The health department uses the numbers of the Thembisa model as the “basis for the HIV estimates that South Africa submits to UNAids [for their] Global Aids Monitoring [process],” says Leigh Johnson, the lead developer of the data project, based at the University of Cape Town.
That’s because, although measured numbers are real, they can be counted only in a sample of people, while a model can give an estimate for a country’s whole population, making it easier to use the numbers to plan ahead.
Actual numbers tell us what things look like now, or have looked like in the past, but they can’t tell us what things will look like in the future.
A data model is a set of calculations in which symbols, for example x, y or z, are used as placeholders for numbers of which the values can change.
These placeholders are called variables (because their values aren’t fixed). By putting known numbers to these variables and then adding, subtracting, dividing or multiplying them in specific ways, you can work out the value of an unknown variable.
This means a model is a way to use maths to find out how things in a system work together to bring about a specific result.
The numbers that come from a model are calculated rather than measured, though, so they’re estimates of what’s going on and how numbers will look like a few years ahead.
Yogan Pillay, who was the deputy director general at the national health department between 2008 and 2020 and who managed the department’s HIV targets during this time, says the agreement with UNAids is that South Africa uses the Thembisa model, rather than the Spectrum system like many other countries in sub-Saharan Africa because it is “based on South African data that is routinely collected, and over the years, we’ve found it to be more reliable [for us] than Spectrum”.
Spectrum is also a mathematical model, but its calculations and variables are set up somewhat differently from that of Thembisa’s.
Pillay is currently the head of HIV and TB delivery at the Bill & Melinda Gates Foundation.
The Thembisa model estimates that 95% of adults with HIV — people of 15 years and older — have been diagnosed. Of these, 78% are on ARVs, and of the ones on treatment 91% are virally suppressed.
READ MORE | One in four people with HIV not on treatment – new estimates
So, according to these numbers, the country is a long way from reaching the second of the three targets.
But real-life studies are another way to get an idea of how well — or how bad — things are going.
Two such studies in South Africa are the Human Sciences Research Council’s (HSRC) National HIV Prevalence, Incidence and Behaviour Survey and the 2022 Antenatal HIV Sentinel Survey.
From the HSRC’s survey, in which about 70 000 people took part and about a third of them gave a blood sample to be tested, it looks like South Africa is in a better position than what the Thembisa model shows: of the 90% of people with HIV who have been diagnosed, 91% are on treatment and 94% are virally suppressed.
Numbers from the antenatal survey, in which just under 38 000 pregnant women across South Africa participated, are looking even better. According to those figures, the country has already hit both the first two 95 targets, and is at 91% for the third.
But the numbers are different from those of the Thembisa model, because in a survey, data is collected from a group of people from a certain population.
For instance, the antenatal survey only counts HIV infections, diagnoses and treatment numbers in pregnant women, whereas the HSRC survey’s participants were from general households.
However, says Johnson, “models and surveys both have their place”.
Putting it all together
Real numbers from surveys such as that of the HSRC’s are used to calculate the Thembisa model’s figures.
Because models are used to make predictions about the future, it’s important to make sure that the data that’s used to calculate estimates is realistic and accurate, he says.
“Using many different data sources in a single model helps to lower uncertainty [in the estimate]”.
That’s why the model uses data not only from the latest HSRC survey, but also numbers from the five previous ones, going back to 2005. The survey was conducted in 2005, 2008, 2012, 2016 and 2017.
In turn, the HIV Investment Case, which looks at how to get the best bang for our buck when trying to prevent new HIV infections, uses numbers from the Thembisa model, explains Johnson.
According to these figures, South Africa has hit the first target, but sits only at 84% on the second and at 93% on the third.
Because the Investment Case usually comes out six to nine months after the Thembisa figures, on which it’s based, Johnson explains, “it might [therefore] not match the most recent [model] estimates”.
How do the estimates help the health department and health workers?
While there are different numbers in South Africa for the 95-95-95 targets, what we do know is that “we have saturated the number of people who have tested for HIV at least once, but people need to be retested”, says Pillay.
Pillay added:
The big challenge has been to ensure that everybody who’s tested positive is initiated on ARVs and that they’re supported to stay on their medicine so they can remain virally suppressed.
He says the best way to use 95-95-95 data to decide how the health department should use its resources is to also look at what the data says about progress in specific groups of people.
“The national averages do cover the discrepancies between women and men, and women, men and children, with the latter doing the worst,” Pillay explains.
For example, in 2023 almost twice as many new infections were in women than in men, and about 6 500 of the total number of new cases were among children.
This means that if provincial health departments don’t spend more effort and money on preventing infections in, for instance women, it will hold us back from reaching the UN targets.
We’re not doing well with curbing new infections in general, says Pillay.
ALSO READ | New HIV prevention injection to be made available to low and middle income countries
The target is to, by 2030, have reduced new infections by 90% compared with 2010 figures. In 2010, South Africa had about 350 000 new infections, so the goal should be to have only 35 000 in 2030. In 2023, though, the country had 150 000 new cases — almost five times more than where we want to be.
Pillay says: “South Africa is failing on the second 95 target [getting people on treatment], and we see many still with advanced HIV disease. Data also shows us that people cycle in and out of treatment, which is another challenge because if they aren’t virally suppressed [from staying on their medicine] they can still transmit HIV to others.”
