Introduction: Thoracic trauma is, depending on severity, a frequent cause of mortality, morbidity, hospitalization and incapacity. Minor closed thoracic injuries are often discharged from Emergency Departments (ED) and treated on an outpatient basis. One potential complication is the development of a delayed hemothorax (DHx). Currently, there exists no consensus on the best method for classifying DHx. The goal of this study is to evaluate the level of interrater and intrarater agreement with respect to three classification schemes for DHx. Methods: This was a secondary analysis drawn from a prospective multicenter cohort study of consecutive patients presenting to one of four Canadian ED for minor closed thoracic trauma over a four-year period. Using intraclass correlation (ICC), chest radiographs of 50 patients previously diagnosed with new DHx within 2 weeks of discharge were randomly selected and subjected to analysis by emergency physicians, radiologists, surgeons and family physicians using three different methods of classification to study their reliability, both between raters and for the same rater on two separate evaluations, at determining hemothorax severity. Results: Analysis of ICC values demonstrates poor interrater agreement (Global ICC 0.44, 0.35-0.52) for the current classification method, based on professional experience and opinion. The second method, based on hemothorax / total thorax ratio calculation, showed good Global ICC (0.58, 0.49-0.67) on lateral films. The third method, based on presence / absence of overflow from the costo-phrenic angle, showed equally good Global ICC (0.56, 0.47-0.64) on postero-anterior films and was more homogenous across the four different groups of physicians. Conclusion: Our results demonstrate that the current method used to classify DHx, based on gestalt, shows poor interrater agreement. Two innovative classification methods achieved good interrater agreement. Future studies, analyzing possible correlation of this more reliable classification method to objective, clinical measures would be of value to management decisions.