Statistical issues in serial killer nurse cases

Suspicions about medical murder often arise due to a surprising or unexpected series of events, such as an unusual number of deaths among patients under the care of a particular professional.  The RSS has published a report outlining major concerns about use of this kind of evidence in a criminal investigation: first, over the analysis and interpretation of such data, and secondly over whether it can be guaranteed that the data have been compiled in an objective and unbiased manner. in this conference talk I discuss two cases: the famous case of Lucia de Berk (Netherlands) and the current case of Lucy Letby (UK)

1. Richard Gill, CPS Causality 1, 14 August 2024.
Statistical issues in cases
of investigation of possible
medical misconduct
A tale of two Lucy’s
https://rss.org.uk/news-publication/news-publications/2022/section-group-reports/rss-publishes-
report-on-dealing-with-uncertainty-i/

2. A tale of two Lucy’s
Left: Lucia de Berk reading letters, Christmas 1993, (c) Carole Edrich
Right: Lucy Letby with friends at Coldplay concert, Etihad Stadium Manchester, 10 June 2012
Lucia de Berk
Criminal investigation & arrest 2001 into events at several hospitals in recent years; life
sentences definitive 2005; exonerated 2010
Lucy Letby
Criminal investigation 2017-2020 into events at CoCH 2015–2016; trials 2022-2024; life
sentences definitive 2024 …

3. Statistical issues in cases
of investigation of possible
medical misconduct
https://rss.org.uk/news-publication/news-publications/2022/section-group-
reports/rss-publishes-report-on-dealing-with-uncertainty-i/

4. Peter Green Julia Mortera Richard Gill
William Thompson Jane Hutton Neil Mackenzie

5. Published a month before Lucy’s trial started

6. Suspicions about medical murder often arise due to a surprising or unexpected series of
events, such as an unusual number of deaths among patients under the care of a particular
professional.  
 
The RSS has major concerns about use of this kind of evidence in a criminal investigation:
first, over the analysis and interpretation of such data, and secondly over whether it can be
guaranteed that the data have been compiled in an objective and unbiased manner.  
 
When interpreting such data, investigators need to consider: 
•Could the deaths have occurred for reasons other than murder?  
•If murder was the cause, is the person under suspicion responsible?  
The report also calls for more care to be taken by experts to avoid drawing erroneous
inferences from such data, by properly controlling for plausible causal factors. 
 
The RSS’s concerns about the compilation of the data used in such investigations are that
attention is rarely given to ensuring that unconscious bias has not influenced the selection of
cases. Such innocent cognitive biases are prevalent throughout society and control of these
needs active steps such as blinding. For medical misconduct cases, the report recommends
that investigations should be supervised by expert panels independent of both the suspect
and their employer. 
 
RSS CEO, Stian Westlake, commenting on the report, said: ‘We have seen cases across the
world of people wrongfully charged based on incorrect statistical analysis. We’re calling for
better collaboration between the legal and statistical communities to prevent such
miscarriages of justice happening in the future.’

7. 3, 3, 8, 7, 2, 2
An unexpected cluster of bad outcomes at CoCH NICU
Source: FOI request by Peter Elston
Lucy Letby: fully qualified and full time on unit, January 2015–June 2016

8. ✝
✝
✝
✝
✝
✝
✝
✝
G
G
G
G
G
G
G
G
NG
–
G
–
–
–
G
–
–
G
G
–
The Police spreadsheet
(Lucy Letby marked in violet)

9. 9
Nurses ordered by “event of first appearance”
(marked in green)
✝
✝
✝
✝
✝
✝
✝
✝
G
G
G
G
G
G
G
G
NG
–
G
–
–
–
G
–
–
G
G
–

10. •Should those events ever have been labelled “unexpected”?
•Who, how and when were those events labelled unexpected?
•Are nurses exchangeable?
•Fortunately we have access to a lot of data: court transcripts contain
detailed information about medical condition of all the babies; various
NHS organisations publish data at various levels of aggregation; FOI
requests can supply more.
•Several persons have published fascinating statistical analyses: John
O’Quigley; Peter Elston; Scott McLachlan; anon @ https://
triedbystats.com
Statistical issues

11. •The number of early neonatal deaths is much what you would
expect given the acuity of the patients (MBRRACE-UK statistical
modelling using log linear models – Scott Maclachlan)
•Clusters like this are quite common – John O’Quigley
•Maternity care in the UK is scandalously bad, and every year
there is another major scandal
•Serial killer nurses in neonatal units are extremely rare - once in
100 years in the UK?
•Data inconsistencies, errors, …
Statistical findings so far

12. •Jan 2015. Lucy fully qualified, & most qualified nurse on unit
despite v. young age. Older nurses had been fired to save £££
•Lucy fills in many DATIX forms. Conflict develops between her and
four consultant paediatrians esp. Brearey and Jayaram. They ask
management to take her off the ward for “spoiling the atmosphere”
•Everyone is surprised by number of unexpected deaths
•Lucy is worried because bad events kept happening when she is
there (she is working longest hours and hardest shifts)
•B & J compile spreadsheet v.1 and gossip that Lucy is maybe
killing babies, inform management but not police
(But never revealed in court)
Run-up to police involvement (1/2)

