We combine COVID-19 case data with mobility data to estimate a modified susceptible-infected-recovered (SIR) model in the United States. In contrast to a standard SIR model, we find that the incidence of COVID-19 spread is concave in the number of infectious individuals, as would be expected if people have inter-related social networks. This concave shape has a significant impact on forecasted COVID-19 cases. In particular, our model forecasts that the number of COVID-19 cases would only have an exponential growth for a brief period at the beginning of the contagion event or right after a reopening, but would quickly settle into a prolonged period of time with stable, slightly declining levels of disease spread. This pattern is consistent with observed levels of COVID-19 cases in the US, but inconsistent with standard SIR modeling. We forecast rates of new cases for COVID-19 under different social distancing norms and find that if social distancing is eliminated there will be a massive increase in the cases of COVID-19.

我们将 COVID-19 病例数据与流动性数据相结合，以估计美国修改后的易感者感染者康复 (SIR) 模型。与标准 SIR 模型相反，我们发现 COVID-19 传播的发生率在感染者数量上是凹的，如果人们拥有相互关联的社交网络，这也是可以预料的。这种凹形对预测的 COVID-19 病例有重大影响。特别是，我们的模型预测，COVID-19 病例数只会在传染事件开始时或重新开放后的短时间内呈指数增长，但很快就会稳定下来，并在很长一段时间内保持稳定，疾病传播水平略有下降。这种模式与美国观察到的 COVID-19 病例水平一致，但与标准 SIR 模型不一致。我们预测了不同社会距离规范下的 COVID-19 新病例发生率，并发现如果消除社会距离，则 COVID-19 病例将大幅增加。