Disclosed herein, among other things, are methods and apparatus related to identification of apnea type. One aspect of the present subject matter provides a method for real-time apnea discrimination. The method includes sensing an impedance-based tidal volume signal to monitor a respiratory cycle of a patient, and detecting a reduction in tidal swing using the sensed impendence to detect an apnea event. When the apnea event is detected, a shape of the sensed signal is compared to a stored signal shape to determine whether the apnea event is primarily an obstructive sleep apnea (OSA) event or primarily a central sleep apnea (CSA) event, in various embodiments.