Dealing with uncertainty in medical “murder” cases
Professor Richard Gill, Emeritus Professor of Statistics, Leiden University is a well-known figure in the field of forensic statistics and has made significant contributions to the field of criminal justice. One of his most notable cases was his involvement in the overturning of the guilty verdict of Dutch nurse Lucia de Berk. I mentioned Gill in another newsletter, published last April.
In 2002, de Berk was convicted of seven counts of murder and three counts of attempted murder of patients in her care, based on statistical evidence presented in court. The prosecution argued that the odds of the deaths and incidents happening by chance were astronomical, and therefore, de Berk must have been responsible.
Gill, however, took a closer look at the statistics and found several flaws in the prosecution's arguments. He pointed out that the prosecution had ignored important factors such as the patients' illnesses and the fact that the hospital where de Berk worked had a generally high mortality rate.
Gill's expert testimony and statistical analysis helped to cast doubt on the prosecution's case and ultimately led to de Berk's conviction being overturned. The court found her not guilty, and she was released from prison in 2008.
Gill's work on this case and many other cases has been widely recognised by the legal community, and has helped to improve the field of forensic statistics and the use of statistics in the criminal justice system.
In September 2022, a report co-authored by Richard Gill was peer-reviewed and distributed by the Royal Statistical Society (RSS):
"Healthcare serial killer or coincidence? Statistical issues in investigation of suspected medical misconduct".
The report is a comprehensive study of the use of statistics in medical "murder" cases, and it makes several recommendations on how to deal with uncertainty in such cases. It emphasises the importance of considering all relevant factors and not relying solely on statistics to prove guilt. It also calls for the use of multiple experts to provide different perspectives on the evidence and for greater transparency in the presentation of statistical evidence in court.
In addition, the report recommends the use of Bayesian statistics, allowing for the incorporation of prior knowledge and the explicit representation of uncertainty. This approach, the authors argue, would provide a more realistic and nuanced understanding of the evidence and would help to avoid the problems that occurred in the Lucia de Berk case.
Furthermore, the report suggests that the use of a pre-trial conference, where the prosecution and defense experts meet to discuss the evidence and the statistical methods to be used in the trial, could help to avoid confusion and misunderstandings that can occur during a trial.
Gill and his co-authors’ recommendations were widely read and discussed among legal and statistical experts and were seen as a significant contribution to the field of forensic statistics. They have been taken into account in several high-profile cases, and helped to improve the use of statistics in criminal trials.
Could your future as a statistician lead you into the field of criminal justice?
Find out more about Richard Gill and his fascinating work here.