Machine Learning Predicts Diuretic Response in Acute Decompensated Heart Failure

A recent study led by Dr. Ambarish Pandey of the University of Texas Southwestern Medical Center and co-authored by Dr. Matthew Segar, a third-year cardiovascular disease fellow at The Texas Heart Institute, used a machine learning-based approach to identify, understand, and predict diuretic responsiveness in patients with acute decompensated heart failure (ADHF). 

Decades of clinical and registry datasets are utilized in the study “A Phenomapping Tool and Clinical Score to Identify Low Diuretic Efficiency in Acute Decompensated Heart Failure,” which was published in JACC: Heart Failure. 

The BAN-ADHF score is a prediction tool that researchers developed using machine learning techniques. It showed promising results in effectively predicting the diuretic response. Implementing this technique could result in customized methods for efficiently managing the congestion of patients hospitalized with ADHF after it has been validated in other clinical populations. 

Expert opinion on the best way to treat diuretic resistance in heart failure patients with stable hemodynamics and high fluid retention is still lacking. Before thinking about combination therapy, it is usually advised to optimize the dosage of loop diuretics; however, opinions differ as to how much the dosage should be raised before starting another diuretic. 

“Ineffective diuretic response in hospitalized patients can impede the course of treatment and raise the risk of mortality and readmission after discharge. Dr. Segar states that early identification of patients with low diuretic efficacy is essential to customize decongestion techniques and enhance clinical outcomes. 

Concerns around ADHF, a public health risk, are growing. Hospital stays, ER visits, and the ensuing high medical expenses are brought on by the disease. The hallmark of ADHF is an overabundance of fluid in the body, which frequently necessitates hospitalization or a modification of the patient’s present treatment regimen. 

“Today, employing loop diuretic medications to alleviate congestion is one of the main goals of treating ADHF. The ideal dose of these drugs to use is still up for debate, though. Dr. Joseph G. Rogers, President and CEO of The Texas Heart Institute, added that a more individualized approach to forecasting effective dosing techniques is required due to the variety of ADHF patients. 

Machine learning (ML) algorithms were employed in the study by researchers from various institutions in the United States to determine patient subgroups with acute heart failure according to their responsiveness to diuretic therapy. Using publicly available and deidentified data from multiple clinical trials and registries, such as DOSE, ROSE-AHF, CARRESS-HF, ATHENA-HF, ESCAPE, and the American Heart Association Precision Medicine Platform Get with the Guidelines-HF (GWTG-HF) registry, the researchers specifically developed a diuretic efficiency phenomapping approach for patients with ADHF. 

The researchers were able to create a diuretic efficiency score and phenomapping technique thanks to the participant-level pooled data. Although the patients in each category had some traits in common, they differed from one another clinically, especially in how they responded to diuretic medication. 

The predictive usefulness of the phenogrouping fact that the patient subgroups had disparities in their diuretic response in addition to meaningfully different clinical outcomes highlighted the phenogrouping strategy’s predictive usefulness.  The fact that the patient subgroups had disparities in their diuretic response and meaningfully different clinical outcomes highlighted it. 

We are aware that the BAN-ADHF score can quantitatively reliably identify, describe, and forecast diuretic resistance in ADHF patients. With this medical information, we now need to investigate if adding the BAN-ADHF score to our care regimens helps hospitalized patients with acute decompensated heart failure,” Dr. Segar stated in clinical research. 

Journal Reference  

Matthew W. Segar et al, A Phenomapping Tool and Clinical Score to Identify Low Diuretic Efficiency in Acute Decompensated Heart Failure, JACC: Heart Failure (2023). DOI: 10.1016/j.jchf.2023.09.029.