FRANÇAIS :
S'appuyant sur une étude comparative dans l'Est du Connecticut, l'article présente les premières données comparant l'efficacité d'un groupe d'auto support avec un programme traditionnel de proximité fonctionnant avec des travailleurs sociaux salariés. L'efficacité est évaluée en termes de nombre d'usagers touchés, d'éducation efficace de la communauté des usagers de drogues et de coûts d'intervention.

ENGLISH :
AIDS prevention efforts for injection drug users (IDUs) since 1988 in over 60 inner-city areas within the United States have been based on a "provider-client" model called "street-based outreach." We document the research showing that these traditional outreach projects operate under conditions that cause hierarchy and supervision to break down easily. The result is an array of organizational problems that push outreach projects toward inertia, and invite high levels of mal- and nonperformance by outreach workers (OWs) as rational adaptations to their work conditions. Nevertheless, research has also documented that IDUs responded very impressively to the outreach services they received. Based on IDUs' responsiveness, we outline the operational features of a "Peer-Driven Intervention" (PDI) that relies on an active collaboration with IDUs in place of a "provider-client" model. Based on a comparative study in eastern Connecticut, we present preliminary impact data comparing the effectiveness of a PDI with a traditional outreach intervention in terms of recruitment power, educational effectiveness of IDUs in the community, and comparative intervention costs.