Since 1985 in the United States, federally-funded HIV prevention efforts for injecting drug users (IDUs) have been based on a providerclient model called streetbased outreach. In this model, a few individuals - usually exdrug users or people with "street" credentials - are hired as outreach workers to work with local drug injectors. Rather than offer IDUs any direct rewards or incentives to play active roles in HIV prevention, IDUs are encouraged to give freely of their time and energy in helping outreach workers carry out the core activities of the prevention effort.

Research demonstrates that IDUs respond positively to such traditional outreach in reducing drugrelated risk behaviours. Traditional outreach interventions have nevertheless been found to suffer from a host of organizational problems that lead projects to drift towards stagnation and invite high levels of mal and nonperformance among outreach workers. An innovative social network model for accessing IDUs, the peerdriven intervention, is now going further than outreach models. It also draws upon and strengthens the sharing rituals and norms of reciprocity that underlie and sustain social networks among users.

In contrast, the peerdriven intervention provides active IDUs with guidance on carrying out - in their own drug-using networks - the same tasks traditionally performed by outreach workers. This model offers them direct, nominal, pertask monetary rewards to carry out such tasks.

Findings from a multiyear experiment begun in 1994 indicate that a peerdriven intervention reaches a larger and more diverse set of IDUs at a substantially lower cost than traditional interventions. The peerdriven intervention is also significantly more effective in reducing IDUs' drugrelated risk behaviours. With regard to research methodology, the peer-driven intervention draws upon recent developments in snowball and other forms of chainreferral sampling mechanisms; these innovations are designed to overcome limitations that have traditionally caused chain-referral samples to be seen merely as convenience samples.

Participants are given coupons for recruiting other IDUs, who are then also offered the opportunity to recruit. Finally, they return to the storefront to be rewarded for their education and recruitment efforts; they are also debriefed and given another knowledge test. Thus, the recruitment process in a peerdriven intervention has the potential of expanding geometrically.

In the process of educating and recruiting peers in the community, IDU-recruiters repeat the prevention message with each point of contact for a total of six or more exposures upon completion of the recruitment cycle. In contrast, the traditional outreach intervention provides only two points of contact and exposure of the prevention message. The first contact occurs when an outreach worker recruits and educates an IDU in the community; the second occurs when the IDU goes to the storefront. Storefront activities for both interventions are similar, except that in the peerdriven intervention the IDUs are trained to educate and recruit their peers.

In terms of combating HIV/STDs and other drugrelated diseases, the peerdriven intervention invites IDUs to become personally involved in prevention efforts, both by encouraging their peers to behave in certain ways and by changing their behaviour for their own sake and the sake of others. Whereas subjects are active recipients in the peerdriven model, IDUs remain passive recipients of services in the traditional model.

Further, in the peerdriven model the number of contacts with the prevention message is three times that of the traditional outreach model. Finally, the peerdriven intervention combines six factors that have been found to promote behavioural change: increases in knowledge, skillsbuilding, incentives, peer pressure, social norms and repetition.

In the multiyear study mentioned above, the costs of running the storefronts - which primarily consisted of the health educators' salaries - were identical for the traditional outreach and peer-driven interventions. The cost for each respondent in the peer-driven intervention to access, educate and recruit other users to the storefront was US$ 16. On the other hand, in the traditional outreach intervention with salaries for three fulltime outreach workers, the recruitment and education effort resulted in a cost of $470 per recruit. In addition, the peerdriven intervention recruits scored significantly higher on the education knowledge test than the IDUs recruited by outreach workers.

Overall, the peerdriven intervention out-performs traditional outreach efforts in recruiting a large and diverse set of subjects. It is nevertheless not a replacement for outreach work. Traditional outreach workers not only engage in community-based HIV/STD prevention education and recruitment, they also are involved in case management, referrals to community agencies and community presentations.

Furthermore, the peerbased intervention requires a staff to operate the incentive system, educate recruits in the project's storefront and provide HIVtest counselling. By re-educating and reassigning outreach workers to serve as health educators, the peer-driven intervention enhances their talents by economizing and reallocating their efforts in working with IDUs, allowing for a working collaboration with IDUs to locate and educate their peers in the community about AIDS prevention.