More than 40,000 deaths predicted: behind SA’s Covid-19 models

After endless pleas by respected scientists, health minister Zweli Mkhize has finally released a model of what the Covid-19 epidemic could look like in SA.

The numbers, revealed on Tuesday night by the SA Covid-19 modelling consortium, predicts that by November, SA will face between 40,000 Covid-19 deaths in an optimistic scenario and 48,000 in a pessimistic one.

This is significant, not least because it reveals that the fatalities are likely to exceed the average number of deaths per year (about 30,000) from tuberculosis, SA’s biggest killer.

But it also opens a window into the health challenges.

The number of beds in intensive care units (ICU) — a measure which includes the staff needed to manage them in SA — could be insufficient by June; at the peak of the epidemic, the number of beds needed would be 10 times more the number of beds available. The model says there are 3,300 ICU beds available in the private sector and the state. Even before Covid-19, there were instances when there weren’t enough ICU beds for patients in state hospitals.

However, the modellers were at pains to point out that there is considerable “uncertainty” in the models, as they rely on assumptions rather than hard facts, and these assumptions are sometimes made by considering the global evidence and coming to a “consensus”.

As for infections, the pessimistic model expects a peak of 1.2-million people at a peak in July in the pessimistic scenario. In the optimistic scenario, SA will have 1-million at a peak in August. The number of infected cases in total over the year will run into many millions, but 75% of people will have no symptoms and the fatality rate is very low, says the model.

The consortium was convened by Dr Harry Moultrie, an epidemiologist at the National Institute for Communicable Diseases (NICD), and includes Prof Juliet Pulliam, the head of the SA Centre for Epidemiological Modelling & Analysis at Stellenbosch University, doctors and statisticians Gesine Meyer-Rath and Brooke Nichols, and Dr Sheetal Silal of the University of Cape Town’s modelling and simulation hub.

Their model shows that the lockdown pushed the peak of cases from May 30 (without any lockdown) to late July, the middle of winter.

It predicts that of the 25% of people who experience symptoms, 95% of them will have mild symptoms, 2.28% will have severe symptoms, and 1.8% will be considered critical. Ultimately, between three and four people per 1,000 infected will die — a fatality rate of 0.3%-0.4%.

In her presentation, Silal stressed that models change often, as new global and local data becomes available, while behaviour can also alter outcomes. “There is considerable uncertainty [about] how it is going to fully play out in future,” she said. “The model is likely to be updated many times.”

Mkhize also stressed this point: “The numbers are not fixed. They are giving us an idea of how bad things are and, as long as you know how bad they can be, you put mechanisms in place to try and reduce the severity.”

The short-term models, which look at monthly infection numbers, are updated weekly. Currently, there is a projection that SA will have around 30,000 cases by the end of May — but even this scenario ranges from 18,710 to 54,540 infections.

Either way, to reach this number, the number of cases would have to nearly double from Tuesday’s level of 17,200 infections.

While this consortium isn’t the only group providing data, it is one of the few that has shared its projections with the media. The ministerial advisory committee, the National Ventilator Project, the National Treasury, the health department and the National Health Laboratory Service all have sight of this model, the consortium says.

Asked directly by journalists if the country should move out of lockdown, Sheetal said: “It is not up to modellers to make the decision to move in and out of lockdown. It is an economic and social decision. It requires a consultation aspect.”

Still, the new fatality number is much lower than the upper estimate of 88,000 deaths calculated by the Actuarial Society of SA; its optimistic scenario predicts about 48,300 deaths.

Asked to explain how the consortium’s model calculated deaths, Silal said it applied the death rates of different ages of Wuhan patients to the infected SA population. As more data on SA deaths becomes available, the model will be updated.

“When the epidemic [progresses] and we have our own death data with respect to age, we can adjust the model. This may lead to updated projections,” she said.

Still, the model has to adapt to how people behave: if social distancing, frequent hand washing and the wearing of masks become more commonplace, transmission could reduce by 10% in the pessimistic scenario, and to as much as 20% in the optimistic scenario, from June.

Moultrie said the models would be made publicly available in the interests of democracy, “so that the public can engage and see and criticise”. Already, the model has been reviewed by a number of groups abroad and locally. Moultrie said it was important that it remains open for scrutiny and robust engagement. “We look forward to … constructive engagement,” he said.

Though Moultrie said he’d been warned he’d need a thick skin, a transparent debate would help strengthen the model’s projections.