Abstract
Aim. To compare the effectiveness of the POAF, PAFAC, COM-AF, HATCH, ms2HEST and CHA2DS2-VASc scores for predicting new-onset atrial fibrillation (AF) in patients with ST-elevation myocardial infarction (STEMI) after percutaneous coronary intervention (PCI), as well as to develop novel prognostic models based on machine learning methods.Material and methods. This single-center retrospective study was conducted using data from 3449 electronic health records of patients with STEMI. Two groups of individuals were identified, the first of which included 312 (9%) patients with new-onset AF in the postoperative period of PCI, and the second — 3139 (91%) patients without cardiac arrhythmia. To predict AF, univariate and multivariate logistic regression (ULR and MLR), decision tree (DT), artificial neural networks (ANN) were used.Results. The study results showed that of the 6 analyzed scores, only 4 (mc2HEST, COM-AF, CHA2DS2-VASc and HATCH) have an acceptable prognostic potential for new-onset AF after PCI, which was documented by the AUC metrics in the ULR models developed on the basis of the sum of points of each score (AUC — 0,736, 0,731, 0,71 and 0,702, respectively). The integrative ANN model, combining the prognostic resource of 6 scores, demonstrated higher accuracy than the DT and MLR models (AUC — 0,759 vs 0,745 and 0,755, p-valueConclusion. Further studies aimed at improving the quality of AF prognostic models in patients with STEMI after PCI may involve searching for novel predictors characterizing severity of coronary involvement and effectiveness of its restoration, inflammatory response, myocardial electrophysiological status, etc.

This publication has 12 references indexed in Scilit: