open access publication

Article, 2023

Derivation and Validation of a Risk Prediction Model for Waitlist Mortality in Left Ventricular Assist Device Patients

In: The Journal of Heart and Lung Transplantation, ISSN 1053-2498, 1557-3117, Volume 42, 4, Page s237, 10.1016/j.healun.2023.02.536

Contributors (11)

Tibrewala, Anjan (Corresponding author) [1] Hu, M. [1] Petito, Lucia C (0000-0002-0156-1219) [1] Rich, J. [1] Pham, D. [1] De By, Theo M M H (0000-0001-8991-3348) [2] Gustafsson, Finn (0000-0003-2144-341X) [3] Veen, Kevin M (0000-0003-2806-8384) [4] Vanderheyden, Marc (0000-0003-3050-5221) [5] Lloyd-Jones, D. [1] Shah, S. [1]


  1. [1] Northwestern University
  2. [NORA names: United States; America, North; OECD]
  3. [2] Euromacs, S-Gravenhage, ZH, Netherlands
  4. [3] Rigshospitalet
  5. [NORA names: Capital Region of Denmark; Hospital; Denmark; Europe, EU; Nordic; OECD]
  6. [4] Erasmus MC
  7. [NORA names: Netherlands; Europe, EU; OECD]
  8. [5] Onze Lieve Vrouwziekenhuis Hospital
  9. [NORA names: Belgium; Europe, EU; OECD]


Purpose Left ventricular assist devices (LVAD) are often used as a bridge to heart transplant (HT). Given limited availability of donor organs, discrimination of risk of waitlist mortality is important to inform prioritization on the wait list. Thus, we sought to derive and validate a risk prediction model for waitlist mortality in LVAD patients. Methods Adult, continuous-flow, centrifugal durable LVAD patients listed or likely to be listed for HT in the Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) who were implanted between 2012-2017 and survived at least 3 months were included. The outcome was time to all-cause mortality. We considered 41 candidate predictors including demographics, clinical presentation, laboratory, and diagnostic values at 3-months post-implant. Univariate Fine-Gray models and four logistic regression techniques (logistic, LASSO, random forest, gradient boosting) were used for predictor selection. Variables were selected if their variable importance rank was 1-10 by at least 2 techniques. A final survival model was estimated with all selected predictors using the Fine-Gray method to account for competing risk of transplant. The model was validated in the same patient population from an independent cohort: the European Registry for Patients with Mechanical Circulatory Support (EUROMACS). Results The derivation cohort had 2364 LVAD patients (76% male, 62% white), of whom 394 (16.7%) died within 1 year. A risk prediction model for wait list mortality at 1 year was derived with area-under-the-curve (AUC) of 0.726 (Table 1). The validation cohort had 577 patients with 37 (6.4%) deaths within 1 year; the associated AUC was 0.626. Conclusion We derived and validated a risk prediction model for waitlist mortality in LVAD patients using two independent cohorts. As the transplant prioritization schemes evolve, studies like ours can provide additional data to inform waitlist prioritization for LVAD patients awaiting HT.


AUC, European Registry, Fine-Gray method, Fine-Gray model, Interagency Registry, LVAD patients, Mechanically Assisted Circulatory Support, Methods Adults, additional data, adults, area, assist device, associated AUC, availability, bridge, candidate predictors, cause mortality, circulatory support, clinical presentation, cohort, conclusion, curves, data, death, demographics, derivation, derivation cohort, device patients, devices, diagnostic value, discrimination, discrimination of risk, donor organs, evolve, final survival model, heart transplant, importance rank, independent cohort, laboratory, left ventricular assist device patients, limited availability, list, list mortality, logistic regression techniques, mechanical circulatory support, method, model, months, mortality, organs, ours, outcomes, patient population, patients, population, prediction model, predictor selection, predictors, presentation, prioritization, purpose, rank, registry, regression techniques, results, risk, risk of transplant, risk prediction model, same patient population, selection, study, support, survival models, technique, time, transplant, validation, validation cohort, values, variables, ventricular assist device, ventricular assist device patients, wait list, wait-list mortality, waitlist mortality, waitlist prioritization, years