Concussion heterogeneity remains a major challenge for clinical and scientific research. Yet, most studies continue to employ group comparison approaches that assume consistent one-to-one relationships between brain structure and symptoms. To parse concussion heterogeneity, we employed a double-multivariate approach using diffusion MRI of children with histories of concussion. Multivariate white-matter structural features captured more information about symptoms than all individual features. Results also revealed how different white-matter abnormalities led to similar symptom profiles. Lastly, multivariate features significantly predicted adverse psychiatric outcomes. These results suggest that concussion heterogeneity arising from complex structure/symptom relationships can be well captured by our double-multivariate approach.