Periprosthetic joint infections (PJIs) are uncommon but are devastating complications of total knee replacement (TKR). We analysed the risk factors of revision for PJI following primary TKR and their association with PJI at different post-operative periods. Primary TKRs and subsequent revision surgeries performed for PJI from 2003–2014 were identified from the National Joint Registry (NJR). Multilevel piece-wise exponential non-proportional hazards models were used to estimate the effect of the investigated factors at different post-operative periods. Patient, perioperative and healthcare system characteristics were investigated and data from the Hospital Episode Statistics for England were linked to obtain information on specific comorbidities. The index TKRs consisted of 679,010 primaries with 3,659 subsequently revised for PJI, 7% within 3 months, 6% between 3–6months, 17% between 6–12months, 27% between 1–2years and 43% ≥2 years from the index procedure. Risk factors for revision for PJI included male sex, high BMI, high ASA grade and young age. Patients with chronic pulmonary disease, diabetes and liver disease had higher risk of revision for PJI, as had patients who had a primary TKR for an indication of trauma or inflammatory arthropathy. Surgical procedure, fixation method, constraint and bearing type influenced the risk of revision for PJI. Their effects were period-specific. No or small associations were found with the operating surgeon grade, surgical volume and hospital surgical volume. These findings from the world's largest joint replacement registry show a more complex picture than the meta-analyses published to date with specific time-dependent effects for the identified risk factors