Hoping that some of you have come across this I'd like to ask:

In a descriptive study of the incidence of a certain per-operative and potentially harmful complication, (Answering the question How common is it?) 

What level of precision would you like too see?

What range would you expect the confidence interval for the incidence to be?

Depending of the actual number?

Depending of how serious the complication is?

I mean, we have a commonly agreed level of significance: 95%, of power: 80 or 90%, of acceptable loss to follow up 20% etc

What is the "rule of thumb" here?

Cheers