I'm sure Scott was in a rush and meant to suggest something like:
AND NOT EXISTS(SELECT 1 FROM TEST t1 WHERE t1.SAMPLE_NUMBER = t.SAMPLE_NUMBER)
If you include COUNT(*) in a sub-select SQL will actually go to the trouble of counting up all the rows ... and if you then don't USE that COUNT(*) value somewhere it was a waste of time.
EXISTS or NOT EXISTS just needs to establish if there is a single row that matches (and not worry if there is more than one row). If there is a suitable index available SQL will just pull the first matching entry from that (or fail to find one) and that takes a very short amount of time.