Academic background
Professor John O’Quigley is an Emeritus Professor of Statistics at University College London with a research career spanning survival analysis, proportional-hazards modelling, and biostatistics. His methodological specialisms are directly relevant to the Letby case: the prosecution’s central statistical exhibit was a chart purporting to show that Letby was present at each adverse event, and the question of whether that correlation constitutes evidence of causation or merely reflects her work-pattern exposure is precisely the kind of problem that survival-analysis methods are designed to address.
The shift-statistics chart and its flaw
The prosecution’s shift-statistics chart compared Letby’s presence to adverse events by cross-referencing duty rosters with incident records. The chart was presented to the jury as demonstrating that no other nurse was present at every adverse event — a form of visual argument for unique association. The methodological problem, identified by Prof. O’Quigley and others including Prof. Jane Hutton and Prof. Richard Gill, is that the chart does not control for the variable of hours worked. A nurse who works more shifts than her colleagues will, by that fact alone, appear more often in any roster-adverse-event overlay, regardless of whether her presence has any causal relationship to the events. The chart was, in statistical terms, an unadjusted analysis presented as if it were an adjusted one.
Proportional-hazards and beta-binomial methodology
Prof. O’Quigley’s published analysis applies a proportional-hazards framework — the standard tool for analysing time-to-event data with exposure covariates — alongside a beta-binomial model to account for the overdispersion inherent in small-unit neonatal event data. The central result is that when hours-worked is introduced as the exposure variable and the presence-at-adverse-event statistic is evaluated conditionally on that exposure, the probability that Letby’s association with adverse events exceeds what would be expected from her roster alone reduces to approximately 10%. That figure does not exculpate or implicate anyone; it is a statement about the statistical strength of the presence-correlation evidence. A 10% residual probability means the chart evidence does not carry the inferential weight the prosecution assigned to it.
Peer review and the significance of publication
The O’Quigley analysis is notable not just for its conclusions but for its form: it is peer-reviewed and published in an academic statistics journal, placing it within the normal scientific-validity framework rather than the more variable standards of expert-witness evidence. The existence of a formally peer-reviewed quantitative challenge to the prosecution’s primary statistical exhibit is a significant feature of the post-conviction landscape. It provides a documented foundation for the argument that the shift-statistics evidence should not have been placed before the jury in its unadjusted form, or should have been accompanied by adjusted analysis and a clear explanation of the limitations.
Read alongside
- Evidence: shift statistics
- Analysis: Poisson cluster analysis
- Prof. Jane Hutton — statistician
- Prof. Richard Gill — statistician
- Commentary library
Source
Published peer-reviewed analysis by Prof. John O’Quigley on shift-statistics and the Letby case; UCL academic profile; trial transcript references to prosecution statistical exhibits; published commentary by O’Quigley, Hutton, and Gill.