[Possibilites and limitations of retrospective research on cause of death within the framework of a nationwide epidemiological study].

Cause-specific mortality is an important endpoint in longitudinal epidemiological studies to research causal health links and carry out individual process analyses. As in Germany no national mortality register exists, state-specific data-protection conditions for and approaches to cause of death studies were researched and tested within the framework of a pilot project. The database was provided by a sample of deceased study participants from the 1998 nationwide health survey. Under the given conditions, cause of death research in a nationwide epidemiological study is possible in principle, but requires a great deal of time and effort. In Rhineland-Palatinate research needs to be carried out in a timely manner as the health authorities' retention period for death certificates is limited to minimum of 5 years. In Berlin the research situation is particularly difficult as, while the death certificates are held 3 years by the statistics bureau, this department cannot release any pseudonymous data. In Berlin it is possible to research the cause of death for some of the deceased through the doctors and hospitals that issued the death certificates and through the GKR Berlin-Brandenburg (in keeping with the special data protection measures of the cancer registry). In North Rhine Westphalia the written consent obtained during people's lifetimes, including an explicit release from medical confidentiality, is required to carry out cause of death research. In all other German states exceptions are possible, in all cases requiring the consent of the state's data protection commissioners. The results of the pilot project underline the need for a national mortality register. Until this is established the approach used in the pilot study can be used and, working together with the state and national data protection authorities and with the highest national health authorities, can be further optimized in order to avoid losing data and to use resources efficiently.