マイクロブログを用いたインフルエンザsurveillanceの有効性について

概要

論文の詳細を見る
Microblog services have potential to collect epidemic information in real time, which is hard to achieve with traditional surveillance approaches. We made a preliminary analysis of Twitter posts by general users, to evaluate whether it can detect early stage of influenza outbreak. First, we investigated search terms to efficiently extract incidence of influenza patient, and identified several terms, which appear in related tweets, namely, "influenza", "fever", and "hospital". Second, from "week 30 of 2011" to "week 4 of 2012", we collected Twitter posts relating influenza patient using the selected search terms. Lastly, after manual cleaning of the collected posts, the resulting tweet statistics was compared against the national influenza sentinel surveillance data of the National Institute of Infectious Diseases. The analysis revealed significant correlation between the number of hospital visits in a week per sentinel and the number of Twitter posts (R = 0.83). Because so many people use the microblog services, for timeliness and for convenience, they would be useful tools for information gathering of public health authorities. Nevertheless, toward practical utilization of such data in infectious disease surveillance, there must be a means to reliably collect demographic and location information of the users, such as, gender and age.
2012-10-00