Same Influenza, Different Responses: Social Media Can Sense a Regional Spectrum of Symptoms

TL;DR

This paper demonstrates that social media data, specifically from Weibo, can effectively sense regional differences in influenza symptoms and patient emotions, providing a real-time, comprehensive supplement to traditional epidemiological data collection methods.

Contribution

It introduces a novel approach using social media reports to infer regional influenza symptom spectra and emotional states, enhancing surveillance accuracy and timeliness.

Findings

Regional symptom differences between northern and southern China detected.

Patients in the south are more optimistic than those in the north.

Social media data improves influenza monitoring regression performance.

Abstract

Influenza is an acute respiratory infection caused by a virus. It is highly contagious and rapidly mutative. However, its epidemiological characteristics are conventionally collected in terms of outpatient records. In fact, the subjective bias of the doctor emphasizes exterior signs, and the necessity of face-to-face inquiry results in an inaccurate and time-consuming manner of data collection and aggregation. Accordingly, the inferred spectrum of syndromes can be incomplete and lagged. With a massive number of users being sensors, online social media can indeed provide an alternative approach. Voluntary reports in Twitter and its variants can deliver not only exterior signs but also interior feelings such as emotions. These sophisticated signals can further be efficiently collected and aggregated in a real-time manner, and a comprehensive spectrum of syndromes could thus be inferred. Taking Weibo as an example, it is confirmed that a regional spectrum of symptoms can be credibly sensed. Aside from the differences in symptoms and treatment incentives between northern and southern China, it is also surprising that patients in the south are more optimistic, while those in the north demonstrate more intense emotions. The differences sensed from Weibo can even help improve the performance of regressions in monitoring influenza. Our results suggest that self-reports from social media can be profound supplements to the existing clinic-based systems for influenza surveillance.