[H+T--Friends] A Data Question

It was nice to see everyone at TRB. While there I was in a discussion about the probably that bike crashes that result in injuries might be under represented in our crash data. Especially when you consider that a high percentage of the crashes do not involve a motor vehicle and never make it to police records. If i were trying to get a handle on this at a regional or state level are there any "non-traditional" sources of data like hospital records that can be used. Anyone doing any work in this area? This is becoming ever so important as we see more and mode shifting going on.