I should have been braver
On the Letby case and its flaws. And why I should have spoken publicly more, earlier.
A couple of months ago I wrote a piece arguing that intellectual courage is the scarcest source. It was motivated not merely by the fact that I notice intellectual cowardice in other people and that bothers me but, perhaps more crucially, because I see it in myself more often than I would like. Since I consider myself “an intellectual”, broadly speaking, this is not great. I am not doing anything practical like landing planes or doing cardiac bypass surgeries, so the least I can do to be more than someone that indulges in nice intellectual pursuits and not be a leech on society is defend truth. And there’s an important recent instance where I did not do enough of it, despite being well placed to do so. That is the case of Lucy Letby, a former neonatal nurse in the UK, who was convicted in 2023 of murdering seven babies and attempting to murder six others at the Countess of Chester Hospital between 2015 and 2016.
I watched the Panorama documentary about her right after she was convicted in August 2023. I still remember the day: I was tired from work and did not feel like doing anything intellectually stimulating like reading a book, so I turned to watching TV with my boyfriend. Despite our usual disinterest in true crime, we were drawn in by a morbid fascination: what could drive a young woman to allegedly murder multiple infants in her care? This curiosity was in large part instigated by the constant stream of sensationalistic headlines that were all over the place in the British press.
Initially, we expected the documentary to present some strong evidence of guilt. However, as we watched, we realized no such thing was coming. The Panorama documentary, while clearly designed to be sensational and paint her in a negative light, failed to deliver the damning proof we anticipated. The case seemed to hinge primarily on statistical analyses and inconclusive circumstantial evidence. The medical professionals interviewed, who were among the first to accuse her at the hospital, did not inspire a lot of confidence. Their handling of scientific and statistical concepts appeared superficial, and their reasoning seemed flawed. As for the behavioural evidence that was presented as damning and proving she was a psychopath, I saw it as neither here not there: if you started with the premise she was a murderer, it looked weird. If you started with more charitable assumptions, it looked exactly like how an innocent and confused woman who was accused of murder would have behaved.
Nevertheless, it was a fact that I had not been at the trial. I had not gone through all the evidence brought during it (although the question remained: how could a documentary clearly made to portray her in a bad light and constantly announcing the strength of evidence, only come up with such weak stuff?) So I started reading on my own, digging into what I could find about the case from other news source portraying her in a negative light and trying to separate facts from opinion. My doubts were not at all quelled by this, quite the opposite.
Timidly, I tweeted my concerns about the safety of the conviction, making it very clear I was not making a judgement on her innocence, but the safety of the conviction. I must say I never got so much hate on social media as I did for this. People accusing me of “loving baby murder” and being supportive of serial killers and so on. This did not just awaken my “I am socially shunned” ancestral instinct, but perhaps even more importantly, raised my internal doubts: what if I was wrong? What if I was missing something the jury was aware of? What if I was defending a baby murderer? These doubts made it easier for me to shut up and stop talking about this in public. But inside me I knew this was not about her innocence per se, but about the justice system more broadly. The scientific evidence seemed shaky. If such evidence is allowed in court, it is not simply wrong, but it does mean that someone will be wrongfully convicted at some point. These simple facts did not cease to be true and they gnawed at my conscience.
Fast forward to September 2024, more than a year later. A comprehensive New Yorker article, which interviewed a bunch of experts and went through the case in much detail, has been published (note that in UK it’s still not accessible so I cannot link to it.) A bunch of experts have publicly started to raise concerns on the case. On statistics: Professor Jane Hutton and Professor Richard Gill (who was on the case very early on and whose blogs I had read in 2023.) On the credibility of the scientific evidence: top immunologist Carola Vinuesa from the Crick Institute, who has been involved in overturning a wrongful conviction based on flawed science in Australia. The list could go on, with top neonatologists and other medics. The critics also come from the legal profession and from the forensic psychology side (the latter mainly questioning the validity of the behavioural evidence.)
Some of my doubts
This is not a post that is aimed at carefully going through the case. Others have done it much better than me. But I’ll briefly go through why I find the pillars of the case to be, to the best of my knowledge, flawed:
The statistical evidence: The simple, provable fact in all this was a spike in deaths at the Counter of Chester Hospital between 2015 and 2016. At the time, an investigation was carried out and no foul play was suspected. The autopsy reports for all the babies did not indicate suspicion of murder. Instead, the hospital was found to have a bunch of issues, related to overworked staff, lack of organization etc. As a result, it was downgraded. Subsequently though, cases were reanalyzed and alternative explanations that rely on Letby having killed many of the babies that were part of the spike were found by the prosecution’s witnesses and presented to the jury.
I recommend this website that explains how one could be misled by the core plot presented to the jury, a plot that I believe really instilled a strong “guilty prior” in the jury’s mind. There are many issues with this plot, including the fact that the deaths seem to have been cherry picked and that there are other babies that were part of the same spike in deaths that Letby was not in charge of and which were left out. There was no base rate calculation in terms of the probability of observing a spike in deaths in a poorly run hospital just by chance. For example, John O'Quigley, honorary professor in the Department of Statistical Science at University College London (UCL), argues that the chance of such a spike is quite high in a longer article that I recommend: “There was in fact no evidence of anything extraordinary having taken place and that the belief in an inexplicable spike is based on an elementary, and very common statistical misconception.“The inexplicable spike is in fact perfectly explicable.”
