You can imagine a function like this:
TOTAL EXPENSES = function(HEALTHCARE BILLS, DEDUCTIBLE)
y = f(x)
Then for each deductible you can draw a chart, which I did already once in this forum, July 2018 (oh how quickly the stuff gets buried here). Then you can see which deductible gives you the lowest expenses for each level of healthcare bills, which as you said, you can’t predict. And then you will notice, that the middle options are never the best.
Now when I think about it, in order to get the best deductible for you, you should try to figure out the probability distribution for your predicted healtcare bills. Then you plug it into the function for each deductible and you get the expected total cost. Maybe then the middle deductibles would start making sense? Like, what if your healthcare bills would be: 50% 0 CHF, 50% 5000 CHF. Then it would be for some years the best to take the 500 deductible, and for the others the 2500 one. But since you don’t know what to expect, maybe the 1500 one would average out as the best one? Dont’ know, would need to calculate.