Health problems are among the most common topics for a web search. A study has found that data on web searches can accurately predict the risks of non-infectious chronic diseases in the United States.
Researchers at Deakin University in Geelong, Australia, used Google Trends to identify search terms for a one-year period concerning risk factors that can predict non-communicable diseases and compared those terms to state-by-state data from the U.S. Centers for Disease Control and Prevention. The terms included topics such as exercise frequency, tobacco use, high blood pressure, cardiovascular disease, and diabetes. The search trends from the previous year were found to be strongly correlated with the CDC's estimates of disease risk based on population data.
They created a web search model to determine the accuracy of using search terms to predict the incidence of disease. Using the model on search terms for 2011, they predicted that 11.2% of people in Alabama, 9.4% in New Jersey and 8.1% in Nevada had diabetes. The actual values for diabetes as measured by the CDC were 11.8%, 8.8% and 10.3%, respectively, which is not far off. The average difference between predicted and measured values ranged from 0.03% to 2.16% for a given state.
Being able to use web search data to predict the incidence of health problems would be valuable in making healthcare policy decisions. Such a model would be used in addition to collecting data about health problems the traditional way, but the traditional way can take several years and is more costly. This kind of a predictive tool could provide useful information in real time.
Web search data has been used to predict and track outbreaks of the flu in the past, but this study worked with non-communicable disease.
The study was published in the Journal of Epidemiology and Community Health.
