Condom use, earlier treatment and increased education have gone a long way to reducing HIV spread in the U.S. Nonetheless, some 4,000 inhabitants of New York City still became infected with HIV in 2009.

Injection drug users make up a small portion of the new infections (just over 4 percent in NYC, and about 9 percent nationally), but they represent a finite and targetable population that can benefit from low-cost and well-vetted programs, such as needle exchanges.

Establishing even better needle exchange programs or more widespread substance-abuse treatment opportunities might help to limit these new infections among drug users. But finding out how effective these prevention programs truly are with scientifically controlled studies can take years—and lots of money. If only researchers could run computer simulations to come up with some answers, as they do to model other complex systems…

Now they might just be able to, with the help of a high-power, automated version of what you could call Sims for the urban class. The goal of the computer model, conceived of in part by Brandon Marshall, an epidemiologist at Brown University, is to "identify the ideal combination of interventions to reduce HIV most dramatically in injection drug users," he said in a prepared statement. The new model was described July 27 at the 2012 International AIDS Conference in Washington, DC.

People, it turns out, are relatively predictable—at least when you study them in groups. The researchers focused on New York City's intravenous drug users (and those who might have sex with users), who are at especially high risk for contracting HIV.

The investigators collected decades of data on the city's HIV prevention programs and HIV infection rates. They then constructed "an artificial society of drug users and non-drug users," that each had substance and sexual behavior that reflected a part of the general population, the researchers described. Next, they refined the model so that it accurately predicted infection rates for a full decade of known rates (1992 to 2002). Then they created six different "scenarios" featuring different HIV-prevention policies: ramping up needle exchanges; enrolling more people in substance abuse treatment programs; increasing the rate of testing; starting people on medication earlier; a combination of these four intervention strategies; and sticking with the current policies.

Each of the six computer scenarios included 150,000 individual hypothetical "agents" or individuals. Each agent makes a series of behavioral decisions, such as having protected sex one day but unprotected sex the next or starting drug treatment and then dropping out. "With this model you can really look at the micro-connections between people," Marshall said. "It reflects what's seen in the real world."

To play out each of these minute decisions for each individual over every year, the researchers needed serious computing power, so they turned to a supercomputing cluster at Brown University. Even with all of this crunching ability, each "scenario" took 72 hours to run.

They played out the scenarios from now until 2040 several times each to make sure they were getting accurate readings. Increasing the number of people who get tested for HIV by 50 percent reduced new infections in intravenous drug users only by about 12 percent over the next three decades, according to the model. The most effective single intervention was to start treatment earlier, which lowered new infections by 45 percent. Combining all four of the interventions would cut infections by 62 percent.

"I actually expected something larger," Marshall said of the effects of these interventions—even when used all together. The modest reductions "show that a comprehensive set of proven interventions must be scaled up immediately if we are to substantially reduce the spread of HIV among drug users," Marshall noted. "That speaks to how hard we have to work."