Developing a predictive nomogram for AMI in elderly patients with AHF: a retrospective analysis

TL;DR

This study creates a prediction model to identify elderly heart failure patients at risk of heart attacks, helping doctors make better decisions.

Contribution

A novel nomogram model using logistic regression is developed to predict AMI risk in elderly AHF patients.

Findings

Age, diabetes, and heart disease are key predictors of AMI in elderly AHF patients.

The model achieved 91.3% accuracy and an AUC of 0.780 in predicting AMI risk.

Decision curve analysis confirmed the model's clinical usefulness for early intervention.

Abstract

This study focuses on the clinical issue of acute myocardial infarction (AMI) in the context of acute heart failure (AHF), particularly among the elderly population. Elderly patients with AHF experiencing AMI represent a severe cardiac condition with poor prognosis. Hence, this research aims to analyze potential risk factors and establish a clinical prediction model using logistic regression to facilitate early assessment and guide clinical decisions.

A retrospective analysis design was employed, selecting elderly AHF patients hospitalized in the Cardiovascular Department of Qinhuangdao City First Hospital from October 2019 to December 2023. Patient history and clinical data were analyzed using LASSO regression and logistic regression to identify and analyze predictors of AMI, leading to the construction of a nomogram. The model’s predictive performance was evaluated using the concordance index, receiver operating characteristic curve, decision curve analysis, and clinical impact curves to gain insights into the nomogram’s accuracy and clinical utility.

The study included 1,904 patients. Logistic regression analysis identified age, coronary heart disease, diabetes, pulmonary infection, ventricular arrhythmia, hyperlipidemia, hypoalbuminemia, left ventricular diastolic diameter (LVDD), and left ventricular ejection fraction (LVEF) as independent risk factors for AMI during hospitalization. The predictive model was formulated as follows: Logit(P) = −7.286 + 0.065 × Age + 0.380 × Coronary heart disease + 0.358 × Diabetes + 0.511 × Pulmonary infection + 0.849 × Ventricular arrhythmia + 0.665 × Hyperlipidemia + 0.514 × Hypoalbuminemia + 0.055 × LVDD - 0.131 × LVEF. The model demonstrated an AUC of 0.780 (0.741–0.819), with an accuracy of 91.3%, and a specificity of 91.4%, indicating good predictive performance. Further validation through decision curve analysis and clinical impact curves confirmed the model’s effectiveness in clinical decision support.

The study successfully developed a multivariate analysis-based prediction model capable of effectively predicting the risk of AMI in hospitalized elderly AHF patients. This model provides a powerful tool for clinicians, facilitating early identification and intervention in high-risk patients.

Figures