AI organizes heart failure patients into 5 distinct groups, helping cardiologists manage care
Researchers have used advanced artificial intelligence (AI) models to identify five different heart failure subtypes, sharing their findings in The Lancet Digital Health.[1] This development, the group wrote, could go a long way toward improving patient care.
“We sought to improve how we classify heart failure, with the aim of better understanding the likely course of disease and communicating this to patients,” lead author Amitava Banerjee, DPhil, a professor with the UCL Institute of Health Informatics in London, said in a prepared statement. “Currently, how the disease progresses is hard to predict for individual patients. Some people will be stable for many years, while others get worse quickly. Better distinctions between types of heart failure may also lead to more targeted treatments and may help us to think in a different way about potential therapies.”
Banerjee et al. focused on electronic health record data from more than 322,000 patients with incident heart failure who were treated from 1998 to 2018. All patients were age 30 or older. The group examined key patient factors before and after they developed heart failure, compiling a list of 645 different variables. The group then turned to four unsupervised machine learning methods—k-means clustering, hierarchical clustering, k-medoids clustering and mixture model clustering—to organize those variables into five clusters. The team was ultimately able to identify and validate five heart failure subtypes: early-onset heart failure, late-onset heart failure, atrial fibrillation (AFib)-related heart failure, metabolic heart failure and cardiometabolic heart failure.
AFib-related heart failure was associated with the highest one-year all-cause mortality rate (61%), followed by late-onset heart failure (46%) and cardiometabolic heart failure (37%). Early-onset heart failure (20%) and metabolic heart failure (11%) were associated with the lowest all-cause mortality rates after one year.
The team also developed a smartphone app for identifying which subtype of heart failure any given patient is experiencing. Clinicians interviewed as part of the analysis said the app was “a feasible use of the identified clusters” that could potentially make a big impact on patient care.
Additional studies are still required to learn more about the potential impact of these AI-based clusters.
“The next step is to see if this way of classifying heart failure can make a practical difference to patients—whether it improves predictions of risk and the quality of information clinicians provide, and whether it changes patients’ treatment,” Banerjee said. “We also need to know if it would be cost effective. The app we have designed needs to be evaluated in a clinical trial or further research, but could help in routine care.”
Read the full analysis here.
Michael has more than 18 years of experience as a professional writer and editor. He has written at length about cardiology, radiology, artificial intelligence and other key healthcare topics.