I am would like to use 8-node quadrilateral serendipity elements to model a problem. However it seems to me that mid-nodes are not required at the boundary, as shown in the diagram below i.e. elements 3 and 4 should be 7-node elements, but the elements 1 and 2 should be 8-node elements. (The nodes A, B and C are fixed boundary nodes). Is this correct?
If correct is there any disadvantage to this approach?