This paper introduces a new basic risk model that could also be utilized by Covid-19 warning apps a priori, before an action is performed. Today the common warning apps estimate risk a posteriori and give no advice on particular scenarios. The new model also has the advantage that the individual risks behind the decision-making process would be uniform (in contrast to some current regulations) and it could help to understand the risks better and could also help to reduce risks a priori. It could be easily implemented on a single app screen, needing only some individual preferences to be set and a handful of adjustments to the particular scenario that shall be assessed. The disadvantage as of any simplified semi-quantitative risk models is that calibration is not easy (as some calibration points may even contradict) and that cumulative effects are hard to integrate e. g. the joint effect of combined scenarios. But, in principle calibration is feasible and it may be a good decision to calibrate the model conservatively.
Considerations, Good Practices, Risks and Pitfalls in Developing AI Solutions Against COVID-19
Université de Montréal(蒙特利尔大学);Mila Québec Artificial Intelligence Institute(米拉魁北克人工智能研究所);United Nations Global Pulse(联合国全球脉动);Institute for Data Science, Durham University(杜伦大学数据科学研究所);NYU Stern School of Business(纽约大学斯特恩商学院);Global Coordination Mechanism on NCDs, WHO(世界卫生组织非传染性疾病全球协调机制)
CRISP: A Probabilistic Model for Individual-Level COVID-19 Infection Risk Estimation Based on Contact Data
Zalando(扎兰多);Amazon(亚马逊)
我们提出了CRISP(COVID-19风险评分预测),这是一种基于SEIR模型的COVID-19感染在人群中传播的概率图模型,其中我们假设可以访问(1)跨时间跨各种渠道(例如,蓝牙接触痕迹),以及(2)感染、暴露和免疫测试在给定时间的测试结果。我们的微观模型跟踪每个人在每个时间点的感染状态,从易感,暴露,传染到康复。我们开发了Monte Carlo EM和消息传递算法来推断接触通道特定的感染传播概率。我们的蒙特卡罗算法使用吉布斯抽样来抽取每个人在整个分析时间段内潜伏感染状态的样本,给定所有接触者的潜伏感染状态和测试结果数据。模拟数据的实验结果表明,我们的CRISP模型可以参数化的再生因子$R_0$,并表现出人口水平的传染性和恢复时间系列类似的经典SEIR模型。然而,由于个人接触数据,该模型允许对广泛的COVID-19缓解和抑制政策措施进行精细控制和推断。此外,区块吉布斯抽样算法能够支持测试跟踪隔离方法中的有效测试,以遏制COVID-19感染传播。据我们所知,这是第一个基于个人层面接触数据对COVID-19感染传播进行有效推断的模型;大多数流行病模型都是宏观层面的模型,对整个人群进行推理。在https://github.com/zalandoresearch/CRISP上可以用Python和C++实现CRISP。
Exposure Density and Neighborhood Disparities in COVID-19 Infection Risk: Using Large-scale Geolocation Data to Understand Burdens on Vulnerable Communities