Machine learning predicts long-term mortality after acute myocardial infarction using systolic time intervals and routinely collected clinical data

Abstract

BackgroundPrecise estimation of cardiac patients’ current and future comorbidities is an important factor in prioritizing continuous physiological monitoring and new therapies. Machine learning (ML) models have shown satisfactory performance in short-term mortality prediction of patients with heart disease, while their utility in long-term predictions is limited. This study aims to investigate the performance of tree-based ML models on long-term mortality prediction and the effect of two recently introduced biomarkers on long-term mortality. MethodsThis study utilized publicly available data from the Collaboration Center of Health Information Application (CCHIA) at the Ministry of Health and Welfare, Taiwan, China. The data pertained to patients admitted to the cardiac care unit for acute myocardial infarction (AMI) between November 2003 and September 2004. Included in the dataset were patients over 20 years old diagnosed with type 1 AMI. Exclusion criteria encompassed patients with missing data for brachial pre-ejection period (bPEP) and brachial ejection time (bET), as well as those with atrial fibrillation or extremity amputations, resulting in a final cohort of 139 AMI patients. We collected and analyzed mortality data up to December 2018. All patients had provided informed consent, in accordance with the Declaration of Helsinki. Medical records were used to gather demographic and clinical data, including age, gender, body mass index (BMI), percutaneous coronary intervention (PCI) status, and comorbidities such as hypertension, dyslipidemia, ST-segment elevation myocardial infarction (STEMI), and non-ST segment elevation myocardial infarction (NSTEMI). Using medical and demographic records as well as two recently introduced biomarkers, bPEP and bET, collected from 139 patients with acute myocardial infarction, we investigated the performance of advanced ensemble tree-based ML algorithms (random forest, AdaBoost, and XGBoost) to predict all-cause mortality within 14 years. A nested cross-validation was performed to evaluate and compare the performance of our developed models precisely with that of the conventional logistic regression (LR) as the baseline method. ResultsThe developed ML models achieved significantly better performance compared to the baseline LR (C-Statistic, 0.80 for random forest, 0.79 for AdaBoost, and 0.78 for XGBoost, vs. 0.77 for LR) (PRF< 0.001, PAdaBoost< 0.001, PXGBoost< 0.05). Adding bPEP and bET to our feature set significantly improved the algorithms’ performance, leading to an absolute increase in C-Statistic of up to 0.03 (C-Statistic, 0.83 for random forest, 0.82 for AdaBoost, and 0.80 for XGBoost, vs. 0.74 for LR) (PRF< 0.001, PAdaBoost< 0.001, PXGBoost< 0.05). ConclusionThe study indicates that incorporating new biomarkers into advanced machine learning models significantly improves long-term mortality prediction in cardiac patients. This advancement may enable better treatment prioritization for high-risk individuals.