Techniques are provided for tracking patient respiration based upon intracardiac electrogram (IEGM) signals or other electrical cardiac signals. Briefly, respiration patterns are detected based upon cycle-to-cycle changes in morphological features associated with individual electrical events with the IEGM signals. For example, slight changes in the peak amplitudes of QRS-complexes, P-waves or T-waves are tracked to identify cyclical variations representative of patient respiration. Alternatively, the integrals of the morphological features of the individual events may be calculated for use in tracking respiration. In any case, once respiration patterns have been identified, episodes of abnormal respiration, such as apnea, hyperpnea, nocturnal asthma, or the like, may be detected and therapy automatically delivered. In particular, techniques for detecting abnormal respiration based on various respiratory parameters derived from the IEGM--such as respiration depth and respiration power--are described.