A team of researchers has developed a machine learning (ML)-based calculator capable of predicting one-year mortality in patients who suffer ST-elevation myocardial infarction (STEMI). The study published in BMC Cardiovascular Disorders was conducted across Rabin Medical Center in Israel and Turin Hospital in Italy, and analyzed data from 3,340 patients hospitalized between 2004 and 2020.

The ML model, built using the CatBoost algorithm, demonstrated remarkable predictive power. In external testing, it achieved 95.6% accuracy (AUC 0.95) in Israeli patients and 93.2% accuracy (AUC 0.90) in Italian patients. Compared to traditional risk calculators such as TIMI and GRACE, which rely on regression analyses, the ML tool analyses a wider array of clinical, laboratory, and imaging data. The most remarkable factor is, it can continuously improve as more datasets are incorporated.

Key predictors of survival included left ventricular ejection fraction (LVEF), a measure of heart pumping efficiency, and glomerular filtration rate (GFR), an indicator of kidney function. Patients with lower LVEF and GFR were significantly more likely to face mortality within a year of their heart attack.

Beyond population-level predictions, the model provides personalized risk profiles, displaying how individual factors influence outcomes. This feature could help cardiologists tailor treatment strategies and improve patient engagement in secondary prevention measures.

Although this tool is quite promising, the relatively small external validation cohort and focus on one-year all-cause mortality hinder to full establishment of its significance. Future studies with broader datasets are needed before clinical integration. Still, the study highlights the growing potential of artificial intelligence to enhance cardiovascular care and refine prognosis after a major heart attack.