Between the known, unknown, and the unknowable: How mathematical modeling of COVID-19 helped South Africa plan and shift scarce inpatient resources

Gesine Meyer-Rath, MD, PhD August 4, 2021

scientists use data to create modeling tactics to help manage Covid-19 treatment.

Similar to many African countries, South Africa—the country where I work as a BU associate professor in global health—reported its first COVID-19 case on March 5, 2020, imported by a traveler returning from Europe. Cases spread rapidly across the country during a first wave from May to July 2020. This was followed by an early second wave in November 2020, fuelled by one of the first variants of concern, Beta. A third wave, dominated by the Delta variant, is currently underway and overtaking all previous versions in terms of cases, hospitalizations, and deaths.

At the time of this writing, the country has reported more than 2.1 million cases (about 35,000 per million population) and more than 63,000 laboratory-confirmed COVID-19 deaths (just above 1,000 per million population). In contrast, only 5 million vaccines have been administered. What this means is, about 3.8% of the population are currently fully vaccinated. A lack of COVID-19 PCR testing kits, especially at the beginning of the pandemic, and a severe lack of intensive-care beds with specialized staff and oxygen, mean that a lot of the deaths due to COVID-19 happened out of a hospital and were not officially attributed to COVID-19. Additionally, people sick from the virus might have found it difficult to access care in time, even where it was available. Excess deaths from natural causes (ie, excluding trauma-related deaths) in persons above one year of age are estimated to have been more than 180,000 since May 2020— about three times the reported COVID-19 death toll. Other countries, such as the US, are finding similar results in terms of excess deaths exceeding reports.

Attempting to get out in front

This article summarizes how mathematical modeling of both the epidemic and the budget needed to contain COVID-19’s impact has helped South Africa plan and shift scarce inpatient resources and ultimately move the needle on survival a little.

In late March 2020, I was part of a group of infectious disease modelers and health economists from local universities who came together to form the South African COVID-19 Modelling Consortium (SACMC). Our aim was to support the South African government in planning and budgeting for COVID-19 related healthcare. For the next 18 months, we developed several tools in response to the most pressing needs of decision-makers in the different stages of the epidemic. This allowed the South African government to plan several months ahead of time.

Putting the tools in place

Our tools included epidemic projection models, several budget impact models, and online dashboards to help government and the public by:

Allowing government departments to quantify the need for inpatient resources, including ICU and general ward beds, oxygen, drugs, personal protective equipment, and staff (by far the largest bottleneck)
Visualizing long-term case projections during the first wave
Monitoring resurgence and tracking developments in cases and admissions during the second and third waves in real-time, and
Providing short-term projections of cases and admissions in the next two weeks during the third wave.

Continuous Covid-19 data monitoring

Given the rapidly changing nature of the outbreak globally and in South Africa, all model projections have been updated regularly. The updates reflected the availability of new data—particularly from South African health surveillance data systems whose coverage was improving continuously—and the evolving government response to COVID, such as:

Changes in lockdown levels
Mobility and contact rates
Testing policy
Contact tracing strategies, and
Hospitalization criteria.

Our growing insights into population behavior required another set of model projection updates— for example, the differences between various population groups’ ability to retreat indoors, work from home, and reduce social interactions. Another example of this is behavioral heterogeneity, i.e., differences in exposure to COVID-19 risk that result from people’s underlying contact rates, which helped us explain the earlier-than-anticipated peaking of the case curve during the first wave in May to August 2020. Results of our models were, in turn, used by policymakers in the National and Provincial Departments of Health and in Treasury to plan and shift resources, and quantify the need for resources such as:

Drugs
Oxygen
Ventilators
Hospital beds
Staff, and
Mortuary containers and graveyard space

We realized early on that the main impact of our work would be through effective communication and dissemination of our findings to not just policymakers, but also to the public. For example, ordinary South Africans could use our real-time analyses to make decisions about when and where to travel, especially as the most severe lockdown restrictions were lifted to allow the heavily impacted economy to regain strength in early May 2020. For this, we developed two online applications. The National COVID-19 Epi Model Dashboard visualized the most important model outputs during the first wave. During the second wave, we created the SACMC Epidemic Explorer, which enabled case tracking and visualization of case resurgence driven by the new Beta variant during the second wave, and the Delta variant during the third wave.

At the beginning of the third wave, we developed additional methodology to forecast the need for hospital beds over the next two weeks. This enabled the addition of hospital capacity where it was most needed.

Exemplary health data infrastructure

South Africa has the benefit of good health data and planning systems, all of which are locally funded, developed, and maintained. They also have an excellent science infrastructure. Data from a multitude of SARS CoV-2 seroprevalence studies in various South African locales over the last few months suggest that 40-50% of the population had been exposed to the Beta variant by May 2021. A network of genomic sequencing labs allowed us to ascertain that by the end of June, almost 100% of SARS CoV-2 positive cases were now due to the Delta variant. We were able to incorporate data from these sources into projections of the timing and size of third waves across all nine provinces.

Through these processes, we were able to make sure our models:

Developed rapidly in an emergency setting and updated regularly with local data
Supported the national and provincial government to plan several months ahead of time
Expand hospital facilities
Allocate budgets, and
Procure and move additional resources where possible.

Moving forward

As the country is in the throes of a third wave that started in June 2021, our modeling consortium continues to serve the planning needs of the government, tracking cases, admissions, and deaths, while also developing models to support the national vaccine rollout.

Dr. Gesine Meyer-Rath is a Research Associate Professor at the Center for Global Health and Development of the Boston University School of Public Health. She is a physician and health economist working on the economics of HIV and antiretroviral treatment in low- and middle-income countries. Her focus is on modeling methods for economic evaluation, including infectious disease modeling, decision analysis, and translating research into recommendations for public policy.