Bioengineering

With lives on the line, professor informs HIV policymakers
Margaret Brandeau flew to St. Petersburg, Russia, last month on a mission to save lives with math. The professor of management science and engineering sought to convince a conference of AIDS prevention officials from across Eastern Europe to look at the numbers. Her research team’s model of AIDS in St. Petersburg shows that the official policy of leaving HIV-positive heroin addicts untreated is the worst option for slowing the epidemic’s rapid and tragic spread.

“I strongly believe in using mathematics for the public good,” says Brandeau, an expert at the nexus of epidemiology and health care policy. “I’m kind of a crusader. I want to make things better.”

An important way to make health care better is to supply officials who allocate limited public resources with information about the costs and effectiveness of their options. Such officials can draw on the expertise of doctors, social workers and economists, but the unique contribution of engineers is in developing sophisticated mathematical models to predict the health and economic consequences of their choices.

City under siege
Unfortunately, St. Petersburg is an ideal place for Brandeau’s work. The city, like much of Russia, has one of the fastest-growing HIV epidemics in the world. Many experts estimate that about a million Russians — about 1 percent of the adult population — have the virus. The epidemic coincides with a similarly big increase in the abuse of injectable drugs.

The connection between increased drug use and increased HIV infection is real, but Russian officials, in St. Petersburg and elsewhere, have so far not included drug users in their limited program of HIV treatment. The 5,000 afflicted Russians who receive the HIV-suppressing treatment known as highly active antiretroviral therapy (HAART) are almost exclusively not drug users. Officials have assumed that drug users wouldn’t take their medicine and that the money would therefore be wasted. The numbers Brandeau and her co-authors published in the journal AIDS late last year indicate that both the assumption and the resulting policy choice are a prescription for disaster.

The research began in May 2005, when a multidisciplinary team of Stanford researchers visited Russia for two weeks on a whirlwind fact-finding mission.  Brandeau and lead author and MS&E doctoral student Elisa Long traveled with several members of the Center for Primary Care and Outcomes Research at Stanford Medical School, including Douglas Owens, a professor of medicine. Their goal was to gather data for a model that would determine the number of new infections that could be prevented by expanding HAART to different mixes of drug users and non-drug users in St. Petersburg. In addition to improving life expectancy in infected patients, HAART substantially reduces the chance of infecting others by reducing the virus levels in the body.  

The team visited clinics and hospitals to speak with doctors, officials, and members of non-governmental organizations to gather as much data as they could about a variety of behaviors. How often do addicts inject drugs?  How many sex partners do non-drug users have?  How frequently do individuals use condoms? 

The next step was to combine this behavioral data with facts about how HIV infection progresses to AIDS and death, how much protection HAART offers, and how the virus is spread by sex or needle sharing. The researchers then factored in cost estimates for the treatment, including an extra $10 a week per drug user for a program to ensure they take their medicine. They used their model to determine the cost effectiveness of three types of strategies: focusing treatment on drug users, on non-users, or a mix of the two.

Their answer is that the most cost-effective policy is to do pretty much the opposite of the status quo: to focus treatment on drug users. With that strategy, an expenditure of $10.3 billion over 20 years would prevent an estimated 40,377 infections in the city, of which 75% are among non-drug users. A strategy focusing on non-users cost $10.4 billion and prevented only an estimated 9,463 infections.  The reason for this difference, Brandeau says, is that drug users are the key drivers of the epidemic, and they often have sexual partnerships with non-users, so reducing drug users’ infectivity significantly reduces the HIV epidemic in the general population.

That said, the team’s recommendation is to spend more money to treat users and non-users alike.  Although not quite as cost-effective as a strategy focused solely on injection drug users, such a strategy is still highly cost-effective.

Eager to get their message out to the audience of policymakers last month, the team had their paper translated into Russian.

“There’s no use to doing this kind of work unless you can influence policy, so that’s our goal,” Brandeau says.

Helping at home
Brandeau has also applied her expertise to combating the domestic HIV epidemic, which is still unfolding, although not as dramatically as in Russia. In the United States about 40,000 new cases are reported every year.

Brandeau has looked at whether policy makers at the federal and local levels are allocating prevention funds optimally. In a paper published earlier this year in the journal Medical Decision Making, she and former student Gregory Zaric asked whether funds should be distributed in proportion to the prevalence of the epidemic in an area, or based on which programs would be most efficient in achieving results. For example, who should receive funding: a low-prevalence area with very successful HIV prevention programs, or a region heavily afflicted with HIV that lacks effective programs?  Before any money reaches patients, this decision is made twice: the Centers for Disease Control and Prevention allocate money to 65 HIV Prevention Community Planning Groups (CPG) around the country and those CPGs then allocate money to competing local programs.

In a rigidly utilitarian way, it may seem obvious that all money should be allocated to the most cost-effective programs, but there are a lot of reasons — often important social and political ones — why the more typical practice is to allocate money proportional to overall prevalence. Hoping to give policymakers a clear picture of what the competing options mean in terms of infections prevented, Brandeau and Zaric (now a professor at the University of Western Ontario) created a model based on prevalence and efficacy data for three types of prevention programs and three risk groups in 38 states, Puerto Rico and Washington DC.

They found that allocation based on efficacy at the federal and local levels saves the most lives and that allocation based on proportion at both levels saves the fewest (almost 25 percent fewer). But their model also went further to uncover some important nuances. For example, assigning 50 percent of funds for proportional allocation and distributing the rest based on efficiency saved 97 percent as many lives as focusing completely on efficiency, so there is room for proportional allocation. It just shouldn’t be the predominant practice.