AIMS Electrocardiographic changes, e.g. arrhythmias causing syncope or palpitations, are often transient and therefore difficult to diagnose. Systematic and symptom-activated ECG recordings can increase diagnostic yield in such patients. We evaluated the diagnostic accuracy of a simple, leadless, patient-operated ECG device compared with a standard 12-lead ECG. METHODS AND RESULTS We recorded a standard 12-lead surface ECG and a patient-activated ECG in direct succession in 508 consecutive patients enrolled in four centres. All ECGs were analysed by a single, blinded observer. ECGs were analysable in 505 (99.4%) patients (66% male, age 61 +/- 15 years, and body mass index 27 +/- 4). Analysis of the patient-activated ECG adequately detected a normal ECG (sensitivity 91% and specificity 95%), atrial fibrillation (AF) (sensitivity 99% and specificity 96%), and even T-wave abnormalities (sensitivity 90% and specificity 75%). Diagnostic accuracy for atrioventricular nodal block was moderate (sensitivity 79% and specificity 99%). Continuous parameters correlated well: (r(2) = 0.89 for heart rate, 0.83 for PR interval, 0.78 for QRS duration, and 0.89 for QTc). CONCLUSION Recordings made by this patient-operated ECG device allow to detect arrhythmias and other ECG changes with high accuracy compared with a standard ECG. It may help to improve accurate diagnosis of transient ECG changes such as paroxysmal AF in palpitations or other unexplained cardiac symptoms.