The present invention relates to a method of predicting cardiac events using physiological data relating to a patient and comprises: inputting physiological data relating to the patient 22, evaluating a property of multiple heartbeats within said physiological data optionally using an analysis module 24； determining a value associated with the number of said multiple heartbeats that exceed an abnormality threshold set for said property； and comparing said value against a predetermined value, thereby to indicate a probability of said patient experiencing a cardiac event and； providing an output based on the comparison； where the abnormality threshold is determined based on a dataset of a plurality of physiological data obtained from multiple sources. In embodiments, the data is cardiac data, such as RR intervals of multiple heartbeats provided by an electrocardiogram (ECG). Also disclosed is a hybrid classifier formed from a combination of two cardiac classifiers.
