Aims. The proportion of arthroplasties performed in the ambulatory setting has increased considerably. However, there are concerns whether same-day discharge may increase the risk of complications. The aim of this study was to compare 90-day outcomes between inpatient arthroplasties and outpatient arthroplasties performed at an ambulatory surgery centre (ASC), and determine whether there is a learning curve associated with performing athroplasties in an ASC. Methods. Among a single-surgeon cohort of 970 patients who underwent arthroplasty at an ASC, 854 (88.0%) were matched one-to-one with inpatients based on age, sex, American Society of Anesthesiologists (ASA) grade, BMI, and procedure (105 could not be adequately matched and 11 lacked 90-day follow-up). The cohort included 281 total hip arthroplasties (THAs) (32.9%), 267 unicompartmental knee arthroplasties (31.3%), 242 primary total knee arthroplasties (TKAs) (28.3%), 60 hip resurfacings (7.0%), two revision THAs (0.3%), and two revision TKAs (0.3%). Outcomes included readmissions, reoperations, visits to the emergency department, unplanned clinic visits, and complications. Results. The inpatient and outpatient groups were similar in all demographic variables, reflecting successful matching. The reoperation rate was 0.9% in both cohorts (p = 1.000). Rates of readmission (2.0% inpatient vs 1.6% outpatient), any complications (5.9% vs 5.6%), minor complications (4.2% vs 3.9%), visits to the emergency department (2.7% vs 1.4%), and unplanned clinic visits (5.7% vs 5.5%) were lower in the outpatient group but did not reach significance with the sample size studied. A learning curve may exist, as seen by significant reductions in the reoperation and overall complication rates among outpatient arthroplasties over time (p = 0.032 and p = 0.007, respectively), despite those in this group becoming significantly older and heavier (both p < 0.001) during the study period. Conclusion. Arthroplasties performed at ASCs appear to be safe in appropriately selected patients, but may be associated with a learning curve as shown by the significant decrease in complication and reoperation rates during the study period. Cite this article: Bone Joint J 2021;103-B(7 Supple B):84–90