Diverse local epidemics reveal the distinct effects of population density, demographics, climate, depletion of susceptibles, and intervention in the first wave of COVID-19 in the United States

Abstract

<jats:p>The SARS-CoV-2 pandemic has caused significant mortality and morbidity worldwide, sparing almost no community. As the disease will likely remain a threat for years to come, an understanding of the precise influences of human demographics and settlement, as well as the dynamic factors of climate, susceptible depletion, and intervention, on the spread of localized epidemics will be vital for mounting an effective response. We consider the entire set of local epidemics in the United States; a broad selection of demographic, population density, and climate factors; and local mobility data, tracking social distancing interventions, to determine the key factors driving the spread and containment of the virus. Assuming first a linear model for the rate of exponential growth (or decay) in cases/mortality, we find that population-weighted density, humidity, and median age dominate the dynamics of growth and decline, once interventions are accounted for. A focus on distinct metropolitan areas suggests that some locales benefited from the timing of a nearly simultaneous nationwide shutdown, and/or the regional climate conditions in mid-March; while others suffered significant outbreaks prior to intervention. Using a first-principles model of the infection spread, we then develop predictions for the impact of the relaxation of social distancing and local climate conditions. A few regions, where a significant fraction of the population was infected, show evidence that the epidemic has partially resolved via depletion of the susceptible population (i.e., "herd immunity"), while most regions in the United States remain overwhelmingly susceptible. These results will be important for optimal management of intervention strategies, which can be facilitated using our online dashboard.</jats:p>

Overdispersion in COVID-19 increases the effectiveness of limiting nonrepetitive contacts for transmission control

“…By adjusting the mean infection rate, we varied the initial epidemic growth rate from 16 to 30% per day ( SI Appendix , Fig. S2 ), an interval that covers the range of premitigation growth rates observed in Europe and North America ( 22 – 24 ). We found, as expected, that a faster-growing epidemic is more difficult to mitigate; however, the enhanced effect of random sector mitigation when superspreading is present remains.…”

“…Here, k is the dispersion parameter, which determines the CV of the distribution according to CV = . The rate constant β is calibrated to reproduce the observed initial exponential growth rate of 23% per day of an unmitigated COVID-19 epidemic ( 22 – 24 ).…”

“…Contacts that occur less frequently belong to the random sector. To simulate superspreading, we assigned infectivity according to a gamma distribution with dispersion parameter k = 0.1 and adjusted the overall infectivity to produce an initial growth rate of 23% per day, as observed for COVID-19 in Europe and North America ( 22 – 24 ), which corresponded to a basic reproductive number of 2.5. In the unmitigated case, contacts were allowed in all three sectors; we then simulated two additional scenarios in which the regular and random contacts were restricted.…”

A strategy for finding people infected with SARS-CoV-2: optimizing pooled testing at low prevalence

“…During this period, the prevalence p of the virus grows exponentially. The doubling time for the virus is somewhat uncertain but it has been estimated, using data from China [3], to be τ 2 ≈ 2.4 days (a recent analysis gives τ 2 ≈ 1.9 days for New York and 2.9 days for Los Angeles [15]). If each RT-PCR test takes τ days, the viral prevalence grows by during an adaptive search.…”

Variability of Individual Infectiousness Derived from Aggregate Statistics of COVID-19

“…Naturally, a municipality with a large population will have a larger number of cases per day than a municipality with a small population. This is because there will be more imported cases in large regions (there may also be variations in R between cities and rural areas [18], but this is a second-order effect that we ignore). As most imported cases will come from other municipalities, we ignore effects of international travel.…”