Arming the Fight Against COVID-19 With Reliable Data

QNRF-funded project uses multiple variables to forecast the spread of COVID-19

Earlier this year, policymakers and institutions around the world were faced with a myriad of unexpected and daunting challenges caused by the rapid spread of COVID-19.

To counter these challenges, access to the right data is crucial for governments, healthcare systems, and various other sectors. While accessing data is important in sudden and emergent situations like the COVID-19 pandemic, we also need to be aware that a lot of data produced is often unreliable, hence counting on it could add to the crisis.

To solve this challenge, a team of scientists from Qatar Environment and Energy Research Institute (QEERI) at Hamad Bin Khalifa University (HBKU) led by Dr. Antonio Sanfilippo, has developed a COVID-19 analytic platform that achieves situation awareness by estimating and forecasting the spread of COVID-19 as a function of epidemiological, socioeconomic, environmental, and global health indicator variables.

Titled ‘DEFCOV’, the analytic platform has been made possible by a grant awarded through QNRF’s Rapid Response Call (RRC). The datasets and models produced through DEFCOV provide a multifaceted, holistic approach that helps assess and predict daily COVID-19 rates.

The research team observed that, while there are a growing number of studies suggesting that meteorological factors affect the spread of the COVID-19 pandemic, understanding about the extent of this correlation remains limited.

To fill this gap, the team started working on DEFCOV, to show that such an understanding is attainable through the development of regression models. These models can verify the contribution of meteorology to the modeling of COVID-19 transmission by using techniques that assess the relative weight of meteorological variables as compared to other factors.

The models developed in the project integrate the epidemiological data publicly available from Johns Hopkins University with meteorological and environmental data from the Global Forecast System and the Copernicus Atmosphere Monitoring Service provided by QEERI’s partner Transvalor S.A.. Moreover, the team also analyzed additional data on socioeconomic factors, environmental indicators, and health indicators.

The team began by applying machine learning techniques to the selected data, so as to understand the relative importance of climatic and other variables in the spread of COVID-19. They then developed regional and global regression models correlating socioeconomic, environmental, health, epidemiological, and meteorological factors with COVID-19 daily cases, followed by performing a feature importance analysis on the global regression model. Figure 1 shows that the regression models that include meteorological factors provide a close fit between estimated and observed COVID-19 rates.

The contribution of meteorological factors is detailed in the feature importance analysis shown in Figure 2, where climatic and environmental factors both appear to have a significant impact on the estimation of COVID-19 transmission rates.

After recording the observations, the team performed the projection of COVID-19 transmission rates using time series forecasting techniques. Figure 3 shows the outcome in relation to Qatar. The blue line shows actual COVID-19 cases registered up to 19 August 2020, while the orange line shows the daily cases predicted by the forecasting model developed within the project. The model can predict COVID-19 transmission cases with a median daily error of 5%, which corresponds to 5.3% when averaged over one week.

These outcomes indicate that climate plays a meaningful role in modulating the dynamics of the pandemic, which will help us better prepare for Qatar’s changing weather patterns. The team is also in the final stages of developing a web-enabled application that will provide user-friendly access to the models developed in the project, which will allow the user to analyze COVID-19 data worldwide in real-time and provide forecasts of COVID-19 transmission rates for diverse time horizons.