•Patients with late acute PJI have a better outcome when treated with revision surgery instead of DAIR.•Patients with late acute PJI can be selected for revision surgery according to the preoperative risk of DAIR failure defined by the CRIME80 score.•The causative microorganism and its susceptibility to antibiotics should ideally be taken into account as well to decide for the best surgical approach. We evaluated the treatment outcome in late acute (LA) periprosthetic joint infections (PJI) treated with debridement and implant retention (DAIR) versus implant removal. In a large multicenter study, LA PJIs of the hip and knee were retrospectively evaluated. Failure was defined as: PJI related death, prosthesis removal or the need for suppressive antibiotic therapy. LA PJI was defined as acute symptoms <3 weeks in patients more than 3 months after the index surgery and with a history of normal joint function. 445 patients were included, comprising 340 cases treated with DAIR and 105 cases treated with implant removal (19% one-stage revision (n = 20), 74.3% two-stage revision (n = 78) and 6.7% definitive implant removal (n = 7). Overall failure in patients treated with DAIR was 45.0% (153/340) compared to 24.8% (26/105) for implant removal (p < 0.001). Difference in failure rate remained after 1:1 propensity-score matching. A preoperative CRIME80-score ≥3 (OR 2.9), PJI caused by S. aureus (OR 1.8) and implant retention (OR 3.1) were independent predictors for failure in the multivariate analysis. DAIR is a viable surgical treatment for most patients with LA PJI, but implant removal should be considered in a subset of patients, especially in those with a CRIME80-score ≥3.