Social media analytics for early detection of foodborne disease

食源性疾病已成为一个严重的公共卫生问题，报告不足，健康和财政成本高。世界卫生组织（WHO）将食源性疾病暴发确定为21世纪世纪全球公共卫生的主要威胁。传统的监测系统，如加拿大可报告疾病监测系统，只捕获加拿大每年估计400万食源性疾病病例的一小部分。它们依赖于收集大量指标，包括临床症状、病毒学实验室结果、住院和死亡率统计数据，导致临床医生向卫生部门报告病例之间的平均延迟时间为6.5天。公共卫生决策者认为延迟通知是调查食源性疾病的障碍，因为它可能在地理上分布很远。食源性疾病的早期检测可以通过从零售和食品服务网点清除受污染的产品、提高公众意识以及为暴露个体提供更及时的预防和治疗措施来减少暴露个体的数量。我们建议将社交媒体数据作为传统监测系统的补充。通过大型用户社区传播的信息的一致性和高方差提供了挖掘食源性疾病活动信号的数据的机会;定性和定量分析疾病的传播;并预测未来的爆发。我们提出了一系列基于社交媒体的预测模型，通过对社交媒体中用户对话的环境跟踪和监测来表征和检测即将发生的食源性疾病爆发。其目的是推进非传统来源食源性疾病检测的研究，为卫生决策者提供情境意识。

Foodborne Disease has emerged as a serious and underreported public health problem with high health andfinancial costs. The World Health Organization (WHO) identifies foodborne illness outbreaks as a major globalpublic health threat in the twenty-first century. Traditional surveillance systems such as Canadian NotifiableDisease Surveillance System capture only a fraction of the estimated 4 million annual cases of foodborneillness in Canada. They rely on the collection of numerous indicators including clinical symptoms, virologylaboratory results, hospital admissions and mortality statistics resulting in a median delay of 6.5 days betweencase report from clinicians to the health departments. Public health decision-makers consider the delayednotification as a barrier to investigating foodborne disease, as it can potentially distribute geographically acrossgreat distances. Early detection of foodborne disease can reduce the number of exposed individuals byremoving contaminated product from retail and foodservice outlets, increasing public awareness, and offering amore timely preventative and therapeutic measures to exposed individuals.We propose that social media data should be exploited as a complementary component of the traditionalsurveillance systems. The enormity and high variance of the information that propagates through large usercommunities presents the opportunity to mine the data for signals of foodborne disease activity; analyze illnesspatterns qualitatively and quantitatively; and to predict future outbreaks. We propose a host of socialmedia-based predictive models to characterize and detect upcoming foodborne illness outbreaks throughambient tracking and monitoring over users' conversations in social media. The objective is to advance researchon foodborne disease detection from non-traditional sources to supply health decision makers with situationalawareness.