lancet-header
Preprints with The Lancet is part of SSRN´s First Look, a place where journals and other research experts identify content of interest prior to publication. These preprint papers are not peer-reviewed and are posted here as part of a 12-month trial. Authors have either opted in at submission to The Lancet family of journals to post their preprints on Preprints with The Lancet, or submitted directly via SSRN. The usual SSRN checks and a Lancet-specific check for appropriateness and transparency have been applied. These papers should not be used for clinical decision making or reporting of research to a lay audience without indicating that this is preliminary research that has not been peer-reviewed. For more information see the Comment published in The Lancet, or visit The Lancet´s FAQ page, and for any feedback please contact preprints@lancet.com

Predicting Atrial Fibrillation in Primary Care Using Machine Learning

22 Pages Posted: 25 Sep 2018

See all articles by Nathan R. Hill

Nathan R. Hill

Bristol-Myers Squibb

Daniel Ayoubkhani

affiliation not provided to SSRN

Phil McEwan

affiliation not provided to SSRN

Daniel M. Sugrue

affiliation not provided to SSRN

Usman Farooqui

affiliation not provided to SSRN

Steve Lister

affiliation not provided to SSRN

Matthew Lumley

affiliation not provided to SSRN

Ameet Bakhai

affiliation not provided to SSRN

Alexander T. Cohen

affiliation not provided to SSRN

Mark O’Neill

affiliation not provided to SSRN

David Clifton

affiliation not provided to SSRN

Jason Gordon

affiliation not provided to SSRN

More...

Abstract

Background: Atrial fibrillation (AF) is the most common sustained heart arrhythmia. Difficulties in diagnosing AF - particularly when paroxysmal and asymptomatic - and lack of routine screening contribute to underdiagnoses. 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: Machine learning and conventional statistical models were evaluated using patients identified from UK healthcare records, aged ≥30 years, without a history of AF, from January-2006 to December-2016. A matched case-control study evaluated published risk models (Framingham, ARIC, CHARGE-AF), machine learning models using baseline and time-updated information (neural network, LASSO, random forests, support vector machines) and Cox regression. Models were ranked with maximal discrimination between AF and non-AF cases (area under the receiver operating characteristic curve, AUROC) and potential number-needed-to-screen (NNS) to detect a case of AF.

Findings: Analysis of 2,994,837 patients (3.2% AF) identified time-varying neural networks as the optimal model. Compared to the best existing model CHARGE-AF, the optimal model achieved an AUROC of 0.827 vs. 0.725, with NNS of 9 vs. 13 patients at 75% sensitivity. The optimal model confirmed known baseline risk factors (age, previous cardiovascular disease, antihypertensive medication usage) and identified additional time-varying predictors previously unknown (recency of cardiovascular events, body mass index (both levels and changes), pulse pressure, and the frequency of blood pressure measurements).

Interpretation: The optimal time-varying machine learning model exhibited greater predictive performance than existing AF risk models and reflected known and new relationships between patient risk factors for AF. If utilised, the model may help improve the precision of AF screening, leading to better patient outcomes and cost-effective use of healthcare resources.

Funding: Bristol-Myers Squibb and Pfizer.

Declaration of Interest: NR Hill, U Farooqui, & S Lister are employees of Bristol-Myers Squibb Company. M Lumley is an employee of Pfizer Inc. D Ayoubkhani, P McEwan, D M Sugrue, J Gordon are employed by HEOR ltd., which provides consulting and other research services to pharmaceutical, medical device, and related organizations. In their salaried positions, they work with a variety of companies and organizations, and are precluded from receiving payments or honoraria directly from these organizations for services rendered. AT Cohen reports grants and personal fees from Bristol-Myers Squibb Company and Pfizer Inc. during the conduct of the study; personal fees from Boehringer Ingelheim, grants and personal fees from Bristol-Myers Squibb Company, grants and personal fees from Daiichi-Sankyo Europe, personal fees from Johnson & Johnson, grants and personal fees from Pfizer, Inc., personal fees from Portola, personal fees from Sanofi, personal fees from XO1, personal fees from Janssen, personal fees from ONO Pharmaceuticals, and grants and personal fees from Bayer AG, outside the submitted work. A Bakhai reports personal fees from Bristol-Myers Squibb Company and Pfizer Inc. during the conduct of the study; personal fees from Boehringer Ingelheim, personal fees from Bristol-Myers Squibb Company, personal fees from Daiichi-Sankyo Europe, personal fees from Johnson & Johnson, personal fees from Pfizer, Inc., personal fees from Novartis, personal fees from Sanofi, personal fees from MSD, personal fees from Janssen, personal fees from Roche, and personal fees from Bayer AG, outside the submitted work. DA Clifton reports personal fees from Bristol-Myers Squibb Company during the conduct of the study; and outside the submitted work, personal fees from Drayson Health (now Sensyne Health), personal fees from Ferrovial plc., personal fees from Quanta Dialysis, and personal fees from BioBeats Ltd. M O’Neill reports personal fees from Bristol-Myers Squibb Company and Pfizer Inc. during the conduct of the study; grants and personal fees from Biosense Webster, grants and personal fees from Abbott, personal fees from Siemens, personal fees from Vytronus, personal fees from Medtronic outside the submitted work.

Ethical Approval: The study protocol was reviewed and approved by the Independent Scientific Advisory Committee for MHRA database research (ISAC, reference number 17_151).

Suggested Citation

Hill, Nathan R. and Ayoubkhani, Daniel and McEwan, Phil and Sugrue, Daniel M. and Farooqui, Usman and Lister, Steve and Lumley, Matthew and Bakhai, Ameet and Cohen, Alexander T. and O’Neill, Mark and Clifton, David and Gordon, Jason, Predicting Atrial Fibrillation in Primary Care Using Machine Learning (August 20, 2018). Available at SSRN: https://ssrn.com/abstract=3236366

Nathan R. Hill (Contact Author)

Bristol-Myers Squibb ( email )

United States

Daniel Ayoubkhani

affiliation not provided to SSRN

Phil McEwan

affiliation not provided to SSRN

Daniel M. Sugrue

affiliation not provided to SSRN

Usman Farooqui

affiliation not provided to SSRN

Steve Lister

affiliation not provided to SSRN

Matthew Lumley

affiliation not provided to SSRN

Ameet Bakhai

affiliation not provided to SSRN

Alexander T. Cohen

affiliation not provided to SSRN

Mark O’Neill

affiliation not provided to SSRN

David Clifton

affiliation not provided to SSRN

Jason Gordon

affiliation not provided to SSRN

Click here to go to TheLancet.com

Go to TheLancet.com

Paper statistics

Abstract Views
297
Downloads
59