A mathematical model has been constructed to estimate the prevalence of parkinsonism from register information on the use of a tracer drug in a defined geographical area. The prevalence of parkinsonism was estimated for several European countries using data on rates of L-dopa derivatives in 1980. The estimates were compared with data from community-based surveys of parkinsonism. The precision of the estimates as calculated by the coefficient of variation ranged between 1.6 and 6.9% and the accuracy as measured by the standard deviation of the percent error was 9.1 %. Up-to-date community-based data are not available. Using this calibrated model, the prevalence of parkinsonism in Sweden was estimated as 167.9/100,000.