Development and validation of influenza forecasting for 64 temperate and tropical countries

Accurate forecasts of influenza incidence can be used to inform medical and public health decision-making and response efforts. However, forecasting systems are uncommon in most countries, with a few notable exceptions. Here we use publicly available data from the World Health Organization to generate retrospective forecasts of influenza peak timing and peak intensity for 64 countries, including 18 tropical and subtropical countries. We find that accurate and well-calibrated forecasts can be generated for countries in temperate regions, with peak timing and intensity accuracy exceeding 50% at four and two weeks prior to the predicted epidemic peak, respectively. Forecasts are significantly less accurate in the tropics and subtropics for both peak timing and intensity. This work indicates that, in temperate regions around the world, forecasts can be generated with sufficient lead time to prepare for upcoming outbreak peak incidence.Author summary: Influenza is responsible for an estimated 3–5 million cases and 300–650,000 deaths each year worldwide. If produced early enough, accurate forecasts of influenza activity could guide public health practitioners and medical professionals in preparing for an outbreak, reducing the subsequent morbidity and mortality. For example, hospitals could use these forecasts to determine how many beds will be needed when an outbreak is most intense. Despite this potential impact, influenza forecasts are primarily generated for the United States, with forecasts for other countries being comparatively rare. Here, we use publically available influenza data to forecast influenza activity in 64 countries. We find that accurate forecasts can be produced several weeks before the outbreak’s peak in temperate countries, where influenza outbreaks occur regularly during the winter. Forecast accuracy is lower in the tropics and subtropics, where outbreaks occur more sporadically. Overall, our results suggest that forecasts have potential as an important public health tool in many countries, not only in the US.