/ 21 September 2001

HIV time bomb implodes life expectancy in South Africa

Belinda Beresford

The release this week of the most up-to-date projection analysing the impact of HIV/Aids on South Africa shows that cumulative deaths from Aids-related illnesses are expected to exceed one million in just three years time if there are no interventions.

And it confirms that the epidemic may be too firmly entrenched for its spread to be drastically curtailed in the medium term.

The Actuarial Society of South Africa’s latest statistical model ASSA 2000 predicts that without major behavioural or medical changes, life expectancy is likely to fall to 41 years by 2009 calendar year. By then it expects that 16% of the total population will be infected with HIV.

Life expectancy is only forecast to start rising in 2015, when cumulative Aids deaths will have reached nine million South Africans and there will be more than 1,8-million Aids orphans.

University of Cape Town Professor Rob Dorrington, who is the guiding force behind the ASSA model, says that when run retrospectively ASSA 2000 suggests that early estimates of HIV infection among pregnant women were overstated. This implies that the increase of the epidemic has been swifter than previously thought.

The ASSA Aids Committee has produced a summary table which includes behavioural changes such as a doubling in condom usage over the next five years and a 15% decrease in sexually transmitted diseases over the short term.

Depressingly, such changes make little difference to the “no change” scenario. It would mean life expectancy in 2008 of 43 years while accumulated Aids-related deaths in 2015 would be just 10% lower.

All these figures come from the “full” version which aggregates across the country as a whole. Attempts to model the impact of the epidemic at a provincial level have been hampered by the refusal of the national government to release the disaggregated provincial figures.

Dorrington says that he has been given provincial data up to 1999 and has been told he can have the later information one month after the Department of Health publishes its full report. “The report has been at the printers for the last three months,” he said.

In common with most models, ASSA 2000 extrapolates from the annual surveys of pregnant women attending public antenatal clinics. Unlike many other countries, South Africa presently has no other coordinated nationwide studies into HIV prevalence.

Countries such as Thailand for example, produce their estimates of the Aids epidemic by testing a number of different groups, such as students and soldiers.

The ASSA model checks its forecasts against real data as it becomes available for example the predictions of Aids related deaths against real mortality data.

Some of the refinements in the latest model are that it can incorporate behavioural responses such as condom use into its forecasts.

Statistical analyses on the same data can differ dramatically depending on the underlying assumptions made in constructing each model.

A quick demonstration of this was given by Dorrington at the Aids conference in Durban last year. He showed that for last year the model produced by life assurer Metropolitan Life predicted HIV prevalence of 9%, compared to 13% for the United States Bureau of Census model and 11% from the United Nations model.

Looking a decade later the differences are even more extreme: 16% by the UN, 25% by the US Bureau of Census, and 15% from Metropolitan Life.