🤖 AI Summary
This study investigates leveraging multinational online search behavior data to augment traditional public health surveillance for early warning and cross-national transmission forecasting of COVID-19. Method: We propose an unsupervised symptom modeling framework, revealing—novelly—that rare symptom queries exhibit higher predictive power than common ones; introduce news coverage as a proxy for public attention bias; and develop a search-augmented autoregressive mortality prediction model integrating NHS symptom taxonomy with media-based calibration. Contribution/Results: Search signals lead confirmed case and death reports by median lags of 16.7 and 22.1 days, respectively. The model significantly improves short-term mortality forecasting accuracy across multiple countries (average MAE reduction: 23.6%). Empirical analysis confirms strong stage-dependent associations between symptom query patterns and epidemic progression, supporting cross-national transfer learning and generalizability.
📝 Abstract
Previous research has demonstrated that various properties of infectious diseases can be inferred from online search behaviour. In this work we use time series of online search query frequencies to gain insights about the prevalence of COVID-19 in multiple countries. We first develop unsupervised modelling techniques based on associated symptom categories identified by the United Kingdom’s National Health Service and Public Health England. We then attempt to minimise an expected bias in these signals caused by public interest—as opposed to infections—using the proportion of news media coverage devoted to COVID-19 as a proxy indicator. Our analysis indicates that models based on online searches precede the reported confirmed cases and deaths by 16.7 (10.2–23.2) and 22.1 (17.4–26.9) days, respectively. We also investigate transfer learning techniques for mapping supervised models from countries where the spread of the disease has progressed extensively to countries that are in earlier phases of their respective epidemic curves. Furthermore, we compare time series of online search activity against confirmed COVID-19 cases or deaths jointly across multiple countries, uncovering interesting querying patterns, including the finding that rarer symptoms are better predictors than common ones. Finally, we show that web searches improve the short-term forecasting accuracy of autoregressive models for COVID-19 deaths. Our work provides evidence that online search data can be used to develop complementary public health surveillance methods to help inform the COVID-19 response in conjunction with more established approaches.
