Using past pandemics to guide COVID-19 predictions

随着2019冠状病毒病的到来，省级和联邦政府都在使用数学模型来指导决策，这种模型的可靠性变得至关重要。我们建议分析最近三次大流行的建模预测——2003年的SARS（也是一种冠状病毒）、2009年的H1N1“猪流感”爆发和2012年开始的MERS（另一种冠状病毒）——以便评估这些模型在长期内的准确性。该提案还将探讨从超级传播者（比大多数人更有可能传播疾病的个体）或第二（或第三）波爆发等随机事件中吸取的经验教训和预测极限。这些模型所依据的疫情发生在“大数据”时代，当时人们利用数学模型做出了各种预测。这使他们成为这项研究的理想人选。虽然最好是等待进一步的数据，以验证当前的模型，但在快速蔓延的大流行的早期阶段，我们没有足够的时间。然而，过去流行病建模的性质可以作为当前流行病的指南。利用这些数据和过去的预测，我们将根据过去最成功的模型的最佳做法，开发和分析平行的COVID-19模型。这些COVID-19模型将使决策者能够获得进一步浪潮以及未来其他大流行的早期预警，并了解哪些模型可能是可靠的，在什么情况下是可靠的。

With COVID-19 upon us and mathematical modelling being used by both provincial and federal governments to guide decision-making, the reliability of such models becomes paramount. We propose to analyse the modelling predictions done in three recent pandemics - SARS in 2003 (also a coronavirus), the H1N1 "swine flu" outbreak of 2009 and MERS, starting in 2012 (another coronavirus) - in order to assess how accurate these models were in the long term. The proposal will also explore the lessons learned and the predictive limits of random events such as superspreaders (individuals who are vastly more likely to transmit the disease than most people) or the onset of a second (or third) wave. The outbreaks on which the models are based happened during the era of "big data", and various predictions were made at the time using mathematical models. It makes them ideal candidates for this research. While it would be optimal to wait for further data in order to validate current models, in the early stages of a fast-moving pandemic, we do not have the luxury of time. However, the nature of modelling in past pandemics can serve as a guide for the current one. Using the data and past predictions, we will develop and analyse parallel COVID-19 models based on best practices from the most successful past models. These COVID-19 models will allow decision-makers to gain early warning of further waves, as well as other future pandemics, with the knowledge of which models are likely to be reliable and under what circumstances.