Ankle injuries are one of the most common musculoskeletal injuries seen in athletics. Identifying factors that contribute to the diagnosis of the severity of the ankle injury could benefit patient care and reduce the time spent collecting extraneous variables. The purpose of this study was to construct a decision tree to predict the severity of ankle injuries. Ankle injuries are routinely graded as mild, moderate, and severe (Grade 1, 2, or3). Decision tree, a data mining technique was employed to produce decision models. While decision trees have been applied in research studies for some time, it is a relatively new method being used in data mining. The data was provided by a large university°¯s health center, which included 194 ankle injuries excluding fractures for the construction of the decision trees. Eighteen variables were collected and analyzed ranging from detail about the type of injury (e.g., sports related, re-injury, pain severity and duration) to physical observations (e.g., amount of edema and ecchymosis) and clinical tests (e.g., bone tenderness and anterior drawer). The data analyses were completed using SPSS Answer Tree and the CHAID and CART methods were utilized. The ankle injuries included were 51 mild, 120 moderate, and 2 severe. Severe ankle sprains (Grade 3) are rare without fracture; therefore, the amount in the study was low. Multiple trees were constructed, validated and cross-validated using 90% and 10% of the sample, respectively. The tree with the lowest risk estimate and least levels and nodes was chosen. The tree that had the best results was a CHAID constructed tree consisting of three levels and four factors as predictors of the physician°¯s diagnosis. The first level had one factor: edema (mild, moderate, and severe) present. The second level contains two subjective factors: whether it was a re-injury (Yes/No) and if it was sports/exercise related (Yes/No). The final level included re-injury and the amount of ecchymosis (mild, moderate, and severe) present. The Misclassification Matrix showed that 28/51 (54.90%) of Grade1, 111/120 (92.50%) of Grade 2, and 1/2 (50%) of Grade 3 ankle sprains were correctly predicted. Overall, the model correctly predicted 80.9% of the ankle injuries, which could be a useful aid that may free up the valuable time of the healthcare provider. The decision tree has demonstrated a useful method for identifying predictors and streamlining the amount of variable needed in the proper classification of an injury or illness.