Abstract

Twenty-five years ago in The Hague saw the first events in a ten-year lasting saga with long-lasting after-effects. A disturbing and apparently inexplicable spike in deaths at a children’s hospital in 2001 was associated by hospital staff with the presence of a particular nurse. The nurse’s name was Lucia de Berk. She worked long hours and was a striking personality; she certainly stood out in the crowd. Police were called in and in a short time this led to her conviction for serial murder; the conviction being almost entirely based on statistical evidence. Nobody had ever seen Lucia do anything wrong and there was no clear medical proof of criminal activity. The decisive evidence was a statistician’s p-value of 1 in 342 million, supposed to be the probability that such a large number of bad events could have happened by chance in Lucia’s shifts. Still, many people were unhappy with a conviction based only on a probability calculation. At an appeal one year later, the prosecution changed tactics and focussed on what they now called “irrefutable medical scientific evidence”. Lucia de Berk was convicted again; the Netherlands worst ever serial killer.

However, in 2005 certain whistleblowers started gaining the attention of the media, and this evolved into a campaign to have the case reviewed. Statisticians played a huge role in that campaign. The 1 in 342 million had left an indelible impact on everyone’s minds. Closer inspection showed that the prosecution’s reasoning was still statistical in nature, just informal instead of formal. The coincidence of Lucia’s presence with deaths and collapses was agreed by everyone, including the defence, to be “impossible”. It was important to explain the coincidence (confounding factors and biased medical judgements) and to explain the spike (changes in admissions and transfer policies within the hospital). In 2010, Lucia was finally exonerated after a retrial. The case is now seen by many as the worst ever miscarriage of justice in the Netherlands.

Many lessons were actually learnt from the case, and this will be the focus of the lecture. Parallels with the presently controversial UK case of Lucy Letby are inescapable. Yet again, statisticians in the UK are in the vanguard of a campaign to give a convicted serial killer nurse a retrial.