There was a study published in 2007 which took FARS (for traffic fatalities) and GES (for injuries) data, and combined with NHTS (National Household Transp Survey. 2001) data to try to quantify relative risk of fatality/injury by travel mode per trip. Full text and full .pdf are both available free online:
Motor Vehicle Crash Injury Rates by Mode of Travel, United States: Using Exposure-Based Methods to Quantify Differences
Am J Epidemiol. 2007 Jul 15;166(2):212-8. Epub 2007 Apr 21.
The headline result was:
Relative to passenger vehicle occupants, motorcyclists, bicyclists, and pedestrians are 58.3, 2.3, and 1.5 times, respectively, more likely to be fatally injured on a given trip. Bus travel is the safest travel mode
There were some interesting observations (quirks?) in the stratified results, e.g. male/female vs. fatals/non-fatals; whereas male:female ratios were high for both car fatalities and bicyclist fatalites, the ratio for injuries was close to 1:1
There was some discussion on CaD! regarding, in particular, the shortcomings of attempting to measure exposure based on a survey, such as the NHTS. As opposed to, for example, the elaborate system of direct measurement used for motor vehicles.
The HPMS (Highway Performance Management System) claims to measure (motor vehicle) traffic on all public roads, using of course statistical sampling methods.