Navigation

For PROFESSIONALS

# in SCIENCE

# Lectures in Statistics by Richard Gill

EPFL, Swiss Federal Institute of Technology Lausanne
Thursday, 19 February 2009

"Statistical Coincidence in Court: the Case of Lucia de Berk"
Abstract

The case of Lucia de B. is a highly controversial legal case in the Netherlands, in which a statistically significant correlation between the presence of a particular nurse and the occurrence of suspicious medical incidents on her ward played a central role in getting her a life conviction for serial murder. However, recent re-investigation of the meagre medical evidence for wrong-doing, and re-investigation of the statistics, makes it very plausible that no murders were committed at all, by anybody. The Dutch supreme court has recently overturned the conviction and a re-trial is about to start. I will discuss various statistical approaches to analysing the data and argue that given the lack of empirical knowledge about the ‘normal situation’ on comparable hospital wards, statistics should not have been brought into the court at all – at best, they could be seen as an exploratory tool in medical or police investigations.


Lucia dances

RCN 2008 Annual International Nursing Research Conference
8-11 April 2008 — Liverpool, England, UK

Health Care Serial Killers (HCSKs):  a world epidemic, or just bad statistics?
Friday, 11th April 2008
Presentation: Professor Richard Gill, University of Leiden, with other scientists involved in the Dutch Lucia de Berk case

Recent years have seen a world-wide spate of murder cases, triggered by the occurrence of clusters of ‘unexplained’ incidents on a nursing ward, together with the presence of one particular nurse at many of those events. Many of these cases collapse again, because finally no hard evidence could be found to prove that the cases were murders or murder attempts, let alone that that particular nurse had any connection with them.

The presenter believes that there is a simple statistical explanation for the phenomenon of innocent co-clustering of events and nurses’ shifts. He believes that research is urgently needed into the phenomenon, since if he is right, bad statistics are leading to the initiation of criminal proceedings, devastating to all concerned, when the case should never have left the stage of an internal medical investigation.

The target audience consists of all those who are concerned with the risks which are run by anyone who works in a place where people can die. He wishes to initiate discussion in the nursing community, and he wishes to stimulate multidisciplinary research into the phenomenon, for which nursing researchers and statisticians are going to need to learn one another’s languages.


The case of Lucia de Berk: Science versus Justice

Tuesday, 29 January 2008, 16:00 - 17:30
IWI - University Groningen

Tuesday, 15 January, 13:00, "lunch-colloquium", FREE LUNCH 12:30
J.H.Oortgebouw, University of Leiden

I will analyse the scientific evidence in the case of the Dutch nurse Lucia de Berk, serving a life sentence for 7 murders and 3 attempts in hospitals in the Hague during 1996–2001. I will argue that the medical evidence for the conviction was for one part wrong, and for the other part contaminated by bad statistics; the statistical evidence for the conviction was for one part wrong, and for the other part contaminated by incorrect medical information. No other serious evidence exists. This logical circle is locked down by the legal institutions of the Netherlands, and it appears that they have thrown away the key. The principle of trias politica means that nothing whatsoever can be done about this.

Naturally I shall give special attention to the mathematical details on the statistical side. The keywords and phrases here are: hidden confounders, unmeasured heterogeneity, overdispersion. The case of Lucia de Berk does not stand alone, there is in fact a world-wide epidemic of (collapsed) murder cases against nurses who seemed far too often for it to be a coincidence, present during all of an unexplained cluster of incidents.

Sheets of the Lucia de B. lecture [PDF]


Who Done It?

Tuesday, 22 January 2008, 20:00, Studium Generale
Wageningen University
"Do DNA fragments form a withstanding tool for nailing criminals?
Does a DNA-match, with a theoretical chance of 1 in billions, practically prove you did it?"


