This article makes the following contributions: (1) a multimodal workload prediction algorithm in the RAS domain; (2) exploration of the Dempster-Shafer Theory of uncertainty for sensor fusion and modeling; (3) investigation of the relationship between objective workload ground-truth (task difficulty) and subjective workload ground-truth (NASA-TLX); (4) evaluation of the algorithm's generalization across multiple surgeons and tasks; and (5) examination of the algorithm's resilience to noise.