Africa's HIV rate may be grossly exaggerated

[Guest guest]

HIV Rate Still Pretoria's Secret

http://www.wired.com/news/medtech/0,1286,62114,00.html?tw=wn_story_top5

multiple links

 

By Megan Lindow | Also by this reporter Page 1 of 2 next »

 

02:00 AM Feb. 03, 2004 PT

 

JOHANNESBURG, South Africa -- This country's latest HIV/AIDS battle is not

about getting the government to provide anti-retroviral drugs, or about

breaking the pharmaceutical giants' stranglehold on the drug patents. It's

about computer-generated statistics.

 

Measuring the extent of Africa's AIDS epidemic never has been easy,

particularly in the most underdeveloped and war-torn regions of the

continent. Lacking hard data, United Nations' demographers usually rely on

computerized modeling programs to estimate the mortality and infection rates

for most countries.

But while these numbers were generally believed by experts to be reasonably

reliable -- if not exactly accurate -- prominent South African writer Rian

Malan has sparked outrage by claiming that the number of AIDS deaths has

been wildly, and deliberately, inflated.

 

An army of professional " doomsayers, " he implies, are sitting at their

computers in Geneva, gleefully churning out ever-more-dire AIDS estimates

for the continent, and ignoring all evidence that the much-vaunted

apocalypse is simply not happening.

 

Africa's AIDS epidemic has been reduced to " something of a computer game, "

Malan wrote in an article published last December in the British magazine

The Spectator. " When you read that 29.4 million Africans are 'living with

HIV/AIDS,' it doesn't mean that millions of living people have been tested.

It means that modelers assume that 29.4 million Africans are linked via

enormously complicated mathematical and sexual networks (to the epidemic). "

 

The story has generated a steady stream of outraged opinion pieces in the

country's newspapers. Nathan Geffen, national manager for the Treatment

Action Campaign, South Africa's most prominent AIDS advocacy group, wrote a

19-page summary attacking Malan's arguments and pointing out a number of

errors and distortions in the story.

 

Geffen accuses the celebrity author, best known for his 1991 confessional

memoir, My Traitor's Heart, of flirting with pseudoscience and trying to

make a name for himself as a " whistle-blower on exaggerated epidemiological

estimates. "

 

Nonetheless, some recent studies have suggested that HIV/AIDS might not be

as widespread as previously believed. For example, a Kenyan survey of 8,561

households, which was released earlier this month, found the prevalence of

HIV among adults to be around 6.7 percent, as opposed to the 9.4 percent

predicted by UNAIDS. Previous surveys in Mali and Zambia have shown similar

patterns.

 

South African scientists, meanwhile, have been refining their computerized

model for producing HIV/AIDS estimates, the ASSA 2000, and predict that when

the updated version is released in late February, it will generate numbers

about 10 percent lower than current figures.

 

Malan's skepticism was born when he began examining a UNAIDS estimate that

250,000 South Africans died of AIDS in 1999. Later on, a more sophisticated

precursor to the South African ASSA 2000 model reduced that number to 92,000

deaths.

 

In contrast to the more generalized UNAIDS model, which has to be simple so

it can be applied in different countries where data is limited, the ASSA

model was calibrated using more detailed input specific to South Africa.

 

Researchers are quick to admit that the numbers they produce are only

estimates and should be interpreted as such. These inconsistencies are still

relatively minor, they argue, and reflect an ongoing process of refining and

updating their models, rather than any conspiracy to inflate the numbers in

order to gain funding and prestige, as Malan implies.

 

" I'm thinking of writing an article called 'Explaining Computer Modeling to

Rian Malan,' " quips one researcher, who accuses the author of distorting the

truth by misinterpreting the data.

 

" The nature of statistics is that we don't know, " said Mary Crewe, director

of the Centre for the Study of AIDS at the University of Pretoria. " Modeling

is to some extent guesswork ... and in a way it doesn't matter if you're

working on a figure of 10 percent or 20 percent of the population. It's

still an appalling number of people who are dying. "

 

 

 

 

While rough estimates may be good enough to show the broad patterns of the

epidemic, professor Carel van Aardt, research director of the marketing

research bureau at the University of South Africa, emphasizes that more

precise data is needed to plan treatment for those who are infected, and to

anticipate and respond to the disease's impact on the economy.

 

South Africa has an advantage over most of its neighbors in that the country

tracks data such as public surveys and death records against which it can

compare the output of computer models. For much of the rest of Africa,

however, the World Health Organization and UNAIDS provide the only data,

largely in the form of computer-modeled estimates. Using a program called

EPP (Estimation and Projection Package), demographers enter results gathered

from testing pregnant women at clinics in order to calculate an estimated

prevalence among the broader population.

By necessity, researchers acknowledge, a number of assumptions are thrown

into the equation about peoples' sexual behaviors, how long they will

survive with the virus and other considerations.

 

The prevalence figures are then combined with these assumptions in a model

called Spectrum, which produces estimates on the number of people infected,

AIDS deaths and orphans, said John Stover, vice president of The Futures

Group, and one of the designers of the model.

 

" We do consider the uncertainty associated with each of the assumptions used

in this work, and combine these sources of uncertainty into a final figure, "

Stover said.

 

As more hard data on HIV/AIDS becomes available, van Aardt said, researchers

increasingly are able to test the accuracy of their models and improve the

design of these models by incorporating the new knowledge.

 

" A computer model is only as good as the data and the assumptions, " he said.

" With HIV/AIDS in South Africa, you've got a lot of hard data sets, but

often the data is flawed to some extent ... and the chain is only as strong

as the weakest link. "

 

Just as meteorologists, once notoriously inaccurate at predicting the

weather, have incorporated an improved understanding of weather systems into

new forecasting models, demographers now are gradually improving their

models through trial and error, van Aardt said.

 

The ASSA 2000 model, for example, is being updated to reflect improved data

on South African fertility rates, as well as a 2002 survey based on saliva

samples from nearly 9,000 South Africans, conducted by the Human Sciences

Research Council of South Africa.

 

At the end of the day, van Aardt says that despite some inaccuracies in

Malan's story -- most notably, van Aardt contends, some misstated data on

deaths -- the writer has done the country's researchers a favor.

 

" When a lot of people start to believe a series of ideas, the best things is

for one person to start chiseling away at those assumptions, " he said. " With

him asking, 'How sure are we that our statistics are correct?' it's forcing

a lot of guys to go have another look at their models. "

 

Story Tools

 

See also

Legal Battle Over Chat-Room STDs

Cool Science, for Adults Only

Reaping New Meds From Old Cures

Check yourself into Med-Tech

Today's Top 5 Stories

Stacking the Deck Against Science

HIV Rate Still Pretoria's Secret

The Place Where Tech Stands Still

Matrix Plan Fuels Privacy Fears

SCO Fights Back With New Site

Tech Jobs Partner

never settle.