Developing in hospital mortality prediction model tools for patients with acute myocardial infarction (STEMI) using
Seyedeh Mahdieh Namayandeh1, Mohsen Mohammadi2, Masoud Mirzaei3
1Afshar Research Development Center, Center for Healthcare Data Modeling, Departments of Biostatistics and Epidemiology, School of Public Health, Afshar Hospital, Shahid Sadoughi University of Medical Sciences, Yazd, Iran.
Insights
This study compared logistic regression and decision tree models for predicting hospital mortality in ST-elevation myocardial infarction (STEMI) patients. The decision tree model demonstrated higher accuracy, while logistic regression offered better sensitivity and ROC curve performance.
Area of Science:
- Cardiology
- Medical Informatics
- Predictive Analytics
Background:
- Acute ST-elevation myocardial infarction (STEMI) is a critical condition involving myocardial ischemia and potential necrosis.
- Assessing and predicting hospital mortality risk in STEMI patients is crucial for effective clinical management.
- Existing models require evaluation and comparison for improved risk stratification.
Purpose of the Study:
- To develop and evaluate predictive models for hospital mortality in STEMI patients.
- To compare the performance of logistic regression and decision tree algorithms in risk prediction.
- To identify key clinical and laboratory variables associated with STEMI mortality.
Main Methods:
- Utilized data from the Yazd Cardiovascular Diseases Registry (YCDR) including 1,861 STEMI patients.
- Employed decision tree analysis (rpart package) and logistic regression (glm2 package) for model development.
- Compared model effectiveness using accuracy measures and Receiver Operating Characteristic (ROC) curves.
Main Results:
- Clinical, laboratory, and combined models were developed. Variables like blood sugar, triglycerides, and low ejection fraction increased mortality risk.
- Logistic regression achieved ROC performance of 86.5%-90.2% and accuracy of 88.3%-93%.
- Decision tree models showed higher accuracy (92.5%-95.8%) but lower ROC performance (69.8%-81.7%) compared to logistic regression.
Conclusions:
- The decision tree algorithm exhibited superior accuracy across all developed models.
- Logistic regression demonstrated higher sensitivity and superior Receiver Operating Characteristic (ROC) curve performance.
- Both models offer valuable insights, with distinct strengths in predicting STEMI mortality risk.
Background:
An acute ST-elevation myocardial infarction (STEMI) is a medical event characterized by transmural myocardial ischemia that leads to myocardial injury or necrosis. This study was undertaken to develop, evaluate, and compare models for assessing the risk of hospital mortality in patients with acute myocardial infarction.
Methods:
The study made use of data from the Yazd Cardiovascular Diseases Registry (YCDR), which is a cohort study of inpatient records of ischemic heart disease in Yazd province, Iran. A total of 1,861 patients who had experienced a STEMI were included in the analysis. Decision tree analysis was conducted using R software with the rpart package. Additionally, to analyze the data and adjust for any confounding variables, logistic regression was performed using the glm2 package. To compare the effectiveness of the two models, accuracy measures were used, and the Receiver Operating Characteristic (ROC) curve was applied.
Results:
In this study, three clinical, laboratory, and clinical-laboratory models were created. In a clinical-laboratory model, variables such as blood sugar (BS), triglycerides, HDL cholesterol, peak myocardial band (MBpick), CVA history, and low ejection fraction (EF) were found to increase the risk of in-hospital mortality in patients with ST-elevation myocardial infarction (STEMI). Conversely, higher levels of hemoglobin, low HDL-C, and previous myocardial infarction (MI) were associated with a protective effect against the risk of in-hospital mortality from acute myocardial infarction.The performance of the models in terms of Receiver Operating Characteristic (ROC) curve was 86.5%, 79.5%, and 90.2% for logistic regression model in three different models: clinical, laboratory, and combined clinical-laboratory. The accuracy of these models was calculated to be 88.3%, 81.3%, and 93%, respectively. Important variables influencing the prediction of in-hospital mortality in STEMI patients included Killip class, triglycerides, blood sugar, creatinine levels, the need for treatment due to ventricular fibrillation or ventricular tachycardia (VF/VT), age, and hemoglobin (HB). In the ROC curve analysis of the decision tree algorithm across the clinical, laboratory, and combined clinical-laboratory models, the performance levels were 74.6%, 69.8%, and 81.7%, respectively. The accuracy of the decision tree was 93.0%, 92.5%, and 95.8%.
Conclusion:
The findings of this study indicated that the decision tree algorithm had higher accuracy across all three models: clinical, laboratory, and combined clinical-laboratory compared to logistic regression. However, logistic regression showed higher sensitivity and better ROC curve performance than the decision tree algorithm.
More Related Videos
04:05Author Spotlight: Unveiling Prognostic Indicators in Heart Failure - The Role of Phase Angle and Bioelectrical Impedance Analysis
Published on: June 30, 2023
04:41Digital PCR for Quantifying Circulating MicroRNAs in Acute Myocardial Infarction and Cardiovascular Disease
Published on: July 3, 2018
