Airfinity forecasts the ‘tripledemic’ burden of COVID-19, flu and RSV will be less severe this winter as outbreak peaks are not predicted to overlap as extensively as last year.

Our methodology

The forecasts for each disease were generated using a variety of techniques. Influenza and RSV forecasts were developed using primarily machine learning (ML) techniques; each of both flu and RSV were created from an ensemble of three ML models. These ML models factored in historic trends, as well as currently circulating variants and anticipated vaccine uptake. COVID-19 forecasts were generated using a compartmental SIR model, with parameters to the model being fitted to historic data. COVID forecasts did not use ML techniques in the same way flu and RSV forecasts did due to the dramatically changing data landscape for COVID, with testing capacity significantly reduced in recent months, which disrupted ML models.

To discuss the methodology, as well as our wider forecasting and simulation capabilities, get in touch via infectiousdiseases@airfinity.com.