Artificial Intelligence System More Accurately Predicts Prevalence, Location of Flu

A new study published in Nature Communications combines two forecasting methods with artificial intelligence to estimate local flu activity.

Although the CDC regularly monitors patients visits for flu-like illnesses in the United States, the information often lags up to two weeks behind real time. Researchers at the Computational Health Informatics Program (CHIP) at Boston Children’s Hospital aimed to develop an approach that would accurately and effectively predict influenza activity. 

The researchers found that when the approach, known as ARGONet, was applied to flu seasons from September 2014 to May 2017, it made more accurate predications than the team’s earlier high-performing forecasting approach, ARGO, in more than 75% of the states studied.

The ARGONET uses machine learning and two robust flu detection models. The first model is known as ARGO (AugoRegression with General Online Information), which amalgamates information from electronic health records, flu-related Google searches, and historical flu activity in a particular location. The second model draws on spatial-temporal patterns of flu spread in neighboring areas. According to the study authors, the model uses the fact that the presence of flu is in nearby locations to inform the likeliness of experiencing a disease outbreak in a given location.

In the study, ARGO alone outperformed Google Flu Trends, the previous forecasting system operating in 2008 to 2015. The machine learning system was trained by feeding it flu predictions from both models as well as actual flu data, helping to reduce errors in the predictions.

The researchers believe their approach will become more accurate and will set a foundation for precision public health in infectious diseases.

Reference

• Fred S. Lu, Mohammad W. Hattab, Cesar Leonardo Clemente, Matthew Biggerstaff, Mauricio Santillana. Improved state-level influenza nowcasting in the United States leveraging Internet-based data and network approaches. Nature Communications, 2019; 10 (1) DOI: 10.1038/s41467-018-08082-0. https://www.nature.com/articles/s41467-018-08082-0. Accessed August 19, 2019.