Automated Real-Time Percutaneous Coronary Intervention Risk Model Leveraging Electronic Health Records
Authors' conclusion
Real-time, automated, point of care PCI risk assessment is feasible in an EHR environment.
Study context
- Population
- Patients who underwent PCI from 2016 to 2024 (Mayo Clinic PCI registry)Events: Death 157 (1.8%), stroke 43 (0.5%), bleeding 157 (1.8%), acute kidney injury (AKI) 682 (7.6%)
- Predictors
- Variables in the Mayo Clinic PCI risk score, automatically extracted from EHR
- Predicted outcome
- In-hospital death, stroke, bleeding, acute kidney injury (AKI)
Methodological detailsLeast absolute shrinkage and selection operator regression, training and testing data sets, bootstrap analysis
- Least absolute shrinkage and selection operator regression
- Training and testing data sets
- Bootstrap analysis
Model performance
As reported in the abstract
C-statistic
In-hospital deathtraining
0.83(95% CI: 0.80-0.86)
In-hospital deathtesting
0.84(95% CI: 0.78-0.89)
Stroketraining
0.76(95% CI: 0.65-0.84)
Stroketesting
0.77(95% CI: 0.65-0.86)
Bleedingtraining
0.80(95% CI: 0.75-0.83)
Bleedingtesting
0.75(95% CI: 0.68-0.81)
AKItraining
0.82(95% CI: 0.80-0.84)
AKItesting
0.80(95% CI: 0.77-0.84)
Further findings
Overfitting (bootstrap analysis)
- Models were not overfit to the available data set
Calibration
- Estimated probabilities matched observed data well (calibration curve slope and intercept), across subgroups including women, acute coronary syndrome, cardiogenic shock, and diabetes mellitus
Original publication
Automated Real-Time Percutaneous Coronary Intervention Risk Model Leveraging Electronic Health Records. JACC Adv. Published 25 June 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →