Using artificial intelligence (AI) tools, a recent study has identified five subtypes of heart failure that have the potential to predict future risk for individual patients. Published in The Lancet Digital Health journal, the study highlights the use of machine learning to analyze heart failure subtypes, encompassing large population-based datasets, diverse causes and presentations, and validation through different machine learning methods.
The researchers analyzed detailed anonymized patient data from individuals aged 30 years or older who were diagnosed with heart failure in the United Kingdom over a 20-year period. The data included pre-heart failure and post-heart failure factors such as demographic information, medical history, examination results, laboratory values, and medications. Four unsupervised machine learning methods were employed (K-means, hierarchical, K-Medoids, and mixture model clustering) using 87 factors in each dataset.
The study identified five incident heart failure subtypes: early onset, late onset, atrial fibrillation-related, metabolic, and cardiometabolic. These subtypes were evaluated for external validity, prognostic validity (predictive accuracy for 1-year mortality), genetic validity (UK Biobank), and association with polygenic risk scores (PRS) for heart failure-related traits and single nucleotide polymorphisms. Late-onset and cardiometabolic subtypes exhibited the highest similarity and were strongly associated with PRS for hypertension, myocardial infarction, and obesity. The one-year all-cause mortality risks for each subtype were early onset (20 percent), late onset (46 percent), atrial fibrillation-related (61 percent), metabolic (11 percent), and cardiometabolic (37 percent).
In addition, the researchers developed an app that clinicians could use to identify the heart failure subtype a patient falls into and predict their survival. Clinicians who tested the app found it feasible for routine care, enabling them to classify patients into specific clusters during consultations. Further prospective studies are necessary to test the effectiveness and cost-effectiveness of the app.
Lead author Professor Amitava Banerjee from University College London (UCL) explained that the goal was to improve heart failure classification in order to better understand the disease’s progression and effectively communicate it to patients. By distinguishing between types of heart failure more accurately, targeted treatments can be developed, potentially leading to new therapeutic approaches.
Post Your Comments