Noninvasive Risk Prediction Models for Heart Failure Using Proportional Jaccard Indices and Comorbidity Patterns

Area of Science:

• Digital health
• Precision preventive medicine
• Cardiovascular disease research

Background:

• The post-coronavirus disease 2019 (COVID-19) era highlights the need for remote diagnosis and personalized preventive medicine.
• Heart failure (HF) management requires advanced tools for risk stratification and early intervention.
• Electronic Medical Records (EMRs) offer a rich data source for analyzing complex disease patterns.

Purpose of the Study:

• To develop a digital health-monitoring tool for predicting heart failure risk.
• To analyze comorbidity patterns associated with heart failure using non-random correlation.
• To establish a foundation for machine learning models in HF risk prediction.

Main Methods:

• Utilized novel similarity indices: proportional Jaccard index (PJI), multiplication of the odds ratio proportional Jaccard index (OPJI), and alpha proportional Jaccard index (APJI).
• Constructed machine learning models for heart failure risk prediction based on EMR data.
• Developed and validated models across different age groups and sexes.

Main Results:

• The prediction models demonstrated accurate identification of high-risk heart failure patients across various demographics.
• The optimal prediction model achieved an accuracy of 82.1%.
• The area under the curve (AUC) for the optimal model reached 0.878, indicating strong predictive performance.

Conclusions:

• A noninvasive heart failure risk prediction system was developed using historical EMR data.
• The proposed indices offer practical, straightforward indicators for comorbidity pattern matching.
• Source codes for the prediction models are publicly available on GitHub for further research and application.