Quality control in medicine is a hot topic. Any publication in non-scientific media guarantees immediate interest and corresponding reactions from the alternative media platforms. Der Spiegel cites on the very front of their journal nr 44 of October 10th 1995 the words "Gefahrliche Chirurgen: Pfusch am Herz". The title inside is even more shocking: "Jeder hat seine Toten". The journalist was not interested to hear about the uncertainty of an observation, confidence limits and variability of risk. These fundamental bases of science become irrelevant when quality control is concerned in non-scientific media. The Nouvel Observateur is not much better in their journal nr 1626 of January 4th 1996 where they cite, of course on their front page "Hôpitaux, ceux qu'il faudrait fermer. Ce n'est pas une question d'argent mais de sécurité des malades.
The drama is not that these non-scientific journals publish conclusions, without any scientific basis.
Even scientific journals and media, or health authorities prefer to discuss quality control without taking into considerations the fundamental elements of science. Some very expensive programs of quality control (see the STS database) focus on the first four days, have unverified data, no idea about completeness or accuracy.
This project brings forward the fundamental elements and limitations of quality control. Starting from an index quality model, defined on the long term follow up of 6000 patients and structured in a series of time-related equations. An independent test sample of several thousand patients with five years of follow-up is analyzed using this quality model and several thousand individual prediction lines are constructed. Based on these different lines a line is created (with its uncertainty) of the test sample behaviour. The variation of this test sample versus the index sample is identified.
The variation over time is studied and a univariate analysis identifies the incremental variables structuring this variation. Surgeon X is identified as one of these variables. Let us drop this name to one of the journalist and the fourth power will do the rest.
Multivariate time-related analysis identifies that it is not the surgeon but some new variability of patients that is the cause of the variability. The complete analysis is redone, excluding some patients with a rare variability. No more residual differences are observed and the test sample has the same quality performance versus the index quality sample. Too bad the surgeon has just been hanged.
Finally the essential elements and limitations of quality control will be discussed.
Eur Journal of Cardio-thoracic Surgery 1997;11:2-9