Vertebral fracture assessment (VFA) is a potential screening tool for vertebral fractures, but it is uncertain how to optimize the selection of women for VFA. We investigate the use of a probability score (VFscore) to select women for VFA screening and compare this to other means of targeting VFA. We identified 1,572 treatment-naive women over age 65 who had undergone routine VFA screening. Risk factors for fracture on VFA were identified using multivariate logistic regression, and a VFscore was created. Different thresholds of VFscore were examined and compared to using BMD as a means of targeting screening. After multivariate logistic regression, the risk factors significantly associated with the presence of a fracture on VFA were age, femoral neck BMD, prior clinical fracture, and height loss/kyphosis. The VFscore derived from these factors had a 65.5% sensitivity and a 65.5% specificity for determining vertebral fracture status. For equal resource requirements, the VFscore identified more women with fracture than using BMD category to target VFA. Compared to routinely screening all women, VFscore enabled a 30% reduction in the number of women undergoing VFA while still identifying >90% of women with a vertebral fracture. Overall, a large proportion of the population is required to undergo VFA in order to ensure that the majority of women with a vertebral fracture are selected for screening. The VFscore increased the efficiency of VFA screening to a modest degree compared to screening routinely or according to BMD category.