The scientific case against Lucy Letby (part two): the explicable spike

By Peter Elston

This article was published in The Telegraph on 6 September 2024

In his previous article, the statistician Prof John O’Quigley made the important point that there had been no statistically significant spike in neonatal deaths at the Countess of Chester Hospital (CoCH) in 2015-16. In other words, there was nothing on the face of it that was unusual about the increase in infant mortality.

Understandably, doctors at CoCH may not have seen it that way. They may have seen the three deaths in June 2015, compared them with the hospital’s usual rate of about three per year in previous years, and assumed there was something unusual going on. They may well not have looked at mortality rates at other NHS trusts to put the June 2015 deaths and others in 2015 and 2016 into a broader context.

As noted in O’Quigley’s article, several other NHS trusts showed similar year-on-year increases in mortality rates to that of CoCH at some point during the period from 2013 to 2018.

Not putting events into a broader context is a cognitive error known as the lottery fallacy. If someone wins the lottery, they are not investigated for cheating. The probability of a particular person winning the lottery is very small, but someone winning is a certainty. The lottery fallacy has featured in several so-called cluster cases (for all intents and purposes a spike is a cluster) that later were found to have been miscarriages of justice.

Of course, the absence of a statistically significant spike does not preclude there being underlying factors pushing up mortality.

The prosecution successfully argued that there was an underlying factor, and that their underlying factor was Lucy Letby. However, is it possible that there is evidence of non-malevolent factors that were instead to blame?

Deaths are rarely simple. They will often be the result of a combination of general and specific factors. Analogously, a hurricane may kill many people because of its general destructive character, but whether a particular individual dies depends also on their specific vulnerabilities – for example. their health, the strength of their home, when the emergency services reach them.

In all neonatal deaths, individual babies will have faced their own specific medical issues, as well as general or systemic ones.

In the Letby case, it is becoming better understood in the public arena just how unwell the 17 babies were. In the case of one of the babies who died, Baby A, there were so many issues in relation to his health and care that the pathologist was unable to ascertain a cause of death.

As for systemic issues, it is also becoming clear that in 2015-6, CoCH’s neonatal unit was a disaster zone.

The Royal College of Paediatrics and Child Health (RCPCH) service review into elevated mortality commissioned by CoCH in July 2016, dated November 2016, and published in February 2017 appeared to be a litany of hospital failings. Bizarrely, the review was not used at court. Furthermore, in the past year and a half or so, reports of sewage and infection during the so-called Letby period have come to light.

The RCPCH review noted that staffing was inadequate. Indeed, in an email to Tony Chambers, the CoCH chief executive at the time, in December 2015, Alison Timmis, a paediatric consultant, wrote that staff were “stretched thinner and thinner and are at breaking point. When things snap, the casualties will either be children’s lives or the mental and physical health of our staff.”

Specifically, the RCPCH report noted the need to employ two advanced neonatal nurse practitioners (ANNPs), though review panel members may not have known that CoCH had got rid of eight highly qualified senior nurses, including two ANNPs, in the years leading up to 2015 and replaced them with poorly qualified agency and nursery nurses.

Other systemic issues noted in the review included:

• gaps in both medical and nursing rotas

• poor decision making; delays in seeking advice

• delayed retrieval of infants to tertiary units

• indecision around integration of the three network transport services

• non-compliance on nurse and medical staffing levels, environment and accommodation for parents, support from the community neonatal team and postnatal follow-up

• the ‘hot week’ system being insufficient to safely cover both the paediatric and neonatal wards

• insufficient storage space resulting in many pieces of equipment being stored in corridors

• poor direct visibility from one area to another

• infants being moved regularly to accommodate acuity, an extra risk in the system

• the locum recruitment process not being sufficiently robust, and there having been no documented learning/action in relation to one particular locum

• nurses reporting that external escalation was not always as timely as it could have been, and them not feeling empowered to participate

• non reporting of deaths and it not being clear who was responsible for DATIX entry

• while other areas in the hospital reported well, the neonatal unit had for some time apparently been less systematic in reporting

• the higher activity not having been raised as a risk since data had not been formally reviewed and in response the neonatal team had just worked harder

• concern at whether there were sufficient staff for the unit to care for triplets

