Panda on a crusade to force disclosure of SA’s Covid models
The group wants the models to be examined to see whether the predictions they made about the potential number of deaths from the pandemic were reasonable and based on information that was available at the time
In a bid to keep modellers of the Covid-19 pandemic accountable for their fatality predictions, the group of actuaries, economists, statisticians and doctors known as Pandemic Data & Analytics (Panda) has issued a number of requests relating to the promotion of access to information (Paia) to multiple academic groups.
The group wants to see the calculations and workings of the models that predicted between 20,000 and 351,000 Covid-19 deaths in SA, though the current consensus in the newer models is about 40,000 fatalities.
Panda believes the fatality predictions for Covid-19 in these models played some part in the decision to lock down SA and that at the very least the calculations need to be openly accessible.
Panda argues that these models should be scrutinised to see whether they were reasonable and based on the science of the time, or made wild claims, such as one model, shared with the president in March, which predicted 350,000 deaths.
Shayne Krige, a lawyer working with Panda, says academics can’t just walk away without accountability from outrageous fatality claims that may have led to a lockdown.
The Paia requests were sent to the SA Centre for Epidemiological Modelling & Analysis at Stellenbosch University (Sacema) as well as to the Health Economics & Epidemiology Research Office (HE2RO) at the University of the Witwatersrand and the Modelling & Simulation Hub, Africa (Masha) at the University of Cape Town. Also included was the Actuarial Society of SA (Assa), which predicted a range of 48,000 to 88,000 deaths.
Assa has emailed Panda’s Nick Hudson, discussing its model and answering questions, but the society has not yet released all the calculations behind its model, which is now being updated.
Juliet Pulliam, the head of Sacema, responded to Panda this week and shared information. But as the three models she shared are now outdated, Pulliam said the current model used by government must be accessed through the National Institute for Communicable Diseases (NICD), which co-ordinates the work done by the UCT, Wits and Stellenbosch teams.
Masha and HE2RO have yet to acknowledge the Paia application, which was sent on June 2, says Krige. He expects some parties will use the legally allowed 60 days before responding.
Panda has yet to approach the NICD for the model now being used, but will do so this week.
Regarding Sacema’s release of its early models, Hudson writes in an open letter: “Tucking the models away in obscure parts of the internet does not, in our view, constitute making the models “publicly available”. No links to the models appear on Sacema’s website or in any public documents that we are aware of.”
Speaking of Sacema’s earliest model, which predicted between 87,900 and 351,000 deaths, Hudson says: “Given the lack of sophistication of your models and the results they produced, it is hardly surprising that Sacema has turned its back on these models and has had to be forced to endure the embarrassment of public disclosure.”
Sacema has explained that those numbers were based on the best information at the time – something Panda disputes.