OBJECTIVE: Useful Field of View (UFOV) refers to the area of the visual field where one can extract information in a brief glance. Previous studies have shown that reductions in UFOV lead to an increase in automobile crash risk. It is also known that UFOV performance can be improved using computer-based adaptive training. It was recently reported that UFOV training reduced at-fault crash rates in the 5 year period following training. The long-term goal of the current study is to investigate whether these findings generalize to a large population of older automobile insurance policyholders. The first stage of this study was to determine if baseline visual performance on a range of tasks is correlated with previous crash history and to confirm that training improves performance on these tasks. METHODS: Allstate Insurance partnered with Posit Science to offer the InSight visual training program to a subset of members residing in Pennsylvania. InSight is a computer-based program developed based on findings from brain plasticity research. It was designed to improve the speed and accuracy of visual processing and expand visual attentional capacity. It contains five exercises including the latest version of UFOV training. A mass mailing was sent to 100,000 Allstate automobile policyholders aged 50 and over. Approximately 8% requested the software and 51% of those receiving the software started the training (N = 4036, mean age 60.7, 54% male). Usage data was reported electronically to Posit Science via the internet. Driving history was obtained from Allstate records. Data from Posit Science and Allstate were matched and anonymised by a third party for analysis. RESULTS: Pre-training baseline performance on the InSight tasks was correlated with crash history over the previous 3 years. Lowest quintile performance was associated with a twofold increase in crash incidents. This result is consistent with previous research from randomized-controlled studies. Almost half of the participants had at least one assessment in the highest risk quintile. Training improved performance on the InSight tasks, thus reducing an established crash risk factor. CONCLUSIONS: This study has shown that the findings from randomized controlled studies generalize to a broad population of automobile policyholders. It established an association between poor performance on In-Sight visual training tasks and crash history, and demonstrated that people who train at home can improve performance on such tasks and thus lower this particular measure of crash risk. The prospective relationship between reduction of this measure of crash risk and observed reduction of actual recorded crashes is now underway. The study highlights the potential for improving driving safety in the community of older drivers.