PhD Scientific Days 2026

Poster Session 1.P - Cardiovascular Medicine and Research

Dynamic Changes in Predictors of Mortality Following Heart Transplantation Assessed by Machine Learning

Name of the presenter

Koritsánszky, Kinga Bianka

Institute/workplace of the presenter

Department of Anesthesiology and Intensive Therapy, Semmelweis University

Authors

Kinga Bianka Koritsánszky¹, Rita Szentgróti², Ádám Szijártó², Márton Tokodi², Alexandra Vereb², Andrea Kőszegi², Balázs Sax², Attila Kovács², Béla Merkely², Andrea Székely²
1: Department of Anesthesiology and Intensive Therapy, Semmelweis University
2: Heart and Vascular Center, Semmelweis University

Text of the abstract

Introduction: Orthotopic heart transplantation (OHT) remains the standard treatment
for end-stage heart failure, yet individualized prediction of postoperative mortality
remains difficult.

Aims: We aimed to develop interpretable machine learning models to
predict 30-day and 1-year mortality and to assess whether predictors differ between
early and later outcomes.

Methods: We analyzed 581 patients undergoing OHT between 2012 and 2024, with
30-day and 1-year mortality rates of 9.9% and 17.6%. Eighty-seven preoperative and
48 postoperative variables were evaluated. Random forest models were trained using
five-fold cross-validation. Model interpretability was assessed using SHapley Additive
exPlanations (SHAP).

Results: Using preoperative variables only, models achieved AUCs of 0.62 (95% CI
0.48–0.75) for 30-day and 0.67 (95% CI 0.56–0.78) for 1-year mortality. SHAP
analysis showed that early mortality was mainly associated with hepatic dysfunction,
inflammation, and hemodynamic instability, whereas longer-term risk was more
influenced by renal function, metabolic reserve, and frailty. Adding postoperative
variables markedly improved performance (AUC 0.98 [95% CI 0.97–0.99] and 0.86
[95% CI 0.80–0.92], respectively). Short-term predictions were dominated by hepatic
injury and circulatory instability, while 1-year risk reflected persistent hepatic and
renal dysfunction, metabolic resilience, and duration of circulatory support.

Conclusions: Random forest models integrating pre- and early postoperative data
enable accurate prediction of short- and mid-term mortality after OHT. Temporal
changes in key predictors highlight the importance of dynamic, data-driven risk
assessment in transplant care.

Funding: KKB was supported by the 2025-2.1.1-EKÖP-2025-00014 (EKÖP-2025-445) University Research Scholarship Programme of the Ministry for Culture and Innovation from the source of the National Research, Development and Innovation Fund.