Development and validation of a prognostic model for predicting post-discharge mortality risk in patients with ST-segment elevation myocardial infarction (STEMI) undergoing primary percutaneous coronary intervention (PPCI)

Abstract

Background: Accurately predicting post-discharge mortality risk in patients with ST-segment elevation myocardial infarction (STEMI) undergoing primary percutaneous coronary intervention (PPCI) remains a complex and critical challenge. The primary objective of this study was to develop and validate a robust risk prediction model to assess the 12-month and 24-month mortality risk in STEMI patients after hospital discharge.

Methods: A retrospective study was conducted on 664 STEMI patients who underwent PPCI at Xiangtan Central Hospital Chest Pain Center between 2020 and 2022. The dataset was randomly divided into a training cohort (n = 464) and a validation cohort (n = 200) using a 7:3 ratio. The primary outcome was all-cause mortality following hospital discharge. The least absolute shrinkage and selection operator (LASSO) regression model was employed to identify the optimal predictive variables. Based on these variables, a regression model was constructed to determine the significant predictors of mortality. The performance of the model was evaluated using receiver operating characteristic (ROC) curve analysis and decision curve analysis (DCA).

Results: The prognostic model was developed based on the LASSO regression results and further validated using the independent validation cohort. LASSO regression identified five important predictors: age, Killip classification, B-type natriuretic peptide precursor (NTpro-BNP), left ventricular ejection fraction (LVEF), and the usage of angiotensin-converting enzyme inhibitors/angiotensin receptor blockers/angiotensin receptor-neprilysin inhibitors (ACEI/ARB/ARNI). The Harrell's concordance index (C-index) for the training and validation cohorts were 0.863 (95% CI: 0.792-0.934) and 0.888 (95% CI: 0.821-0.955), respectively. The area under the curve (AUC) for the training cohort at 12 months and 24 months was 0.785 (95% CI: 0.771-0.948) and 0.812 (95% CI: 0.772-0.940), respectively, while the corresponding values for the validation cohort were 0.864 (95% CI: 0.604-0.965) and 0.845 (95% CI: 0.705-0.951). These results confirm the stability and predictive accuracy of our model, demonstrating its reliable discriminative ability for post-discharge all-cause mortality risk. DCA analysis exhibited favorable net benefit of the nomogram.

Conclusion: The developed nomogram shows potential as a tool for predicting post-discharge mortality in STEMI patients undergoing PPCI. However, its full utility awaits confirmation through broader external and temporal validation.

Keywords: All-cause mortality risk; Decision Curve Analysis (DCA); Least Absolute Shrinkage and Selection Operator (LASSO); Predictive model; ST-segment elevation myocardial infarction (STEMI).

Fig. 1

Flow diagram for participant screening, eligibility, and analysis. (Note: The flow diagram in Fig. 1 outlines the process of participant screening, eligibility assessment, and data analysis in the study. The diagram illustrates the sequential steps followed from the initial screening of participants to the final analysis of the collected data)

Fig. 2

LASSO Regression Coefficient Path and CV LASSO Regression Coefficient Path. A LASSO Regression Coefficient Path. B CV LASSO Regression Coefficient Path. (Note:The LASSO regression coefficient path displays how the coefficients of each variable change with increasing regularization parameter λ.The CV LASSO regression coefficient path illustrates the coefficients' behavior with λ tuned through cross-validation. Both paths provide insights into the impact of regularization on variable selection and coefficient estimation in the LASSO regression model)

Fig. 3

Area under the Receiver Operating Characteristic (ROC) curve and Bootstrap validation. A ROC curves for the training set at 12 months and 24 months. B ROC curves for the validation set at 12 months and 24 months. C Comparison of model stability between the original model and 500 rounds of Bootstrap validation on the training set

Fig. 4

Calibration curves at different time points. A Calibration curves for the training set at 12 months and 24 months. B Calibration curves for the validation set at 12 months and 24 months. (Note: The calibration curves depict the agreement between the predicted probabilities and the observed outcomes at different time points. The curves represent the performance of the predictive model in terms of calibration, indicating how well the model's predicted probabilities align with the actual probabilities)

Fig. 5

Decision Curve Analysis (DCA) with Time and DCA Nomogram. A1: DCA with Time for the Training Set. B1: DCA with Time for the Validation Set. A2: DCA Nomogram for the Training Set. B2: DCA Nomogram for the Validation Set. (Note: The DCA curves in A1 and B1 illustrate the net benefit of the predictive model over a range of threshold probabilities at different time points for the training and validation sets. These curves provide insights into the clinical usefulness and added value of the model compared to alternative decision strategies. Additionally, the DCA nomograms in A2 and B2 provide a graphical representation of the decision curves, allowing for a more intuitive interpretation and application of the model's results)

Fig. 6

Nomogram for all-cause mortality risk prediction. (Note: The nomogram presents a visual tool for predicting the risk of all-cause mortality. It combines various predictors or risk factors into a comprehensive model that provides an individualized risk assessment. The nomogram allows for a simple and intuitive estimation of the probability of mortality based on the values assigned to each predictor. Clinicians can use this nomogram as a practical aid in risk assessment and shared decision-making with patients regarding appropriate interventions and management strategies)

Fig. 7

Rationality Analysis: Kaplan–Meier Survival Curves of High-Score and Low-Score Groups. A Rationality Analysis for the Training Set. B Rationality Analysis for the Validation Set. (Note: The Kaplan–Meier survival curves depicted in A and B demonstrate the differences in survival outcomes between the high-score and low-score groups. These curves serve as a rationality analysis to evaluate the predictive performance of the scoring system or model. The separation of the survival curves indicates the ability of the scoring system to stratify patients into distinct risk groups. This analysis provides insights into the reliability and validity of the scoring system in predicting survival outcomes and aids in assessing its clinical utility)

Fig. 8

Central Illustration