In the workplace accident and injury space, breakdowns in return-to-work can be costly to workers’ compensation systems and detrimental to recovery of workers. The ability to predict whether a claim is on the trajectory towards a breakdown will allow for early intervention and additional support, resulting in improved return-to-work outcomes. The main goal of this project is to create a data science tool that will alert claim managers about claims that have a high likelihood of experiencing a re-open before a breakdown occurs. Further, it aims to answer three questions: which features are most influential in predicting breakdowns, which kind of model performs best overall, and is the predictive capability of the final model sufficient to justify integration into the claims management system.