If you dont believe in stats then how about this approach:
The Safety guy walks in and hands you a paper with the following:
"The next aircraft accident for our company will be a ________________ involving (pilot) ______________."
Interesting this was given to us in an air assault unit. Since you know your peers, you may know who is at risk, weak aviator, hot dog...etc. If your name came up, the Safety guy pulled you aside, to make you aware of concerns and hopefully a chance to change it.
What data is driving this statement?
What have we identified as the biggest contributors to these events?
What data/methods did you use to identify these risks/threats?
What is our current exposure to these risks/threats?
What mitigation have we implemented to reduce exposure to these risks?
Are there additional forms of mitigation/elimination we have yet to implement?
Which aircraft/locations do you feel are at the biggest risk?
What is it about these aircraft/locations that increase our exposure?
Is there anything we can do to minimize/mitigate exposure in these specific locations?
One source is data gathered from reports submitted from the bases (EMS company). You still would have flawed data in the reporting because someone will not submit a report if they feel it will be used against them later, not worth reporting or "better not say anything, self protection". Worst case nothing will change.