No. I'm not sure how I would get that research, or if it's even possible. The problem with preparing for war is that war is fairly uncommon, and even when it happens, certain activities are rare enough with such a huge number of alternate causalities for failure that you might not have enough data for a useful natural experiment.

What I can say with some confidence is that there are perverse organizational incentives for some highly accident-prone training exercises. So, even if these exercises are required for readiness, that's not the (only) reason they're done in such high numbers.