Validating the Use of Google Trends to Enhance Pertussis Surveillance in California

We examined the possible role of a delay to pertussis diagnosis (from symptom onset) limiting the model performance, but an extended models using lagged observed pertussis time series did not show improvement of model fit. This may indicate that the model is detecting real-time Google activity performed around the time patients receive a pertussis diagnosis (rather than detecting Google activity about pertussis symptoms before seeking healthcare). Since the launch of Google Flu Trends in 2009, there have been numerous other applications of Google Trends to a broad number of communicable diseases as diverse as Ebola and methicillin-resistant Staphylococcus aureus 18,21,22. It remains unclear, to our knowledge, whether any Google disease models are being implemented in actual public health practice or have ever triggered any public health response beyond that prompted by conventional surveillance. Concerns regarding the more recent performance of Google Flu Trends despite updated models emphasizes caution in the implementation of any Google-based epidemic prediction system8. The very recent decision by Google to cease public access to their Google Flu Trends and Google Dengue Trends predictions further underscores the need for further, robust study in this field16. Our findings help to further define the limitations and potential of Google-based epidemiological methods for enhanced communicable disease surveillance. Competing Interest The authors have declared that no com...
Source: PLOS Currents Outbreaks - Category: Epidemiology Authors: Source Type: research