The Society of Thoracic Surgeons National Intermacs Database Risk Model for Durable Left Ventricular Assist Device...

Area of Science:

• Cardiology
• Medical Statistics
• Health Services Research

Background:

• Statistical risk models are crucial for left ventricular assist device (LVAD) implantation.
• These models aid in candidate selection, quality improvement initiatives, and provider performance evaluation.
• This study focused on developing a 90-day mortality risk model using The Society of Thoracic Surgeons National Intermacs Database (STS Intermacs).

Purpose of the Study:

• To develop and validate a statistical risk model for predicting 90-day mortality after durable LVAD implantation.
• To assess the model's performance in terms of discrimination and calibration.
• To enable the calculation of observed-to-expected mortality ratios for individual centers.

Main Methods:

• Utilized The Society of Thoracic Surgeons National Intermacs Database (STS Intermacs) for primary durable LVAD implants (January 2019 - September 2023).
• Employed multivariable logistic regression on derivation (2019-2021) and validation (2022-2023) cohorts to build the risk model.
• Assessed model performance using C-statistics (AUC), Brier scores, and calibration plots; calculated observed-to-expected (O/E) ratios for centers.

Main Results:

• The study included 11,342 patients, with 90-day mortality at 8.0% overall (9.2% derivation, 7.4% validation).
• The derived logistic regression model demonstrated similar discrimination (AUC 0.714 and 0.707) and calibration (Brier score 0.08 vs. 0.07) in both cohorts.
• Observed-to-expected analysis identified 22 centers with significantly worse and 14 centers with significantly better than expected mortality.

Conclusions:

• The STS Intermacs risk model exhibits satisfactory discrimination and calibration for predicting 90-day mortality after LVAD implantation.
• This validated tool can inform patient selection for LVADs.
• The model serves as a valuable resource for quality improvement and assessing provider performance in LVAD centers.