Researchers created a surveillance model that uses machine learning to provide highly accurate estimates of local flu activity, according to a study published in Nature Communications.
For the study, researchers from the Computational Health Informatics Program at Boston Children's Hospital combined two forecasting methods with machine learning to estimate flu activity.
The first model, ARGO, uses data from EHRs, flu-related Google searches and historical flu activity for a specific location. The second model analyzes information on spatial-temporal flu activity for nearby locations. Researchers trained the new machine learning model, called ARGOnet, using flu predictions from both models and actual flu data.
"The system continuously evaluates the predictive power of each independent method and recalibrates how this information should be used to produce improved flu estimates," senior author Mauricio Santillana, PhD, a CHIP faculty member, said in a press release.
Researchers used ARGOnet to analyze flu seasons from September 2014 to May 2017 and found the model made more accurate predictions than ARGO in more than 75 percent of states included in the analysis.
More articles on clinical leadership and infection control:
Cancer comorbidities reduce clinical trial enrollment, study finds
Flu has already sickened more than 6 million this season
How CRISPR can help with antibiotic development