This retrospective study over a 15-year period is to identify independent predictors of malignancy in Bosniak III (BIII) renal lesions and to build a prediction model based on identifiable clinical variables. These cysts, characterised by contrast-enhanced computed tomography (CECT), have thickened, irregular or smooth walls, or septa with measurable enhancement and has a 50% malignancy risk. In the study, patients with Von Hippel-Lindau (VHL) disease were excluded. Of the 107 lesions in 101 patients, 59 were malignant and 48 benign. The following data were recorded – gender, age, race, symptoms, body mass index (BMI), smoking, history of renal cell carcinoma (RCC), largest diameter of the lesion in CT, coexisting Bosniak IV (BIV) and solid renal masses, number and size of BIII lesions and time interval from diagnosis to excision. The malignant lesions were sub-classified into aggressive and non-aggressive. On univariate analyses, the strongest predictors of malignancy were African American race (P=0.043), history of RCC (P=0.026), co-existing BIII lesions (P=0.032), co-existing BIV lesions (P=0.104), BMI (P=0.078), and lesion size (P<0.001). Sixty-four percent of lesions <5cm were malignant vs. 34.4% of lesions >5cm. Eighty percent of African Americans had malignant lesions vs. 50.6% of whites. A model with lesion size, history of RCC and BMI offered the best performance with estimated probability of malignancy of >80%, the positive predictive value of the model is 92% (CI 78-100%). The model suggests that for every 1cm increase in BIII lesion size, the likelihood of malignancy decreases by a factor of 1.5 (95% CI 1.2-1.7), for every five-unit increase in BMI, the likelihood of malignancy increases by a factor of 2.0 (95% CI 1.3-3.0) and a history of RCC confers nine times (95% CI 0.99-82.15) the additional risk. The model has only a modest c-index of 0.719. Higher risk of malignancy in smaller BIII lesions are unexpected. Among the variables tested, lesion size was the strongest independent predictor of malignant potential of BIII lesions. Multivariate analysis could not be performed due to the small sample size.