13. •End of June 2016: management calls in RCPCH; Lucy transferred to
desk duties; NICU demoted to “Level 1”
•Lucy formally objects to move and complains of harassment by doctors.
Management decide doctors’ complaints are unfounded and demands
that they desist and apologise both in person and in a signed letter. They
acquiesce.
•RCPCH visits CoCH and produces devastating report blaming
management and doctors and exonerating nurses in general and Lucy in
particular
•February 2017: “redacted” report is published (references to nurses and
to allegation of suspicion of LL removed)
•Brearey and Jayaram go to the police. Say that deaths on the unit were
totally abnormal, give spreadsheet v.2 to the police. Chief inspector is
convinced in 10 minutes. B & J give medical dossiers of 30? 60? patients
to police
(But never revealed in court)
Run-up to police involvement (2/2)

14. Stephen Brearey Ravi Jayaram
Dewi Evans

15. •Postmortem reports (natural deaths) of all babies who died
•The RCPCH report
Evidence which did not make it to court
•Only two ward rounds per week, there should have been one or
two per day
•A neonatologist needed to be recruited, immediately
•Insufficient nurses and insufficient senior nurses
•Insufficient investigations into causes of surprising bad events
•Junior doctors reluctant to bring in senior, and senior doctors
reluctant to come in
•Some events seemed hard to explain and a forensic investigation
was needed … it happened, and concluded no evidence of foul
play
•The unit should remain a Level 1 NICU

16. •A long-retired paediatrician from South Wales suddenly turned
up in Chester and offered to help the police, saying he had read
about the case in the Sunday papers
•Evans went through the stack of medical documents and
fantasised murderous attacks when he saw something odd, e.g.
a bubble of air or gas on a post-mortem X-ray
•Later Brearey came by and also found anomalous immunoassay
results for insulin/C-peptide ratio of babies being treated for
hypoglycaemia (he and his colleagues had never seen them
before)
Amateur sleuth, charlatan, and gun-for-hire
Enter Dewi Evans

17. •Showed for each of the 25 indictments that there was no evidence
that Lucy did anything wrong and indeed that the “murderous
attacks” fantasised by Evans were indeed fantasies.
•Unfortunately police had recruited more dodgy experts to support
Evans’ guesses – there were 8 prosecution medical experts versus
the defence’s 0
•The defence called one expert only: the plumber who made weekly
visits to combat sewage back-flow and leaking sewage pipes
•Many babies exhibited symptoms of sepsis and some clearly died
of it but this was not recognised by prosecution experts (it had
been recognised in some autopsies)
Did not use statistical evidence, did not criticise NHS
What about the defence

18. •A handful of outsiders started criticising the trial scientifically
(both statistically – Gill, Elston, McClachlan, and medically –
Sarrita Adams, founder of “Science on Trial”)
•We were suppressed by Cheshire Constabulary and made out to
be sick and deluded conspiracy theorists by newspapers and
in social media (*)
•Individual citizens with doubts did not dare express their doubts
to their friends or family
•It became ‘bon ton’ to call for restoration of the death penalty for
Lucy.
The media enjoyed the witch-hunt
What happened outside the courtroom
(*) Most recently by The Times, Monday 12 August

19. •Rachel Aviv published a very long article in The New Yorker. It
was geo-blocked in the UK
•The Guardian published another long article based in part on
Aviv’s work and in part on their own investigative journalism
•The Daily Telegraph did the same, adding information from a
whistleblower about a Pseudomonas aeruginosa infection at
CoCH
•Channel 5 documentary, “Part 1”, was screened
•More blogs, YouTube channels, X groups, …
First signs of opposition
Lucy was given full life sentences for 7 murders and 8 murder
attempts.
[Most verdicts were not unanimous and one juror left before the end of the
trial.]
The media cheered, social media cheered: the witch will die in jail!
Hopefully, very soon!

20. •Lucy’s request to appeal is turned down. The arguments are ludicrous.
For instance: science which was known pre-trial but not used by the
defence may not be used to appeal against the convictions. [The science
in question proves that no babies died of air embolism – Evans’ favourite
fantasy murder method]
•The prosecution appealed against the “no verdict” in the case of Baby K.
Judge tells jury that they need not take any notice of the evidence they
heard in that case, since she was already convicted of 17 counts of
murder or attempted murder, and they probably don’t have an objection
to that. So count is upped to 18.
•These rulings meant that going to CCRC will not help since CCRC will
conclude that (in absence of new evidence), and in view of court rulings,
a new trial is not possible
•The up-coming Thirlwall Inquiry is designed to blame CoCH management
for stopping the consultants from going to the police
•UK legal injunctions are decreasing Google rankings of sites calling for a
re-trial or arguing for innocence
The enemy fights back

21. •If you are British, write to your MP and to relevant ministers or
state secretaries
•Talk about the case to your friends and relatives
•Sign the various existing petitions for a retrial and for release of
the “missing data”
•Statisticians: take a look at existing data and statistical analyses
and try to do better ones. Publish academic papers using
material from the case as an example. Discuss pro’s and con’s of
different methodologies, bring in new ones
What can you do? (1/2)
What will happen next?

22. •Calling all causality experts: Develop a suitable causal model
which allows definition of interesting counterfactual variables.
•For example: how many deaths would have been avoided, if high
risk babies had been treated in Level 2 or Level 3 NICU’s as
appropriate?
What will happen next?
What can you do? (2/2)