PiCCO hemodynamic parameters in cardiogenic shock - prediction of LVEF, NT-proBNP and MACE based on XGBoost machine learning model

Aug 5, 2026

This study applies an Extreme Gradient Boosting (XGBoost) machine learning model to examine associations between pulse index continuous cardiac output (PiCCO) hemodynamic parameters and key prognostic outcomes in cardiogenic shock. The analysis targets prediction of left ventricular ejection fraction (LVEF), N-terminal pro-brain natriuretic peptide (NT-proBNP), and 30-day major adverse cardiovascular events (MACE). Multiclass ROC performance evaluates predictive ability, while SHAP value analysis quantifies the contribution of individual PiCCO parameters to model decisions and supports evidence for individualized risk assessment.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages11

Reading Time28 min

AccessFree to download

Introduction

Cardiogenic shock challenges and need for dynamic assessment

PiCCO monitoring and advantages

Methods

XGBoost modeling and evaluation strategy

ROC analysis and SHAP interpretability

Results

Predictive performance for LVEF and NT-proBNP

Feature contributions for LVEF/NT-proBNP and MACE

Discussion

Implications for prognostic prediction and individualized decision-making

OPEN ACCESS EDITED BY Ling Sun, Nanjing Medical University, China REVIEWED BY Weipeng Jiang, Albury-Wodonga Health, Australia Jiawen Chen, Max Planck Institute for Heart and Lung Research, Germany *CORRESPONDENCE Qi Zhang [zhangqnh@hotmail.com](zhangqnh@hotmail.com)[ ](zhangqnh@hotmail.com)Wei Guo [guowei70@163.com](guowei70@163.com) †These authors have contributed equally to this work RECEIVED 11 August 2025 ACCEPTED 29 September 2025 PUBLISHED 15 October 2025 CITATION You J, Wei T, Yu Y, Huang J, Sun Y, Guo W and Zhang Q (2025) PiCCO hemodynamic parameters in cardiogenic shock: prediction of LVEF, NT-proBNP and MACE based on XGBoost machine learning model. Front. Med. 12:1683425 . doi: 10.3389/fmed.2025.1683425 COPYRIGHT © 2025 You, Wei, Yu, Huang, Sun, Guo and Zhang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. TYPE Original Research PUBLISHED 15 October 2025 DOI 10.3389/fmed.2025.1683425 PiCCO hemodynamic parameters in cardiogenic shock: prediction of LVEF, NT-proBNP and MACE based on XGBoost machine learning model Jieyun You†, Tianwen Wei†, Yue Yu, Jing Huang, Yuxiao Sun, Wei Guo * and Qi Zhang * Department of Cardiology, Shanghai East Hospital, School of Medicine, Tongji University, Shanghai, China Introduction: This study used the Extreme Gradient Boosting (XGBoost) machine learning model to conduct an in-depth analysis of the potential relationship between pulse index continuous cardiac output (PiCCO) and multiple clinical prognostic indicators, including left ventricular ejection fraction (LVEF), N-terminal pro-brain natriuretic peptide (NT-proBNP) levels, and 30-day major adverse cardiovascular events (MACE), in patients with cardiogenic shock. The aim of this study was to investigate the predictive ability of PiCCO hemodynamic parameters and the relative contribution features based on the XGBoost model. Methods: Multi-class receiver operating characteristic (ROC) curves explored that the XGBoost prediction model performed extremely well about LVEF and NT-proBNP. Further SHapley Additive explanation (SHAP) value analysis revealed the contributions of different PiCCO hemodynamic parameters. Results: Features such as CI (cardiac index), CPI (cardiac power index), and SVRI (systemic vascular resistance index) showed significant positive effects on the prediction of LVEF and NT-proBNP. In terms of MACE, dPmax (index of the left ventricular contractility), CFI (cardiac function index), and GEDVI (global enddiastolic volume index) showed significant predictive value. Discussion: Overall, XGBoost machine learning model based on PiCCOhemodynamic parameters provide evidence that effectively predict key clinical prognostic indicators in the patients with cardiogenic shock. These results provide important theoretical basis for further individualized clinical decisionmaking in cardiogenic shock patients. KEYWORDS XGBoost machine learning model, pulse index continuous cardiac output, left ventricular ejection fraction, N-terminal pro-brain natriuretic peptide, major adverse cardiovascular events 1 Introduction Cardiogenic shock is a critical condition characterized by cardiac dysfunction resulting inan inadequate cardiac output ( 1–3). It is associated with substantial morbidity and mortality (2–4). Despite significant advances in etiological treatment, pharmacological therapy, and mechanical circulatory support, the diagnosis and treatment of cardiogenic shock continue to present formidable clinical challenges. Previous studies have mainly guided the management

FAQs

What does the study predict using PiCCO hemodynamic parameters?

The XGBoost model predicts left ventricular ejection fraction (LVEF), N-terminal pro-brain natriuretic peptide (NT-proBNP) levels, and 30-day major adverse cardiovascular events (MACE) in patients with cardiogenic shock.

How is the prediction model evaluated in this research?

Multiclass receiver operating characteristic (ROC) curves are used to assess how well the XGBoost model performs for LVEF and NT-proBNP, and SHAP value analysis is applied to interpret feature contributions.

Which PiCCO features show significant predictive effects according to the results?

For LVEF and NT-proBNP, CI, CPI, and SVRI show significant positive effects. For MACE, dPmax, CFI, and GEDVI show significant predictive value.

Aug 5, 2026

Upload by Genevieve