While studying the original literature of the Hausdorff-Measure "Dimension und äußeres Maß" (1918), on page 168 there is given the inequality $$f(x-y+z) > f(x) - f(y) + f(z) \text{ for } x < y < z.$$ Since the publication is pretty long ago and it's written in a strange style, I'm not quite sure which conditions $f$ needs to fulfill. I guess it needs to be a concave (convex upwards) function, at least for simple examples it seems to work. But I can't find a proof for this statement.
I would be very happy if someone could say whether this statement is correct and in the best case could provide proof for it.
It is correct that the inequality holds for concave function. It is a special case of Karamata's inequality.
In both cases, for a concave function $f$ it follows that $$ f(z) + f(x) \le f(y) + f(x+z-y) \, . $$ If $f$ is strictly concave then the inequality is strict.
You can also derive it directly by writing down the concavity condition for $x < y < z$ $$ f(y) \ge \frac{z-y}{z-x}f(x) + \frac{y-x}{z-x}f(z) $$ and for $x < x+z-y < z$ $$ f(x+z-y) \ge \frac{y-x}{z-x}f(x) + \frac{z-y}{z-x}f(z) $$ and add these two inequalities. Again, the inequalities are strict if $f$ is strictly concave.