L'analisi statistica delle storie individuali di utenti dei servizi per il recupero delle tossicodipendenze. Considerazioni metodologiche

Purpose of this paper is to examine some methodological issues related to the statistical analysis of the histories of heroin addicts in charge of rehabilitation centers. Histories of addicts who seek for treatment can be regarded as sequences of episodes of tretament and non-treatment, the latter being characterised by the use of heroin or not. Proper statistical analysis of these event histories may be carried out with duration and transition data models if appropriate longitudinal data is available. The main objects of interest are the length of treatment spells and the length of the episodes of no drug use after treatment. Treatment evaluation is a difficult task in this context: commonly used methods based on aggregate measures are not appropriate in this context. Alternative definitions of "effect of treatment" are discussed. The assessment of the effect of treatment should be carried out with models for individual data, and selection bias should be taken into account.