Predicting long-term survival after coronary artery bypass graft surgery

Research output: Contribution to journalArticleResearchpeer-review

Abstract

OBJECTIVES To develop a model for predicting long-term survival following coronary artery bypass graft surgery. METHODS This study included 46 573 patients from the Australian and New Zealand Society of Cardiac and Thoracic Surgeons (ANZCTS) registry, who underwent isolated coronary artery bypass graft surgery between 2001 and 2014. Data were randomly split into development (23 282) and validation (23 291) samples. Cox regression models were fitted separately, using the important preoperative variables, for 4 'time intervals' (31-90 days, 91-365 days, 1-3 years and >3 years), with optimal predictors selected using the bootstrap bagging technique. Model performance was assessed both in validation data and in combined data (development and validation samples). Coefficients of all 4 final models were estimated on the combined data adjusting for hospital-level clustering. RESULTS The Kaplan-Meier mortality rates estimated in the sample were 1.7% at 90 days, 2.8% at 1 year, 4.4% at 2 years and 6.1% at 3 years. Age, peripheral vascular disease, respiratory disease, reduced ejection fraction, renal dysfunction, arrhythmia, diabetes, hypercholesterolaemia, cerebrovascular disease, hypertension, congestive heart failure, steroid use and smoking were included in all 4 models. However, their magnitude of effect varied across the time intervals. Harrell's C-statistics was 0.83, 0.78, 0.75 and 0.74 for 31-90 days, 91-365 days, 1-3 years and >3 years models, respectively. Models showed excellent discrimination and calibration in validation data. CONCLUSIONS Models were developed for predicting long-term survival at 4 time intervals after isolated coronary artery bypass graft surgery. These models can be used in conjunction with the existing 30-day mortality prediction model.

LanguageEnglish
Pages257-263
Number of pages7
JournalInteractive Cardiovascular and Thoracic Surgery
Volume26
Issue number2
DOIs
Publication statusPublished - 1 Feb 2018

Keywords

  • Cardiac surgery
  • Coronary artery bypass graft
  • Coronary revascularization
  • Long-term survival
  • Risk prediction model
  • Risk stratification

Cite this

@article{054de6671380426a908a39f0a5b9b20e,
title = "Predicting long-term survival after coronary artery bypass graft surgery",
abstract = "OBJECTIVES To develop a model for predicting long-term survival following coronary artery bypass graft surgery. METHODS This study included 46 573 patients from the Australian and New Zealand Society of Cardiac and Thoracic Surgeons (ANZCTS) registry, who underwent isolated coronary artery bypass graft surgery between 2001 and 2014. Data were randomly split into development (23 282) and validation (23 291) samples. Cox regression models were fitted separately, using the important preoperative variables, for 4 'time intervals' (31-90 days, 91-365 days, 1-3 years and >3 years), with optimal predictors selected using the bootstrap bagging technique. Model performance was assessed both in validation data and in combined data (development and validation samples). Coefficients of all 4 final models were estimated on the combined data adjusting for hospital-level clustering. RESULTS The Kaplan-Meier mortality rates estimated in the sample were 1.7{\%} at 90 days, 2.8{\%} at 1 year, 4.4{\%} at 2 years and 6.1{\%} at 3 years. Age, peripheral vascular disease, respiratory disease, reduced ejection fraction, renal dysfunction, arrhythmia, diabetes, hypercholesterolaemia, cerebrovascular disease, hypertension, congestive heart failure, steroid use and smoking were included in all 4 models. However, their magnitude of effect varied across the time intervals. Harrell's C-statistics was 0.83, 0.78, 0.75 and 0.74 for 31-90 days, 91-365 days, 1-3 years and >3 years models, respectively. Models showed excellent discrimination and calibration in validation data. CONCLUSIONS Models were developed for predicting long-term survival at 4 time intervals after isolated coronary artery bypass graft surgery. These models can be used in conjunction with the existing 30-day mortality prediction model.",
keywords = "Cardiac surgery, Coronary artery bypass graft, Coronary revascularization, Long-term survival, Risk prediction model, Risk stratification",
author = "Karim, {Md N.} and Reid, {Christopher M.} and Molla Huq and Brilleman, {Samuel L.} and Andrew Cochrane and Lavinia Tran and Baki Billah",
year = "2018",
month = "2",
day = "1",
doi = "10.1093/icvts/ivx330",
language = "English",
volume = "26",
pages = "257--263",
journal = "Interactive Cardiovascular and Thoracic Surgery",
issn = "1569-9293",
publisher = "Oxford University Press",
number = "2",

}

TY - JOUR

T1 - Predicting long-term survival after coronary artery bypass graft surgery

AU - Karim, Md N.

