The two investigative frames
When a cluster of unexplained deaths is referred to police, the investigation can be scoped two ways:
- Cluster-first. Start from the cases themselves. Systematically investigate each death. Rule out natural causes through forensic-pathology review, blinded differential diagnosis, and environmental investigation (outbreak, staffing, infrastructure). Only after those alternatives are adequately examined does the investigation narrow toward potential suspects.
- Suspect-first. Start from a named suspect whose shift-presence or other pattern has been identified by referring authorities. Investigate the cases through the lens of whether they fit the suspect’s pattern of harm. Natural-cause alternatives are tested against the suspect hypothesis.
Both frames are investigative choices. They produce different evidence. A cluster- first investigation produces evidence about what happened to the patients. A suspect-first investigation produces evidence about a relationship between the suspect and the cases.
Why the choice is load-bearing
The evidence produced by a suspect-first investigation is, by construction, biased toward confirming the suspect hypothesis:
- Case selection. Cases where the suspect was not present are excluded from the dataset. The chart produced at trial is partly a function of this selection.
- Expert instruction. Experts are instructed to consider whether specific findings are consistent with the alleged mechanism. A blinded differential-diagnosis expert, instructed to identify the most likely cause of death without reference to suspect hypothesis, is a different kind of expert evidence.
- Exhibit retention. Retention decisions are made based on what the suspect hypothesis suggests should be kept. TPN bags that the suspect hypothesis does not obviously implicate may not be retained.
- Witness framing. Statements from colleagues are taken with the suspect hypothesis already in the interviewer’s frame. Questions are shaped accordingly.
- Cumulative bias. Confirmation bias is cumulative across an investigation. By the time the case reaches trial, the evidence has been filtered through many successive suspect-hypothesis-aligned decisions.
Why Operation Hummingbird went suspect-first
When Cheshire Police were engaged in May 2017, they received a briefing from the Trust executive team that had been developed over the previous two years of internal review. That briefing identified Letby as the common factor and framed the cluster as a possible “another Allitt” case. The investigation opened on that frame.
A cluster-first investigation would have required Cheshire Police to treat the executive briefing as one hypothesis among several and systematically investigate each death without that frame. Commissioning blinded differential-diagnosis expert review. Independently examining whether the outbreak, staffing or infrastructure explained the cluster. Only later narrowing to potential suspects.
The anonymous 150-page Hummingbird whistleblower report (see our summary) and former DCS Wenham’s Thirlwall Inquiry evidence (see our summary) together indicate Hummingbird did not run this way.
The Shoo Lee Panel as cluster-first retrospective
The Shoo Lee International Expert Panel’s February 2025 case-by-case review is effectively the cluster-first investigation Hummingbird did not conduct. Fourteen senior international neonatologists applying blinded differential-diagnosis methodology to each death, systematically considering natural causes and systems failure, without starting from the suspect hypothesis. Its conclusion — no medical evidence of deliberate harm in any case reviewed — is the conclusion the cluster-first investigation at the time would have produced.
Why the scoping choice affects conviction safety
A conviction produced by a suspect-first investigation carries specific confirmation-bias risks. Those risks are corrected by a cluster-first retrospective review. The CCRC’s function, on cases of this character, is in part to examine whether the cluster-first retrospective produces a different picture from the suspect-first original. In the Letby case, it does.
This is not a claim that Cheshire Police officers acted in bad faith. It is a structural claim about how confirmation bias operates in complex investigations. Modern UK policing standards increasingly require cluster-first scoping precisely because the confirmation-bias failure mode of suspect-first scoping is now well documented.