You could add the blank record as you are doing now and add an index on that field. If it's a relatively small dataset, that should be fast.