The HIV/AIDS Epidemic in Africa

In 2003 2.3 million people in Sub-Saharan Africa died of AIDS and 3.1 million more were infected with HIV, bringing the total number of HIV infections to 25.4 million. Though AIDS is a worldwide problem, Sub-Saharan Africa has been much more heavily affected than elsewhere.

The Human Immunodeficiency Virus (HIV) epidemic is one of the primary threats to continued development in Africa. No country has remained untouched by HIV, making the pandemic an issue of global concern. At the same time, the difference across areas in the percentage of the population infected is staggering.

In the United States, estimates suggest that HIV infection among heterosexual, adult women is about two-tenths of one percent. In Sub-Saharan Africa, this figure is 10 to 15 percent. Even within Africa, HIV infection rates vary dramatically: from less than 1 percent in Madagascar to more than 20 percent in South Africa, Botswana, and Zimbabwe.

In the study "Sexually Transmitted Infections, Sexual Behavior, and the HIV/AIDS Epidemic," Emily Oster, postdoctoral Becker Fellow for the Initiative on Chicago Price Theory at the University of Chicago Graduate School of Business, addresses the question of why there is such significant variation in HIV rates, considering both the difference between the United States and Africa, and differences within Africa.

Oster built a statistical model that simulated the heterosexual HIV/AIDS epidemic, then used the model to measure the importance of different factors in explaining differences in infection rates.

"In order to make informed public-policy decisions about HIV prevention or treatment programs in Africa, you have to know the cost of different options," says Oster. "Understanding the path of the epidemic is the key to identifying the most cost-effective policy interventions."

Oster finds that nearly all the variation in HIV prevalence between the United States and Sub-Saharan Africa can be explained by the fact that the transmission rate of HIV (the chance that an uninfected individual becomes infected during a sexual partnership with an infected individual) is much higher in Sub-Saharan Africa than in the United States.

The estimated HIV transmission rate from men to women in Sub-Saharan Africa is approximately 25 percent, versus approximately 10 percent in the United States. One explanation, which is largely supported in the medical literature, is that the HIV transmission rate is much higher for individuals who have other sexually transmitted infections (STIs), particularly those which cause open genital sores. The rate of untreated STIs in Africa is approximately 11.9 percent, versus approximately 1.9 percent in the United States.

Differences in the HIV rate across countries in Africa can be traced to differences in sexual behavior and in when the epidemic first appeared in each area.

Oster's results suggest that cost-effective public policy interventions in Africa should focus on decreasing HIV transmission rates. Substantial strides could be made toward curbing the HIV epidemic in Africa by simply treating other STIs using inexpensive, off-patent drugs.

Transmission Patterns

Oster's primary analysis uses a model of the HIV epidemic to predict the growth of the epidemic across areas. The model incorporates information about sexual behavior across groups and the biological parameters of the epidemic, transmission rates in particular. The data on sexual behavior suggests how often sexual partnerships occur. The data on transmission rates indicates the chance of infection in cases where partnerships do occur. Together, these two pieces of information provide information about how the HIV rate in the model changes over time.

The model is based on groups of individuals, with a group being defined as all people of the same gender, age, and marital status. Behavior for these groups is observed and tracked over time. The model simulates the life span of the epidemic (on average, about 20 years), with each period being one year. At the end of each period, predicted HIV rates are observed for each group. The model then simulates the epidemic again, using the new predicted HIV rates as the starting point.

Four primary pieces of data on sexual behavior were used in the model: 1) data on the share of people having premarital or extramarital sex, 2) data on the number of partners for those having extramarital sex, 3) data on condom use, and 4) data, for men only, on the share of nonmarital partners who are commercial sex workers.

Data for Africa came from the Demographic and Health Surveys (DHS) for 14 countries, which have been conducted since the late 1980s. Individuals in these surveys were asked about their sexual behavior, including questions about premarital or extramarital sex, and about their number of partners. U.S. data comes from the General Society Survey, which has been conducted since 1972. Besides basic demographic information, individuals were asked detailed questions about their nonmarital sexual partners.

Comparing the United States and Africa indicates that, with the exception of condom use, sexual behavior does not systematically vary across the two areas, suggesting it may not play a major role in explaining the difference in HIV infection rates. What does vary significantly is the transmission rate.

