Objective: To assess our results of a prospective algorithm applied to patients with thoracic esophageal perforation. Methods: A retrospective review of a prospective algorithm. Patients with esophageal perforation underwent an esophagram. If there was a contained esophageal perforation they were admitted, kept nothing by mouth, and restudied in 3-5 days. If the leak was not contained, they underwent operative repair. Results: From 1/1998 to 6/2009 there were 81 patients. The gastrograffin swallow showed 56 patients had contained perforations and 25 did not. Twenty-two of the 25 patients with noncontained perforation underwent immediate operative repair (one patient refused surgery, two were not stable enough for the operating room); their morbidity was 68% and there were six (24%) operative mortalities. Median hospital length of stay (LOS) was 11 days (range, 2-120). Of the 56 patients with contained perforations, 26 were managed successfully without surgery. However, 30 of the patients initially treated nonoperatively eventually required operations due to new pleural effusion, mediastinal abscess, or conversion to noncontained perforation. Their morbidity was 41% and there were three operative mortalities (5%). On univariate analysis, these patients were more likely to have undergone previous esophageal procedures (surgical or dilation) (p=0.03), had new or increased pleural effusion (p=0.04), and had greater than 24. h between diagnosis and treatment (p=0.02). Only greater than 24. h between diagnosis and treatment remained a significant predictor on multivariate analysis. Their median hospital LOS was 21 days (range, 7-77). Conclusion: Contained thoracic esophageal perforations can usually be safely managed nonoperatively without significant morbidity or mortality. However, careful in-hospital monitoring is needed if surgery is not chosen. © 2011.