I've seen the assertion made that we can statistically measure how many people died due to (for example) a heat wave, but we can't say for sure which ones.

I'd imagine something similar applies here. You'd have some number of deaths specifically attributable to lose of power, plus countless other deaths caused or prevented in non-obvious ways. This might be visible at a high level as a statistical outlier in the total number of deaths during the time period of an outage.