American Journal of Epidemiology Advance Access originally published online on July 13, 2005
American Journal of Epidemiology 2005 162(4):390-391; doi:10.1093/aje/kwi210
American Journal of Epidemiology Copyright © 2005 by the Johns Hopkins Bloomberg School of Public Health All rights reserved
TWO OF THE AUTHORS REPLY
Anne Gallay, Dr.1 and
Jean-Claude Desenclos1
1 Institut de Veille Sanitaire, 12 rue du Val d'Osne, 94415 Saint Maurice, France
We thank Prof. Dr. Dankmar Böhning for his comments (1
) on our article (2). Dr. Böhning used the information provided in the article to reconstitute the data for a training course on the capture-recapture method. Doing this exercise, he proposed two ways to reconstruct the frequency information for column 4 in table 1 (1, p. 389), which resulted in two different data sets. When he used the first set of data presented in brackets, he obtained different results and conclusions. However, when he used the second set, he obtained similar results consistent with those provided in the published article, after correction of the total number of matches of any kind (88 instead of 108 as given in the published article (2)). Indeed, the figure of 108 was an error of transcription and should be replaced by 88 (see figure 1; n1 + n2 + n3 + n5 = 88). In the first data set, Dr. Böhning used the total number of matches of any kind (n = 108) between the free sources given on page 173 and the results shown in our table 2 (2, p. 173) to reconstitute the number of unmatched outbreaks in each source. Since the number given for the total number of matches of any kind (n = 108) is wrong, the reconstituted number of outbreaks in each source is also wrong.
View larger version (13K):
[in this window]
[in a new window]
|
FIGURE 1. Numbers of Salmonella outbreaks in France in 1995 and matches made between three data sources: the National Public Health Network (NPHN), the Ministry of Agriculture (MA), and the National Salmonella and Shigella Reference Center (NRC).
|
|
We agree with Dr. Böhning that when using capture-recapture
analysis, complete capture-recapture data should be shown. However,
in many published articles, the number of tables and figures
that can be presented is limited, and a choice has to be made.
In our article (2
), except for the error of transcription in
the text, all of the data given allowed the reader to reconstitute
the data set, and there was no error in the analysis.
Finally, the exercise proposed by Dr. Böhning is very relevant. It shows how sensitive capture-recapture analysis is to the observed data, mostly when log-linear modeling is used for frequency data and when the size of the data set is small. Small variations in the numbers of matches and events could have major consequences in the results and conclusion.
 |
ACKNOWLEDGMENTS
|
|---|
Conflict of interest: none declared.
|
References
|
|---|
- Böhning DA. Re: "How many foodborne outbreaks of Salmonella infection occurred in France in 1995? Application of the capture-recapture method to three surveillance systems." (Letter). Am J Epidemiol 2005;162:389–90.[Free Full Text]
- Gallay A, Vaillant V, Bouvet P, et al. How many foodborne outbreaks of Salmonella infection occurred in France in 1995? Application of the capture-recapture method to three surveillance systems. Am J Epidemiol 2000;152:171–7.[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
TOP
References