Statistician Richard D. Gill seeks cooperation with nursing staff

18 January 2008 in the Dutch web-magazine Nursing
There are, on a nursing department, many factors which can influence the work process, such as the regularity of (night) shifts and the personnel policy of hospitals...
This is why statistician Richard Gill seeks cooperation with nursing staff. He wishes to come into contact with researchers in the area of nursing, with the aim of forming an international multidisciplinary project team. He will gather together as many practical accounts as possible from nursing staff...

Please respond by email via his own website:

Richard D. Gill homepage
"Ceterum censeo Luciam esse liberandam"

# Mistakes made by Prof. Dr. H. Elffers, expert witness in the case of Lucia de B.

  • Elffers made a beginner's mistake of multiplying three p-values and implying the result was a p-value.
  • He should have insisted on an explicit and "intersubjective" definition a) of a shift with an incident, and b) of a shift with Lucia. And such that the shifts could now be classified on both criteria, on the one hand not knowing which nurses were there, and on the other hand, not knowing what happened. He didn't and the hospital exploited his omission with disastrous results.
  • He should have checked that the judge used his results in a valid way. The judge does use the results but uses them improperly. His statement "Your honour, it was not chance. The rest is up to you" was totally irresponsible.
  • He should have realised that his approach only made sense if one KNOWS that there are murders by one person and that that person is a nurse. Otherwise it is a senseless exercise because of the post-hoc problem (hypothesis being tested within the Neyman-Pearson framework on the same data which suggested it).
  • He gave wrong arguments why you should not look at data of other years. But data of other years in fact is very relevant.
  • He wrote that the data from the other hospitals where Lucia worked (Pen., Leijenburg) would only make things worse for Lucia but that does not have to be so, and anyway, no one has ever seen that other data.
  • He does not take account of relevant confounding factors in his statistics, and does not check that the judge takes account of them in some other way.
  • He continues to spread disinformation concerning his role in the case and has never admitted his errors publicly (only in private).
  • His reports should be seen as one side of a confrontation. He assumes there is a statistician working for the defence, who makes sure that the judge takes account of the confounding factors which he did not mention. But no statistician took on that role.
  • He worked first for the police and hospital authorities. But in the court, he was given the role of an independent scientific witness. In the Dutch inquisatorial system, an official scientific witness to the court is supposed to be completely independent, and his advice is taken to be perfectly reliable. It is completely incorrect that the same experts first advise the police and later advise the judge.

8-5-2002 Elffers [PDF - EN] 29-5-2002 Elffers
8-5-2002 Elffers [PDF - NL] 29-5-2002 Elffers

Elffers' reports have been officially released for scientific purposes.


# in LAW

# Ministry of Justice – temporary suspension of Lucia's sentence

(from our own translation team)

Ministry of Justice
Directorate General for Prevention, Youth and Punishment
Department of Prevention and Punishment Policy

To the Chairman of the Lower House
Postbus 20018
2500 EA Den Haag

Section: Punishment and Rehabilitation
Date: 2 April 2008
Our ref.: 5538809/08/DSP
Subject: Lucia de B


Supplementary to the Undersecretary for Justice's letter dated 14 January 2008 (TK 31200 VI, nr. 100) to the House, we now inform you as follows:

As stated in the aforementioned correspondence, on 13 December 2007 the Undersecretary denied a petition (dated 29 October 2007) for a temporary suspension of sentence by Lucia de B. Lucia de B. founded this petition on two grounds: the findings of the CEAS report (Commission for Evaluation of Closed Criminal Cases) and the possibility of a subsequent order for judicial review, and her health. The denial was partly based on the conclusion of the Attorney General of the Supreme Court of the Netherlands that the findings of the CEAS report in themselves provided insufficient cause for an order for review and that further investigation would need to be instigated. As stated in the aforementioned correspondence, on 20 December 2007 Lucia de B. lodged an appeal against the Undersecretary's decision with the Appeal Committee of the Board for the Administration of Criminal Justice and Child Protection (Raad voor Strafrechtstoepassing en Jeugdbescherming - RSJ). On 26 February 2008, the Appeal Committee declared the appeal unfounded (07/3544/GV).

