Study: Internet Search Data Might Help Predict Disease Prevalence
April 1, 2015 in News
Details of Study
For the study, researchers used Google Trends to identify and study search terms for a one-year period. Those data then were compared with CDC data on the prevalence of risk factors that can be used to predict non-communicable disease.
Risk factors included:
- Cardiovascular disease;
- Exercise frequency;
- High blood pressure diagnosis; and
- Tobacco use.
The data were broken down by state.
The researchers found that previous-year search trends were strongly tied to CDC’s disease risk estimates based on population data, according to the authors.
Specifically, the researchers’ model predicted a diabetes prevalence of:
- 11.2% in Alabama, compared with CDC’s 11.8% estimate;
- 8.1% in Nevada, compared with CDC’s 10.3% estimate; and
- 9.4% in New Jersey, compared with CDC’s 8.8% estimate.
Svetha Venkatesh — study co-author and director of the Center for Pattern Recognition and Data Analytics at Deakin University in Geelong, Australia — said using search trends to obtain data could help health care policies and interventions change more quickly. She said traditional data collection and processing can take up to three years.
Venkatesh added that the Web search method would not replace hard data measurements but could support them at a low additional cost (Doyle, Reuters, 3/31).