Exploring risk factors for new-onset atrial fibrillation in elderly patients with myocardial infarction based on serum NT-proBNP, RDW, Hcy, and clinical parameters: an exploratory study using LASSO-logistic regression

ObjectiveTo investigate the association of serum N-terminal pro-brain natriuretic peptide (NT-proBNP), red cell distribution width (RDW), homocysteine (Hcy), and clinical parameters with new-onset atrial fibrillation (NOAF) in elderly patients with acute myocardial infarction (AMI), and to identify potential risk factors in an exploratory analysis.MethodsThis retrospective exploratory study enrolled 204 elderly AMI patients (January 2020—December 2022), divided into NOAF (n = 41) and non-NOAF (n = 163) groups. Clinical and laboratory data were collected. LASSO regression was used for initial feature selection from candidate variables. Multivariate logistic regression was then employed to identify factors associated with NOAF. A nomogram was constructed to visualize the risk factor model. Given the exploratory nature and limited number of events (n = 41), model performance metrics including area under the ROC curve (AUC), calibration, and decision curve analysis were evaluated with caution, accompanied by internal validation using 1,000 bootstrap resamples to assess overfitting.ResultsSignificant differences were found between the NOAF and non-NOAF groups in gender, smoking status, hospitalization duration, Killip class, NT-proBNP, RDW, and Hcy (P < 0.05). LASSO regression identified six key predictors: gender, hospitalization duration, Killip class, NT-proBNP, RDW, and Hcy. Multivariate logistic regression confirmed these as independent risk factors: female gender (OR = 3.367, 95% CI: 1.193–9.498), longer hospitalization duration (OR = 1.110 per day, 95% CI: 1.021–1.206), Killip class II-IV (OR = 3.171, 95% CI: 1.157–8.689), higher NT-proBNP (OR = 4.677 per ng/mL increase, 95% CI: 1.774–12.328), higher RDW (OR = 1.358 per % increase, 95% CI: 1.163–1.585), and higher Hcy (OR = 1.814 per µmol/L increase, 95% CI: 1.414–2.328); all P < 0.05. The model showed good fit (Hosmer-Lemeshow χ² = 6.957, P = 0.541). The final combined nomogram model achieved an AUC of 0.926 (95% CI: 0.880–0.973) in the development set. Internal validation via 1,000 bootstrap resamples yielded an optimism-corrected AUC of 0.901. Calibration was assessed visually and statistically (Hosmer–Lemeshow χ² = 6.957, P = 0.541), showing good agreement between predicted and observed risks (Figure 4). Internal validation indicated good calibration (Cox-Snell R² = 0.593, Nagelkerke R² = 0.376), and decision curve analysis confirmed a high net clinical benefit.ConclusionThis exploratory study identifies gender, hospitalization duration, Killip class, NT-proBNP, RDW, and Hcy as factors significantly associated with NOAF in elderly AMI patients. However, due to the modest sample size, low events-per-variable ratio, and lack of external validation, these findings should be considered hypothesis-generating. The proposed model is not ready for clinical application. Further validation in larger, prospective cohorts is essential to confirm these associations and to develop a robust predictive tool.