American policymakers have a potent new tool to combat the spread of HIV. It’s not a vaccine or a miracle cure, but instead a newfound way to implement the strategies already known to work.
A team of epidemiologists, including University researchers, published a paper this month in the journal Health Affairs, which outlined a statistical model designed to determine the optimal way to curb the spread of HIV from needle-sharing. The results from this model indicate that combining a particular set of pre-existing interventions could produce a 62 percent reduction in the number of drug-injecting New York City residents who test positive for HIV by 2040.
The computer program was specifically designed to model HIV transmission behavior of adult residents of NYC, said Brandon Marshall, assistant professor of epidemiology and the study’s lead author.
The model is made up of 150,000 “agents” — or simulated individuals — that are assigned behaviors that determine their risk of contracting and transmitting HIV. Agents can be male or female, gay or heterosexual, non-drug users, non-injecting drug users or injecting drug users. The makeup of the agent population at the simulation’s initial stage mirrors NYC’s actual demographics with respect to these categories. For instance, the model features far more female non-drug users than female injecting drug users to reflect the city’s population statistics.
The simulation, which is run on a supercomputer at the University’s Center for Computation and Visualization, begins in the year 1992 and ends in the year 2040. As time progresses, agents interact with one another and form connections. Agents sometimes perform disease-transmitting behaviors — such as needle-sharing and unprotected sex — together.
To faithfully model how HIV is actually transmitted, the frequency of these behaviors is different for each type of agent and is based on actual data and precise algorithms. For instance, in the virtual reality, two non-drug users will never share needles.
The virtual reality also incorporates historical HIV treatment programs, different stages of the virus and other real-world nuances that affect how the disease spreads through a population, Marshall said. “As interventions changed in the real world, we effectively turn them on in the model.”
The researchers also have the ability to “turn on” novel treatment combinations. Marshall and his colleagues took four HIV prevention strategies — increasing HIV testing, improving substance abuse treatment, bolstering needle exchange programs and increasing medical treatment — and tested different combinations of them.
According to results of different simulations, the most effective intervention strategy is to pour more resources into each of the four interventions simultaneously. The model predicts this “high-impact” intervention would cause the 62 percent decrease in HIV incidence.
Each of the combination strategies cut HIV incidence to some degree by 2040. Though the “high-impact” strategy was shown to be the best way to curb HIV, three out of the four next most effective intervention programs included a bolstered needle exchange program.
Marshall noted that, though it’s clearly the most effective, the “high-impact” strategy may be the least feasible of the seven due to its high cost. He said he plans to work with an economist to compare the cost-effectiveness of different interventions.
Associate Professor of Public Health and Associate Dean of Master’s Education Don Operario, who works in HIV prevention but was not involved in creating the model, said he believes the evidence is good news regardless of whether or not the “high-impact strategy” would be cost-effective. Operario, who described himself as an “idealist” on this subject, said the success of the intervention in the model provides evidence that the United States needs to “coordinate and harmonize (its) services better.”
Marshall has worked to perfect the computer program since 2011, according to a 2012 University press release. Marshall said a belief that virtual models are vital to the future of disease prevention inspires his research.
In the past, controlled trials testing the efficacy of individual disease intervention strategies have been the gold standard of epidemiology. But these controlled experiments fall short when researchers aim to look at more than one intervention at a time, Marshall said.
A virtual model such as this one is a beneficial tool because it is capable of examining “the effect of multiple interventions acting in concert,” he added. “They provide a way to integrate all our existing data and look at how interventions affect each other.”
Marshall said he hopes politicians will take note of the model’s capabilities and the study’s findings. “Policymakers have to choose between a whole menu of partially effective interventions. (The model) is a nice way to provide them with information on where resources could be directed.”