A Parsimonious Machine Learning Model for Predicting Invasive Coronary Angiography Referral After Coronary Computed Tomography Angiography: Development and Multicentre External Validation
4 affiliations
- 1Departments of Radiology and Cardiac Sciences, Libin Cardiovascular Institute, Cumming School of Medicine, University of Calgary, Calgary, AB T2N 4Z6, Canada
- 2Department of Biomedical Engineering, Schulich School of Engineering, University of Calgary, Calgary, AB T2N 4Z6, Canada
- 3Department of Medicine (Division of Cardiology and Nuclear Medicine), University of Ottawa Heart Institute, Ottawa, ON K1Y 4W7, Canada
- 4Department of Medicine and Mazankowski Heart Centre, University of Alberta, Edmonton, AB T6G 0M9, Canada
Abstract
Background/Objectives: Coronary computed tomography angiography (CCTA) is recommended as a first-line test for suspected coronary artery disease. Many patients undergo invasive coronary angiography (ICA) after CCTA, and no prediction model exists to estimate the probability of this downstream testing. We aimed to develop and validate a parsimonious extreme gradient boosting (XGB) model for predicting ICA referral after CCTA. Methods: Adults undergoing a CCTA were identified at three Canadian centres. An XGB model was trained on 156 variables in the internal cohort (Calgary, n = 8823) and tested in an external cohort (Edmonton, n = 10,596). The full model and a reduced 15-variable model were compared using discrimination and calibration. Results: ICA referral occurred in 1161 (13.2%) internal and 1269 (12.0%) external patients. Internally, the reduced model matched the full model (AUC 0.698 vs. 0.701; p = 0.16) with equivalent calibration (Brier 0.1077 vs. 0.1076; slope 1.21 vs. 1.18). Externally, the full model retained a small but significant advantage (AUC 0.701 vs. 0.687; p < 0.001). Age, anginal nitrate use, and lowest HDL cholesterol were the strongest predictors. Conclusions: A 15-variable XGB model preserved most of the predictive performance of a 156-variable model for ICA referral after CCTA. Clinical use would require evidence that it improves patient selection for ICA.
Related
- Test-bolus peak attenuation for predicting aortic enhancement in coronary computed tomography angiography: an exploratory subgroup analysis
- Development and internal validation of an oxygen saturation-based prehospital cardiac risk assessment index (P-CRAI) for predicting 30-day mortality in acute myocardial infarction: a multicentre registry study from the Mie ACS Registry, Japan
- Development and validation of an interpretable clinical-radiomics model for predicting 3-year postoperative recurrence in clinical stage I lung adenocarcinoma: a multicentre study.
- Development and Validation of a Parsimonious Financial Well-Being Index in the Health and Retirement Study
- OCT-guided versus angiography-guided stent optimisation for post-interventional FFR in patients with complex coronary artery lesions: protocol for a multicentre, open-label, randomised controlled trial (PRECISION-COMPLEX)
BibTeX
@article{chunta2026parsimonious,
title = {{A Parsimonious Machine Learning Model for Predicting Invasive Coronary Angiography Referral After Coronary Computed Tomography Angiography: Development and Multicentre External Validation}},
author = {Alec Chunta and Caitlyn Gilbert and Dina Labib and Augustine Amakiri and Yuanchao Feng and Atefeh Alavi Fili and Derek Chew and Benjamin JW Chow and Michelle Graham and Julio Garcia and Mariana Bento and James White and Robert JH Miller},
journal = {Journal of clinical medicine},
year = {2026},
month = oct,
doi = {10.3390/jcm15197641},
url = {https://europepmc.org/article/PMC/PMC13665986},
note = {CC BY},
}
RIS
TY - JOUR
TI - A Parsimonious Machine Learning Model for Predicting Invasive Coronary Angiography Referral After Coronary Computed Tomography Angiography: Development and Multicentre External Validation
AU - Alec Chunta
AU - Caitlyn Gilbert
AU - Dina Labib
AU - Augustine Amakiri
AU - Yuanchao Feng
AU - Atefeh Alavi Fili
AU - Derek Chew
AU - Benjamin JW Chow
AU - Michelle Graham
AU - Julio Garcia
AU - Mariana Bento
AU - James White
AU - Robert JH Miller
PY - 2026
DA - 2026/10/01
JO - Journal of clinical medicine
DO - 10.3390/jcm15197641
UR - https://europepmc.org/article/PMC/PMC13665986
AB - Background/Objectives: Coronary computed tomography angiography (CCTA) is recommended as a first-line test for suspected coronary artery disease. Many patients undergo invasive coronary angiography (ICA) after CCTA, and no prediction model exists to estimate the probability of this downstream testing. We aimed to develop and validate a parsimonious extreme gradient boosting (XGB) model for predicting ICA referral after CCTA. Methods: Adults undergoing a CCTA were identified at three Canadian centres. An XGB model was trained on 156 variables in the internal cohort (Calgary, n = 8823) and tested in an external cohort (Edmonton, n = 10,596). The full model and a reduced 15-variable model were compared using discrimination and calibration. Results: ICA referral occurred in 1161 (13.2%) internal and 1269 (12.0%) external patients. Internally, the reduced model matched the full model (AUC 0.698 vs. 0.701; p = 0.16) with equivalent calibration (Brier 0.1077 vs. 0.1076; slope 1.21 vs. 1.18). Externally, the full model retained a small but significant advantage (AUC 0.701 vs. 0.687; p < 0.001). Age, anginal nitrate use, and lowest HDL cholesterol were the strongest predictors. Conclusions: A 15-variable XGB model preserved most of the predictive performance of a 156-variable model for ICA referral after CCTA. Clinical use would require evidence that it improves patient selection for ICA.
SN - 2077-0383
ER -