Proposed model and estimated preliminary value for SARS-CoV-2 seroconversion time


Seroconversion is defined as the period necessary for antibodies to be detectable in the blood in reaction to a viral infection. Good estimate of serocoversion time from an infection by SARS-CoV-2 is of critical importance for epidemic modelling and for interpreting correctly various serology surveys now underway in various countries. From published evidence and simple considerations, we propose a first estimate of 15.4 days. This value was implemented to model seroprevalence in our CovModel version 1.5.


The reference for studying seroconversion of SARS-CoV-2 antibodies is Bin Lou et al. “Serology characteristics of SARS-CoV-2 infection since the exposure and post symptoms onset”

A key document in this paper is the visual timeline of antibodies appearance and detection, copied in Figure 1 below.

Figure 1: Antibodies dynamics visual timeline and data source


The objective was to interpret Figure 1 in terms of a mean time to achieving seroconversion, counted from exposure (infection) time.

In Figure 1, events are timed from exposure (infection) at day 0. The event of interest is the start of dark orange period (first occurence of “1”, i..e. the time when antibodies get detectable). We observed that in only 16 patients the precise timing was observed. For example, for Patient 14 the Date After Exposure (of seroconversion) is DAE=15. But for Patient 50, DAE can not be determined. We only know this was Day 8 at latest, because no sample results are provided for Days 1 to 7. We also observed (visually) that the 16 available for DAE were biased towards long incubation times (four last samples at bottom of Figure 1). In search of an unbiased mean value for DAE, we successfully correlated DAE to the incubation time – it is apparent in Figure 1 that time of onset of disease or incubation time is driving DAE. We applied the correlation to the known Covid-19 incubation time.


Seroconversion DAE and Date of Onset stats for 16 Patients are in Figure 2. The correlation between the two is DAE = 0.8869 DO + 10.802.

By applying the correlation to a published mean incubation period for Covid-19 of 5.2 days (*), we estimate the mean seroconversion time is 15.4 days.

Figure 2: Stats for 16 Patients with seroconversion time data (DAE). DO, Date of onset of disease is indicated (stats clearly show the subset is biased towards high inclubation, with mean 8.8 days). Plot shows DAE vs DO and correlation line. Calculation of mean time to seroconversion is provided.

(*) Source: The incubation period of 2019-nCoV from publicly reported confirmed cases: estimation and application.


Our workflow is very simple, and relies on only one publication. As often with Covid-19, time and efficiency is of the essence. We’ve already observed that the handful of serology surveys that were done as of today 19 April 2020 (in Denmark, Scotland, Finland, Germany and Santa Clara, CA, U.S.A) were not fully interpreted in terms of epidemic dynamics. Seroprevalence was oftern mixed up with infection rate. Our estimate of 15.4 days seroconversion time simply means that a serology survey is a “photography” of the epidemic two weeks earlier, in terms of the number of infected plus recovered.

Our value is preliminary, adapted to the philosophy of CovModel 1.5 to model the epidemic with manageable simple parameters, in a deterministic controlled way. For more advance usage, we recommend more work to refine our preliminary value and determine intervals of confidence.


To José Lourencço for having shared knowledge and the existence of Lou et al. paper in a tweet below. For his original paper on the importance of serology testing to assess the true number of infected and better define Covid-19 dynamics.