Spike in infant deaths at Lucy Letby unit warranted investigation, says statistician
The “unusual” spike in infant deaths at the neonatal unit where Lucy Letby worked had a probability of less than 1% given the previous mortality rate, a public inquiry has found.
Sir David Spiegelhalter, emeritus professor of statistics at the University of Cambridge, said the sudden rise in deaths in 2015 was “surprising” and enough to warrant an internal investigation.
However, he added that from a national perspective the increase was “not very surprising at all” and “not extreme enough to be considered an outlier.”
The Thirlwall inquiry found that three babies died in the Countess of Chester’s neonatal unit in 2012, followed by two in 2013 and three in 2014.
This rose to eight in 2015 and there were a further five deaths in the first six months of 2016. Letby was removed from the neonatal unit in July 2016.
The former nurse, now 35, is serving 15 life sentences after being convicted of murdering seven babies and attempting to kill seven others between June 2015 and June 2016. She maintains her innocence.
The use of statistics in the first trial against Letby is one of the main causes of concern among experts who believe she is the victim of a miscarriage of justice.
Critics have suggested that the jury was misled because the prosecutor highlighted Letby’s name in a shift schedule to suggest she was the only doctor on duty for each of the murders and attempted murders.
Letby’s presence at a series of ‘unexpected and unexplained’ deaths and collapses was what first raised the suspicions of senior doctors, who raised alarms with their superiors, while acknowledging they had no further evidence of deliberate harm.
During his testimony at Liverpool Town Hall on Wednesday, Spiegelhalter did not comment directly on the prosecution’s interpretation of the team chart or use of statistics.
However, he stressed the need for rigorous analysis and caution before drawing conclusions based on data alone.
“The reason why statistics are so important is that countless studies and personal experiences have shown that people are not very good at evaluating data,” he said.
“They may overlook long-term trends and accumulate changes slowly… and on the other hand, they may pay too much attention to sporadic sets of bad outcomes due to unknown factors and try to find patterns that may not actually exist.
“These are known human characteristics and that is why we need statistical analyses.”
Spiegelhalter, a member of the statistical team for the Harold Shipman Inquiry, said the increase in the number of neonatal deaths at the Countess of Chester hospital – from less than three before 2015 to eight that year – was a “surprising event within the Countess or Chester Hospital. but for a national event that is not surprising at all.” He added: “We expect this to happen at some point every year.”
He said the probability of the unit recording eight deaths per year, given the previous mortality rate, was “0.008, which is less than 1%”.
“That would be generally enough to trigger a warning, but not extreme enough to be considered an outlier,” he said, adding that it would be enough to warrant an internal investigation.
However, Spiegelhalter urged caution because “bad things tend to cluster” and that “the fact that the numbers have increased does not necessarily mean that there is a special reason for it.”
Questioned by Peter Skelton KC, who represents some of the bereaved families, the statistician said it was important that those interpreting the data show ‘humility’ in dealing with uncertain situations.
“Humility is trying to think that I might be wrong, not having too much confidence in someone’s judgment, and … not being so confident in your judgment that you blind yourself to evidence that might point in a different direction.”
The inquiry, by Lady Justice Thirlwall, will conclude the hearing on Friday before the closing speeches begin in March. The findings are expected to be published in the autumn.