ECRAID-Base POS-VAP

A clinically actionable risk score for ventilator-associated pneumonia using interpretable machine learning

Published in

  • CHEST Critical Care

Authors

  • Yeleen Fromage
  • Ana Catalina Hernandez Padilla
  • Julien Vaidie
  • Hamza Sayadi
  • Aleksandra Barac
  • Olaf L. Cremer
  • Denis Garot
  • Anahita Rouzé
  • Thomas Daix
  • L.E.M Vintcent
  • Holly Jackson
  • Marlieke E.A. De Kraker
  • C.H. van Werkhoven
  • Jean-Baptiste Woillard
  • Bruno François
  • POS-VAP Study Group and the ECRAID-Base Consortium

Abstract

Background

Ventilator-associated pneumonia (VAP) remains the most common Intensive Care Unit (ICU)-acquired infection, with high morbidity and substantial mortality. Diagnosis is often delayed, reducing the potential benefit of timely treatment.

Research question

Can a machine learning-based VAP Risk Score (VRS) enable dynamic daily risk prediction of VAP in patients ventilated for ≥48 hours?

Study design and Methods

Data were obtained from the prospective POS-VAP cohort, which enrolled 3571 adults under mechanical ventilation ≥48 hours across 25 European ICUs between 2022 and 2024. VAP was defined according to FDA criteria. Thirty-eight candidate predictors were selected based on clinical relevance. Several machine learning algorithms were trained, and a reduced model was used to derive a continuous VRS ranging from 0 to 100. This score was stratified into three risk categories (low, medium, high), each associated with increasing VAP prevalence. Sensitivity analysis included the use of physician-diagnosed VAP and evaluation of the predictive performance. External validation was done with data from four geographically distinct ICUs.

Results

The reduced model used five clinical variables collected at intubation (SOFA, APACHE II, Charlson index, age and number of VAP prevention strategies applied) and the ongoing duration of MV, enabling a daily updated VRS. In the external validation cohort (n=556), VAP prevalence was 5.5% [95%CI: 3.4%–8.7%] in the low-risk (n=18/326), 16.8% [11.9%–23.0%] in the medium-risk (n=32/191), and 61.5% [44.7%–76.2%] in the high-risk VRS category (n=24/39). Likelihood ratios for VAP were 0.4, 1.3, and 10.4 respectively, confirming meaningful separation. The VRS achieved a precision-recall area under the curve of 0.75 [0.71–0.79]. Similar results were observed for sensitivity analysis.

Interpretation

The VRS allows daily stratification of VAP risk in patients ventilated ≥48h, using six clinical variables. It provides a reproducible and externally validated tool that may help clinical decision-making.

Clinical trial registration

NCT05719259