Background: Current models to estimate renal cell carcinoma (RCC) recurrence risk following surgery are derived from populations containing primarily low-risk patients. The objective of this study was to evaluate risk factors for recurrence among high risk non-metastatic RCC patients following attempted curative surgery. Methods: Data from 3 independent centers was analyzed for consecutive ≥ pT3a RCC patients without evidence of lymph node or distant metastases who were treated surgically from 2000-2016. Univariate and multivariate Cox proportional hazard models were used to evaluate associations of common clinical and pathological variables with recurrence risk. A risk model was constructed using independent predictors and recurrence risk was evaluated using Kaplan-Meier analysis. Results: Of 771 patients, 190 (24.6%) had RCC recurrence following attempted curative surgery at median 10.2 months (IQR 4.4-20.7). Median overall follow-up interval was 21.4 months (IQR 6.6-53.5). After multivariate Cox proportional hazard analysis, significant associations with RCC recurrence were not identified with: age, gender, race, systemic symptoms, local symptoms, pT stage, perinephric fat invasion, tumor thrombus, sinus fat invasion, serum hemoglobin, or serum albumin. Independent predictors included grade 4 HR 3.27 (95% CI 2.17, 4.92); tumor diameter > 7cm HR 1.70 (95% CI 1.18, 2.45), tumor necrosis HR 1.47 (95% CI 1.06, 2.02), and sarcomatoid/ rhabdoid features HR 1.86 (95% CI 1.12, 3.09) An unweighted risk model was created by assigning one point for each independent predictor. Estimated 3-year recurrence risk was 14%, 25%, 40%, 49%, and 69% for patients with 0,1,2,3 and 4 risk factors (p < 0.001) respectively. Conclusions: Independent predictors of recurrence for high risk non-metastatic RCC include: tumor diameter, necrosis, nuclear grade 4, and sarcomatoid/ rhabdoid features. This model may be used to estimate individual risk in RCC adjuvant therapy clinical trials.