The reason for carrying out this research project into the registration rate by the police of those victims admitted to hospital (the target population) is that it is well known, that they do not record every accident. Both the politicians and policymakers wish to know the actual number, as far as possible. One of the ways of responsibly extrapolating the recorded numbers is to link the police data, as processed by the Dutch Department for Statistics and Data Management (VOR), with the Dutch National Medical Registration (LMR). The project was based on this linkage. This linkage had already been carried out. However, this time certain differences between the two databases were tolerated, and the complete VOR database was used together with a larger LMR subset. The validity of both databases was judged using a sample set of the files and searching for coding differences of the so-called `link variables'. Discovered coding mistakes were then given a certain weighting. By permitting a certain maximum distance, the linkage was made unsusceptible to common mistakes. The knowledge obtained about the coding mistakes has been collected together in a catalogue. Recommendations have been made for improving the quality of both databases.
Abstract