Several methods can be used to identify the medical causes of death among participants in an epidemiological study. In a study of EDFGDF workers, the causes of death were obtained from the INSERM national database, but occupational health data were also available for 1,330 deaths that occurred between 1989 and 1994. The information from the two sources was compared to assess whether they were generally equivalent, and the potential impact of their differences on the results of epidemiological analyses was evaluated. The concordance rates between the causes reported by occupational health services and those reported by INSERM were calculated using ICD-9 codes and different groupings of causes. The cause descriptions were also examined to shed light on the observed discrepancies. Finally, age-standardized mortality rates (ASMRs) were calculated using each source and compared. The causes from both sources belong to the same group of causes of death in 81% of cases, but the specific cause differs in more than half of the cases. Agreement is better for deaths from AIDS and cancer, and poorer for deaths from diseases of the respiratory and digestive systems. Occupational medicine reports more deaths from unknown causes than INSERM data. SMR calculations show results that can vary considerably depending on which data source is used. The causes of death reported by occupational medicine and INSERM are mostly different. Consequently, it is not valid to calculate SMRs using causes from occupational medicine and reference rates established by INSERM; moreover, it is observed that these SMRs can differ significantly from those calculated using causes of death obtained from INSERM. Generally speaking, for reasons of comparability, it is preferable to use the INSERM database to describe mortality in different populations, as it is a common data source. However, this database has limitations, and it may be preferable, for internal analyses, to use another source if it is considered more accurate for certain conditions. (R.A.)