Breeding programs that select for scab resistance aim to reduce Deoxynivalenol (DON), the primary trait. Because evaluating DON is costly, the secondary scab resistance traits: incidence, severity and Fusarium damaged kernels (FDK) are typically evaluated on large numbers of lines while DON is measured on fewer lines at later stages of selection. Genomic selection (GS), selection based genomic estimated breeding value, can improve plant breeding efficiency in two ways. GS can accelerate rates of genetic gain by enabling parent selection prior to phenotyping. At later stages of selection, GS can improve phenotypic selection accuracy or achieve the same level of selection accuracy with less phenotypic data by tapping into information from relatives, and from correlated traits in the case of multi-trait GS. Using datasets from Purdue University and the University of Illinois, we tested whether multi-trait GS models could be used to reduce phenotyping costs for scab resistance without sacrificing selection accuracy. We considered a scenario where selection for scab resistance among breeding candidates is done based on traits other than DON. We compared multi-trait GS models for DON which included DON phenotypes on the training set and other scab resistance traits on both the training set and the selection candidates. Multi-trait GS models which included FDK on the selection candidates were the most predictive of DON. Furthermore, we discovered that once FDK was included in the model, adding data on severity and incidence did not improve accuracy. These results indicate that once breeding programs begin using GS, phenotyping severity and incidence may no longer be necessary. Regardless of the selection method, phenotyping FDK and DON will remain critical.