Digital prescriptions in real-time reveal flu trends from 21 million transactions

Real-time digital prescriptions have unlocked new insights into the dynamics of influenza outbreaks thanks to evidence from over 21 million transactions analyzed over a two-year period. This analysis from China’s largest on-demand medication delivery platform across all 31 provinces between 2022 and 2024 reveals that digital prescription data can serve as a validated epidemic proxy, offering a significant improvement over traditional influenza surveillance methods plagued by 1-2 week reporting delays.

The study found that prescription data has the advantage of a predictive lead time of up to two weeks, a substantial improvement over environmental predictors while on par with online search indices. Moreover, digital prescriptions showed bidirectional causal coupling with laboratory-confirmed influenza positivity in the majority of provinces, a dynamical signature not found in online search or environmental variables. This distinction helps differentiate validated disease signals from confounded correlates, making digital prescriptions a superior tool for epidemic surveillance.

Furthermore, the analysis revealed that digital prescriptions exhibit heightened sensitivity to environmental factors compared to laboratory surveillance data, providing a more comprehensive view of the relationship between influenza dynamics and external influences. Leveraging this validated proxy, researchers developed a spatiotemporal deep learning framework that integrates Graph Neural Networks (GNN), Mamba, and Long Short-Term Memory (LSTM) models to achieve a remarkable 96-day forecasting accuracy of daily prescription rates across most provinces.

This innovative approach showcases the potential of digital prescriptions not only for immediate epidemic detection with 24-hour data availability but also for actionable long-range forecasting. By providing an additional validated data stream for multi-source epidemic surveillance, digital prescriptions offer a novel perspective on influenza dynamics. The study’s findings highlight the importance of integrating emerging digital technologies with traditional disease surveillance methods to enhance epidemiological insights and response strategies.

Overall, this research underscores the transformative power of real-time digital prescriptions in unlocking valuable information about influenza dynamics and enhancing our ability to understand and respond to outbreaks in a timely and effective manner. By embracing digital innovations and leveraging new data sources, public health authorities can better prepare for and manage future epidemics with greater precision and efficiency.