Machine Learning Reveals Prognostic Value of Medical History in Suspected Coronary Artery Disease
Model performance · MACEResult — go to the resultΔAUROC 0.014 95% CI 0.002–0.027Authors' conclusion
In patients with suspected CAD, machine learning models using 10-year historical registry data have the potential to improve prognostic risk prediction beyond demographics and stress perfusion CMR.
Study context
- Population
- Patients with suspected coronary artery disease (CAD) undergoing stress perfusion cardiovascular magnetic resonance (CMR)65 ± 12 years, 44% women
- Participants
- 2,159 patientsEvents: 212 patients (9.8%)
- Predictors
- Demographics, CMR findings, and registry diagnoses and procedures from the prior 10 years; Random Forest, Random Survival Forests (RSF) and penalized Ridge Cox regression
- Predicted outcome
- major adverse cardiovascular events (MACE), defined as cardiovascular death, myocardial infarction, or coronary intervention
Model performance
MACE
AUROC or C-index
Full model
AUROC 0.82(95% CI: 0.79-0.85)
Demographics (Random Forest, binary prediction)
AUROC 0.66(95% CI: 0.62-0.69)
Demographics plus historical registry data
AUROC 0.72(95% CI: 0.68-0.75; P = 0.001)
Demographics-plus-CMR model
AUROC 0.81; 95% CI: 0.78-0.84
RSF, demographics and CMRtime-to-event analysis
C-index 0.80
Cox regression, demographics and CMRtime-to-event analysis
C-index 0.80
RSF, including registry datatime-to-event analysis
C-index 0.81(P = 0.144)
Cox, including registry datatime-to-event analysis
C-index 0.77(P < 0.001)
ΔAUROC
0.014
95% CI 0.002–0.027
P = 0.029
Full modelvs Demographics-plus-CMR model
ΔAUROC 0.014 (95% CI 0.002–0.027); P = 0.029
As reported in the abstract
Further findings
Major adverse cardiovascular events (MACE), time-to-event analysis
- RSF and Cox performed similarly using demographics and CMR; P = 0.546
Follow-up
- Follow-up
- 2.7 ± 0.9 years
Original publication
Machine Learning Reveals Prognostic Value of Medical History in Suspected Coronary Artery Disease. JACC Adv. Published 10 November 2026. doi:10. 1016/ j. jacadv. 2026. 103443
Composed by AI from the published abstract and checked against it. How Briefs are made →