Machine learning for prediction of newly diagnosed atrial fibrillation after emergency percutaneous coronary intervention during hospitalization in patients with acute ST-segment elevation myocardial infarction: a multi-center prospective study
The present study developed and compared eight machine learning models for the prediction of in-hospital NDAF among acute patients with STEMI treated with emergency PCI and found the Gradient Boosting model achieved optimal predictive performance.
Abstract
Background In-hospital complications after emergency PCI in acute STEMI are important, especially newly diagnosed atrial fibrillation and are associated with hemodynamic instability, heart failure, stroke and mortality. There are complex nonlinear interactions between clinical, inflammatory, cardiac remodeling and procedural factors that traditional risk scores may not adequately capture. Thus, this study aimed to develop and validate an interpretable machine-learning model for early prediction of in-hospital NDAF in patients with acute STEMI undergoing emergency PCI. Methods From January 2021 to December 2024, this study collected data on patients with acute STEMI after emergency PCI from the five tertiary general hospitals in Chongqing. Important clinical variables were identified using the selection operator and least absolute shrinkage. Based on the area under the curve, the best predictive model was selected from eight machine learning methods. The predictive model’s results were interpreted using Shapley Additive explanations. Results Eighteen variables were chosen for model building, and a total of 154 patients with acute STEMI following emergent PCI were included. With the largest area under the curve of 0.991 (95% CI: 0.963–1), sensitivity of 0.846, specificity of 0.914, and Brier score of 0.2356, the Gradient Boosting model was selected. The Shapley value analytical framework was systematically applied to decode feature importance patterns and illuminate individual prognostic determinants Conclusion The present study developed and compared eight machine learning models for the prediction of in-hospital NDAF among acute patients with STEMI treated with emergency PCI. The Gradient Boosting classifier achieved optimal predictive performance, highlighting its potential clinical utility for early risk stratification. By identifying high-risk individuals in advance, this model may support optimized clinical management and contribute to better prognosis in this high-risk population. Machine learning computations were implemented via a publicly accessible web server integrating the pre-trained Gradient Boosting prediction model. All analytical steps were performed following the platform’s default settings to maintain methodological standardization and reproducibility.
Abstract Background Despite advances in understanding and treating non–ST-elevation acute coronary syndrome (NSTE-ACS), patients continue to experience high rates of adverse outcomes, particularly those with non–ST-segment elevation myocardial infarction, which remains a leading cause of cardiovascular mortality. Existing risk models may not fully reflect contemporary patient populations due to substantial changes in clinical profiles. Developing new machine learning (ML)–based risk calculators may improve the prediction of in-hospital mortality (IHM) at different stages of the diagnostic process, and ultimately improve patient outcomes. Objective This study aimed to develop predictive models for IHM in patients with NSTE-ACS using ML methods and predictor sets obtained during the diagnostic process. Methods This retrospective observational study included 1144 patients with NSTE-ACS admitted between 2019 and 2021. IHM occurred in 94 (8.1%) of 1144 patients. Predictive models were developed using multivariable logistic regression, Random Forest, XGBoost (Extreme Gradient Boosting), and CatBoost algorithms. Model performance was evaluated using the area under the receiver operating characteristic curve (ROC-AUC), area under the precision-recall curve, calibration metrics, and decision curve analysis. Results A key feature of the developed models was their applicability at different stages of the diagnostic process, using predictors available at each specific stage. At admission, ML models achieved ROC-AUC values up to 0.93. After incorporating laboratory and echocardiographic data, predictive performance increased to ROC-AUC values of 0.94 to 0.95. The best-performing model demonstrated an ROC-AUC of 0.951 (95% CI 0.946‐0.956), a sensitivity of 0.902 (95% CI 0.889‐0.916), and a specificity of 0.891 (95% CI 0.886‐0.896). The area under the precision-recall curve reached 0.685, and the Brier score was 0.0331, indicating good calibration. Random Forest models demonstrated greater clinical utility than the Global Registry of Acute Coronary Events score (P<.001). Shapley Additive Explanations analysis identified the most significant predictors of mortality risk, including Killip class of acute heart failure, age, Charlson comorbidity index, creatinine level, and hematocrit level. Conclusions ML-based models enabled accurate prediction of IHM in patients with NSTE-ACS at different stages of the diagnostic process and may improve risk stratification and clinical decision-making in real-world practice.
V. V. Ryabov, A. Nesova, A. Vishnevskiy et al.· JMIR Medical Informatics· 0 citations
The occurrence of NOAF was associated with increased in-hospital mortality, which was 2–3 times higher in patients with arrhythmia, and most NOAF prediction models developed specifically in STEMI cohorts undergoing PCI demonstrated higher discriminative ability.
R. L. Pak, B. I. Geltser, E. Kokarev et al.· Siberian Journal of Clinical...· 0 citations
A five-factor nomogram is developed and validated that effectively predicts the individualized risk of AHF following EVT in patients with AIS-LVO and has excellent discriminative ability.
