European Network to Develop Policy Relevant Models and Socio-Economic Analyses of Drug Use, Consequences and Interventions Final report: Part 4 – Time trends and incidence of problem drug use

In the 1st 12-month period Kaplan Meyer and Cox Regression models were used to estimate the latency time between first drug use and first treatment. The group applied a Back-Calculation Model to estimate the incidence of problem drug use from the observed incidence in treatment, using the latency time. Latency time was remarkably similar in Amsterdam, London and Rome (on average between 6 and 8 years). However, this depended strongly on age at first drug use, latency time being longer in drug users who started at a younger age. This has important consequences for treatment centres, which might not be reaching young drug users sufficiently. Incidence curves were also estimated in the three cities on the basis of treatment data, using an updated version of the Empirical Bayesian Back Calculation procedure, showing important differences in the dynamic of the drug use epidemic. Data for Lisbon were analysed as well and showed similar results, despite being not from drugs treatment but from other services and therefore difficult to compare.

In the 2nd 12-month period Kaplan Meyer and Cox Regression models were used to estimate the latency time between first drug use and first treatment introducing epidemiological information about the phase of the sub-epidemics concerning the sub-groups defined by discrete covariates (ethnicity, gender, route...) in order to correct possible biases due to the different starting point of the sub-epidemics. The group generalised the Empirical Bayesian a Back-Calculation Model to estimate the incidence of problem drug use from the observed incidence in treatment, using the latency time as incubation distribution. User friendly interfaces were produced for the BC procedure written in S+ language. Latency time was remarkably similar in Amsterdam, London and Rome (on average between 6 and 8 years). However, this depended strongly on age at first drug use, latency time being longer in drug users who started at a younger age. Incidence curves were estimated for London, Amsterdam and Italy, showing important differences in the dynamic of the drug use epidemic. For Italy it was also possible to estimate some incidence curves at local (regional level). Data for Lisbon were analysed as well and showed similar results, despite being not from drugs treatment but from other services and therefore more difficult to compare. The incidence curve for Lisbon was estimated using a special snapshot method, namely the onset delay adjustment method (ODAM). Recent work from Dublin indicates that the situation might be very different, latency time on average about 2 years and not related to age at first use.

This might however be related to the stage of the drugs epidemic, which appears to be still in its early phase in Dublin, while already endemic in most other European cities. Unfortunately data from Dublin presently comprises only one year of treatment, thus it is not possible to check for the phase of the epidemic. Data from the French Community of Belgium were also analysed estimating the latency period with results comparable to the other sites (except Dublin). The incidence curve was estimated by the ODAM model. Similarly data from Budapest were also analysed producing results similar to those of Dublin. In this latter case, however, the phase of the epidemic could be considered due to the completeness of the data-set, which comprises more years of treatment. During the period 4 papers were written to be submitted to International Journals (2 of them are in press on UN Bulletin on Narcotics). Other papers related to the local analyses, results and policy implications were also produced. Preprints related to 3 of the 4 International papers can be presently downloaded from the website: http://mat.uniroma2.it/biometria/. A system dynamic model was developed and used to make some preliminary scenario analyses.