Taking ACTION: A Prognostic Tool for Pediatric Ventricular Assist Device Mortality

Katerina Boucek, Anaam Alzubi, Farhan Zafar, Matthew J. O'Connor, Mary Mehegan, Deepa Mokshagundam, Ryan R. Davies, Iki Adachi, Angela Lorts, David N. Rosenthal

Research output: Contribution to journalArticlepeer-review

1 Scopus citations

Abstract

We sought to develop a contemporary risk assessment tool for use in pediatric ventricular assist device (VAD) candidates to estimate risk for mortality on the device using readily available preimplantation clinical data. Training and testing datasets were created from Advanced Cardiac Therapies Improving Outcomes Network (ACTION) registry data on patients supported with a VAD from 2012 to 2021. Potential risk factors for mortality were assessed and incorporated into a simplified risk prediction model utilizing an open-source, gradient-boosted decision tree machine learning library, known as random forest. Predictive performance was assessed by the area under the receiver operating characteristic curve in the testing dataset. Nine significant risk factors were included in the final predictive model which demonstrated excellent discrimination with an area under the curve of 0.95. In addition to providing a framework for establishing pediatric-specific risk profiles, our model can help inform team expectations, guide optimal patient selection, and ultimately improve patient outcomes.

Original languageEnglish
Pages (from-to)602-609
Number of pages8
JournalASAIO Journal
Volume69
Issue number6
DOIs
StatePublished - Jun 1 2023

Keywords

  • congenital heart disease
  • machine learning
  • pediatrics
  • ventricular assist devices

Fingerprint

Dive into the research topics of 'Taking ACTION: A Prognostic Tool for Pediatric Ventricular Assist Device Mortality'. Together they form a unique fingerprint.

Cite this