Xiuling Yang, Wenfei Liang, Shaoan Ge et al.· Frontiers in Cardiovascular...· 0 citations
Background Patients with concurrent hypertension, heart failure with preserved ejection fraction (HFpEF), and unstable angina pectoris (UAP) represent a high-risk population with heterogeneous long-term cardiovascular outcomes. However, practical tools for individualized risk stratification in this clinically complex population remain limited. This study aimed to develop and validate a machine learning-assisted nomogram for predicting long-term major adverse cardiovascular events (MACEs) in patients with concomitant hypertension, HFpEF, and UAP. Methods This multicenter retrospective cohort study included 1,669 patients with hypertension, HFpEF, and UAP from Six medical centers in China. Patients were assigned according to treating hospital to a development cohort, a validation cohort, and an independent test cohort. The primary endpoint was MACEs, defined as cardiac death, unplanned coronary revascularization, or rehospitalization for acute heart failure. Candidate predictors were selected using least absolute shrinkage and selection operator regression, the Boruta algorithm, and random forest feature importance analysis. Variables consistently identified across these methods were entered into a multivariable Cox proportional hazards model, and a Cox-based nomogram was constructed to estimate 2-, 3-, and 4-year MACE risk. Model performance was assessed using receiver operating characteristic curves, time-dependent area under the curve, calibration curves, decision curve analysis, and Kaplan-Meier survival analysis. Results During a median follow-up of 48 months, 254 patients experienced MACEs. Five predictors were retained in the final model: diabetes mellitus, previous myocardial infarction, systemic immune-inflammation index, triglyceride, and N-terminal pro-B-type natriuretic peptide. In the development cohort, the nomogram yielded area under the curve values of 0.785, 0.787, and 0.760 for predicting 2-, 3-, and 4-year MACEs, respectively. The corresponding values were 0.674, 0.697, and 0.682 in the validation cohort, and 0.685, 0.711, and 0.718 in the test cohort. Calibration curves showed acceptable agreement between predicted and observed risks, and decision curve analysis suggested potential net benefit across clinically relevant threshold probabilities. Patients classified as high risk according to the nomogram had a significantly higher incidence of MACEs than those classified as low risk across all cohorts. Conclusion A machine learning-assisted, Cox-based nomogram incorporating five routinely available clinical and laboratory variables provided acceptable discrimination in the development cohort and modest-to-acceptable performance in geographically distinct validation and test cohorts. This model may support individualized long-term risk stratification in patients with concurrent hypertension, HFpEF, and UAP, although prospective validation and clinical impact studies are warranted before routine implementation.
Long Tang, Xiaomeng Yang, Zongchao Zuo et al.· Frontiers in Endocrinology· 0 citations
BackgroundSepsis-associated acute kidney injury (SA-AKI) is a common and severe complication in critically ill patients, with poor prognosis. Diabetes may further increase adverse outcomes through infection susceptibility, immune dysfunction, and renal vulnerability. However, mortality prediction models for patients with diabetes complicated by SA-AKI remain limited. This study aimed to develop and validate a machine learning-based model for early in-hospital mortality prediction in this population.MethodsA total of 6929 patients with SA-AKI and diabetes were identified from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database and randomly divided into training and validation sets at a ratio of 7:3. Ninety-four variables, including demographics, diagnoses, clinical parameters, and medication records within the first 24 h after ICU admission, were extracted. Twelve machine learning algorithms were developed and compared, and the optimal model was selected. Recursive feature elimination was used to identify key predictors, while SHapley Additive exPlanations were applied for model interpretation. The final model was deployed as a web-based tool and externally tested using the eICU Collaborative Research Database.ResultsThirty-two key predictors were ultimately selected, including urine output rate, platelet count, lactate, weight, blood glucose, SOFA score, pH, blood urea nitrogen, vital signs, coagulation indices, vasopressor use, and other clinically relevant variables. The categorical boosting algorithm model presented better predictive performance [receiver operating characteristic (AUC): 0.828] than other models [accuracy (ACC): 70.9%, sensitivity: 78.7%, specificity: 69%, F1 score: 0.509, positive predictive value (PPV): 33.7%, and negative predictive value (NPV): 93.1%]. External testing using data from the eICU database was also well validated (AUC: 0.793).ConclusionsA CatBoost-based machine learning model incorporating 32 clinically accessible variables showed good predictive performance for in-hospital mortality in patients with diabetes and SA-AKI, supporting early risk stratification and clinical decision-making.
Yan-Ni Wang, Hongjie Shen, Sheng-Ze Wu et al.· Journal of Intensive Care Me...· 0 citations
Stress hyperglycemia is prevalent in acute ST-elevation myocardial infarction (STEMI) patients and associates with adverse outcomes. The stress hyperglycemia ratio (SHR), which standardizes admission glucose against estimated average glucose from HbA1c, reduces chronic glycemic interference. However, SHR's predictive value for heart failure (HF) during hospitalization following emergency percutaneous coronary intervention (PCI) remains unclear.
This single-center retrospective cohort study enrolled 1,068 consecutive STEMI patients undergoing emergency PCI from January 2020 to June 2025. SHR was calculated as the ratio of admission plasma glucose to estimated average glucose. The primary endpoint was hospitalization HF. Predictive efficacy was evaluated using multivariate logistic regression, propensity score matching (PSM), receiver operating characteristic (ROC) analysis, net reclassification improvement (NRI), and integrated discrimination improvement (IDI).
Overall, 158 patients (14.8%) developed hospitalization HF. The HF group showed significantly higher SHR than non-HF patients [1.52 (1.25–1.88) vs. 1.10 (0.92–1.30);
P
< 0.001]. Each 1-unit SHR increase associated with 2.31-fold HF risk elevation (adjusted OR=2.31, 95% CI 1.68–3.18;
P
< 0.001). Following 1:1 PSM, high SHR patients maintained significantly higher HF incidence (20.4% vs. 10.2%;
P
< 0.001). SHR's ROC area under curve was 0.748, superior to admission glucose (0.728;
P
= 0.031). Adding SHR yielded NRI = 0.318 and IDI = 0.035 (both
P
< 0.001).
SHR independently predicts hospitalization HF in STEMI patients post-emergency PCI, demonstrating superior discriminative performance to admission glucose with significant incremental predictive value. SHR may facilitate early risk stratification and management.
Fachao Shi, Wenlong Ding, Cun-Ming Fang et al.· Frontiers in Cardiovascular...· 0 citations