After comparing the HIV rate predicted by the model to the HIV rate in 14 countries in Sub-Saharan Africa and the United States, Oster found that her estimates closely parallel actual infection rates in the two regions.

For Africa, Oster's model estimates an HIV infection rate of 12.7 percent, whereas the actual estimated prevalence is 11.8 percent. For the United States, her estimate is 0.23 percent, whereas the actual estimated prevalence is 0.15 percent. In addition, Oster's model predicts that if the United States had the same transmission rate as Africa, the predicted HIV rate would be around 12 percent, whereas if the two regions had the same sexual behavior the HIV rates would be similar to their true levels. This strongly suggests that transmission rates play the more important role in explaining the differences between the two areas.

Oster then used the same model to predict HIV rates for 14 countries within Africa separately. While transmission rates of the virus do not seem to vary across Africa, the data indicate that sexual behavior varies substantially. In addition, there is significant evidence that the virus began to affect some countries earlier than others. In particular, countries in East Africa experienced high rates of the virus by the early 1980s, while there were virtually no cases in West Africa until the mid-1980s.

Intervention Models

Differences in sexual behavior and differences in epidemic timing were important in explaining the variations across countries. Once differences in sexual behavior and differences in epidemic timing are taken into account, the model predicts HIV rates that are close to the actual HIV rates estimated for each country.

Oster used her model to evaluate two types of interventions that might prevent further epidemic spread: 1) interventions to decrease HIV transmission rates; and 2) interventions to decrease the frequency of risky sexual behavior. Both interventions would slow the course of the epidemic, but interventions focusing on reducing transmission rates would be substantially more effective.

Oster first considered an intervention that would decrease HIV transmission rates by treating untreated (bacterial) STIs. An intervention of this type conducted in Mwanza, Tanzania decreased HIV transmission rates by 25 percent in male-tofemale transmissions, and 36 percent in female-to-male transmissions at a yearly cost of $59,060 (in 1993 dollars) for 150,000 individuals. For 14 countries, such an intervention over 10 years would cost approximately $1.07 billion (in 2000 dollars).

The second type of intervention addresses risky sexual behavior. Oster analyzed this approach using a scaled-up version of an educational intervention in Uganda. Survey data from Uganda indicate that the large-scale intervention run during the 1990s decreased the percentage of women having premarital sex (from 35 to 22 percent), women having extramarital sex (from 6 to 3 percent), and men having extramarital sex (from 23 to 16 percent). There also was evidence of decreases in the number of partners for people having premarital or extramarital sex. This reduction in risky sexual behavior was achieved at a cost of $180 million over 10 years. Aggregating up to a population of 14 countries, this suggests an overall 10 year cost of approximately $2.8 billion (in 2000 dollars).

How cost-effective are these interventions for Africa overall? Oster estimates that an intervention focused on treating STIs would save 291 million life-years and avert 13 million new HIV infections (approximately 25 percent of total infections over the next decade). This type of intervention would cost $3.67 per year of life saved, or about $78 per infection. Interventions designed to alter sexual behavior are slightly less effective, preventing 6 million infections at the cost of $16.82 per life-year, or $436 per infection.

Feasibility

In 2002, total spending on HIV/AIDS interventions in low and middle-income countries was $3.7 billion worldwide, half of which went to Sub-Saharan Africa. Expanding the STI treatment program to cover the entire continent would cost approximately $300 million per year. Fully a quarter of new infections in 2002 could have been prevented for just 15 percent of the total spending in Africa, suggesting that this type of intervention is financially feasible.

Treating other STIs also may be feasible because it may be easier to convince people to take part. Encouraging changes in sexual behavior is challenging. Since many STIs are highly uncomfortable, there may be much higher levels of compliance.

Oster's results suggest that preventing HIV, either through treating other STIs or through educational intervention, is more cost effective than treatment of HIV using antiretrovirals. Although treatment of HIV in the developing world is a popular cause, even generic versions of the treatment can cost as much as $1 per day in drug costs alone.

"There is no question in my mind that not enough money is being spent on AIDS in Africa," says Oster. "With unlimited funds, treatment is a feasible and important intervention. However, with the limited existing budget, more lives will be saved with prevention than with treatment."