[BostonCard]

If you look at figure 2 in this paper, it provides the cumulative distribution function of the incubation period.

Based on the figure, the median incubation period is 7 days (table 2 estimates it at between 5 and 7 days).  Depending on the model, the 95th percentile (that is, 95% of the time, the incubation period is that length or shorter) is 10 - 13 days depending on which model is chosen.  The 99th percentile 12 to 18 days, again depending a bit on the model.

Here's a review paper that finds something similar.

Results The incubation period distribution may be modelled with a lognormal distribution with pooled mu and sigma parameters (95% CIs) of 1.63 (95% CI 1.51 to 1.75) and 0.50 (95% CI 0.46 to 0.55), respectively. The corresponding mean (95% CIs) was 5.8 (95% CI 5.0 to 6.7) days. It should be noted that uncertainty increases towards the tail of the distribution: the pooled parameter estimates (95% CIs) resulted in a median incubation period of 5.1 (95% CI 4.5 to 5.8) days, whereas the 95th percentile was 11.7 (95% CI 9.7 to 14.2) days.

So, reducing the quarantine period would increase the risk that someone on the tail end of the distribution makes it through.  However, assuming the more optimistic model is correct, you can probably reduce the quarantine period to 10 - 12 days and still be able to catch about 95% of secondary cases.  If you add asymptomatic testing at the end, you will probably catch some asymptomatic and pre-symptomatic secondary cases that may not have become symptomatic until the tail end of the distribution.

On that basis, I think a 10-day quarantine period with testing at the end probably is supported.  A seven day quarantine period is probably too short.

BC

[BostonCard]

If you look at figure 2 in this paper, it provides the cumulative distribution function of the incubation period.

Based on the figure, the median incubation period is 7 days (table 2 estimates it at between 5 and 7 days).  Depending on the model, the 95th percentile (that is, 95% of the time, the incubation period is that length or shorter) is 10 - 13 days depending on which model is chosen.  The 99th percentile 12 to 18 days, again depending a bit on the model.

Here's a review paper that finds something similar.

Results The incubation period distribution may be modelled with a lognormal distribution with pooled mu and sigma parameters (95% CIs) of 1.63 (95% CI 1.51 to 1.75) and 0.50 (95% CI 0.46 to 0.55), respectively. The corresponding mean (95% CIs) was 5.8 (95% CI 5.0 to 6.7) days. It should be noted that uncertainty increases towards the tail of the distribution: the pooled parameter estimates (95% CIs) resulted in a median incubation period of 5.1 (95% CI 4.5 to 5.8) days, whereas the 95th percentile was 11.7 (95% CI 9.7 to 14.2) days.

So, reducing the quarantine period would increase the risk that someone on the tail end of the distribution makes it through.  However, assuming the more optimistic model is correct, you can probably reduce the quarantine period to 10 - 12 days and still be able to catch about 95% of secondary cases.  If you add asymptomatic testing at the end, you will probably catch some asymptomatic and pre-symptomatic secondary cases that may not have become symptomatic until the tail end of the distribution.

On that basis, I think a 10-day quarantine period with testing at the end probably is supported.  A seven day quarantine period is probably too short.

BC