Reported infectious period (IP) for symptomatic cases (T3+T5 parameter) from virological studies where serial diagnostic tests were undertaken to infer IP (exposure to ≥2 neg. tests); tracking studies where IP is inferred from patient histories from onset to recovery or death; modelling studies where IP is reported as a prior (assumed parameter value) or an posterior estimate
| Study | Location | Parameter (days) | Central tendency reported | Variation (days; inclusion) | Comment |
| Tracking studies | |||||
| He et al 29 | Vietnam, Malaysia, Japan, China, Taiwan, USA and Singapore | 9.3 days | Mean | 7.8–10 (95% CI*) | The paper reported on 77 infector–infectee pairs which were sequential/serially tested, using publicly available data. Viral dynamics (Guangzhou, China; n=94) interpreted by the authors suggested an infectious period starting 2.3 (95% CI 0.8 to 3.0 days) days prior to symptoms, peaking 0.7 days (95% CI −0.2 to 2.0 days), continuing up to 7 days from onset |
| Ma et al 8 | Various | ~5 days | Median | Range 0–24 | The authors estimated the infectious period as latent minus the serial interval, using a dataset of 1155 cases. Range 0–24; IQR: 2–9; calculated from data presented within the paper. |
| Modelling studies | |||||
| Li et al 27 | China | 3.45 days (posterior estimated from model for documented cases) | Median | 95% CI for the mean: 3.19, 3.72 | Mathematical model. Priors for mean documented infectious period was a flat (uniform) distribution 2–5. ‘Documented’ cases were defined as those severe enough to be confirmed. This corralling of the infectious period relative to other studies should take into account that the distribution is used for the central tendency, not the whole distribution. |
| Tuite et al 26 39 | Canada | 6–6.5 days (prior; fixed parameter within a deterministic model) | Fixed parameter | Mathematical model (deterministic), with a fixed parameter estimate of 6.5 days (a) and 6 days (b), respectively. Important to note that duration for ‘mild’ was equal to severe cases. | |
| Lourenço et al 89 | UK | ~3–5 days (posterior; approximate depending on scenario tested) | Mean | 95% CI of 3–6 days | Mathematical model. The prior used was given a Gaussian distribution (normal curve); mean 4.5; SD 1; approximate 95% CI of 3–6 days. The reported posterior of this parameter was presented graphically and depended on R0 and proportion at risk. Depending on the scenarios tested, mean duration of infectiousness appeared to vary from 3 to 5 days. |
| Zhu et al 37 | Wuhan, China | 12.5 days (posterior estimated from model) | Mean | 11.4 variance | Mathematical model. The parameter was estimated using a Weibull distribution. The prior for this parameter was 10 days. The posterior variance around the mean was 11.4, and therefore the distribution had a long tail. This study was a modelling (SEIR extended model). |
| Davies et al 15 | UK | 7 days (prior) | Mean | Model with asymptomatic infection compartment. Modelled with a gamma distribution, beta 1.4; alpha 5. Despite the subclinical aspect of this parameter, it could be considered analogous to total infectious period without intervention. | |
| Davies et al 14 | UK | 5 days (prior) | Mean | Model with asymptomatic infection compartment. Modelled with a gamma distribution, k=4. Authors: ’Assumed to be the same duration as total infectious period for clinical cases, including preclinical transmission’. | |
| Piccolomini and Zama38 | Italy | 20 days (prior) | Fixed | Parameter estimate assumed for the infectious period within an SEIRD Susceptible, Exposed, Infectious, Recovered and Dead) model, fitted to data from the epidemic in Italy. | |
*95% CI from presymptom infectious period only.