Lucia de B. was sentenced to life imprisonment by the Amsterdam Court of Justice on 13 July 2006 (LJN AY3864). As a result of the CEAS report dated 19 October 2007, the Procurator General of the Supreme Court of the Netherlands instituted further investigation into a specific part of the criminal case.

By letter dated 26 March 2008, the Board of Procurators General informed the Minister for Justice of the provisional conclusion of the Procurator General of the Supreme Court that the investigation casts serious doubt on the safety of Mrs. De B's conviction in this case. The Procurator General proposes to submit an order for review to the Supreme Court within one month. Furthermore, he also recommends that the Board of Procurators General consider the temporary suspension of Mrs. De B's sentence, in accordance with the provisions of article 570b of the Penal Regulations (Sv). In his letter, the Board of Procurators General abide by the recommendation, referring to the findings of this investigation.

In view of the recommendation of the Board of Procurators General, the Undersecretary for Justice, by virtue of her office, has decided to grant Lucia de B. a temporary suspension of sentence, commencing 2 April 2008, for a period of three months. After this period has lapsed, it will be considered whether grounds exist for an extension.

The minister for Justice,         The Undersecretary for Justice,


Letter from the Minister [DOC - EN]

Letter from the Minister [PDF - NL]


# Important legal documents for the case of Lucia de B.


Press Release 1 – Evaluation-investigation completed 2007

From our own translation team – some gaps in the text still need to be filled:
Grimbergen Report
Press Release 2

# The function of the ‘Novum’ in Dutch Law

Deliberation on grounds for review

Dutch law has as its point of departure that a judgement pronounced by a Dutch criminal judge becomes irrevocable once the periods allowed for lodging an appeal, to have recourse to ordinary legal remedies, and possibly also those required for those remedies to be exhausted, have lapsed. Under exceptional circumstances only can the irrevocability principle be breached – if a petition for review of a judgement that found an accused guilty as charged is submitted, examined and granted.
This is expressed both in the introduction to the first clause and under 2º of art. 457 Strafverordening (Penal Regulations) in such a fashion that the basis for a review, insofar as this applies to the matter in hand, can only consist of a circumstance which, in the examination of the facts during the trial, was not known to the judge and which gives rise to a serious suspicion that, had the circumstance been known, the examination of the case would have resulted in a judgement finding the accused not guilty. Thus, applying the extraordinary remedy of a review can only lead to a reopening of a criminal procedure concluded with an irrevocable judicial decision in very exceptional cases, indeed.

It is inherent in the nature of this remedy that the grounds for review adduced must not have been apparent during the earlier trial. After all, this would negate the notion of a new circumstance in the sense of art. 457 Sv (hereafter referred to as novum) and merely entail a circumstance, which the sentencing judge was able to incorporate in his considerations.

An equally inadequate ground for review is that the new circumstance might have resulted in a different conclusion to the criminal trial, here meaning the exculpation of the appellant. Art. 457 Sv imposes a more stringent requirement in this respect, stipulating as it does that the new circumstance must give rise to ‘a serious suspicion’ that the judge would have found the appellant not guilty had he had cognizance of the novum at that time.

Nor does the singular circumstance of the preliminary investigation or the examination of the facts during the trial having been incomplete constitute a ground for review. This is only excepted if the information submitted for review engenders a legal opinion that – insofar as is pertinent here – the court would have found the appellant not guilty as charged had it had cognizance of the new information at that time.

The above also applies when, upon the judgement having become irrevocable, the Procurators General request an ‘exploratory inquiry’ the results of which are laid down in a report, as is the case here. An assertion of the incompleteness of the ‘exploratory inquiry’ and/or of the inquirers having drawn incorrect conclusions from their findings cannot, in itself, constitute a ground for review.

