Methods for population size estimation of problem drug users using a single registration (2013)

Background: The number of problem drug users is used as a key indicator to monitor the drug situation in the European Union. An alternative approach to estimate the number of problem drug users is given by ‘the one-source capture-recapture analysis' that uses a single registration.
Methods: Two variants of the one-source capture-recapture analysis were applied to a single registration: the truncated Poisson regression model (TPR) and the Zelterman regression model. These models were applied to data about clinical drug-related hospital admissions derived from the Dutch Hospital Registration (LMR). The TPR accounts for heterogeneity in capture probabilities by allowing for covariates and the Zelterman regression model relies on the problem drug users that were seen only once or twice in the hospital; the latter model is known to be robust against unobserved heterogeneity.
Results: The TPR model was found to have a bad fit due to unobserved heterogeneity. The Zelterman regression model estimated the population size at 10,415 problem drug users (95% CI is 8400-12,429). This figure is an estimate of the number of problem drug users who are at risk of a clinical hospital admission due to the medical consequences of their drug use. The model can also provide estimates of different subgroups of problematic drug users.
Conclusion: The method presented here offers a promising alternative for estimating the number of problem drug users, including different subgroups of drug users. In addition, observed and unobserved heterogeneity can be accounted for in these estimates.