• concern that the Child Death Overview Panel (CDOP) did not appear to have been alerted to the elevated mortality

• significant capacity pressures on the Cheshire and Merseyside Neonatal Transfer service, which contributed to delays in transferring infants out promptly

• reports of doctors waiting too long before escalating concerns about an infant, both from junior to consultant and also to the network and when they do seek tertiary level advice, the transport team is not informed sufficiently early to be on 'standby'

• there being no use of a ‘conference call’ system employed at other hospitals to inform transport team of status of infants which may require transfer; inadequate liaison between COCH clinicians and the transport team

• there being no mechanism to trigger closure of a unit when it has reached capacity; in cases requiring surgery, there was confusion among clinicians regarding protocols

• consultants not exploring possible factors behind the elevated mortality in a systematic way, nor following sound governance and root cause analysis processes

• staffing levels being inadequate when mapped to the actual activity and acuity of a level 2 unit under the British Association of Perinatal Medicine (BAPM) standards, both from a nursing and a medical perspective

• the need for a lower threshold for escalation of concerns to tertiary units for advice or transport

• issues with umbilical venous catheter (UVC) insertion that required new protocols to be introduced in February 2016

The review also noted that the proportion of late gestation admissions had fallen in 2015-16 to 7.8 per cent from 10.7 per cent previously. By implication, the proportion of higher risk early gestation admissions had risen. It also noted that there had been “higher activity and lower admission birthweight than average during the period corresponding to the increase in mortality”.

However, it went on to say that “this was not however considered to have been significant enough to explain the increase in mortality”.

Just because this may not have solely explained the increase in mortality does not mean it was not a contributory factor. To the statistician, a contributory factor is just as important as one that explains all variation. Indeed, rarely is variation in empirical data explained by a single factor.

Unsurprisingly, gestational age at birth and birthweight, which are naturally highly correlated, are the biggest contributors to mortality risk. A small shift towards lower birth weights or lower gestational ages can have a huge impact on mortality.

Some simple statistical modelling – which it seems the RCPCH report did not do – illustrates that the increase in actual mortality at CoCH in 2015 and 2016 can be reproduced with a tiny shift towards lower birth weights or lower gestational ages in those years.

The chart above is a sensitivity analysis. It illustrates how a small shift towards lower birth weights or lower gestational ages at birth can result in a large increase in expected mortality/mortality risk.

It is based on the relationship between mortality risk and birth weight and gestational age at birth as set out in various empirical studies. This data is then combined with CoCH births by birthweight and gestational age at birth that assumes correspondence with national data (adjusted to include a small shift to lower birth weights and gestational ages at birth in 2015-16 per the RCPCH report) and with actual CoCH number of births data to derive expected mortality.

Throw in the other factors listed in the RCPCH report and even smaller shifts are required.

Why were birth weights and gestational ages lower during the period that the RCPCH reviewed? It is possible that the answer lay across the border. Capacity changes in hospitals in North Wales in 2015 may well have resulted in mothers who went into labour prematurely ending up at CoCH, with the longer travel distance presenting a further risk (a study in the King’s Fund report in November 2014 found that babies forced to travel more than 28 miles during labour were more at risk from neonatal death).

The BBC had reported in November 2015 that maternity units in North Wales had had to close suddenly on 16 separate occasions in just 12 months. The BBC had also reported in February 2015 the plans to suspend consultant care for mothers-to-be at Glan Clwyd Hospital.

Believers in Letby’s guilt say that she took advantage of hospital failings to commit attacks.

However, if you adjust the mortality data for deaths with which Letby was associated, as well as for the lower birth weights, lower gestational ages, and the many other issues set out in the RCPCH review, you end up with a hospital with an impossibly low background mortality rate.

What is remarkable is that Dr Dewi Evans, the chief prosecution expert, and Cheshire Constabulary believe that they have found a killer so prolific that she accounted for the entire spike in neonatal deaths.

On that basis, how do they explain the rise in neonatal deaths at so many other trusts? Did each of them harbour a Letby?

If so, they should be hunted down across the country.

Peter Elston is a fellow of the Royal Statistical Society and an accredited scientist. He spent 30 years in the financial industry applying statistical techniques, and first wrote about the Letby case and its potential to be a miscarriage of justice in October 2022