STS Intermacs 90-Day Mortality Risk Model for Durable Left Ventricular Assist Device Implant: Validation in an Expanded Contemporary Cohort
Related Experiment Video
Updated: Sep 17, 2026
Implantation of Left Ventricular Assist Device (LVAD) in Juvenile Landrace Swine: A LVAD Implantation Model of Pediatric Heart Failure
Published on: January 16, 2026
Ezequiel J Molina1, Katherine Wood2, Ryan S Cantor3
1Department of Cardiac Surgery, Piedmont Heart Institute, Atlanta, Georgia.
Background:
Mortality risk models for durable left ventricular assist device implantation inform candidate selection, provider performance, and quality improvement. We sought to validate (STS) Intermacs mortality risk model in an expanded contemporary patient cohort.
Methods:
STS Intermacs was queried for patients undergoing primary left ventricular assist device implantation between January 1, 2019, and March 31, 2025. Beyond the original derivation and validation cohorts (N = 11,342), 4344 additional eligible patients who underwent implantation between October 1, 2023, and March 31, 2025, were identified, yielding an expanded study population of 15,686 implants. Model performance was evaluated using measures of discrimination and calibration.
Results:
Compared with the original cohort, patients in the new validation cohort were less likely to be White (54.4% vs 58.6%, P < .0001); have prior coronary artery bypass grafting (11.0% vs 13.9%, P < .0001); more likely to have diabetes (16.4% vs 12.6%, P < .0001); require concomitant surgery (58.1% vs 51.8%, P < .0001); and have lower levels of serum glutamic-oxaloacetic transaminase/aspartate aminotransferase (36.7 vs 41.0 IU/L, P = .003), blood urea nitrogen (27.5 vs 28.1 mg/dL, P = .02), and bilirubin (1.1 vs 1.2 mg/dL, P < .0001). Ninety-day mortality was similar between cohorts (P = .99). After model refitting to account for significant baseline differences, logistic regression applied to the original full cohort (derivation) and new contemporary validation cohort produced similar discrimination (area under the curve, 0.6963 vs 0.6928) and calibration (Brier score, 0.074 vs 0.075).
Conclusions:
STS Intermacs 90-day mortality risk model maintains acceptable performance when applied to an expanded cohort. The stability of discrimination and calibration in >15,000 patients support its use as a robust tool for candidate selection, benchmarking, and quality assessment in a high-risk patient population.
