Spoilt for choice: implications of using alternative methods of costing hospital episode statistics.