Association of the CONUT score with mortality in critically ill patients with acute ischaemic stroke: a machine learning study with external validation.

Authors

Ge Li, Wenshi Wei

Published in

Frontiers in neurology. Volume 17. Pages 1889258. Epub Jul 27, 2026.

Abstract

Malnutrition frequently occurs in patients with acute ischaemic stroke (AIS) and is associated with adverse clinical outcomes. Nevertheless, the prognostic utility of the Controlling Nutritional Status (CONUT) score in this population remains unclear. Therefore, we investigated the relationship between the CONUT score and mortality and developed a machine learning model for early risk prediction.
Data from 1,103 patients with AIS in the MIMIC-IV database were retrospectively analyzed. An independent cohort comprising 659 patients from a Chinese tertiary hospital was included in the study. Survival analyses, multivariable Cox models, and subgroup analyses were conducted to evaluate the relationship between CONUT score and both 30-day and 365-day all-cause mortality. Machine learning models for 30-day mortality prediction were developed after Boruta-based feature selection in the training cohort, and model performance was evaluated by discrimination, calibration, decision analysis, and SHAP interpretation.
Elevated CONUT scores were independently associated with higher risks. After multivariable adjustment, a 1-unit increase in CONUT score corresponded to a 14.1% rise in short-term mortality risk and a 15.9% rise in long-term mortality risk. Kaplan-Meier curves further showed progressively reduced survival among patients with poorer nutritional status, and the associations remained stable across subgroup analyses. Among the six machine learning approaches, the CatBoost model achieved the strongest discriminative ability, yielding AUC values of 0.794 and 0.761 in the internal and external validation cohorts. Decision curve and calibration analyses also supported the robustness of the model.
The CONUT score was independently associated with both short- and long-term mortality among critically ill AIS patients. The CatBoost-based prediction model showed favorable predictive performance in individualized risk assessment.

PMID:
42577279
Bibliographic data and abstract were imported from PubMed on 11 Aug 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.