Overall, 608 patients were listed for hip (319) or knee (289) arthroplasty. The mean scores for knees were all better than hips (P < .001). On initial scoring, 324 patients (53%) were given certainty (mean NZOA, 80.5; OHKS, 10.0; RWS, 35.1), 90 (15%) given clinical over-ride (NZOA, 69.6; OHKS, 12.0; RWS, 33.2), and 194 (32%) returned to general practitioner (NZOA, 64; OHKS, 14.2; RWS, 30.8). Knees (38%) were more likely to be returned than hips (26%; P = .002). Fifty (26%) were re-referred during the study period (mean, 5 months) and given certainty or over-ride. The difference at final outcome between patients with certainty and clinical over-ride was NZOA, 10.3 points; Oxford, 1.6 points; and RWS, 1.4 points. The difference between clinical over-ride and returned to general practitioner was NZOA, 7.2; Oxford, 4.4; RWS, 5.3.