Effective pandemic response requires accurate and timely surveillance of disease spread and clinical course understanding. Traditional methods, primarily relying on laboratory testing, face limitations, particularly in the early stages of a pandemic due to testing capacity constraints. Internet search data offers a potential complementary resource, being readily available in near real-time and at population scale. While previous studies explored its use for pandemic spread tracking, this study investigated its potential for characterizing the clinical course of COVID-19. Understanding symptom progression can aid clinicians in patient care planning and public health officials in tracking pandemic stages. Existing clinical studies, while informative, are limited by sample size and time constraints. Internet search data, conversely, offers a broader population view and real-time insights.

Several studies have examined the use of internet search data for tracking various health phenomena, including influenza, MERS, measles, and vaccination compliance. Regarding COVID-19, some studies correlated search terms with reported cases, focusing on general terms like "coronavirus" or specific aspects like mental health impacts. However, a detailed analysis of symptom-specific search terms to reconstruct the clinical disease progression was lacking. Existing clinical case studies on COVID-19 provided insights into symptom progression, typically reporting a 5-day lag between initial symptoms and the onset of shortness of breath. However, these studies were often limited by small sample sizes and publication delays.

This study analyzed internet search data from Google Trends and Weibo Search Trends for 32 countries across six continents. The data included symptom-specific search terms ("fever," "cough," "dry cough," "chills," "sore throat," "runny nose," "shortness of breath") and general terms ("coronavirus," "coronavirus symptoms," "coronavirus test"). Data on reported COVID-19 cases and deaths were obtained from sources like the European Centre for Disease Prevention and Control and the World Health Organization. Temporal correlation analyses determined the lags between search terms and reported cases/deaths. Cross-country ensemble averaging was used to create average temporal profiles for each search term, constructing a search-data-based view of the clinical course of disease progression. Search terms were translated into local languages using native speakers and Google Translate where necessary. Data were smoothed using a 7-day moving average.

The analysis revealed that increases in symptom-related searches consistently preceded increases in reported COVID-19 cases and deaths across all 32 countries. The average lag was approximately 2-3 weeks. Furthermore, the temporal relationships between symptom-specific searches reflected the clinical progression of the disease. The ensemble average curves showed a clear temporal pattern: initial symptoms (fever, cough, sore throat, chills) were followed by shortness of breath, with an average lag of approximately 5 days. This finding aligns with the clinical course reported in the medical literature. While individual country-level data showed variability, the overall temporal order of symptom progression remained consistent across countries.

This study demonstrates the utility of internet search data as a complementary resource for understanding both the spread and clinical course of a novel pandemic like COVID-19. The ability to predict increases in cases and deaths several weeks in advance, alongside the accurate portrayal of symptom progression, highlights its value in public health surveillance, especially during the critical early stages when testing capacity is limited. The findings underscore the real-time, population-scale nature of this data source. The 5-day lag between initial symptoms and shortness of breath, as observed through search data, can inform clinical care and resource allocation. However, the study acknowledges the inherent limitations of using search data, such as potential biases in internet access and the motivations behind individual searches. Future research should explore the use of this approach during later pandemic phases and in combination with other data sources.

This study successfully demonstrated the use of internet search data to track COVID-19 spread and characterize its clinical course across 32 countries. The findings highlight the potential of this real-time, population-level data source as a valuable complementary tool for pandemic surveillance and clinical care planning. Future work should address limitations and integrate search data with other data sources for a more comprehensive understanding of pandemic dynamics.

The study acknowledges several limitations. Internet access and digital literacy vary across countries, potentially introducing bias. The motivations behind individual searches are unknown, meaning that search volume may not solely reflect disease prevalence. The study also used automated translations in some cases, potentially impacting accuracy. Finally, the relationships between search terms and illness may change over time as public awareness grows, limiting the generalizability of the findings to later pandemic phases.