A user on reddit suggested the following, which seems like a good idea to consider:
It seems size and color represent the same thing.
Consider, per region, fitting a regression starting with linear, possibly also exponential (whatever has the highest avg R^2 value overall for all regions). Then, the size of each node could represent the current legend (# cases) but the color would represent growth, with its own legend.
A user on reddit suggested the following, which seems like a good idea to consider: