Unravelling the complex inflammatory landscape of COVID-19 infection: a pathway to biomarker identification in infection-associated delirium in the ICU

Authors: Tristan Born, Matthieu Perreau, Pierre-Paul Axisa, Craig Fenwick, Andrea Pinto, Nawfel Ben-Hamouda, Andrea O. Rossetti, Renaud Du Pasquier, Jean-Daniel Chiche & Raphaël Bernard-Valnet

A retrospective cohort study from Lausanne University Hospital investigated peripheral immune signatures associated with delirium in ICU patients with COVID-19–related ARDS. Among 62 ICU patients, those who developed delirium (n = 39) were compared with delirium-free patients (n = 23), alongside reference cohorts of non-ICU COVID-19 patients and healthy controls. High-dimensional immunophenotyping was performed using multiplex cytokine assays and mass cytometry.

Delirium was associated with elevated levels of CXCL1, CCL11, CXCL13, HGF, and VEGF-A, and reduced IL-1α, IL-21, and IL-22, along with increased exhausted B cells and reductions in activated and memory T- and B-cell subsets. Ridge regression models incorporating these markers achieved an AUC of 0.95 when supported by synthetic data generated using beta-variational autoencoders. The findings identify reproducible immune correlates of ICU delirium and demonstrate the utility of machine learning approaches for biomarker discovery and predictive modelling in critical illness.