Objective: Patient reported outcomes are increasingly being used to evaluate the effectiveness of interventions. Often, patients are asked to assess their current health in relation to their past health. Yet, if their perceptions have shifted, the assessment of improvement may be biased. Our objective was to evaluate response shift in arthritis patients' self-report of health status 3 years after undergoing total joint replacement (TJR). Methods: As part of a longitudinal prospective cohort study on the outcome of TJR in patients with arthritis of the knee or the hip, participants were evaluated before surgery, and 3 years later. In both occasions, pre- and post-surgery, patients were asked to rate their overall health status on the EQ-5D thermometer, a 0 to 100 visual analogue scale (VAS). At year 3, patients were also asked to rate what their health status had been before surgery using the same VAS (‘then VAS'). Response shift was calculated as ‘then VAS' – pre-surgical VAS. Patient satisfaction with the procedure was evaluated using 5-point ordinal scales. Results: 358 patients were included, 204 underwent knee replacement and 154 hip replacement; mean age was 68.3 yrs and 211 (59%) were female. Mean VAS before surgery was 56.1 (±19.9); ‘then VAS' was 45.3 (±24.7), showing a mean response shift of -10.8 (±28.6). Three quarters of the patients had a shift of at least 10 points, and in two thirds of these the shift was negative (thought health was worse than had been reported). No significant associations were found between response shift and joint replaced (hip or knee), gender or age. A statistically significant negative correlation was observed between response shift and current VAS health status (r=-0.23, p<0.001): patients with better self-reported post-surgical health had lower values in their response shift, indicating lower perception of ‘then' health. Response shift was also significantly associated with satisfaction with TJR (p<0.001): higher levels of satisfaction were associated with lower ‘then' health, whereas dissatisfaction was related to better perception of ‘then' health than had been reported presurgically. Conclusion: A significant response shift was observed in patients undergoing TJR with respect to their presurgical health status. This shift was significantly associated with current health and satisfaction with the results of the surgery. Response shift should be considered when evaluating patient reported improvement.