The medical evidence: The most direct piece of medical evidence there is in this case is two high readings of insulin observed in two babies after the prosecution’s experts went through various medical records of babies treated at the hospital. In 2019, they found 2 abnormally high readings of insulin dating from 2015, which they argued were proof that someone attempted the babies with insulin (the babies survived.)
The only problem is that the assay that was used to measure insulin was not designed to show exogenous insulin administration, something that is actually written on the test itself.
Professor O’Rahilly from Cambridge and Carola Vinuesa explain here and here and here the issues with using such tests. In a Telegraph article, Prof Joseph Wolfsdorf, a specialist in child hypoglycemia at Harvard Medical school, said the results “made no sense”:
“When serum insulin is extremely high (caused by injection), the beta cells of the pancreas stop secreting insulin (and C-peptide). One expects, therefore, to have extremely low or unmeasurable serum C-peptide concentration. In one of the babies, serum insulin was extremely high and yet C-peptide was within the reference range, i.e. not extremely low or unmeasurable. Therefore, one has to be highly suspicious that the high insulin level was caused by exogenous administration. If the sample was obtained at the time plasma glucose was low (hypoglycemia), a C-peptide concentration within the reference range makes no sense.”
The insulin cases involving Children 'F' and 'L' were pivotal in Lucy Letby's trial. The defense's acceptance of this evidence as proving someone poisoned babies may have been a critical mistake, potentially influencing the jury's decision. The judge's instruction allowed for conviction based on harm to at least one baby, possibly leading to a domino effect from the insulin cases. Two of the prosecution’s medical expert witnesses described this evidence as the "smoking gun". The other medical evidence is much more indirect and has been contested (this article has a good summary.)
The behavioural evidence: Much has been made of the so-called “confession note.” Among other scribblings, Letby wrote on a piece of paper: “I am evil, I did this”, which was presented as a confession. Except that on the same piece of paper she also wrote: Today is your birthday but you are not here. I am so sorry for that. I'm sorry that you couldn't have the chance at life you should have and for the pain that your parents must experience everyday
We tried our best and it wasn’t enough” and “I didn’t do this”
To me this sounds like it could very well be what a diligent and well-intentioned nurse would write after having several babies die in her care and having being accused of her colleagues of murdering them. What’s more, these notes were written on the advice of the hospital counsellor, who encouraged letby to write down her worst thoughts. David Wilson, a professor of criminology at Birmingham City University, who specialises in serial killers, says in a Guardian article, that in his view the so-called confession notes were “meaningless” and had no value as evidence, particularly if they had been written as part of counselling. “Many people will say things when they are under stress and feeling bereft, that seem to imply one thing but mean nothing at all, other than reflecting the underlying stress.”
It’s about a principle
It is a triumph of our civilization that we have been able to fight tribal instincts and come up with moral innovations like “presumption of innocence” and “right to a fair trial.” In the UK, they can be traced to the 1215 Magna Carta which, while not explicitly mentioning presumption of innocence, established principles of due process that laid groundwork for these concept.
The possibility that an innocent person might be in jail for life is bad enough, but the case is about more than just this. It’s about principles, like the right to a fair trial. While I do not know if Lucy Letby is innocent, I do not think a trial can be fair if scientific evidence is misinterpreted, as was the case for the insulin data and a lot of others. This is a principle that is worth defending in isolation from anything else and I should have done it more vigorously before, regardless of other doubts I had. I do not know how much of a difference me having continued to be vocal about this would have made. But the fact is that I did insist to a friend to read the New Yorker article and this has directly led to David Davis getting involved. My friend told me he wishes he knew about this sooner.
One does not not need full access to transcripts or being certain of someone’s innocence to defend such a principle. It is this kind of cold, complete decoupling that needs to be promoted and not emotional outbursts about '“defending baby killers.” If the justice system is not scrutinized for the level of rigour it treats scientific evidence with, wrongful convictions will happen.
Great post. I don't know the medical details of the case, but I agree that convicting someone based partly on statistics like this is seems really bad. Weird stuff happens in medicine all the time and usually it isn't intentional wrongdoing. Statistical anomalies can be used to alert us and guide further investigation, but we ultimately need to find a real and convincing cause.
Also, generally speaking—and again I don't know all the details of the case—using something someone wrote in the context of a therapy exercise as evidence against them is shockingly inappropriate. The fact that the counselor was employed by the hospital, thus creating even more conflicts of interest, is even crazier. I'm surprised that kind of thing would be admitted in court in the first place.
While statistical evidence, used correctly, can be very strong I fear it is particularly tricky for the adversarial method to handle. Part of it is that all but the most wealthy defendents will struggle to get the same quality experts and the judge won't understand how to evaluate the reliability but the issue goes deeper.
A key aspect of getting useful statistical evidence is unbiased samples but a prosecutor will always investigate the cases they see as suspicious. This creates a fundamental problem that by the time you summon a statistician to perform an analysis the investigation has already selected a deeply biased sample (the suspicious cases) and information about how those cases were selected will have been lost or it's importance not understood.
So the statistician working for the prosecution comes in and gets asked to evaluate how likely it is that these cases occured by chance but that's an impossible task without knowing how those cases were selected. Worst case, the investigator identifies some propoerty discovered after the fact not realizing it matters. Best case the statistician has to just make some vaguely plausible assumptions about the messy process of gut and testing that selected those cases.
I'd go so far as to suggest that in any case the prosecution wishes to introduce statistical evidence/arguments the court should be notified when charges are filed and -- at the defense's consent -- a special master with statistical and forensics expertise appointed to reexamine or reinvestigate the case.