Background: Debilitating pain associated with knee osteoarthritis (OA) often leads patients to seek and complete total knee arthroplasty (TKA). To date, few studies have evaluated the relation of functional impairment to the risk of TKA, despite the fact that OA is associated with functional impairment. Questions/purposes: The purpose of our study was to (1) evaluate whether function as measured by WOMAC physical function subscale was associated with undergoing TKA; and (2) whether any such association varied by sex. Methods: The National Institutes of Health-funded Multicenter Osteoarthritis Study (MOST) is an observational cohort study of persons aged 50 to 79 years with or at high risk of symptomatic knee OA who were recruited from the community. All eligible subjects with complete data were included in this analysis. Our study population sample consisted of 2946 patients with 5796 knees; 1776 (60%) of patients were women. We performed a repeated-measures analysis using baseline WOMAC physical function score to predict the risk of TKA from baseline to 30 months and WOMAC score at 30 months to predict risk of incident TKA from 30 months to 60 months. We used generalized estimating equations to account for the correlation between two knees within an individual and across the two periods. We calculated relative risk (RR) of TKA over 30 months by WOMAC function using a score of 0 to 5 as the referent in multiple binomial regressions with log link. Results: Those with the greatest functional impairment (WOMAC scores 40–68; 62 TKAs in 462 knee periods) had 15.5 times (95% confidence interval [CI], 7.6–31.8; p < 0.001) the risk of undergoing TKA over 30 months compared with the referent group (12 TKAs in 3604 knee periods), adjusting for basic covariates, and 5.9 times (95% CI, 2.8–12.5; p < 0.001) the risk after further adjusting for knee pain severity. At every level of functional limitation, the RR for TKA for women was higher than for men, but interaction with sex did not reach significance after adjustment for covariates including ipsilateral pain (p = 0.138). Conclusions: Baseline physical function appears to be an important element in patients considering TKA. Future studies should examine whether interventions to improve function can reduce the need for TKA. Level of Evidence: Level III, observational cohort study.