Georgia State University alumnus Aleksandr Shishkin (Ph.D. Public Health - Epidemiology, Spring 2025), under the supervision of Dr. Alexander Kirpich and in collaboration with Dr. Pavel Skums, Dr. Gerardo Chowell, Dr. Alexander Perez Tchernov, and other researchers, has published a new article titled “Enhancing the Accuracy of COVID-19 Incidence and Mortality Predictions Using Google Trends Data Across the 50 US States and the District of Columbia” in Data & Policy - Cambridge University Press.
The study explores one of the key challenges in public health forecasting, which is the delay in traditional surveillance reporting. The researchers present an innovative approach that uses Google search query data to improve both the speed and accuracy of predictions for the spread and impact of COVID-19. By analyzing search terms related to disease incidence and mortality, the team identified and ranked keywords according to their predictive value. These keywords were then added to forecasting models such as ARIMA, Prophet, and XGBoost to generate near term predictions of reported cases and deaths.
Including the most predictive search terms in these models significantly improved forecast accuracy, with gains ranging from 50 to 90 percent across different approaches. The improvements were most pronounced for predictions of incidence, while the benefits for mortality forecasts were smaller but still substantial. The authors explain that the use of real time search data can help overcome reporting delays, supporting faster public health decision making and better allocation of resources during outbreaks.
This research shows the promise of digital data sources such as Google Trends in complementing traditional surveillance systems and strengthening preparedness and response strategies for future public health emergencies across the United States.