If you want to model actions that are executed before entering the orthogonal regions, yes: you need to create an enclosing state (See Fig. 15.35 "Orthogonal state with regions" in 2.4.1 superstructure).
From clause 15.3.11:
A composite state either contains one region or is decomposed into two or more orthogonal regions. Each region has a set of mutually exclusive disjoint subvertices and a set of transitions. A given state may only be decomposed in one of these two ways.
On the other side, if all you need are entry/exit pseudostates, you can avoid that additional complexity. See the paragraph about Composite state, its Description and Semantic variation point (default entry rule) in clause 15.3.11.