IBM machine learning tech helps Gauteng track COVID-19 infections

Read time 4min 20sec
Ashley Gritzman, research scientist at IBM Research Africa.
Ashley Gritzman, research scientist at IBM Research Africa.

IBM Research Africa has developed machine-learning-powered technology to help the Gauteng provincial government monitor, track and predict coronavirus (COVID-19) infections, to provide government with response actions on limiting the spread of the pandemic.

As SA braces for the peak infection period, COVID-19 cases have climbed to 118 375, with over 60 000 recoveries and the mortality rate at 2 292, as of this morning.

The Western Cape continues to record the highest provincial numbers, accounting for 48% (56 780) of cases, while Gauteng has the second highest rate, accounting for 24% (28 746) of SA’s positive cases.

To better respond to the rapidly increasing COVID-19 infection rate in the province, the Gauteng Provincial Government collaborated with IBM Research Africa, the University of the Witwatersrand (Wits), private health institutions and other stakeholders to build an epidemiological model of the province’s COVID-19 infection trends and provide a roadmap for containment measures, using the latest technologies.

The announcement was made during the virtual IBM Think Digital Summit Africa this week.

“With the ongoing battle against the spread of COVID-19, timely and accurate data analysis can help governments stay ahead of new transmissions and flatten the curve,” explained Ashley Gritzman, research scientist at IBM Research Africa.

“The platform is being used by the Gauteng Provincial Government in their war room to understand the progression of COVID-19 and then make informed decisions on how to better respond to it, using machine-learning-powered spatial monitoring, reporting and predictive analytics tools focused on various elements of the COVID-19 infection trends.”

The technology, run on IBM’s cloud compute infrastructure, is based on a dashboard, which shows snapshot views of how the pandemic is spreading in Gauteng.

Focusing on three aspects of the infection rate, the technology shows where the hotspots are within the province – to make them a priority in mass screening projects. Secondly, it helps government to understand the risk factors of COVID-19 – such as high-density dwelling areas and lack of access to healthcare facilities – which may contribute to the spread of COVID-19, explains Gritzman.

The third element, dubbed the crown jewel of this technology, is its ability to predict and understand how the virus is likely to spread, right down to the ward level – this enables the Gauteng province to understand how it can allocate medical resources, including ensuring enough beds in hospitals, adequate isolation facilities and ensuring there is enough medical capacity to handle the expected peak period.

IBM scientists, researchers and engineers work on the project on a full-time basis from the tech giant’s research laboratory, based at Wits University’s Braamfontein Campus in Johannesburg.

A screen shot of the dashboard showing Gauteng’s COVID-19 trends.
A screen shot of the dashboard showing Gauteng’s COVID-19 trends.

Bruce Mellado, professor at Wits University’s School of Physics and senior scientist at iThemba Labs, explained that predicting future infection numbers is highly dependent on strong collaborations and constant engagement between epidemiologists, practitioners and healthcare workers across various institutions, who conduct regular monitoring of the evolution of the virus.

“The goal is not to create a very sophisticated mathematical apparatus, but rather to provide a realistic reflection of the province’s ecosystem and what’s going on, on the ground,” noted Mellado.

“While there are many dashboard platforms which track statistics, this system is a major upgrade because it’s a comprehensive suite of visualisation and predictive tools that perform various functions, providing not only the epidemiological model of the infections but also a comprehensive economic model that helps policy-makers understand how controlling the epidemic will impact on SA’s economy.”

Wits University’s contribution is to provide the modelling which has input from epidemiologists, clinicians, practitioners and data analysts who provide guidance to policy-makers on when is the right time to go to the next lockdown phase and other important decisions.

All Gauteng-based laboratories are required to report positive COVID-19 cases to the National Institute for Communicable Diseases, which compiles all the patients’ data, which then goes through the process of geo-coding. Once the patients’ addresses have been mapped, the platform provides an aggregated form of this data on the dashboard, which will provide users with the statistics.

While some aspects of the platform are already in use by government, the long-term vision is to enable Gauteng citizens to also use it to gain insights of the infection trends in their neighbourhood, noted Gritzman.

“Gauteng premier David Makhura has highlighted the importance of putting this technology in the hands of citizens, and that’s our goal. This is the most exciting part because they will be able to monitor in real-time the number of infections in their communities as they increase.

“This is why this technology has been designed with scale in mind, to enable it to be used by tens of thousands, or potentially even millions of concurrent users.”

See also