Development and transfer learning of self-attention model for major adverse cardiovascular events prediction across
Yunha Kim1, Heejun Kang2, Hyeram Seo3
1Department of Medical Science, Asan Medical Center, Asan Medical Institute of Convergence Science and Technology, University of Ulsan College of Medicine, 88, Olympicro 43gil, Songpagu, Seoul, 05505, Republic of Korea.
Abstract:
Predicting major adverse cardiovascular events (MACE) is crucial due to its high readmission rate and severe sequelae. Current risk scoring model of MACE are based on a few features of a patient status at a single time point. We developed a self-attention-based model to predict MACE within 3 years from time series data utilizing numerous features in electronic medical records (EMRs). In addition, we demonstrated transfer learning for hospitals with insufficient data through code mapping and feature selection by the calculated importance using Xgboost. We established operational definitions and categories for diagnoses, medications, and laboratory tests to streamline scattered codes, enhancing clinical interpretability across hospitals. This resulted in reduced feature size and improved data quality for transfer learning. The pre-trained model demonstrated an increase in AUROC after transfer learning, from 0.564 to 0.821. Furthermore, to validate the effectiveness of the predicted scores, we analyzed the data using traditional survival analysis, which confirmed an elevated hazard ratio for a group with high scores.
Insights
This study introduces an advanced AI model using electronic health records to predict major adverse cardiovascular events (MACE) up to three years in advance. The model improves prediction accuracy, aiding early intervention for cardiovascular health.
Area of Science:
- Cardiology
- Artificial Intelligence
- Health Informatics
Background:
- Major adverse cardiovascular events (MACE) pose significant health risks and high readmission rates.
- Existing MACE risk models rely on limited, single-time-point patient data.
- Predictive modeling using comprehensive patient data is needed for improved cardiovascular risk assessment.
Purpose of the Study:
- To develop and validate a novel self-attention-based model for predicting MACE within three years.
- To implement transfer learning strategies for enhancing MACE prediction in hospitals with limited data.
- To improve the clinical interpretability and data quality of electronic medical record (EMR) data for predictive modeling.
Main Methods:
- Utilized time-series data from electronic medical records (EMRs) with a self-attention-based deep learning model.
- Employed transfer learning, including code mapping and feature selection via XGBoost importance, for data-scarce hospitals.
- Established standardized operational definitions for diagnoses, medications, and laboratory tests to streamline EMR data.
Main Results:
- The self-attention model effectively predicted MACE using extensive EMR time-series features.
- Transfer learning significantly improved model performance, increasing the Area Under the Receiver Operating Characteristic curve (AUROC) from 0.564 to 0.821.
- Survival analysis confirmed that high predicted MACE scores correlated with an elevated hazard ratio, validating the model's clinical utility.
Conclusions:
- The developed self-attention model offers a robust method for predicting MACE using EMR time-series data.
- Transfer learning enhances the applicability of the model across different healthcare settings, even with limited data.
- Standardizing EMR data through operational definitions improves data quality and facilitates effective transfer learning for cardiovascular risk prediction.
More Related Videos
Related Concept Videos
01:25Errors occurring during blood pressure monitoring
Several factors...
01:30Pulse rhythm
Conversely, an irregular pulse pattern is termed dysrhythmia, stemming from disruptions in cardiac...