The research lined in this abstract was revealed in medRxiv.org as a preprint and has not but been peer-reviewed.
Key Takeaways
Machine studying recognized 5 subtypes in incident coronary heart failure (HF) with good predictive accuracy for 1-year all-cause mortality.
Future trials and scientific apply may have the ability to use the digital well being file (EHR) and machine studying to determine HF subtypes to assist in administration and prognosis.
Why This Matters
This is the first research to outline and validate HF subtypes throughout a number of machine-learning strategies, nationally consultant datasets, and a number of validation strategies.
The research’s structured framework of inside, exterior, prognostic, and genetic validation might prolong acceptability and generalizability of machine studying to scientific apply and is transferable to different ailments.
Study Design
The research concerned 313,062 sufferers 30 years and older with incident HF and a minimum of 1 12 months of follow-up in The Health Improvement Network (THIN) and Clinical Practice Research Datalink (CPRD).
EHRs from the two main care populations have been linked with hospital admissions and a demise registry in the United Kingdom.
UK Biobank knowledge from 9573 sufferers have been used for genetic validation.
HF subtypes have been recognized utilizing 4 unsupervised machine-learning strategies with 87 (from 645) components (demography, historical past, examination, blood laboratory values, and medicine) included in the algorithm.
Validation was completed internally, externally, prognostically, and genetically.
A HF clustering mannequin and open-access app have been developed to be used in routine apply.
Key Results
Five clusters have been recognized and labeled as subtypes: early-onset, late-onset, atrial fibrillation (AF)-related, metabolic, and cardiometabolic.
The distribution of subtypes 1 to 5 was related throughout THIN (16.5%, 30.9%, 8.9%, 14.0%, and 29.7%) and CPRD (12.6%, 35.6%, 9.3%, 13.8%, 28.7%).
Age assorted throughout subtypes, with the oldest sufferers having late-onset HF and the youngest having early-onset HF.
There have been additionally variations by intercourse, with the most girls in the metabolic subtype and the least in the cardiometabolic group.
Cardiovascular danger components and ailments — reminiscent of hypertension (72.9%), weight problems (34.3%), diabetes (41.1%), and atherosclerosis (59.2%) — have been highest in the cardiometabolic subtype.
In CPRD, 1-year mortality was 2%, 46%, 6%, 11%, and 37% for subtypes 1 to 5, respectively, with C-statistic of 0.68, 0.62, 0.57, 0.71, and 0.68, respectively.
Differences in mortality between THIN and CPRD have been discovered for clusters 2 and 5 solely.
Hypertension, myocardial infarction, stroke, and peripheral vascular illness usually occurred earlier than HF analysis in the cardiometabolic subtype and after HF analysis in the AF-related and early-onset subtypes.
Polygenic danger scores for atrial arrhythmias, diabetes, hypertension, myocardial infarction, weight problems, and steady and unstable angina have been all related to a number of HF subtypes.
Eight single-nucleotide polymorphisms have been nominally related to predicted HF subtype (P = .05), of which 4 have been related to the AF-related subtype.
Five clinicians questioned in the research said that the outcomes had scientific utility and that the app could be helpful.
Limitations
The investigators used EHR phenotypes of HF, which shouldn’t have full biochemical and imaging profiles and, subsequently, some classifications couldn’t be completed.
The danger issue phenotypes relied on the timing and accuracy of the clinicians’ notes in the EHR.
Both datasets used are from the United Kingdom, which could not be consultant of HF in different locations.
The polygenic danger scores have been just for 11 traits and 12 single-nucleotide polymorphisms.
Disclosures
Amitava Banerjee is supported by analysis funding from the National Institute for Health Research (NIHR), British Medical Association, AstraZeneca, and UK Research and Innovation.
Benoit Tyl and Tomasz Dyszynski are staff of Bayer. All different authors declared no competing pursuits.
This is a abstract of a preprint analysis research, Identifying subtypes of coronary heart failure with machine studying: exterior, prognostic and genetic validation in three digital well being file sources with 320,863 people, written by Amitava Banerjee, BMBCh, DPhil, from University College London, and colleagues, on MedRxiv offered to you by Medscape. This research has not but been peer-reviewed. The full textual content of the research may be discovered on medRxiv.org.
Credit:Lead picture: Dreamstime
Cite this: Using the Power of Machine Learning to Hone in on HF Subtypes – Medscape – Jul 22, 2022.
https://www.medscape.com/viewarticle/977754
