The impact of long-term non-pharmaceutical interventions on COVID-19 epidemic dynamics and control

Abstract

Non-pharmaceutical interventions to combat COVID-19 transmission have worked to slow the spread of the epidemic but can have high socio-economic costs. It is critical we understand the efficacy of nonpharmaceutical interventions to choose a safe exit strategy. Many current models are not suitable for assessing exit strategies because they do not account for epidemic resurgence when social distancing ends prematurely (e.g., statistical curve fits) nor permit scenario exploration in specific locations. We developed an SEIR-type mechanistic epidemiological model of COVID-19 dynamics to explore temporally variable non-pharmaceutical interventions. We provide an interactive tool and code to estimate the transmission parameter, β, and the effective reproduction number, R E . We fit the model to Santa Clara County, California, where an early epidemic start date and early shelter-in-place orders could provide a model for other regions. As of April 22, 2020, we estimate an R E of 0.982 (95% CI: 0.849 -1.107) in Santa Clara County. After June 1 (the end-date for Santa Clara County shelter-in-place as of April 27), we estimate a shift to partial social distancing, combined with rigorous testing and isolation of symptomatic individuals, is a viable alternative to indefinitely maintaining shelter-in-place. We also estimate that if Santa Clara County had waited one week longer before issuing shelter-in-place orders, 95 additional people would have died by April 22 (95% CI: 7 -283). Given early life-saving shelter-in-place orders in Santa Clara County, longer-term moderate social distancing and testing and isolation of symptomatic individuals have the potential to contain the size and toll of the COVID-19 pandemic in Santa Clara County, and may be effective in other locations.

Non-pharmaceutical interventions during the COVID-19 pandemic: A review

“…The body of modeling work based on the SEIR scheme touches similar themes to those just mentioned, but due to the larger number of articles, there are also some new interesting angles. In particular, we find papers focusing, among other things, on the effects of face masks [82] , [85] , [86] , [87] , [89] , capturing different aspects of NPIs [64] , [67] , [69] , [70] , [71] , [73] , [74] , [75] , [77] , [79] , [80] , [81] , [83] , [84] , [90] , studying the problem of optimal control of the pandemic [72] , [76] , [78] , [88] , investigating approaches to lift measures put in place in the first wave [62] , [63] , [65] , model the possible exposure to the virus via droplets or fomites [91] , [92] , and consider the impact on the health care infrastructure of COVID-19 and seasonal flu in the winter [68] .…”

A New Approach to the Dynamic Modeling of an Infectious Disease

“…Equation (4) is a six-parameter model – the parameters are m 0 , N, μ 1 , μ 3 , τ 1 and τ 2 . Among these, the last two are well-known – they are about 7 days and 3 days respectively [21]. These are values we shall use throughout the entirety of this Article .…”

“…Rather, we give the values of X and hence of the capture probability P 3 X or 1− μ 3 for some realistic sets of parameter values. With the standard values 7 and 3 for τ 1 and τ 2 and taking τ 4 = 4 (the total incubation period is approximately 7 days on average [21]), the value τ 3 = 2 gives T = 1/6. This implies that contact traced cases remain at large for a relatively short time.…”

Impact of Immediate and Preferential Relaxation of Social and Travel Restrictions for Vaccinated People on the Spreading Dynamics of COVID-19 : a Model-Based Analysis

“…[17] we have shown that contact tracing will indeed capture only a small percentage of total cases if the asymptomatic fraction is high; moreover, contact tracing is managed by healthcare professionals many of whom will now be re-deployed to vaccination drive. We use the parameter values [S1] τ 1 = 7 and τ 2 = 3, so that the function n gets defined as We assume that vaccinated cases have the same μ 1 , τ 2 and τ 1 as non-vaccinated ones (details in §2), so that we can use this function n to count at large cases of both unvaccinated and vaccinated groups. Putting all this together, we have the equation with the second line being a simple rearrangement of the first.…”