AU - Reid, Christopher M.

AU - Huq, Molla

AU - Brilleman, Samuel L.

AU - Cochrane, Andrew

AU - Tran, Lavinia

AU - Billah, Baki

PY - 2018/2/1

Y1 - 2018/2/1

N2 - OBJECTIVES To develop a model for predicting long-term survival following coronary artery bypass graft surgery. METHODS This study included 46 573 patients from the Australian and New Zealand Society of Cardiac and Thoracic Surgeons (ANZCTS) registry, who underwent isolated coronary artery bypass graft surgery between 2001 and 2014. Data were randomly split into development (23 282) and validation (23 291) samples. Cox regression models were fitted separately, using the important preoperative variables, for 4 'time intervals' (31-90 days, 91-365 days, 1-3 years and >3 years), with optimal predictors selected using the bootstrap bagging technique. Model performance was assessed both in validation data and in combined data (development and validation samples). Coefficients of all 4 final models were estimated on the combined data adjusting for hospital-level clustering. RESULTS The Kaplan-Meier mortality rates estimated in the sample were 1.7% at 90 days, 2.8% at 1 year, 4.4% at 2 years and 6.1% at 3 years. Age, peripheral vascular disease, respiratory disease, reduced ejection fraction, renal dysfunction, arrhythmia, diabetes, hypercholesterolaemia, cerebrovascular disease, hypertension, congestive heart failure, steroid use and smoking were included in all 4 models. However, their magnitude of effect varied across the time intervals. Harrell's C-statistics was 0.83, 0.78, 0.75 and 0.74 for 31-90 days, 91-365 days, 1-3 years and >3 years models, respectively. Models showed excellent discrimination and calibration in validation data. CONCLUSIONS Models were developed for predicting long-term survival at 4 time intervals after isolated coronary artery bypass graft surgery. These models can be used in conjunction with the existing 30-day mortality prediction model.

AB - OBJECTIVES To develop a model for predicting long-term survival following coronary artery bypass graft surgery. METHODS This study included 46 573 patients from the Australian and New Zealand Society of Cardiac and Thoracic Surgeons (ANZCTS) registry, who underwent isolated coronary artery bypass graft surgery between 2001 and 2014. Data were randomly split into development (23 282) and validation (23 291) samples. Cox regression models were fitted separately, using the important preoperative variables, for 4 'time intervals' (31-90 days, 91-365 days, 1-3 years and >3 years), with optimal predictors selected using the bootstrap bagging technique. Model performance was assessed both in validation data and in combined data (development and validation samples). Coefficients of all 4 final models were estimated on the combined data adjusting for hospital-level clustering. RESULTS The Kaplan-Meier mortality rates estimated in the sample were 1.7% at 90 days, 2.8% at 1 year, 4.4% at 2 years and 6.1% at 3 years. Age, peripheral vascular disease, respiratory disease, reduced ejection fraction, renal dysfunction, arrhythmia, diabetes, hypercholesterolaemia, cerebrovascular disease, hypertension, congestive heart failure, steroid use and smoking were included in all 4 models. However, their magnitude of effect varied across the time intervals. Harrell's C-statistics was 0.83, 0.78, 0.75 and 0.74 for 31-90 days, 91-365 days, 1-3 years and >3 years models, respectively. Models showed excellent discrimination and calibration in validation data. CONCLUSIONS Models were developed for predicting long-term survival at 4 time intervals after isolated coronary artery bypass graft surgery. These models can be used in conjunction with the existing 30-day mortality prediction model.

KW - Cardiac surgery

KW - Coronary artery bypass graft

KW - Coronary revascularization

KW - Long-term survival

KW - Risk prediction model

KW - Risk stratification

UR - http://www.scopus.com/inward/record.url?scp=85041519945&partnerID=8YFLogxK

U2 - 10.1093/icvts/ivx330

DO - 10.1093/icvts/ivx330

M3 - Article

VL - 26

SP - 257

EP - 263

JO - Interactive Cardiovascular and Thoracic Surgery

T2 - Interactive Cardiovascular and Thoracic Surgery

JF - Interactive Cardiovascular and Thoracic Surgery

SN - 1569-9293

IS - 2

ER -