Orellano Pablo Wenceslao, Reynoso Julieta Itatí, Antman Julián, Argibay Osvaldo
Cad Saude Publica. 2015 Apr;31(4):691-700. doi: 10.1590/0102-311x00072814.
The aim of this study was to find a model to estimate the incidence of influenza-like illness (ILI) from the Google Trends (GT) related to influenza. ILI surveillance data from 2012 through 2013 were obtained from the National Health Surveillance System, Argentina. Internet search data were downloaded from the GT search engine database using 6 influenza-related queries: flu, fever, cough, sore throat, paracetamol, and ibuprofen. A Poisson regression model was developed to compare surveillance data and internet search trends for the year 2012. The model's results were validated using surveillance data for the year 2013 and results of the Google Flu Trends (GFT) tool. ILI incidence from the surveillance system showed strong correlations with ILI estimates from the GT model (r = 0.927) and from the GFT tool (r = 0.943). However, the GFT tool overestimates (by nearly twofold) the highest ILI incidence, while the GT model underestimates the highest incidence by a factor of 0.7. These results demonstrate the utility of GT to complement influenza surveillance.
本研究的目的是找到一种模型，以便根据与流感相关的谷歌趋势（GT）来估算流感样疾病（ILI）的发病率。2012年至2013年的ILI监测数据来自阿根廷国家卫生监测系统。使用6个与流感相关的查询词（流感、发烧、咳嗽、喉咙痛、扑热息痛和布洛芬）从GT搜索引擎数据库下载互联网搜索数据。构建了一个泊松回归模型，以比较2012年的监测数据和互联网搜索趋势。使用2013年的监测数据以及谷歌流感趋势（GFT）工具的结果对该模型的结果进行了验证。监测系统的ILI发病率与GT模型（r = 0.927）和GFT工具（r = 0.943）的ILI估算值显示出很强的相关性。然而，GFT工具高估了最高ILI发病率（几乎高出两倍），而GT模型则将最高发病率低估了0.7倍。这些结果证明了GT在补充流感监测方面的作用。
