CONFIDENT-HFpEF: a machine learning-based risk stratification for mortality and hospitalization using multimodal...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Heart failure with preserved ejection fraction (HFpEF) is a complex condition with significant morbidity and mortality.
• Accurate risk stratification is crucial for effective patient management and the development of targeted therapies.

Purpose of the Study:

• To develop and validate machine learning-based prognostic models for predicting all-cause mortality and heart failure (HF) hospitalization in patients with HFpEF.
• To compare the performance of these novel models against established risk scores like PREDICT-HFpEF and MAGGIC.

Main Methods:

• The CONFIDENT study utilized data from 1208 HFpEF patients across three European and US centers, with a follow-up of at least two years.
• Machine learning models were developed using routinely collected electronic health records, lab tests, echocardiography, and electrocardiography data.
• Model performance was assessed using the C-index and validated in an external cohort, comparing it against PREDICT-HFpEF and MAGGIC scores.

Main Results:

• The machine learning model for all-cause mortality demonstrated good discrimination (C-index: 0.72 in validation cohort), outperforming the PREDICT-HFpEF score.
• The prognostic model for HF hospitalization also showed superior performance compared to existing risk scores, including MAGGIC + natriuretic peptide.
• The models achieved reliable predictive accuracy using readily available clinical data.

Conclusions:

• The developed machine learning models provide reliable risk prediction for mortality and HF hospitalization in HFpEF patients.
• These models have the potential to enhance personalized patient care and optimize recruitment strategies for clinical trials in HFpEF.
• Routine data integration into machine learning models offers a promising avenue for improving outcomes in HFpEF management.