Neither of us was doing that so far. Case mortality seems to be the best basis, adjusted maybe for overall risk of infection when everything is opened up and nobody is using masks anymore. At which point any comparison most likely degenerated into guess work.
The main difference so is that vaccinations reduce the overall risk for the population. And we didn't even talk about severe long term effects of COVID.
That being said, we live with a lot of medications that have severe side effects in the 1/100,000 plus bracket without breaking a sweat.
By the way, I don't see how you come to the conclusion that the AZ vaccine is in any way riskier than COVID.
The challenge with COVID as always is that the risk of death is correlated across the population. So a 25 yo having COVID increases the odds that he dies and that everyone around him dies. But that’s a whole other discussion.
If the covid risk was flat over time, if you offered this vaccine yearly (quite possibly necessary) then you'd be saving ~90 lives in this age bracket per year. 10 would still die of covid, 100 would die from the vaccine, 90 whom would have died of covid would would instead survive.
It would be "better" depending on how you think about better. Telling people to throw away huge chunks of their lives to protect against a risk that you turn around and tell them they need to take a risky vaccine that's only half as bad... it isn't necessarily a good line. Especially because pandemic diseases tend to lower in their negative outcomes at time goes on, the vaccine will probably stay the same while the risk it prevents will get less and less threatening.
All of this based on statistics and assumptions that have a lot of uncertainty in them.
The bottom line is that in order to be a good idea, a vaccine needs to be much safer than the disease it prevents. Half as risky doesn't meet this mark; AZ would be fine for the oldest age bracket because the risk profile wouldn't change much, but the youngest age bracket there is a significant comparison between the two (and perhaps further unknown risks which haven't surfaced)