-letby-and-the-importance-of-und...
Chief among these was the duty roster spreadsheet showing staff shifts, through which the Crown Prosecution Service (CPS) was "able to show the jury that Letby was the one common denominator in the series of deaths and sudden collapses on the neonatal unit".
[3] https://www.dailymail.co.uk/news/article-12286051/Lucy-Letby...
See first image on top, titled: "Staff chart that shows Letby was present at every chilling incident".
They didn't include any examples of the parallel. This goes over some of the notes:
https://www.theguardian.com/uk-news/2022/oct/13/lucy-letby-p...
> On another, she wrote “I am a horrible evil person” and on another in capital letters: “I AM EVIL I DID THIS,” the jury was told.
> She also wrote on the note: “There are no words”, “I can’t breathe”, “kill myself”, “I haven’t done anything wrong”, “how can I get through it?” and “I am an awful person”. She had also written “No Hope” and “Panic Fear” and then circled in capital letters “HATE”.
Personally, I don't think any of this is strong evidence, since she contradicts herself a bit. But the notes more directly address the murders because she wrote them in response to the accusations. It isn't as vague as a diary entry mentioning a secret.
And as far as I know, they had more evidence in the form of testimony from experts suggesting the babies were killed by injecting them with air, but I'm sure there are reasons to be skeptical of that.
Mainly, I don't know how strongly any statistical "evidence" was presented in court, how much the jury depended on it in their guilty verdict, and how the UK court system works, so I don't have a strong opinion. But I also don't believe a not-guilty person would think to write anything like what she wrote in those notes.
One last thing I'm curious about and have never heard definitive information on is whether babies stopped dying when they took Letby out of the NICU or whatever they call it. I suppose that might still be accounted for in this statistical coincidence element the RSS and other statisticians are discussing.
The thing that did for her apparently were messages she had sent some of the parents, the times of which contradicted her testimony she had given the police and put her clearly in the frame when she had given them an alibi.
So I'm not sure the clusters of "unlikely" cases played much role either, other than alerting authorities to there potentially being foul play.
People generally do not read court transcripts, they read media headlines. It's basically a guarantee that some people are already convinced their statistically unlucky coworkers are also serial killers. It can easily lead to witch hunts.
One challenge for the defence was that there are few studies on injecting air in small babies for obvious reasons. Air embolisms are very painful, and such a study wouldn't pass ethics approval, let alone parental consent.
To clarify. Lucy Letby was accused of deliberately administering a high dose of insulin to two babies, in an attempt to murder them.
Those two babies were "Child F", and "Child L" for both of which she was charged , and convicted of attempted murder.
My source for this information is this article in "Nursing Times":
https://www.nursingtimes.net/news/children/breaking-lucy-let...
My point is that the judge awarded her a life sentence on each of the cases:
"on each of the seven offences of murder, and the seven offences of attempted murder, I sentence you to imprisonment for life."
Even if the other cases turn out to be unsafe, she would still be imprisoned for life.
There appears to be an interpretation here by some of a potentially unsafe verdict due to misreading of statistics in her conviction - which has certainly happened in the past in other cases including those mentioned.
The expert input here though is intended as advice for the inquiry into what went wrong at the places Letby worked to lessen the risk of it happening again.
That necessarily includes qualified and responsible analyses of statistics, which will also have the benefit of increasing protection and support of staff as well as improving safety generally.
What I remember from the book is entirely innocent people jailed because the wrong statistical question was asked. IIRC along the lines of two deaths, and the question asked was what is the chance of this happening? As opposed to, assuming the first death, what is the chance of the second? As opposed to, assuming two deaths, what is the chance of bad luck versus malice.
(I am not commenting on the Letby case per se. AIUI there is an appeal pending and UK law can be rather strict. Simply, we should not respond to possible horrible events by convicting other people who are innocent of statistical flukes, or misrepresented statistics.)
The problem with Letby is that a serial killer was able to hide because finding a rare event (a member of staff murdering multiple babies) is hard to do in the noise (the mish mash of data available).
Currently the English NHS is moving away from SIRI (Serious Incident Requiring Investigation) to PSIRF (Patient Safety Incident Response Framework) which focuses a lot more on "no blame" and "open culture", so it's important that NHS Trusts get the stats right to be able to detect and understand these very rare events.
I was going to say that this is the second time I see someone in this thread claim this, but I noticed the other comment is also from your account:
https://news.ycombinator.com/item?id=37814660
As I say in that other comment, statistics were used in Lucy Letby's trial, in the form of the staff duty roster spreadsheet that the prosecution showed to the jurors to convince them that Lucy Letby was a "common denominator" in all the suspicious incidents.
Depends on 'a lot', but I rather suspect there will be/ will have been fewer, now cops have to use cameras, but yes - a lot of miscarriages of justice has happened in the UK. And every other country.
It’s statistically improbably to get 7 sons, but I’m certain that throughout history that’s happened at least once.
If the answer is no, then a follow on question regarding whether or not factors associated with Letby could have resulted in the deaths rather than Letby herself. This could be a fellow worker who always worked in tandem with her etc.
Finding the appropriate set of data and applying statistical techniques for this is not trivial hence the expert mention.
No, the point being made is that a murderer was able to evade detection because her employing organisation did not understand the data they had, because they did not understand statistics.
This isn't a case of relying solely on data, rather the data bear out what everything else suggests.
Child F had a bloodwork which indicated that someone gave them a large amount of insulin.
Child D’s post mortem X-ray is said to show air injected into their blood vessels.
Child L also had insulin in their blood.
I’m not an expert so I can’t say for certain how strong these evidence are. Also my understanding that while these are evidence that a crime happened, they do not tie Lucy directly to the crime. The evidence for that is circumstatial, mainly that she was the only one on duty for all the collapses said to be suspicious.
>> Child L also had insulin in their blood.
Those were not post-mortems because these two children didn't die. Lucy Letby was accused of attempting to murder them, by injecting them with insulin.
See here, towards the end of the article, a summary of accusations and convictions:
https://www.nursingtimes.net/news/children/breaking-lucy-let...