A predictive model for intracranial hemorrhage in adult patients receiving extracorporeal membrane oxygenation.

Abstract

Background: Intracranial hemorrhage (ICH), a severe complication among adults receiving extracorporeal membrane oxygenation (ECMO), is often related to poor outcomes. This study aimed to establish a predictive model for ICH in adults receiving ECMO treatment.

Methods: Adults who received ECMO between January 2017 and June 2022 were the subjects of a single-center retrospective study. Patients under the age of 18 years old, with acute ICH before ECMO, with less than 24 h of ECMO support, and with incomplete data were excluded. ICH was diagnosed by a head computed tomography scan. The outcomes included the incidence of ICH, in-hosptial mortality and 28-day mortality. Multivariate logistic regression analysis was used to identify relevant risk factors of ICH, and a predictive model of ICH with a nomogram was constructed.

Results: Among the 227 patients included, 22 developed ICH during ECMO. Patients with ICH had higher in-hospital mortality (90.9% vs. 47.8%, P=0.001) and higher 28-day mortality (81.8% vs. 47.3%, P=0.001) than patients with non-ICH. ICH was associated with decreased grey-white-matter ratio (GWR) (OR=0.894, 95%CI: 0.841-0.951, P<0.001), stroke history (OR=4.265, 95%CI: 1.052-17.291, P=0.042), fresh frozen plasma (FFP) transfusion (OR=1.208, 95%CI: 1.037-1.408, P=0.015) and minimum platelet (PLT) count during ECMO support (OR=0.977, 95%CI: 0.958-0.996, P=0.019). The area under the receiver operating characteristic curve of the ICH predictive model was 0.843 (95%CI: 0.762-0.924, P<0.001).

Conclusion: ECMO-treated patients with ICH had a higher risk of death. GWR, stroke history, FFP transfusion, and the minimum PLT count were independently associated with ICH, and the ICH predictive model showed that these parameters performed well as diagnostic tools.