Five Forms Of Heart Failure Were Uncovered By AI Techniques

Researchers from the University College London (UCL) have just published a novel study that uses artificial intelligence (AI) technologies to identify five different forms of heart failure that may be used to forecast future risk for specific patients.

“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,” said lead author Professor Amitava Banerjee of the University College London Institute of Health Informatics. For specific patients, it is still difficult to anticipate how the condition will develop. While some people remain stable for many years, others rapidly deteriorate.

Heart failure occurs when the heart is unable to circulate blood throughout the body effectively. Heart failure classification techniques now in use cannot accurately predict how the condition will develop.

A better categorization of the different forms of heart failure may also result in more focused therapy and alter our perspective on potential remedies.

Using several machine learning techniques and various datasets, Banerjee and colleagues “identified five robust subtypes in this new study.”

For the study, which was published in Lancet Digital Health, researchers evaluated comprehensive, anonymized patient data from more than 300,000 patients, age 30 or older, who received a heart failure diagnosis in the UK over a 20-year period.

“The next stage is to determine whether this method of classifying heart failure can actually benefit patients—whether it enhances risk projections, the calibre of information provided by clinicians, and whether it alters patient management. Additionally, we need to determine if it would be economical. The app we created needs to be tested in a clinical trial or through more study, but it might be useful for routine medical treatment, noted Banerjee.

Using a variety of machine learning techniques, the researchers were able to separate the five subtypes into early onset, late onset, atrial fibrillation-related, metabolic, and cardiometabolic. Atrial fibrillation is a heart rhythm disorder that is linked to obesity but has a low prevalence of cardiovascular disease.

They discovered variations in the subgroups’ mortality risks a year after diagnosis.

Early-onset mortality risks were 20%; late-onset mortality risks were 46%; atrial fibrillation-related mortality risks were 61%; metabolic mortality risks were 11%; and cardiometabolic mortality risks were 37%.

The research team also created an app that clinicians may use to ascertain the subtype of heart failure a patient has, which may enhance assessments of future risk and guide conversations with patients.

The researchers grouped heart failure cases using four different machine learning techniques to eliminate bias caused by one particular method. These techniques were used using information from two sizable primary care datasets from the UK, which were typical of the country’s entire population and connected to hospital admission and death records.

The datasets, which covered the years 1998 to 2018, were created by the Clinical Practise Research Datalink (CPRD) and the Health Improvement Network (THIN).

The research team used portions of the data to train the machine learning algorithms, and after choosing the most reliable subtypes, they confirmed these categories using a different dataset.

Age, symptoms, the presence of additional diseases, the drugs the patient was taking, and the results of tests (such as blood pressure measurements) and assessments (such as kidney function evaluations) were some of the 87 (of a possible 635) criteria used to determine the subtypes.

The scientists also examined genetic information from 9,573 heart failure patients from the UK Biobank project. They discovered a connection between specific heart failure subtypes and higher polygenic risk scores (scores of general risk brought on by genes as a whole) for ailments including hypertension and atrial fibrillation.