Background Red blood cell (RBC) transfusion and erythropoiesis-stimulating agent administration are cornerstones of clinical practice, yet concerns exist as to potential increased risk of thrombotic events. This study aims to identify RBC traits most relevant to venous thromboembolism (VTE) and assess their genetically predicted effects on VTE in the general population. Methods and Results We used multivariable mendelian randomization with bayesian model averaging for exposure selection. We obtained genetic variants predicting any of 12 RBC traits from the largest genome-wide association study of hematological traits (173 480 participants of European ancestry) and applied them to the UK Biobank (265 424 white British participants). We used univariable mendelian randomization methods as sensitivity analyses for validation. Among 265 424 unrelated participants in the UK Biobank, there were 9752 cases of VTE (4490 men and 5262 women). Hemoglobin was selected as the plausible important RBC trait for VTE (marginal inclusion probability=0.91). The best-fitting model across all RBC traits contained hemoglobin only (posterior probability=0.46). Using the inverse variance-weighted method, genetically predicted hemoglobin was positively associated (odds ratio, 1.21 per g/dL unit of hemoglobin; 95% CI, 1.05-1.41) with VTE. Sensitivity analyses (mendelian randomization-Egger, weighted median, and mendelian randomization pleiotropy residual sum and outlier test) gave consistent estimates. Conclusions Endogenous hemoglobin is the key RBC trait causing VTE, with a detrimental effect in the general population on VTE. Given men have higher hemoglobin than women, this finding may help explain the sexual disparity in VTE rates. The benefits of therapies and other factors that raise hemoglobin need to be weighed against their risks.