Maravi

(NOSEWEEK) Manipulation of HIV/AIDS Statistics In South Africa

Rian Malan's last ever Aids piece - or so he says...
Issue # 88 - February 2007

I was intrigued to learn that noseweek and I were mentioned at an HIV/Aids conference in Toronto late last year. Treatment Action Campaign honcho Nathan Geffen informed the gathering that “a journalist by the name of Rian Malan wrote a column in noseweek arguing that the Aids statistics were completely wrong, that there wasn’t a serious Aids epidemic in Africa, and that people weren’t dying of Aids in great numbers.

He didn’t dispute that HIV caused Aids. He just thought that this was massively exaggerated. I ended up writing a detailed response, and today I’m glad to say that Malan’s exited the debate. He’s considered to be quite a fool by the South African media, quite correctly.”

Considered a fool, eh? These are fighting words. In truth, I exited the debate because I came under attack by armies of frenzied activists, all bent on portraying me as a grubby, drunken madman whose views on Aids (as expressed in nose52) could not be taken seriously. In truth, I never claimed Aids was not a problem. On the contrary – I described it as a terrible affliction that was claiming countless lives. At the same time, however, it was clear that Aids numbers were being exaggerated and good news suppressed. I stand by that story. Indeed, more good news has emerged in the three years since its publication.

As reported in nose52, newspaper stories stating that millions of Africans die of Aids each year should not be taken literally. These estimates may look authoritative, but they are actually generated by machines entirely vulnerable to the “trash in/trash out” principle – if you put bad data into a computer model you get rubbish out. In the l990s, India repeatedly rejected UNAids’ computer-generated estimates for this very reason. Several African countries were also sceptical. In 2001 I published an article on the subject in an American magazine. My critical views were dismissed as symptoms of derangement and denial. The same would not hold today.

At the time of my last foray into this arena, almost all African Aids statistics were produced by UNAids, which organizes annual HIV surveys in pregnancy clinics across the continent. The results are put into computer models programmed to assume that if a certain percentage of pregnant females come up HIV-positive, it follows that a similar percentage of the general population is probably infected, and that a portion of those hypothetical unfortunates die each year. This form of “sentinel surveillance” suggested an Aids epidemic of dumbfounding proportions in Africa.

In 2004, researchers tried a different sort of survey in Kenya, selecting a representative sample of the population for HIV-testing. The UNAids system had suggested that 15% of Kenya’s adults were HIV-infected. The new method, universally held to be more credible, suggested that Kenya’s real HIV rate was closer to 6.7%. More than half of Kenya’s hypothetical HIV cases disappeared overnight.

Since then, around ten Kenya-style population studies have been carried out elsewhere in Africa, invariably with similar results. In Sierra Leone, estimated HIV prevalence tumbled by 87%. In Ethiopia, by 78%. In Burkina Faso, by 72%. In Burundi, 35%. In Zambia, 27%. In South Africa a 2005 population study pointed to radical overestimation of the HIV rate among racial minorities. Aids researchers had previously claimed that HIV prevalence among whites, for instance, was as high as six percent. The 2005 population study suggested that the real HIV rate was closer to 0.6%. In the coloured community estimated HIV prevalence plummeted from six to 1.9%. In the Western Cape, five out of six previously estimated HIV infections vanished.

This good news was, as usual, ignored by the local media, but some big US newspapers were about to join noseweek in asking awkward questions. The Boston Globe tracked down Jim Chin, creator of the computer model used (until 2001) by UNAids to generate HIV estimates for Africa and Asia. “Chin said he thinks the global rate is inflated by 25% to 40%,” reported the Globe. “Two US health officials working on Aids said they think the global numbers may be 50% inflated.”

Then the Washington Post decided to take a closer look at Rwanda, portrayed in the 1980s as “the epicentre of Africa’s Aids pandemic”, with an adult HIV infection rate as high as 30%. Today, it is acknowledged that the primitive blood tests on which such startling claims rested were defective. When more reliable tests were deployed in the late 1990s, Rwanda’s estimated HIV infection rate came out at 11.2%. And when a population study was carried out in 2005, the rate plummeted to 3%. “Aids deaths on the predicted scale never arrived here,” said the Post, quoting local health officials. “The United Nations has for years overestimated the extent of HIV/Aids in East and West Africa.” “They keep cranking out numbers that you can’t defend,” adds Chin.

On the other hand, the Post harboured no doubts about southern Africa’s Aids cataclysm. Such certainty rests on mortality data from South Africa, the only African country where it is possible to assess the accuracy of Aids estimates against actual death registrations. Elsewhere in Africa, something like one percent of deaths are registered by governments; in South Africa, the equivalent figure is close to 90%. This assessment is accepted by all researchers, who differ only on one issue: when did registration reach its present high levels?

