22.07.2020

COVID-19 infectivity profile correction

Modelization Infectiology
Ashcroft P et al
Pre-prints

Main result

In correcting the method of He et al., the authors obtain a very different infectivity profile. Notably, the infective period starts 5 days before the appearance of symptoms, rather than 2.3 days before. The authors also note a mistake in the normalization function of the likelihood calculation in the original paper, but this does not have a substantial effect on parameter estimation.

Takeaways

This article re-analyzes the data from He et al. (Nat. Med. 2020) on the infectivity profile of an individual infected with SARS-CoV-2. The new analysis shows that the pre-symptomatic infective period could be much longer than initially estimated, up to 5-6 days before the appearance of symptoms.

Strength of evidence Undetermined

It is particularly important to determine precisely the duration of the period before the appearance of symptoms during which an individual is infected with SARS-CoV-2, especially for contact tracing.
In light of these data, it appears necessary to increase the length of the contact history of an infected individual to detect potential secondary infections.
However, the substantial difference between the estimation of He et al. and the estimation in this article is based on the addition of two data points, which may have been outliers.
What may be necessary here is to construct a new dataset of times between infections, and to measure the infectivity profile again.

Objectives

The goal of this study is to determine the infectivity profile of an individual –that is, the temporal distribution of new infections from that individual—around the time of the appearance of symptoms. An earlier analysis by He et al., based on a study of the time to the appearance of symptoms in 77 confirmed cases, showed that individuals infected with SARS-CoV-2 become contagious approximately 2.5 days before showing symptoms. This article redoes the original analysis from the same data, and in correcting an error in the analysis, the authors demonstrate that the pre-symptomatic contagious period could have been underestimated.

Method

To determine the infectivity profile, the authors use the distirbution of incubation times obtained from Li et al. (NEJM 2020). Convolution of the infectivity profile and the incubation times should give the distribution of the times between appearance of symptoms in sequentially infected individuals. The authors use maximum likelihood estimation to infer the translated Gamma distribution that best fit the data.

bibliovid.org and its content are bibliovid property.

Legal Notice