The probability of being in a collision in a transit bus may be low, and the word "safe" covers this idea; but that doesn't speak to the severity of the consequences of the collision.

Statistics about deaths mix the two together are hard to interpret: are there low deaths because the frequency of accidents is low, or because their severity is low? Or what mixture of the two?

In statistics about automobile accidents, I would expect there to be a high preventable component: deaths due to neglected use of seatbelts as well as operator negligence: distracted, drunk or reckless driving.

What is the deaths-per-100M miles figure for automobiles, if we count only the deaths of seatbelt wearers, where the driver wasn't drunk, distracted or otherwise reckless?

Regarding seat belts for instance, according to https://injuryfacts.nsc.org/motor-vehicle/occupant-protectio... 51% of the day-time fatalities in 2020 were confirmed cases of no restraint use.

(That doesn't mean those people would have lived if they had used a seat belt; that can only be estimated. However, the same page cites a high rate of seat belt use for the same period: 90.3%. So 51% of the day-time fatalities are coming from the 9.7% unrestrained passengers/driver population.)