Is single observer identification of wound infection a reliable endpoint?