Established right heart failure risk scores showed limited-to-moderate discrimination after durable left ventricular assist device implantation, with AUC values ranging from 0.394 to 0.729, while an exploratory Firth-penalized model achieved an optimism-corrected AUC of 0.773 but remains hypothesis-generating and requires external validation before clinical implementation.
Moderate-severe right heart failure occurred in 20.4% of patients following durable LVAD implantation.
Among established preoperative right heart failure risk scores, discriminative performance was limited to moderate, with AUC values ranging from 0.394 (Utah) to 0.619 for most preoperative models.
The postoperative EUROMACS score, which incorporates cardiopulmonary bypass time, achieved the highest AUC of 0.729 among established scores.
An exploratory Firth-penalized logistic regression model including five variables showed an apparent AUC of 0.854, which decreased to an optimism-corrected AUC of 0.773 after bootstrap validation.
The exploratory Firth model identified a potentially informative combination of clinical status, hepatic laboratory markers, and pulmonary hemodynamics as predictors of right heart failure.
Perioperative right heart failure was identified as a severe complication after durable LVAD implantation for which accurate preoperative risk assessment is crucial.
This research examined how well existing medical scoring tools can predict which patients will develop right heart failure after receiving a left ventricular assist device (LVAD) — a mechanical pump implanted to help a failing heart. Right heart failure is a serious complication that occurs in roughly 1 in 5 LVAD recipients, and knowing who is at high risk before surgery could help doctors plan more intensive support strategies. The study looked back at 105 patients who received LVADs at a single center over about eight years and tested six established prediction scores against the actual outcomes. The results showed that most established scoring systems performed poorly to moderately at identifying which patients would go on to develop significant right heart failure, with accuracy scores (AUC values) ranging from as low as 0.394 — meaning one score was actually worse than random guessing — to 0.729 for a score that incorporates information only available during or after surgery. The researchers also tested a new experimental model using five factors: a patient's overall clinical status before surgery, two liver function blood tests, and two measures of pressure in the lung's blood vessels. This new model showed better apparent accuracy, but when tested using a statistical technique to correct for the tendency of models to look artificially good on the same data used to build them, performance dropped noticeably, suggesting the model may be over-fitted to this specific group of patients. This research suggests that current tools for predicting right heart failure before LVAD surgery are inadequate, leaving clinicians without reliable guidance for one of the procedure's most serious risks. The experimental five-variable model offers a potentially promising direction — combining information about liver health and lung blood pressure alongside overall patient condition — but the authors emphasize it is hypothesis-generating only and cannot be used clinically until it is tested in independent patient populations. Larger, multi-center studies with standardized outcome definitions are needed to develop and properly validate better prediction tools.
Sales M, Zayat R, Hatam N, Berg T, Tewarie L, Moza A, et al.. (2026). Predicting the right! Validation of right heart failure predictive risk models after primary durable ventricular assist device implantation.. PloS one. https://doi.org/10.1371/journal.pone.0357025