Statistics help shape COVID-19 response
How statistical modelling is predicting the trajectory of COVID-19 in SA
Statistical modelling is being used worldwide to help governments formulate their coronavirus (COVID-19) responses by estimating the rate of the virus transmission.
The models produced by the South African COVID-19 Modelling Consortium, made up of a group of experts from government institutions and several universitybased institutions, are primarily being used by government to estimate the trajectory of the virus in South Africa. This will help the country plan and scale up the response to the pandemic.
According to the consortium, COVID-19 is estimated to infect one million South Africans and result in the death of 40 000 people.
The consortium includes experts from the National Institute for Communicable Diseases; the Modelling and Simulation Hub Africa from the University of Cape Town; the DSI-NRF Centre of Excellence in Epidemiological Modelling and Analysis at the University of Stellenbosch; and the Health Economics and Epidemiology Research Office at the University of the Witwatersrand.
Statistical modelling
Dr Yogan Pillay, the former Deputy Director-General for Communicable and Non-communicable Disease, Prevention, Treatment and Rehabilitation at the National Department of Health (NDoH), says statistical models are used
to estimate or predict what can happen in the future. They are not accurate but, using available data, they are predictions.
He explains that there are three types of models that are useful in planning the country's response to COVID-19. These are predictive, intervention and spatial models.
Mathematical equations are used in conjunction with available data with a number of assumptions to create predictive models, which estimate how many people are likely to become infected and when the epidemic will peak.
Intervention models estimate the impact of various interventions on the rate of transmission that is being modelled.
Spatial models use geospatial maps to show where the objects under review, in our case COVID-19, are found, such as hotspots, he says.
The data used in statistical modelling must relate to the problem that is being modelled. The data used for COVID-19 modelling includes the number of current infections, the age structure of the population, the estimated attack rate and other factors.
Different models that have been made public, therefore, have different estimates of the number of infections at the peak, according to Dr Pillay. This is because the models depend on the assumptions made by the people testing them.“One model suggests that we are likely to reach one million infections – others estimate a much lower number,” he says.
Consortium predictions
The South African COVID-19 Modelling Consortium uses two main models in making its calculated forecasts.
The National COVID-19 Epi
Model estimates the total and reported incidence of COVID-19 in the nine provinces. According to the consortium, the outputs of the model may be used to inform resource requirements and predict where gaps could arise, based on the available resources within the South African health system.
The National COVID-19 Cost Model has been developed using inputs from a range of health economists in South Africa, with contributing data from existing sources that were adapted to represent the type, number and prices of ingredients required in the country's COVID-19 response.
The consortium estimates that government could spend anything between R26 billion and R32 billion in six months (April to September) in the fight against
COVID-19. Of this, between R10 billion and R15 billion could be used by the NDoH alone, while the rest will be used by provincial health departments.
The consortium says it has projected the required budget for the first six months of the COVID-19 response under both pessimistic and optimistic scenarios, covering the incremental cost of personal protective equipment (PPE), additional ICU and hospital beds and staff, additional primary healthcare staff, ventilators, oxygen, drugs at all levels of care, isolation facilities, testing and surveillance and Port Health budgets.
Excluded are the costs of setting up and running field hospitals, oxygen delivery equipment, additional testing platforms beyond the currently planned ones and additional National Health Laboratory Service staff. Stipends for additional community health workers to carry out screening activities are excluded as these are funded by a donor's budget; their PPE and other equipment are, however, covered.
Most of the models estimate that the peak period of the infection rate will be between July and August. However, they also estimate that different provinces are likely to peak at different times, says Dr Pillay.
The provincial outlook
The models expect that at best, the worst affected province will be Gauteng, peaking at more than 150 000 infections in August. KwaZulu-Natal is expected to follow, with more than 100 000 cases while Western Cape, Eastern Cape and Limpopo are each expected have more 50 000 infections. North West, Northern Cape, Mpumalanga and Free State are all projected to have less than 40 000 infections per province.
The more pessimistic models expect that Gauteng will surpass the 200 000 infections mark, followed by KwaZulu-Natal with more than 150 000 infections and Western Cape with less than 100 000. Eastern Cape, Limpopo and Mpumalanga could see more than 50 000 infections in each province.
Northern Cape, Free State and North West are expected to have less than 50 000 infections at the peak.
Dr Pillay says that the lockdown played a very crucial role in slowing down the virus. Had the lockdown not been implemented, estimates are that the infection rate would have hit its peak earlier than is currently being projected with many move people being infected.
“One model estimates that the lockdown decreased the number of infections at the peak of epidemic curve and moved the peak outward by about six weeks,” he says.
Further projections from the consortium show that had South Africa not gone into a strict lockdown, the country's infection rate would have peaked at 1.2 million infected people by May this year.
While the consortium predicts one million infections in South Africa, a more hopeful outlook is that because of the strict lockdown, South Africa's infection rate may peak at an estimated 800 000 infections.
Dr Pillay says: “There are many factors to be taken into account to bring the pandemic under control. The lockdown is only one intervention.
Government has developed a prevention toolbox with many interventions, with the lockdown being one of them. The others include social distancing; hand hygiene; use of face masks; sanitising surfaces; screening and testing; and isolating people with symptoms, among others.