The relation between the number of charges against illegal drug users and the number of drug users, however, varies in time and space, making it difficult to estimate the prevalence of drug consumption from law enforcement data. A dynamic model with two compartments representing not yet charged and charged drug users is presented to detect changes in drug use with no time lag. Using the model, prevalence estimates can be obtained if the charge frequency distribution is known. Data on charge frequency distribution in Switzerland between 1990 and 1994 are compiled, and a theoretical distribution is applied to available data for heroin, cocaine, and other drugs. The result is an estimate of the population of illegal drug users in general and the population of illegal heroin and cocaine users. Mathematical equations used in the model are appended. 17 references, 6 tables, and 5 figures