Automated methods provide leak detection that may be implemented in a respiratory treatment apparatus. In some embodiments, the detection apparatus may automatically determine and score different types of leak events during a treatment session, including, for example, continuous mouth leak events and valve-like mouth leak events. The detection methodologies may be implemented as a data analysis of a specific purpose computer or a detection device that measures a respiratory airflow or a respiratory treatment apparatus that provides a respiratory treatment regime based on the detected leak. In some embodiments, the leak detector may determine and report a leak severity index. Such an index may combine data that quantifies different types of leak events.