Moreover, a novum can only be a circumstance of a factual nature. As a rule, an opinion, conviction or conclusion cannot be deemed a circumstance of a factual nature. Accordingly, barring exceptional circumstances, an expert opinion will only be a valid novum, in principle, provided the expert based his opinion on facts and/or circumstances of a factual nature which were either not known or cannot be considered to have been known to the judge who pronounced the sentence to which the petition for review refers. It should be noted that, in principle, a circumstance of the expert whose findings represent a pillar of the ‘case proven declaration’ subsequently adopting a different view, is accorded more weight than another expert's opinion that differs from the original one.

In accordance with art. 459 Sv, the petition for review must describe the proof that evidences the novum. It does not suffice, therefore, for the appellant to supply a novum conditional upon further investigation by (or at the behest of) the Supreme Court. The onus is on the appellant to demonstrate and argue, within reason, that the judge who previously passed sentence would have found the accused not guilty had he had cognizance at the time of considering the criminal case of the circumstance put forward in the petition for review.

Insofar as a petition for review fails to comply with the principles set out above, it cannot be granted for that reason alone.


Source at the Dutch High Council:
High Council rejects reopening in the Deventer murder case
[server sometimes too busy]

Notes on the content by Richard Gill:

Note the content: An expert's opinion is not a novum if it is founded on facts which were already known to the judge. It makes no difference whether or not any expert at all ever saw those facts before, at all ...

and the novum should not just cause some doubts: it has to be clear that the judge (I repeat: the judge) would have pronounced not guilty.

The way this works in practice: the judges collect a huge dossier. EVERYTHING. Only they see it all, only they know what's in it. The aim is that all possible facts of the case have been entered into the system. Now that they have acquired the facts, they have a monopoly of what to do with it. They can interpret scientific evidence themselves or they can ask half a dozen experts. The reports of the experts also go into the dossier. Everything is secret. Finally: they consult with heaven, they know you did it, and all that remains is a cut and paste job in order to write down the "legal proof" that you did it, too. As little as possible should be written down in the verdict in order to make the verdict as "cassation-proof" as possible.

In the Lucia case the written verdict is actually, unusually, one of the longest in Dutch legal history. One sees now that this alternative strategy pays off, to the judges, in this case: it is hard to see how any one new fact could make much impact against all the facts marshalled against Lucia. The big problem with her case is that there were no murders, so that there can not be found a new murderer.


Deliberation on grounds for review [Dutch] + Notes on the translation


Algemene beschouwing over de gronden voor herziening

Uitgangspunt van het Nederlandse recht is dat een veroordeling die door de Nederlandse strafrechter is uitgesproken, na het verstrijken van de beroepstermijnen voor het instellen van gewone rechtsmiddelen en eventueel na het daadwerkelijk benut zijn van die rechtsmiddelen, onherroepelijk wordt.

Slechts onder bijzondere omstandigheden is een inbreuk op die onherroepelijkheid mogelijk, namelijk ingeval een aanvrage tot herziening van een dergelijke veroordelende uitspraak wordt ingediend en na onderzoek gegrond wordt bevonden.
Dat is in het eerste lid, aanhef en onder 2º, van art. 457 Sv aldus tot uitdrukking gebracht dat als grondslag voor een herziening, voor zover hier van belang, slechts kan dienen een omstandigheid die bij het onderzoek op de terechtzitting de rechter niet is gebleken en die het ernstige vermoeden wekt dat, ware zij bekend geweest, het onderzoek van de zaak zou hebben geleid tot vrijspraak van de veroordeelde. De aanwending van het buitengewone rechtsmiddel van herziening kan daarom slechts in uitzonderlijke gevallen leiden tot heropening van een strafproces dat met een onherroepelijke rechterlijke beslissing was afgerond.

De aard van dit rechtsmiddel brengt mee dat bij de eerdere berechting niet mag zijn gebleken van de aangevoerde grond voor herziening. In dat geval is immers geen sprake van een nieuwe omstandigheid in de zin van art. 457 Sv (hierna aan te duiden als 'novum'), maar van een omstandigheid die de rechter die de veroordeling uitsprak, in zijn oordeel heeft kunnen betrekken.

