To be clear, I mean that hospitals will inevitably be overwhelmed once you see a sharp upswing in hospitalisations; by the time you see that the pipeline is likely already primed and nothing you can do will change the trajectory of the upswing much for days. If you see a sharp upswing in cases, though, swift action can maybe deal with that (assuming you have decent observability). Most the the cases will never see the inside of a hospital, but they provide a somewhat useful proxy for the future of hospitalization.

If you need to pick a metric on which to reinstitute lockdowns, hospitalization isn't really very useful; it's better than nothing but if you use it you're probably going to overshoot.