Background: Late infection after posterior spinal arthrodesis for adolescent idiopathic scoliosis (AIS) is the leading cause of late revision. While implant removal and antibiotic therapy are usually curative, patients may experience deformity progression. The goal of this study was to compare outcomes after implant exchange (IE) or removal (IR) to treat late-onset (≥1 y postoperative) deep surgical site infection (SSI) after spinal arthrodesis in patients with AIS. Methods: Using a multicenter AIS registry, patients who underwent posterior spinal fusion between 2005 and 2019 and developed late deep SSI treated with IE or IR were identified. Radiographic, surgical, clinical, and patient-reported outcomes at most recent follow-up were compared. Results: Of 3,705 patients, 47 (1.3%) developed late infection 3.8±2.2 years (range 1 to 9.7 y) after index surgery. Mean follow-up after index surgery was 6.1 years, with 2.8 years (range 25 to 120 mo) of follow-up after revision surgery. Twenty-one patients were treated with IE and 26 with IR. At the latest follow-up, average major-curve loss of correction (1° vs 9°, P<0.001) and increase in kyphosis (1° vs. 8°, P=0.04) were smaller in the IE group than in the IR group. Two IR patients but no IE patients had reoperation. Patients who underwent IE had higher Scoliosis Research Society 22-Item Patient Questionnaire (SRS-22) total scores (4.38 vs. 3.81, P=0.02) as well as better subscores for self-image, function, and satisfaction at the latest follow-up than those who underwent IR only. There were no significant between-group differences in operative duration, estimated blood loss, length of hospital stay, or changes in SRS-22 total scores. No patient had a subsequent infection during the follow-up period. Conclusions: When treating late-onset deep SSI after posterior spinal fusion for AIS, single-stage IE is associated with better maintenance of major curve correction, sagittal profile, and patient-reported outcomes and fewer reoperations compared with IR, with no significant differences in blood loss, operative duration, or length of stay. No time interval from index surgery to IR was observed where the corrected deformity remained stable. Both techniques were curative of infection. Level of Evidence: Level III