Onvoldoende is ook dat de nieuwe omstandigheid mogelijk zou hebben geleid tot een andere afloop van de strafzaak, in dit geval tot vrijspraak van de aanvrager. Art. 457 Sv is in dat opzicht strikter. Het eist immers dat door de nieuwe omstandigheid het 'ernstig vermoeden' moet ontstaan dat de rechter de aanvrager zou hebben vrijgesproken indien hij destijds met dat novum bekend was geweest.

De enkele omstandigheid dat het voorbereidend onderzoek dan wel het onderzoek op de terechtzitting niet volledig is geweest, levert evenmin een grond voor herziening op. Dit is slechts anders indien de in herziening overgelegde gegevens grond geven voor het oordeel dat – voor zover hier van belang – het hof de aanvrager zou hebben vrijgesproken van het tenlastegelegde indien het destijds bekend was geweest met die nieuwe gegevens.
Een en ander geldt ook indien – zoals in deze zaak – het College van Procureurs-Generaal na het onherroepelijk worden van de veroordeling een "oriënterend vooronderzoek" heeft doen instellen waarvan de resultaten zijn neergelegd in een rapport. De stelling dat dit "oriënterend vooronderzoek" onvolledig is geweest en/of dat de onderzoekers onjuiste conclusies hebben getrokken uit hun bevindingen, kan op zichzelf geen grond voor herziening vormen.

Voorts kan het novum slechts een omstandigheid van feitelijke aard betreffen. Een mening, overtuiging of gevolgtrekking kan in het algemeen niet als een omstandigheid van feitelijke aard worden aangemerkt. Dat brengt mee dat het oordeel van een deskundige in beginsel – behoudens bijzondere omstandigheden – slechts dan als een novum kan gelden indien deze deskundige is uitgegaan van feiten en/of omstandigheden van feitelijke aard welke niet bekend waren dan wel niet geacht kunnen worden bekend te zijn geweest aan de rechter die de uitspraak heeft gewezen waarvan herziening wordt gevraagd.
Daarbij verdient opmerking dat aan de omstandigheid dat de deskundige op wiens bevindingen de bewezenverklaring in belangrijke mate steunt, nadien tot een ander oordeel komt, in beginsel meer gewicht kan worden toegekend dan aan een – van die deskundige afwijkend – oordeel van een andere deskundige.

Op grond van art. 459 Sv moet de herzieningsaanvrage de bewijsmiddelen vermelden waaruit het novum kan blijken. De aanvrager kan dus niet volstaan met het aanvoeren van een novum met het doel dat de Hoge Raad daarnaar een nader onderzoek zal (doen) verrichten. Het is de aanvrager die tot op zekere hoogte aannemelijk moet maken dat en waarom de eerder oordelende rechter tot een vrijspraak zou zijn gekomen indien hij ten tijde van de behandeling van de strafzaak op de hoogte was geweest van hetgeen in de herzieningsaanvrage naar voren is gebracht.

Voor zover de herzieningsaanvrage blijk geeft van miskenning van hetgeen hiervoor is vooropgesteld, kan zij om die reden niet worden ingewilligd.


Notes to the translation by Richard Gill:

The Supreme Court has determined that the Deventer Murder Case is not to be reopened. They give on their news page a short description of what is a "novum". This is something which everyone needs to know.

The awful language is all very very deliberate. The intended meaning is very precise and one must be very very careful to preserve the actual meaning of the actual words. There are a lot of adders hidden in the grass of these words – if you read it quickly and make attractive prose out of it, you miss the adders. the translation beautifully preserves not just the meaning but also the arrogance and meanfulness of the original.

Trying to read it in english, I found it read (for me) more easily, if I mentally set some key words in italics. Only after you have read some of these sentences about three times do you suddenly realise that it does make sense when one puts big emphasis on one little word somewhere. However in the original there is no use of typographical emphasis at all.

But this is a matter of my interpretation so if we did put some words in italics we would have to add a note "translators emphasis".