Attributable risk

When is using attributable risk (AR) a good thing and when is something like a hazard better?

AR can be a time independent thing which gives the proportion of total deaths, say, that are due to an infection, say. Or the proportion of the total population that would have been spared if they hadn't been infected.

A temporal version is available that consists of the cdf of the event times which in the limit equals the independent case.

For rare events i.e. probability of survival approximately 1 then the formula can be posed in terms of hazard functions. It consists of the total hazard and the uninfected conditional hazard.