Machine Learning model helps with contact tracing during the COVID-19 outbreak
During the COVID-19 pandemic in the Netherlands there was overwhelming demand for contact tracing solutions to help identify people who had come into contact with someone who was infected in an attempt to control the spread of the virus. With the number of cases skyrocketing, the capacity of the existing system couldn't keep up.
In response to the demand, researchers developed a prediction model using machine learning to predict COVID-19 cases based on geographical and demographic data. They produced a model that predicted the top 100 postal codes with the highest number of COVID-19 cases with the most accurate predictions being for people aged 20-39 and 40-64 years.
The study included people who tested positive for SARS-CoV-2 between June 2020 and February 2021 in nine Dutch regions. They used data such as population density, household size, and Google search history to train a machine learning model (random forest regression) to predict which postal codes would have the highest number of cases in the following two weeks.
Lead author Max Keuken said, “This model could certainly provide a foundation for targeted contact tracing. By focusing on areas with high predicted transmission rates, public health services can allocate resources more effectively. For instance, they can place mobile testing units or vaccination sites near predicted hotspots. But I must also point out that models like this need a lot of data to be more effective. During early stages of an outbreak, there are usually not enough data available to make reliable predictions.”
Whilst the model can certainly help in managing ongoing outbreaks, it might not be as useful during the initial phase of a new epidemic as more data would be needed.
The study ‘Spatio-temporal forecasting of COVID-19 cases in the Netherlands for source and contact tracing’ published in the International Journal of Population Data Science (IJPDS), highlights the potential of machine learning in public health, particularly in predicting the spread of infectious diseases like COVID-19. By incorporating information on location, time, and demographics, the model could offer a more targeted approach for contact tracing to improve the efficiency and effectiveness of preventive measures, and potentially also enhance public support for government policies and increase adherence to restrictions.
Click here to read the full article