Predictive models of prolonged air leak have relied on information not always available preoperatively (eg, extent of resection, pleural adhesions). Our objective was to construct a model to identify patients at increased risk of prolonged air leak using preoperative factors exclusively.