Statistical issues in Serial Killer Nurse cases

TL;DR

This paper examines the statistical and cognitive biases involved in investigating suspected medical serial killers, highlighting case studies and proposing improved investigative approaches.

Contribution

It provides a detailed analysis of statistical issues and biases in serial killer nurse cases, with new findings on the Ben Geen case and recommendations for future investigations.

Findings

Identification of cognitive biases in investigations

Statistical analysis of the Ben Geen case

Recommendations for reducing bias in future cases

Abstract

We study statistical aspects of the case of the British nurse Ben Geen, convicted of 2 counts of murder and 15 of grievous bodily harm following events at Horton General Hospital (in the town of Banbury, Oxfordshire, UK) during December 2013-February 2014. We draw attention to parallels with the cases of nurses Lucia de Berk (the Netherlands) and Daniela Poggiali (Italy), in both of which an initial conviction for multiple murders of patients was overturned after reopening of the case. We pay most attention to the investigative processes by which data, and not just statistical data, is generated; namely, the identification of past cases in which the nurse under suspicion might have been involved. We argue that the investigation and prosecution of such cases is vulnerable to many cognitive biases and errors of reasoning about uncertainty, complicated by the fact that fact-finders have to determine not only whether a particular person was guilty of certain crimes, but whether any crimes were committed by anybody at all. The paper includes some new statistical findings on the Ben Geen case and suggests further avenues for investigation. The experiences recounted here have contributed to the writing of the hand-book Green et al. (2022), Healthcare Serial Killer or Coincidence? Statistical Issues in Investigation of Suspected Medical Misconduct, commissioned by the Royal Statistical Society, Statistics and the Law section. Submitted to MDPI Laws. This version: 5 August, 2022.