Predicting atrial fibrillation in primary care using machine learning

Nathan R. Hill; Daniel Ayoubkhani; Phil McEwan; Daniel M. Sugrue; Usman Farooqui; Steven Lister; Matthew Lumley; Ameet Bakhai; Alexander T. Cohen; Mark O’Neill; David Clifton; Jason Gordon

doi:10.1371/journal.pone.0224582

Predicting atrial fibrillation in primary care using machine learning

Nathan R. Hill^*, Daniel Ayoubkhani, Phil McEwan, Daniel M. Sugrue, Usman Farooqui, Steven Lister, Matthew Lumley, Ameet Bakhai, Alexander T. Cohen, Mark O’Neill, David Clifton, Jason Gordon

^*Corresponding author for this work

Biomedical Engineering Department

Research output: Contribution to journal › Article › peer-review

93 Citations (Scopus)

Abstract

Background Atrial fibrillation (AF) is the most common sustained heart arrhythmia. However, as many cases are asymptomatic, a large proportion of patients remain undiagnosed until serious complications arise. Efficient, cost-effective detection of the undiagnosed may be supported by risk-prediction models relating patient factors to AF risk. However, there exists a need for an implementable risk model that is contemporaneous and informed by routinely collected patient data, reflecting the real-world pathology of AF. Methods This study sought to develop and evaluate novel and conventional statistical and machine learning models for risk-predication of AF. This was a retrospective, cohort study of adults (aged >30 years) without a history of AF, listed on the Clinical Practice Research Datalink, from January 2006 to December 2016. Models evaluated included published risk models (Framingham, ARIC, CHARGE-AF), machine learning models, which evaluated baseline and time-updated information (neural network, LASSO, random forests, support vector machines), and Cox regression. Results Analysis of 2,994,837 individuals (3.2% AF) identified time-varying neural networks as the optimal model achieving an AUROC of 0.827 vs. 0.725, with number needed to screen of 9 vs. 13 patients at 75% sensitivity, when compared with the best existing model CHARGE-AF. The optimal model confirmed known baseline risk factors (age, previous cardiovascular disease, antihypertensive medication usage) and identified additional time-varying predictors (proximity of cardiovascular events, body mass index (both levels and changes), pulse pressure, and the frequency of blood pressure measurements). Conclusion The optimal time-varying machine learning model exhibited greater predictive performance than existing AF risk models and reflected known and new patient risk factors for AF.

Original language	English
Article number	e0224582
Journal	PLoS ONE
Volume	14
Issue number	11
DOIs	https://doi.org/10.1371/journal.pone.0224582
Publication status	Published - 1 Nov 2019

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1371/journal.pone.0224582

Cite this

@article{d33fba3557aa468ab1ebcf858362a835,

title = "Predicting atrial fibrillation in primary care using machine learning",

abstract = "Background Atrial fibrillation (AF) is the most common sustained heart arrhythmia. However, as many cases are asymptomatic, a large proportion of patients remain undiagnosed until serious complications arise. Efficient, cost-effective detection of the undiagnosed may be supported by risk-prediction models relating patient factors to AF risk. However, there exists a need for an implementable risk model that is contemporaneous and informed by routinely collected patient data, reflecting the real-world pathology of AF. Methods This study sought to develop and evaluate novel and conventional statistical and machine learning models for risk-predication of AF. This was a retrospective, cohort study of adults (aged >30 years) without a history of AF, listed on the Clinical Practice Research Datalink, from January 2006 to December 2016. Models evaluated included published risk models (Framingham, ARIC, CHARGE-AF), machine learning models, which evaluated baseline and time-updated information (neural network, LASSO, random forests, support vector machines), and Cox regression. Results Analysis of 2,994,837 individuals (3.2% AF) identified time-varying neural networks as the optimal model achieving an AUROC of 0.827 vs. 0.725, with number needed to screen of 9 vs. 13 patients at 75% sensitivity, when compared with the best existing model CHARGE-AF. The optimal model confirmed known baseline risk factors (age, previous cardiovascular disease, antihypertensive medication usage) and identified additional time-varying predictors (proximity of cardiovascular events, body mass index (both levels and changes), pulse pressure, and the frequency of blood pressure measurements). Conclusion The optimal time-varying machine learning model exhibited greater predictive performance than existing AF risk models and reflected known and new patient risk factors for AF.",

author = "Hill, {Nathan R.} and Daniel Ayoubkhani and Phil McEwan and Sugrue, {Daniel M.} and Usman Farooqui and Steven Lister and Matthew Lumley and Ameet Bakhai and Cohen, {Alexander T.} and Mark O{\textquoteright}Neill and David Clifton and Jason Gordon",

year = "2019",

month = nov,

day = "1",

doi = "10.1371/journal.pone.0224582",

language = "English",

volume = "14",

journal = "PLoS ONE",

issn = "1932-6203",

publisher = "Public Library of Science",

number = "11",

}

TY - JOUR

T1 - Predicting atrial fibrillation in primary care using machine learning

AU - Hill, Nathan R.

