Dear members of the ESC WG on e-Cardiology,
In the current editorial, we would like to recommend recent articles that have caught the attention of our working group members. Today the focus is on machine learning and prediction of in-hospital mortality and the combination of experimental and in-silico models to determine the impact of accelerated pacing on left-heart filling pressure.
But first, we’d like to encourage you to read the piece by Dr. Sven Knecht (“ECG Analysis to Predict Cardiovascular Diseases - a Critical Appraisal of Dataset Quality and the Need for External Validation”), which comments on a recent paper on artificial intelligence applied to ECG printouts. The piece provides a critical appraisal of dataset quality and highlights the importance of external validation.
Our first recommended paper, “Machine learning score to predict in-hospital outcomes in patients hospitalized in cardiac intensive care unit”, by Weizman et al (European Heart Journal – Digital Health), developed and tested a machine learning model to predict in-hospital major adverse cardiovascular events in patients hospitalized in cardiac intensive care units. The study is based on data from a prospective multicentre study involving 39 French hospitals (data from 31 centres were used for training and the rest for testing) and 1499 patients, 4.3% of which had a major event. The authors implemented a series of decision trees-based methods and found that using 7 out of 28 clinical variables it was possible to predict events with a much higher accuracy (AUC = 0.88) than previous clinical scores. The selected clinical variables include illicit drug use, mean arterial pressure, Killip class, exhaled carbon monoxide level, LVEF, TAPSE value, and peak E/e′ ratio, some of which are not routinely captured in cardiac intensive care units.
On a similar topic, the paper "Validation of machine learning-based risk stratification scores for patients with acute coronary syndrome treated with percutaneous coronary intervention" by Molenaar et al (European Heart Journal – Digital Health) showed that the machine learning based score GRACE 3 predicts in-hospital mortality in patients with acute coronary syndrome treated with percutaneous coronary intervention with excellent discrimination (c-statistics 0.90) and good calibration. The GRACE 3.0 score was previously developed using a large dataset from the UK and was tested in this paper in 2,759 patients from the Netherlands. The model uses ensemble learning and it is based on clinical variables (age, sex, heart rate, systolic blood pressure, Killip class, creatinine concentration, cardiac arrest, presence of ST-segment deviation, and troponin elevation).
Finally, the paper “Accelerated atrial pacing reduces left-heart filling pressure: a combined clinical-computational study” by Van Loon et al (European Heart Journal), elegantly combined an experimental study conducted during catheter ablation of atrial fibrillation and computational modelling to investigate the effect of heart rate on filling pressures. The experimental data shows that accelerated pacing, with optimal rate around 100 bpm, reduces mean left atrial pressure, which suggests that this strategy may alleviate congestion symptoms. Computer simulations support the hypothesis that AV sequential pacing may further optimize this therapy's beneficial effects in patients with heart failure with preserved ejection fraction.
On behalf of the entire ESC WG on e-Cardiology,
Michele Orini, PhD
Our mission: To reduce the burden of cardiovascular disease.