European Network to Develop Policy Relevant Models and Socio-Economic Analyses of Drug Use, Consequences and Interventions Final report: Part 3 – Prevalence of drug use at the local level

The work group was formed to consolidate EMCDDA-funded and other methodological research into estimating the prevalence of drug use at the local level and also to provide advice and scientific support to researchers wishing to apply the methods in their area. New prevalence estimates have been obtained in areas where research has not previously been carried out and updated estimates have been obtained in existing areas. Advice and scientific support was given to researchers wishing to use the capture-recapture method in Copenhagen, Luxembourg and Matosinhos (Portugal). A new prevalence estimation project was initiated in Scotland (UK) where the capture-recapture method was used to provide an estimate for each of the 32 local government areas.

The results of the prevalence studies have been useful for policy makers and those planning the provision of services. For example, the results of the Scottish research have influence the government’s Drug Action Plan including the allocation of funding to local areas of the country. Future prevalence studies will benefit from the experience gained over the course of the project, scientific papers and other reports and from the development of computer routines to assist in analyses.

Members of the group are experienced in applying the methods in their geographical areas and have therefore addressed many of the relevant issues within their research. The work group was therefore a vehicle to exchange ideas about prevalence estimation, particularly the methodological aspects of this type of research, and was a resource available for those interested in undertaking prevalence research in other areas within Europe. Links were made between this workgroup and the National Prevalence workgroup, particularly with respect to analyses in Austria, Finland and Ireland where researchers carrying out local prevalence research were also undertaking research at the national level.

Two main methods have been considered; the capture-recapture method and truncated Poisson. The capture-recapture method employs data on drug users identified from two or more sources, such as hospital admissions, methadone registers, police, prison or treatment agency data. Although prevalence estimates can be obtained using two samples, there are certain assumptions that may be violated leading to biased estimates therefore it is preferable to employ three or more samples. The method examines the overlap between sources and uses log-linear regression modelling techniques to provide an estimate of the hidden population, i.e. those not identified from the available sources. The truncated Poisson method employs data from one source and examines the number of times each individual appears within the source. As the number of times a person appears once, twice, three times etc can be assumed to follow a statistical distribution such as the Poisson distribution, the number of people who appear zero times can be estimated. 95% confidence intervals can be found using either method.

In Amsterdam, the Netherlands, the differing case definitions of contributing sources was examined, particularly with respect to the effect on the estimates. Figure 1 shows how the samples used in a typical capture-recapture analysis relate to the estimate populations. Similar issues were faced within research in Dublin, Ireland. Each source which contributes to a capture-recapture analysis will employ a different case definition of what constitutes the population of interest. This is particularly