We did something like this for the Pedestrian Safety and Access Task Force. You can pull the individual crash reports. We downloaded all of them for - five years? - can't quite remember. Then we parsed them by location, etc. That was most of how we concluded that the crosswalk ordinance wasn't a problem. Most of the crashes didn't happen at mid-block locations, the trends were mostly explained by the 2000s recession, and were similar in Ann Arbor, other Michigan cities, Michigan, and regionally.
Tim's point is great. Without volume information, you can't really tell that much. For instance, most of the crashes happen on 25 mph roads. Well, that's probably mostly because non-arterial streets are mostly neighborhood streets and that's where most people - particularly the least experienced - do most of their biking. If we had volume information, we'd be able to adjust to get per bike/mile traveled, and then see the patterns better.
If we tried to make decisions based on just the crashes, we might conclude that all of our 25mph streets are horribly dangerous for biking, and we need to devote almost all of our resources to making them safer before we do anything on the arterials. But that would probably be the wrong conclusion. No doubt we certainly should be doing more traffic calming in neighborhoods, but the problem in the neighborhoods might also be kids biking into the backs of parked cars and needing to go to the ER. (My daughter did that twice - sans the ER part - before we realized you must teach kids to avoid staring at obstacles or they'll steer right into them!)
If it's of interest, I can probably pull up the code I used to parse the UD-10s. They have a text and PDF version of those reports, and we had some software to go through the text versions to filter out the useful information.
Ken