This has been the subject of bitter backroom arguments since the days of the Rapid Mortality Surveillance project, a multi-agency task force set up in the late l990s to get a better handle on Aids’ impact. The project’s studies showed an ominous rise in registered deaths, with more and more people dying at sexually active ages. Aids researchers blamed Aids, but Dr Sulaiman Bah of Stats SA wasn’t so sure.

Bah noted that during apartheid, Pretoria didn’t want the world to know too much about conditions in the Bantustans, where diseases of poverty were rife and infant mortality rates astronomical. Consequently, death registration in such areas was universally acknowledged to have been abysmal (“grossly incomplete”) when the ANC came to power. According to Bah’s calculations, only 37% of deaths in rural areas were registered in 1996 — as compared to 86% in urbanized areas fully plugged into apartheid’s registration machinery.

This was unacceptable to the ANC, and 1998 saw the start of major campaigns to improve death registration, particularly in apartheid’s former dumping grounds. A simplified death certificate was introduced. Home Affairs opened satellite offices in former homelands. Regulations were amended to allow deaths to be registered with tribal authorities. Undertakers were offered subsidies on condition that deaths were properly registered. And so on. Soon afterwards, as we have seen, registered deaths started rising rapidly. Aids researchers assumed Aids was to blame. Bah believed the rise was to some extent an illusion caused by improving death registration. (See nose30)

The Aids bwanas didn’t want to hear this, not in the winter of 2000. At the time, President Thabo Mbeki’s heretical views on Aids were causing global consternation. Some even accused him of genocide. As far as they were concerned, the data gathered by Rapid Mortality Surveillance provided more than enough evidence to substantiate such charges. They wanted to release it immediately, but Bah refused to go along until the critical registration issue had been addressed. In the end, the Aids faction released the data unilaterally, and it wound up on the front page of the Sunday Times, under a banner headline screaming, “Young, Gifted and DEAD.”

These shocking graphs,” said the Times, “show how the number of South Africans who die before they reach the age of 50 almost doubled over the past 10 years – an increase attributed directly to HIV/Aids.”

Sulaiman Bah had no stomach for this sort of fight and quit his job soon after. His cause was taken up by Dr Rodney M Richards, a Colorado-based microbiologist who continued to argue that Aids researchers were distorting the statistics. Richards’ case rested in part on the fact that the most rapid rise in reported deaths in the years 1997-2004 took place not in provinces with the highest HIV prevalence, but in precisely those provinces where death registration had been “grossly incomplete” before the launch of registration campaigns. Consider rural Limpopo, where, in 1997, HIV prevalence amongst pregnant women was found to be 8%. By 2004, Limpopo deaths among sexually active adults had increased by 212%. But in urbanized Gauteng, although HIV prevalence was twice as high (17%), deaths rose by less than half this amount (99%). Clearly something was wrong with the figures.

Richards didn’t contest that a portion of the rise in deaths was attributable to Aids, but pointed out that deaths caused by conditions entirely unrelated to HIV were rising too. Deaths attributed to stroke, hypertension, breast cancer, and renal failure increased by 47%, 56%, 57% and 69% respectively. Deaths from aplastic anaemia were up 141%. Road accident deaths rose 198%. Assault deaths soared 213%.

To Richards, increases of this magnitude for non-HIV conditions could only be explained by improved registration. The Aids lobby said otherwise. It fell to Statistics South Africa to referee this potentially explosive dispute.

After sitting on the fence for years, the agency published a report in May 2006 that came down solidly on Richards’ side. Completeness of death registration, declared Stats SA, rose from around 68% in 1998 to just under 90% in 2003. The agency declined to spell out the consequences, but Richards was willing. “A shift of this magnitude radically alters interpretation of the raw mortality data,” he said, “because it’s not just death registration that’s improving. With each passing year, the population grows, and deaths should increase accordingly.”

Once you factor all this in, says Richards, the 56% rise seen in raw mortality data for the years 1998 to 2004 shrinks to 14%. In the sexually active age bracket (15-64), an apparent rise of 83% shrinks to 30%. Richards acknowledges that a 30% rise in adult mortality is cause for grave concern, but emphasizes that this is less than a fifth of what local actuaries were predicting just five years ago. “If your government and corporations based resource allocations on those predictions,” said Richards, “millions have been wasted.”

Meanwhile, on the far side of the planet, UNAids was struggling with a credibility crisis. Important newspapers were raising doubts about its estimates. Computer modeler Jim Chin was on record saying: “They keep cranking out numbers you can’t defend.” Remedial action was called for. Next thing, an American researcher named Barbara Anderson showed up in Pretoria to participate in what statistics chief Pali Lehohla described as “a cooperative effort between Stats SA and UNAids”.

