I think in general percentages less than 0.1% should be reported as "1-in-XXX". Certainly when I create pages to measure packet loss I put 1 in 50,000 rather than 0.002%
For example: (chance of getting covid) * (chance of specific side effect) = chance of getting covid and experiencing side effect.
The difference (in "certainty", not "risk") between 90% (1-in-10) and 95% (1-in-20) is the same as the difference between 98% (1-in-50) and 99% (1-in-100).