Article Text

Original research article
Predicting emergency coronary artery bypass graft following PCI: application of a computational model to refer patients to hospitals with and without onsite surgical backup
  1. Zeeshan Syed1,
  2. Mauro Moscucci2,
  3. David Share3 and
  4. Hitinder S Gurm4
  1. 1Department of Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, Michigan, USA
  2. 2Miller School of Medicine, University of Miami, Miami, Florida, USA
  3. 3Department of Family Medicine, University of Michigan, Ann Arbor, Michigan, USA
  4. 4Division of Cardiovascular Medicine, University of Michigan, Ann Arbor, Michigan, USA
  1. Correspondence to Dr Hitinder S Gurm; hgurm{at}


Background Clinical tools to stratify patients for emergency coronary artery bypass graft (ECABG) after percutaneous coronary intervention (PCI) create the opportunity to selectively assign patients undergoing procedures to hospitals with and without onsite surgical facilities for dealing with potential complications while balancing load across providers. The goal of our study was to investigate the feasibility of a computational model directly optimised for cohort-level performance to predict ECABG in PCI patients for this application.

Methods Blue Cross Blue Shield of Michigan Cardiovascular Consortium registry data with 69 pre-procedural and angiographic risk variables from 68 022 PCI procedures in 2004–2007 were used to develop a support vector machine (SVM) model for ECABG. The SVM model was optimised for the area under the receiver operating characteristic curve (AUROC) at the level of the training cohort and validated on 42 310 PCI procedures performed in 2008–2009.

Results There were 87 cases of ECABG (0.21%) in the validation cohort. The SVM model achieved an AUROC of 0.81 (95% CI 0.76 to 0.86). Patients in the predicted top decile were at a significantly increased risk relative to the remaining patients (OR 9.74, 95% CI 6.39 to 14.85, p<0.001) for ECABG. The SVM model optimised for the AUROC on the training cohort significantly improved discrimination, net reclassification and calibration over logistic regression and traditional SVM classification optimised for univariate performance.

Conclusions Computational risk stratification directly optimising cohort-level performance holds the potential of high levels of discrimination for ECABG following PCI. This approach has value in selectively referring PCI patients to hospitals with and without onsite surgery.

This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See:

View Full Text

Statistics from

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.