EXPLAINABLE MACHINE LEARNING FOR CARDIOVASCULAR RISK PREDICTION IN KIDNEY TRANSPLANT RECIPIENTS

Certificate Output Instructions

For best output, select "Paper Size" as "A4" and "Margin" as "0" or "None".

To save or print to PDF, please select Print Destination > Save as PDF, enable Background Graphics under "More Settings", then click "Save".

Presented the abstract " "
(Abstract co-author(s): )

Back

E-Poster Presentation

During the congress, E-Posters will be accessible to all participants on the congress website 24/7, as well as in the E-poster stations in the congress center.

Preparing your E-Poster

Please review the E-Poster format requirements carefully when preparing your E-Poster. Should your E-Poster not meet the mentioned requirements, it may not be displayed as described above.

E-Poster Submission Deadline

Please prepare and upload your E-Poster no later than March 14, 2026 11.59PM CET. After this date, you will no longer be able to prepare and upload your E-poster and it will not be displayed and accessible on the congress website.

E-Poster Format Requirements

PDF file
Layout: Portrait (vertical orientation)
One page only (Dim A4: 210 x 297mm or PPT)
E-Poster can be prepared in PowerPoint (one (1) PowerPoint slide) but must be saved and submitted as PDF file.
File Size: Maximum file size is 2 Megabytes (2 MB)
No hyperlinks, animated images, animations, and slide transitions
Language: English
Include your abstract number
E-posters can include QR codes, tables and photos

E-Poster

Abstract Title *

EXPLAINABLE MACHINE LEARNING FOR CARDIOVASCULAR RISK PREDICTION IN KIDNEY TRANSPLANT RECIPIENTS

Please follow the instructions below to input your abstract title.

Abstract titles should be brief and reflect the content of the abstract.

The title will not be accepted if it exceeds 25 words.
Type in CAPITAL LETTERS.
Lowercase may be used for abbreviations only, for example, mRNA.

Co-author 1

Ömer Tarik Özyilmaz o.t.ozyilmaz@umcg.nl University of Groningen, University Medical Center Groningen Department of Internal Medicine, Division of Nephrology Groningen Netherlands * University of Groningen Bernoulli Institute for Mathematics, Computer Science, and Artificial Intelligence Groningen Netherlands

Co-author 2

Tamas Szili-Torok t.szili-torok@umcg.nl University of Groningen, University Medical Center Groningen Department of Internal Medicine, Division of Nephrology Groningen Netherlands -

Co-author 3

Uwe J.F. Tietge uwe.tietge@ki.se Karolinska Institutet Department of Laboratory Medicine, Division of Clinical Chemistry Stockholm Sweden - Karolinska University Hospital Clinical Chemistry, Karolinska University Laboratory Stockholm Sweden

Co-author 4

Matias Valdenegro-Toro m.a.valdenegro.toro@rug.nl University of Groningen Bernoulli Institute for Mathematics, Computer Science, and Artificial Intelligence Groningen Netherlands -

Co-author 5

Stephan J.L. Bakker s.j.l.bakker@umcg.nl University of Groningen, University Medical Center Groningen Department of Internal Medicine, Division of Nephrology Groningen Netherlands -

Co-author 6

Martin H. de Borst m.h.de.borst@umcg.nl University of Groningen, University Medical Center Groningen Department of Internal Medicine, Division of Nephrology Groningen Netherlands -

Co-author 7

Co-author 8

Co-author 9

Co-author 10

Co-author 11

Co-author 12

Co-author 13

Co-author 14

Co-author 15

Introduction

Cardiovascular disease (CVD) is a leading cause of morbidity and mortality in kidney transplant recipients (KTR). Current population-based CVD risk assessment modules have not proven useful in this specific population. The aim of the present study was to develop an explainable machine learning (ML) model to predict individualized CVD risk in KTR over time, providing interpretable survival curves as output to support clinical decision making.

Methods

Our prospective cohort included 523 KTR at the University Medical Center Groningen (The Netherlands) with a functioning graft >1 year post-transplantation, recruited between August 2001 and July 2003. The primary outcome was a composite of CVD events: cardiovascular death, non-fatal myocardial infarction and revascularization procedures (percutaneous transluminal coronary angioplasty or coronary artery bypass grafting). A random survival forest (RSF) model was trained to predict CVD event risk. Features were selected through forward selection and backward elimination. Model performance was evaluated using a 5-fold cross-validation procedure, with hyper-parameter tuning and feature selection to mitigate bias. The final results were measured by the time-dependent area under the receiver operating characteristic curve (td-AUROC) and Shapley additive explanations (SHAP) visualizations were included in the analyses for explainability.

Results

During a mean (± standard deviation) follow-up of 4.8 ± 1.3 years, 12.8% of KTR had a first CVD event. At baseline, mean age was 50.5 ± 12.2 years and 54% were males. Mean kidney function (CKD-EPI 2009 eGFR) was 47.0 ± 15.8 mL/min/1.73 m². Donors had a mean age of 37.2 ± 15.4 years and 53% were male. The RSF model achieved an average td-AUROC of 0.76 using features selected by backward elimination (Figure 1). SHAP analysis identified high sensitivity troponin T, N-terminal pro-brain natriuretic peptide, and hemoglobin A1c as the most important predictors of CVD event risk.

Figure 1: Time-dependent area under the ROC curve (td-AUROC) for five time-points and averaged over five validation folds, starting 1 year from baseline.

Conclusion

The newly developed explainable machine learning model accurately predicts CVD risk in KTR over 5 years from baseline. Individualized risk stratification, using models such as the one proposed, could help to guide and support early preventive interventions in KTR follow-up.

Kewords