Background: In an epidemiological study, medical causes of death may be obtained from different sources. In a study on French gas and electricity company (EDF-GDF) workers, they were obtained from the national INSERM database. Additionally, the causes collected by the EDF-GDF occupational physicians were available for a subset of 1,330 deaths, which occurred between 1989 and 1994. The data from the two sources were compared with each other, in order to assess whether they were globally equivalent, and the potential impact of their differences on the results of epidemiological analyses.
Methods: Concordance rates between causes of death in the INSERM and EDF-GDF physicians databases were calculated according to the International Classification of Diseases (ICD) ninth revision codes and for various causes groups. Causes of death records were also examined in order to clarify the observed divergences. SMRs were computed in order to evaluate the consequences of using each database in epidemiological analyses. Finally, some SMRs were computed with the two sources and compared with each other.
Results: INSERM and EDF-GDF physicians causes belonged to the same causes group in 81% of cases, but the exact cause was different for more than half of them. The concordance rate was high for the deaths by AIDS and by cancer, and low for deaths by respiratory system and digestive system diseases. More causes of death were coded as "unknown" in EDF-GDF physicians data than in INSERM data. The SMRs varied widely depending on whether the INSERM or EDF-GDF physicians causes of death databases were used.
Conclusions: Causes of death recorded in the INSERM and EDF-GDF physicians databases are very different. Therefore, using the national mortality rates computed by INSERM with the EDF-GDF physicians causes of death to calculate SMRs is not valid, and it is observed that they may be very different from those computed with INSERM data. In a general way, it should be better to use the INSERM database, which is a common source, in order to generate results comparable with other studies. However, this database has its own limitations, and it may be better, in particular situations and for internal analyses, to use another source if it seems to be a better one.