Text this: Pitfalls of using patient recall to derive preoperative status in outcome studies of total knee arthroplasty.