We have read with great interest both the editorial of Dr. Lagasse1and the article of Li et al. ,2which were related to the epidemiology of anesthesia-related mortality in the United States. The study by Li et al.  2used the data from the National Vital Statistics System and the International Classification of Diseases codes (10th revision) to assess mortality related to anesthesia in the United States between 1999 and 2005. A similar study has been performed in France in 1999 and was correctly referenced by the authors.3The major advantage of such a methodology is the completeness of mortality data retrieved from the National Vital Statistics System. However, when trying to identify deaths related to anesthesia, and to describe the precise degree of imputation, some problems arise.

First, the coding system is not detailed enough to capture the precise mechanism(s) that led to death and to ascertain a causal relationship. To assess more precisely the pathophysiological mechanism(s) that led to death, and subsequently to clarify the relation to anesthesia in the sequence of events, an expert analysis remains necessary, which may even be better replaced by a peer discussion with the anesthesiologist in charge of the case and who reported the death.

Second, as Dr. Lagasse noticed in his Editorial, the 10th revision of International Classification of Diseases is curiously quite poor regarding anesthesia. Items are more numerous for anesthesia for pregnancy and labor than for anesthesia in general. They explore mainly the surgical time and are mostly limited to anesthetic medication side effects or overdose. What about aspiration occurring during an emergency procedure, for example? Also, what about hemorrhage and/or delayed blood transfusion? This could be one of the limitations of this method, as the authors have noticed themselves in the discussion section of the article. Maybe the use of specific keywords related to anesthesia practice, in addition of the selection of International Classification of Diseases codes as described, could enhance the sensibility of the filter. However, even adding these suggestions might not be powerful enough to capture all cases. In the experience of the Mortality Research Group of the French Society of Anesthesia and Intensive Care,3we have experienced that in some cases death certificates did not mention any specific International Classification of Diseases code or any previously determined specific keyword. The patients' files could be included in the survey only because researchers had chosen to select also death certificates in which a surgical (or invasive) procedure was mentioned.

Third, although one could manipulate in many ways the method to select death certificates to detect all cases that have a relation to anesthesia, is it the real problem? The specificity of the filter will never reach 100 percent. We only assess the visible part of the iceberg. Maybe it would be more efficient to monitor the same indicator along time as the trend is likely a valuable marker, even if absolute data are very approximate. It could thus be very interesting to choose an indicator both strongly related to anesthesia and reproducible to assess over time the trend of anesthesia-related mortality through a national mortality database, rather than simply obtain punctual data through a great nationwide survey.

*Military Teaching Hospital, Clamart, France. gsmopex@yahoo.fr

Lagasse R: Innocent Prattle. Anesthesiology 2009; 110:698–9
Li G, Warner MA, Lang B, Huang L, Sun LS: Epidemiology of anesthesia-related mortality in the United States, 1999–2005. Anesthesiology 2009; 110:759–65
Lienhart A, Auroy Y, Pequignot F, Benhamou D, Warszawski J, Bovet M, Jougla E: Survey of anesthesia-related mortality in France. Anesthesiology 2006; 105:1087–97