Identifying future high-cost patients allows healthcare organizations to take preventative measures to both reduce future patient costs and lessen the burden of illness. This paper expands upon past risk adjustment strategies to predict the persistently high-cost patients by combining clinical and claims data on patients and assessing risk using machine learning techniques. Our approach not only leads to substantial gains in predictive accuracy, but also reduces the amount of data needed to identify high-risk patients, enabling providers to confidently identify long-term health risk in as little as three months after their initial encounter.