Combining knowledge and data driven insights for identifying risk factors using electronic health records

Area of Science:

• Biomedical Informatics
• Clinical Data Science
• Predictive Analytics

Background:

• Identifying risk factors for heart failure (HF) is crucial for improving care quality and reducing costs.
• Current methods for risk factor identification are either knowledge-driven or data-driven, lacking a combined approach.
• No existing model effectively integrates expert knowledge with data-driven insights for comprehensive risk factor identification.

Purpose of the Study:

• To develop and validate a systematic approach for enhancing known knowledge-based risk factors with potential risk factors derived from data.
• To create a model that effectively combines expert knowledge with data-driven insights for risk factor identification in predicting heart failure.
• To improve the accuracy and scope of heart failure risk factor identification.

Main Methods:

• A systematic approach was developed to augment existing knowledge-based risk factors with additional potential risk factors identified from data.
• The core methodology utilizes a sparse regression model incorporating regularization terms for both knowledge- and data-driven risk factors.
• The approach was validated on a large dataset of heart failure cases and controls using electronic health records (EHRs).

Main Results:

• The proposed method successfully identified complementary risk factors beyond existing known factors, improving the prediction of heart failure (HF) onset.
• Quantitative comparison using the Area Under the ROC Curve (AUC) demonstrated that combined risk factors significantly outperformed knowledge-based factors alone.
• Additional identified risk factors were validated as clinically meaningful by a cardiologist.

Conclusions:

• A systematic framework was presented for integrating knowledge- and data-driven insights for robust risk factor identification.
• The framework proved effective in identifying intuitive and predictive risk factors for heart failure (HF) prediction, surpassing traditional methods.
• This approach offers a powerful tool for discovering novel risk factors and enhancing clinical decision-making in cardiovascular health.