What the chart showed
The shift-presence chart, as presented to the jury, recorded 25 significant clinical events — deaths and collapses among the indicted children — and mapped each against the shift roster to show that Lucy Letby was on duty at the time of each event. Presented in this form, the chart appeared to demonstrate an extraordinary pattern: that a single nurse was present at every significant adverse event over a period of more than a year on the neonatal unit.
The visual and rhetorical force of the chart was substantial. Jurors were invited to treat 25-out-of-25 presence as a statistical anomaly so extreme that it could only be explained by deliberate action. The prosecution characterised the pattern as “remarkable” and invited the inference that chance could not account for it.
What the chart excluded
The chart was constructed by starting with the events rather than with the shifts. The methodology selected the 25 events and then determined who was present. It did not start with all collapses and deteriorations on the unit and then ask what proportion occurred on Letby’s shifts versus other shifts. The collapses, deteriorations and significant clinical events that occurred when Letby was not on duty were excluded from the dataset before the chart was constructed.
This exclusion is the fundamental methodological problem. A chart that records only events at which a person was present will, by construction, show that person present at 100 percent of charted events. The denominator — the total number of adverse events on the unit, including those on shifts where Letby was absent — was not placed before the jury.
The Royal Statistical Society, in its reissued statement of April 2026, named this problem explicitly. The RSS noted that the chart’s construction created a selection effect that precluded any valid probabilistic inference from the 25-out-of-25 figure, because the figure was a product of the selection methodology rather than of an underlying statistical pattern.
Why the exclusion is the whole problem
The Texas-sharpshooter fallacy describes the error of drawing a target around bullet holes after the fact — selecting the evidence that fits the hypothesis rather than testing the hypothesis against all available evidence. The shift-chart methodology is a textbook instance of this error. By including only events at which Letby was present, the chart guaranteed its own conclusion: that Letby was present at all included events.
The relevant statistical question is not “was Letby present at these 25 events?” but rather “given the proportion of unit hours that Letby worked, and the base rate of adverse events on the unit, what is the probability that a nurse working those hours would be present at this proportion of adverse events?” That question requires the full denominator: all adverse events, on all shifts, attributed to all nurses.
Without the denominator, the chart is not a statistical argument. It is a tautology: the selected events show Letby present because the events were selected on that basis.
Prof. Gill’s published critique
Prof. Richard Gill, Emeritus Professor of Mathematical Statistics at Leiden University and a specialist in the statistical analysis of healthcare-worker criminal cases, has published a critique of the shift-chart methodology in the Letby case. Prof. Gill’s analysis identifies the selection-effect problem and quantifies its impact on any probabilistic inference drawn from the chart.
Prof. Gill has noted that the shift-chart argument is structurally similar to the statistical errors made in the Lucia de Berk case in the Netherlands, where a nurse was convicted of serial murder on the basis of a presence-at-adverse-events analysis that excluded the relevant denominator. The Dutch Supreme Court ultimately acquitted de Berk after independent statistical review identified the same methodological flaw. Prof. Gill was among the experts who contributed to the de Berk review.
Prof. O’Quigley’s proportional-hazards reanalysis
Prof. John O’Quigley, a biostatistician with expertise in proportional-hazards modelling, has conducted a reanalysis of the shift-presence data using a proportional- hazards framework that controls for hours worked. This approach treats each shift-hour as an observation and models the hazard of an adverse event as a function of nurse identity, while controlling for known confounders including infant acuity and time period.
When hours worked are properly controlled, the anomalous-presence probability associated with Letby is reported to fall to approximately 10 percent in O’Quigley’s reanalysis — far below the threshold at which any statistical inference of anomaly could be sustained. The residual association, once the denominator is properly specified, is consistent with chance variation in a nurse who worked a substantial proportion of the unit’s shifts during the relevant period.
Royal Statistical Society’s reissued April 2026 statement
The Royal Statistical Society reissued its statement on the Letby case in April 2026, updating its earlier commentary to address the shift-chart methodology specifically. The RSS statement names the Texas-sharpshooter selection effect and calls for any statistical evidence presented in future proceedings to be accompanied by full disclosure of the denominator — the total number of adverse events on the unit, by shift, for all staff.
The RSS also notes that the shift-chart evidence was presented to the jury without the involvement of a qualified statistician in its construction or verification, and recommends that any future appeal or referral proceedings include independent statistical expert evidence on the methodology. Supporters of a CCRC referral are understood to be commissioning that expert evidence as part of the referral application.