Background and purpose: A new periprosthetic joint infection (PJI) definition has recently been proposed by the European Bone and Joint Infection Society (EBJIS). The goals of this paper are to evaluate its diagnostic accuracy and compare it with previous definitions and to assess its accuracy in preoperative diagnosis.Patients and methods: We retrospectively evaluated a multicenter cohort of consecutive revision total hip and knee arthroplasties. Cases with minimum required diagnostic workup were classified according to EBJIS, 2018 International Consensus Meeting (ICM 2018), Infectious Diseases Society of America (IDSA), and modified 2013 Musculoskeletal Infection Society (MSIS) definitions. 2 years’ minimum follow-up was required to assess clinical outcome.Results: Of the 472 cases included, PJI was diagnosed in 195 (41%) cases using EBJIS; 188 (40%) cases using IDSA; 172 (36%) using ICM 2018; and 145 (31%) cases using MSIS. EBJIS defined fewer cases as intermediate (5% vs. 9%; p = 0.01) compared with ICM 2018. Specificity was determined by comparing risk of subsequent PJI after revision surgery. Infected cases were associated with higher risk of subsequent PJI in every definition. Cases classified as likely/confirmed infections using EBJIS among those classified as not infected in other definitions showed a significantly higher risk of subsequent PJI compared with concordant non-infected cases using MSIS (RR = 3, 95% CI 1–6), but not using ICM 2018 (RR = 2, CI 1–6) or IDSA (RR = 2, CI 1–5). EBJIS showed the highest agreement between pre-operative and definitive classification (k = 0.9, CI 0.8–0.9) and was better at ruling out PJI with an infection unlikely result (sensitivity 89% [84–93], negative predictive value 90% [85–93]).Conclusion: The newly proposed EBJIS definition emerged as the most sensitive of all major definitions. Cases classified as PJI according to the EBJIS criteria and not by other definitions seem to have increased risk of subsequent PJI compared with concordant non-infected cases. EBJIS classification is accurate in ruling out infection preoperatively.