Using electronic health records and continuous intra-cardiac electrogram monitoring to understand cardiovascular outcomes in patients with pacemakers and defibrillators
Pacemakers and defibrillators continuously record the heart’s electrical activity, but those recordings are rarely used to predict which patients will develop atrial fibrillation, heart failure, or dangerous arrhythmias. This matters because after a pacemaker is implanted, atrial fibrillation and heart failure are common and need prompt treatment. For defibrillator recipients, most never benefit from the device, and doctors cannot identify who will receive recurrent shocks for ventricular arrhythmia. The intra-cardiac electrogram waveforms stored by these devices have not yet been exploited as biomarkers of arrhythmia risk or heart muscle remodelling. If this research succeeds, it could refine which patients actually need a defibrillator, reducing unnecessary implants and recurrent shocks. It could also turn routine remote-monitoring data into early warnings for stroke, arrhythmia, and heart failure—improving care without extra hospital visits or invasive tests. The work uses machine learning on a large UK device population and validates findings against surface electrocardiograms and external centres, so the results should be robust enough to change clinical guidelines.
View original technical description
Background: After pacemaker implant, atrial fibrillation (AF) and heart failure (HF) are common and require prompt treatment. For implantable cardioverter-defibrillator (ICD) recipients, most do not benefit, nor can we identify who receives recurrent shocks for ventricular arrhythmia (VA). Devices continuously record intra-cardiac electrogram waveforms (EGMs), and their potential as biomarkers of arrhythmogenesis and myocardial remodelling has not yet been exploited. Aim: Identify clinically useful biomarkers for progressive cardiac disease in device recipients, applying machine learning methods to routinely acquired data. Methods: I will adopt a unique approach leveraging national electrophysiology audit data (n=310,974) and device-based physiological sensor measurements from the largest UK device population (n=19,864, >250,000 EGMs). To inform better ICD implant selection, I will evaluate competing arrhythmia and non-arrhythmia mortality, and identify contemporary long-term mortality trends. Detailed rule-based and machine learning analysis will derive EGM biomarkers, validated against surface electrocardiograms. I will identify features predictive of mortality, VA, and precursors to AF, externally validating models with two centres. Using EGM surrogates of mechanoelectrical feedback, I will perform serial assessments to characterise structural and electrophysiological remodelling in pacing-induced HF. Outputs: Refining ICD implant criteria. Biomarkers harnessing ‘wearables’-type data from remote monitoring to reduce VA, stroke and HF morbidity.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know