Anderson and Stats SA’s mortality guru, Dr Heston Philips, reworked the 1997-2004 death registration data, and, in September 2006, published a paper maintaining that completeness of registration among the sexually active actually declined in response to the massive government reforms of l998, from 89% to 82% for females, and far more dramatically from 94% to only 80% for males.

Never mind that this was entirely improbable. It’s the result they were after. The result of this unexplained flip-flop by Stats SA was an instant tripling in perceived mortality rates from 1998 to 2004 – from 14% to 45% for persons of all ages, and from 30% to 83% for the sexually active. “These are huge differences,” says Richards. “This is not science. It is a pathetic exercise that serves no other purpose than to covertly transform SA registered deaths into totals that agree with UNAids’ latest computer model.”

[In fact, much the same sort of statistical manipulation that Transnet’s actuaries indulged in to assist Transnet in defrauding its pensioners. – Ed.]

Needless to say, the world doesn’t care what Richards says. UNAids makes the running in this debate. Within days, Anderson and Phillips’ findings were being quoted as fact in the New York Times and in the speeches of UNAids executives, who could now claim their model’s accuracy had been verified by real-life evidence in the only African country where such an exercise is possible.

I twice wrote to Heston Phillips, Stats SA’s mortality guru, asking him to explain the contradictions between his agency’s May and September papers, and to comment on Richards’ critique. Dr Phillips has not responded, suggesting he is not stupid, just wily.

Any thinking person must find this alarming. The registration of deaths debate may be boring, but its importance cannot be exaggerated. Let’s say you live on a desert island where 100 deaths take place each year, but completeness of registration is only 50%. Statistics will show 50 deaths a year. If government gets its act together and starts recording all deaths, however, registered deaths will double, and naïve observers will say something terrible is happening here.

For twenty years, UNAids’ computer modelers have been telling us that something terrible is happening in Africa – that requires countless billions of dollars in donations to put right. South Africa is the only country where such claims can be checked, provided we are willing to face the truth. The Aids bwanas seem disinclined. The prominent researchers who shouted Sulaiman Bah down in 2000 have been promising for years to publish a scientific explanation of their reasoning. Nothing has materialized. As far as I am aware, the SA government’s post-1998 registration reforms have never, not once, been mentioned in the massive literature generated by Aids researchers. We are simply told rising mortality is caused almost entirely by Aids, and that raising doubts is denialist.

Last May, Stats SA took a position that implicitly trashed this view, only to reverse itself six months later. We would have no case if the first publication was a mistake, but Stats SA has made no such claim. It simply got into bed with UNAids and lent its reputation to an exercise designed, according to Richards, to lend spurious legitimacy to the UN’s computer model.

What’s wrong with a bit of fibbing, if that’s what it takes to raise money for a good cause? Well, apart from conning donors who mght have spent their money better elsewhere, on the scale we’re talking about, it seriously distorts social priorities and government planning. Told that they face appalling problems, governments have diverted pitifully scarce resources from other needs to combating an Aids threat that in several instances has turned out to have been grossly exaggerated. An example: faced with UNAids’ warning in the nineties that their teachers were about to be decimated by Aids, several African governments responded by training armies of replacements. The result, according to UK researcher Paul Bennell, is millions wasted and a glut of unemployed teacher trainees in countries like Botswana and Swaziland.

Meanwhile, the poor continue to die of ordinary diseases that could be cured for a few cents if medicines were available.

But what’s bad for ordinary people is fine for the Aids industry. Computer modelers’ estimates of impending disaster provide the basis for the Aids industry’s funding demands. In Africa, only a fraction of Aids-related aid goes to medicines and health care. Much of the rest is spent on absurdly expensive conferences, endlessly duplicative social research and “soft” interventions like awareness campaigns.

Two years ago, Stats SA staff were privately telling me they found this troubling. Clearly, something has changed. An explanation is awaited.

Meanwhile, I hew to the position set out way back in nose30 (and, in greater detail, in nose52): Aids is the most political disease in human history, and almost all, if not all, Aids statistics are contaminated by the self-interested manipulations of Aids careerists of Nathan Geffen’s ilk. I prefer to trust my bones, which maintain that Aids is a grave problem. But, until we do an honest job of our statistics gathering and analysis, exactly how grave is anyone’s guess.

Estimates on HIV called too high
New data cut rates for many nations
By John Donnelly, Globe Staff | June 20, 2004

WASHINGTON POST FOLLOW-UP (APR 2006)

Labels: HIV/AIDS, RIAN MALAN, SOUTH AFRICA, STATISTICS

posted by MrK @ 7:47 PM