Machine Learning Algorithms for Prediction of Permanent Pacemaker Implantation After Transcatheter Aortic Valve Replacement

HomeCirculation: Arrhythmia and ElectrophysiologyVol. 14, No. 3Machine Learning Algorithms for Prediction of Permanent Pacemaker Implantation After Transcatheter Aortic Valve Replacement Free AccessReview ArticlePDF/EPUBAboutView PDFView EPUBSections ToolsAdd to favoritesDownload citationsTrack citationsPermissions ShareShare onFacebookTwitterLinked InMendeleyRedditDiggEmail Jump toSupplementary MaterialsFree ArticlePDF/EPUBMachine Takahiro Tsushima, MD Sadeer Al-Kindi, Fahd Nadeem, Guilherme F. Attizzani, Yakov Elgudin, MD, PhD Alan Markowitz, Marco A. Costa, Daniel I. Simon, Mauricio S. Arruda, Judith Mackall, Sergio G. ThalMD TsushimaTakahiro Tsushima https://orcid.org/0000-0003-0190-7555 Department Medicine (T.T.), Case Western Reserve University, Harrington Heart Vascular Institute, University Hospitals Cleveland Medical Center, OH. *T. Al-Kindi contributed equally as first authors Search more papers by this author , Al-KindiSadeer https://orcid.org/0000-0002-1122-7695 Division Cardiology, (S.A.-K., F.N., G.F.A., M.A.C., D.I.S., M.S.A., J.A.M., S.G.T.), NadeemFahd Nadeem https://orcid.org/0000-0001-9473-6243 AttizzaniGuilherme Attizzani ElgudinYakov Elgudin Cardiac Surgery, Surgery (Y.E., A.M.), MarkowitzAlan Markowitz https://orcid.org/0000-0003-0101-5047 CostaMarco Costa SimonDaniel Simon https://orcid.org/0000-0002-3386-7650 ArrudaMauricio Arruda MackallJudith Mackall https://orcid.org/0000-0003-4324-0226 ThalSergio Thal Correspondence to: Thal, Electrophysiology, Clinical Associate Professor, Medicine, 11100 Euclid Ave Cleveland, OH 44106. Email E-mail Address: [email protected] Originally published9 Mar 2021https://doi.org/10.1161/CIRCEP.120.008941Circulation: Electrophysiology. 2021;14:e008941Atrioventricular block requiring permanent pacemaker (PPM) implantation remains an important complication after transcatheter aortic valve replacement (TAVR), the risk stratification is essential identify subset patients new PPM beforehand. However, accurate prediction not established yet. Recently, machine learning (ML) technique which a scientific discipline focusing on pattern recognitions utilized developing models in clinical medicine, previously reported ML-based demonstrated significantly high predictive accuracy.1,2 The aim study evaluate performance algorithms predicting post-TAVR implantation.This single-center retrospective consecutive who underwent TAVR from March 10, 2011 October 8, 2018 (derivation cohort, group A) prospective cohort between 9, November 2019 (validation B), at Center. This data extracted research registry that was approved institutional review board All provided signed informed consent collection. Patients with preexisting cardiac implantable electronic device were excluded study. detailed information ML analysis summarized Data Supplement. In had available right ventricular pacing burden (n=132), we also evaluated whether these (trained need) can predict significant (?40%) 1 month combining training testing datasets. We considered ?40% implantations be based prior literature.3 support findings are corresponding upon reasonable request.A total 888 ultimately included A, 272 B. 184 (20.7%) required PPM, major indications complete heart 70.1% left bundle branch subsequent high-grade atrioventricular 23.4%. B, 38 (14.0%) similarly 71.1% 26.3%, respectively. baseline characteristics Table I Both preprocedural associated both groups.Regarding ML-model performances, Figure shows classifier accuracy groups, II Supplement other model parameters. ranged 59% 69%, sequential minimal optimization, simple logistic regression (SLR), locally weighted learner (LWL)–based demonstrating highest results (69%, 68%, respectively). 55% 75%, SLR, LWL, optimization–based classifiers best (75%, 74%, 73%, SLR LWL-based achieved area under curve receiver operating (AUCROC), 0.82. our conventional multivariate it showed diagnostic (AUCROC, 0.81).3 found LWL modestly predicted 0.62 0.66, 61% 67%, respectively).Download figureDownload PowerPointFigure. Accuracy various derivation validation (group A B). indicates learner; REP, reduced error pruning; SMO, optimization.Two unpublished studies ok implantation.4,5 Agasthi et al4 used 964 patients, Gradient Boosting modest discrimination (AUCROC 0.66). Truong al5 701 Random Forest (balanced accuracy, 79%; F1 score, 0.62; AUCROC, 0.88). comparison studies, larger patient sample (n=1390) (n=14). internally validated further supported incidence accurately. one classical methods, most did outperform methods present current ML-algorism still imperfect science clinicians should use appropriate ML-classifiers each dataset characteristic.There some limitations. First, single- center study, older cohort. Second, indication clearly beginning era, limited experience may cause unnecessary implantations. such phenomenon only early cases institution affect entirely. Finally, difference valves adult surgery might outcome.In conclusion, accurately or multicenter external undertaken.Nonstandard Abbreviations AcronymsAUCROCarea characteristicsLWLlocally learnerMLmachine learningPPMpermanent pacemakerSLRsimple regressionTAVRtranscatheter replacementSources FundingNone.Disclosures Dr consultant advisory Medtronic. has received honoraria work course director consulting Abbott. report no conflicts.Footnotes*T. authorsThe https://www.ahajournals.org/doi/suppl/10.1161/CIRCEP.120.008941.For Sources Funding Disclosures, see page 371.Correspondence sergio.[email protected]orgReferences1. Deo RC. Machine medicine.Circulation. 2015; 132:1920–1930. doi: 10.1161/CIRCULATIONAHA.115.001593LinkGoogle Scholar2. Hernandez-Suarez DF, Kim Y, Villablanca P, Gupta T, Wiley J, Nieves-Rodriguez BG, Rodriguez-Maldonado Feliu Maldonado R, da Luz Sant’Ana I, Sanina C, al.. in-hospital mortality replacement.JACC Cardiovasc Interv. 2019; 12:1328–1338. 10.1016/j.jcin.2019.06.013CrossrefMedlineGoogle Scholar3. F, S, Clevenger JR, Bansal EJ, Wheat HL, Kalra GF, Risk Clin Electrophysiol. 2020; 6:295–303. 10.1016/j.jacep.2019.10.020CrossrefMedlineGoogle Scholar4. Mookadam Venepally N, Girardo M, Buras Khetarpal BK, Mulpuru SK, Eleid Greason K, Beohar Abstract 15572: helps requirement post replacement.Circulation. 140:A15572. 10.1161/circ.140.suppl_1.15572LinkGoogle Scholar5. VT, Wigle Bateman E, Pallerla Ngo TNM, Beyerbach D, Kereiakes Shreenivas Tretter Palmer imlantation following TAVR: using optimize stratification.JACC. 75:1478–1478. 10.1016/S0735-1097(20)32105-7CrossrefGoogle Scholar Previous Back top Next FiguresReferencesRelatedDetails 2021Vol Issue 3Article InformationMetrics Download: 169 © 2021 American Association, Inc.https://doi.org/10.1161/CIRCEP.120.008941PMID: 33685208 publishedMarch Keywordsmachine learningatrioventricular blockrisktranscatheter replacementpatientspacemakerPDF download SubjectsAortic Replacement/Transcatheter ImplantationArrhythmiasPacemaker