[Quality of suicide mortality data].
Prevention of suicide is a public health priority in France. Indicators of suicide mortality have been widely used to describe epidemiological situations or to evaluate public health actions. It is therefore essential to examine the quality of suicide mortality data. The purpose of this work was to identify potential biases affecting the quality of such data and their comparability between different countries as well as to determine how they can affect conclusions. Potential biases were identified by studying the characteristics of the death certificate system and analyzing the international literature on data quality. The impact of biases was assessed by analyzing the causes of "concurrent" death with suicide in the official statistics (trauma and poisoning caused in an undetermined way concerning intention and unknown causes). The proportion of suicides listed as "concurrent" causes of death, estimated from specific surveys was extrapolated to official data. This method was also used to correct the international data. Practices concerning death certificates for violent deaths vary considerably from one country to another: type of certifying physician, frequency of medicolegal investigations, frequency of autopsies, suicide definition criteria, confidentiality regulations, religious and culture context. These practical differences lead to variability in undetermined and unknown causes. The corrections made on the mortality data after taking into account for these potential biases showed that the rate of suicide determined from official data is considerably underestimated, but that sociodemographic and geographic factors of suicide change little after correction. Likewise, the order by country was similar after taking into consideration concurrent causes. A reliable evaluation of the rate of suicide for a given country is of course important. However, it is possible to characterize populations at risk and analyze the determinants of suicidal behavior without necessarily recording all suicides, as long as the declaration bias is stable. Statistical analysis of death by suicide in France shows that, despite under-reporting, the principal sociodemographic and geographic features and trends over time can be considered as valid. A series of recommendations is proposed however to improve data quality and homogeneity for death certificate reporting. Designing operational criteria for deciding when to declare suicide as the cause of death would be helpful to guide physicians who report deaths. Classical autopsies could be completed by "psychological autopsies" with friends and family of the deceased. The death certificate form could be improved to include items for complementary information favoring or not suicide.