Survival After Left Ventricular Assist Device: A Society of Thoracic Surgeons National Intermacs Database Risk Model...

Area of Science:

• Cardiovascular Surgery
• Biomedical Engineering
• Medical Device Technology

Background:

• Existing Left Ventricular Assist Device (LVAD) risk models lack current relevance due to device evolution.
• The study addresses the prognostic accuracy of LVAD risk models when incorporating data from diverse durable LVADs versus solely contemporary fully magnetic devices.

Purpose of the Study:

• To assess the prognostic accuracy of 3-year survival models for Left Ventricular Assist Devices (LVADs).
• To compare the performance of models using data exclusively from fully magnetic LVADs versus those including older durable LVADs.

Main Methods:

• Utilized The Society of Thoracic Surgeons (STS) National Intermacs Database (2013-2023) for primary continuous flow LVAD recipients.
• Stratified patients into three cohorts based on LVAD type and era: fully magnetic devices, newer mixed devices, and older mixed devices.
• Employed machine learning (Oblique Random Survival Forests) to compare three models predicting 3-year survival.

Main Results:

• Among fully magnetic LVAD patients, 20.5% experienced mortality within 3 years.
• All three predictive models demonstrated similar discrimination (AUCs ~0.70).
• The model using only fully magnetic LVAD data (Model 1) showed superior calibration (Brier score) compared to models including older devices, especially beyond the first year post-implantation.

Conclusions:

• Risk models incorporating data from only fully magnetic LVADs exhibit superior or comparable performance to those including older device data.
• Findings support the current practice of using fully magnetic LVAD patient data for risk modeling.
• The study provides a framework for evaluating LVAD risk models across different device generations.