[Roaming through methodology. XXXI. Estimating partially hidden populations: heroin addicts in Amsterdan].

To estimate the size of a partly hidden population the capture-recapture method can be used. The estimation is based on analysis of the overlap between several samples of the population of interest. The quality of this capture-recapture estimation highly depends on whether the assumptions of the method are or are not met: samples should be independent, the population should be closed, the overlap should be recognised, and samples should contain only individuals originating from the population of interest. Violation of these assumptions may lead to a serious underestimation or (more often) overestimation. The number of problematic heroin addicted people in Amsterdam was estimated on the basis of registered numbers in the easily accessible methadone programme of the municipal health organisation, in the police station methadone treatment programme, and in the hospital registration in the first quarter of 1997: 4130 (95% confidence interval: 3750-4570).