This study aimed at identifying preoperative predictors of patient-reported outcomes after total knee arthroplasty (TKA) and at investigating their association with the outcomes over time.We used data from 2080 patients from the Knee Arthroplasty Trial who received primary TKA in the United Kingdom between July 1999 and January 2003. The primary outcome measure was the Oxford knee score (OKS) collected annually over 10 years after TKA. Preoperative predictors included a range of patient characteristics and clinical conditions. Mixed-effects linear regression model analysis of repeated measurements was used to identify predictors of overall OKS, and pain and function subscale scores over 10 years, separately.Worse preoperative OKS, worse mental well-being, body mass index greater than 35 kg/m(2), living in the most deprived areas, higher American Society of Anesthesiologists grade, presence of comorbidities, and history of previous knee surgery were associated with worse overall OKS over 10 years after surgery. The same predictors were identified for pain and function subscale scores, and for both long-term (10 years) and short-to-medium-term outcomes (1 and 5 years). However, fitted models explained more variations in function and shorter-term outcomes than in pain and longer-term outcomes, respectively.The same predictors were identified for pain and functional outcomes over both short-to-medium term and long term after TKA. Within the factors identified, functional and shorter-term outcomes were more predictable than pain and longer-term outcomes, respectively. Regardless of their preoperative characteristics, on average, patients achieved substantial improvement in pain over time, although improvement for function was less prominent.