AU - Ayoubkhani, Daniel

AU - McEwan, Phil

AU - Sugrue, Daniel M.

AU - Farooqui, Usman

AU - Lister, Steven

AU - Lumley, Matthew

AU - Bakhai, Ameet

AU - Cohen, Alexander T.

AU - O’Neill, Mark

AU - Clifton, David

AU - Gordon, Jason

PY - 2019/11/1

Y1 - 2019/11/1

N2 - Background Atrial fibrillation (AF) is the most common sustained heart arrhythmia. However, as many cases are asymptomatic, a large proportion of patients remain undiagnosed until serious complications arise. Efficient, cost-effective detection of the undiagnosed may be supported by risk-prediction models relating patient factors to AF risk. However, there exists a need for an implementable risk model that is contemporaneous and informed by routinely collected patient data, reflecting the real-world pathology of AF. Methods This study sought to develop and evaluate novel and conventional statistical and machine learning models for risk-predication of AF. This was a retrospective, cohort study of adults (aged >30 years) without a history of AF, listed on the Clinical Practice Research Datalink, from January 2006 to December 2016. Models evaluated included published risk models (Framingham, ARIC, CHARGE-AF), machine learning models, which evaluated baseline and time-updated information (neural network, LASSO, random forests, support vector machines), and Cox regression. Results Analysis of 2,994,837 individuals (3.2% AF) identified time-varying neural networks as the optimal model achieving an AUROC of 0.827 vs. 0.725, with number needed to screen of 9 vs. 13 patients at 75% sensitivity, when compared with the best existing model CHARGE-AF. The optimal model confirmed known baseline risk factors (age, previous cardiovascular disease, antihypertensive medication usage) and identified additional time-varying predictors (proximity of cardiovascular events, body mass index (both levels and changes), pulse pressure, and the frequency of blood pressure measurements). Conclusion The optimal time-varying machine learning model exhibited greater predictive performance than existing AF risk models and reflected known and new patient risk factors for AF.

AB - Background Atrial fibrillation (AF) is the most common sustained heart arrhythmia. However, as many cases are asymptomatic, a large proportion of patients remain undiagnosed until serious complications arise. Efficient, cost-effective detection of the undiagnosed may be supported by risk-prediction models relating patient factors to AF risk. However, there exists a need for an implementable risk model that is contemporaneous and informed by routinely collected patient data, reflecting the real-world pathology of AF. Methods This study sought to develop and evaluate novel and conventional statistical and machine learning models for risk-predication of AF. This was a retrospective, cohort study of adults (aged >30 years) without a history of AF, listed on the Clinical Practice Research Datalink, from January 2006 to December 2016. Models evaluated included published risk models (Framingham, ARIC, CHARGE-AF), machine learning models, which evaluated baseline and time-updated information (neural network, LASSO, random forests, support vector machines), and Cox regression. Results Analysis of 2,994,837 individuals (3.2% AF) identified time-varying neural networks as the optimal model achieving an AUROC of 0.827 vs. 0.725, with number needed to screen of 9 vs. 13 patients at 75% sensitivity, when compared with the best existing model CHARGE-AF. The optimal model confirmed known baseline risk factors (age, previous cardiovascular disease, antihypertensive medication usage) and identified additional time-varying predictors (proximity of cardiovascular events, body mass index (both levels and changes), pulse pressure, and the frequency of blood pressure measurements). Conclusion The optimal time-varying machine learning model exhibited greater predictive performance than existing AF risk models and reflected known and new patient risk factors for AF.

UR - http://www.scopus.com/inward/record.url?scp=85074347946&partnerID=8YFLogxK

U2 - 10.1371/journal.pone.0224582

DO - 10.1371/journal.pone.0224582

M3 - Article

C2 - 31675367

AN - SCOPUS:85074347946

SN - 1932-6203

VL - 14

JO - PLoS ONE

JF - PLoS ONE

IS - 11

M1 - e0224582

ER -

Predicting atrial fibrillation in primary care using machine learning

Abstract

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this