Scholarlink

Background <p>Delirium is a prevalent and severe complication among critically ill patients. The Stress Hyperglycemia Ratio (SHR), reflecting acute illness-induced hyperglycemia relative to chronic glycemic status, has emerged as a robust prognostic marker. However, its relationship with delirium remains not fully understood. This study aimed to evaluate the association between admission SHR and delirium incidence in critically ill patients.</p> Methods <p>This large-scale retrospective cohort study utilized data from the MIMIC-IV database, including 7,744 adult patients with a first-time ICU admission. The primary outcome was the incidence of delirium. We employed multivariable logistic regression and restricted cubic splines (RCS) to assess the relationship between SHR and delirium. A Boruta-based feature selection procedure and multiple machine-learning algorithms were additionally applied as exploratory analyses to examine the relative importance of SHR.</p> Results <p>The overall incidence of delirium was 26.0%. A significant, nonlinear J-shaped relationship was identified between SHR and delirium risk (P for nonlinearity = 0.005). After adjusting for multiple confounders, restricted cubic spline visualization and piecewise regression suggested a potential inflection at SHR 0.94 (OR = 1.88, 95% CI: 1.69–2.06; <i>P</i> &lt; 0.001). In exploratory machine-learning analyses, SHR was consistently retained as an important feature, and the association was robust across subgroup analyses.</p> Conclusion <p>Admission SHR was independently and non-linearly associated with delirium in critically ill patients. The observed inflection point (SHR 0.94) requires validation in an external independent cohort. SHR may complement existing delirium risk assessment methods, pending external validation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Association of stress hyperglycemia ratio with the incidence of delirium in critically ill patients: a retrospective cohort study with exploratory machine-learning analyses

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月7日

查看全文
  • Yuzeng Xu,
  • Xiaoyong Zhu,
  • Wenying Lu,
  • Shengjie Yao,
  • Lifeng Ni

摘要

Background

Delirium is a prevalent and severe complication among critically ill patients. The Stress Hyperglycemia Ratio (SHR), reflecting acute illness-induced hyperglycemia relative to chronic glycemic status, has emerged as a robust prognostic marker. However, its relationship with delirium remains not fully understood. This study aimed to evaluate the association between admission SHR and delirium incidence in critically ill patients.

Methods

This large-scale retrospective cohort study utilized data from the MIMIC-IV database, including 7,744 adult patients with a first-time ICU admission. The primary outcome was the incidence of delirium. We employed multivariable logistic regression and restricted cubic splines (RCS) to assess the relationship between SHR and delirium. A Boruta-based feature selection procedure and multiple machine-learning algorithms were additionally applied as exploratory analyses to examine the relative importance of SHR.

Results

The overall incidence of delirium was 26.0%. A significant, nonlinear J-shaped relationship was identified between SHR and delirium risk (P for nonlinearity = 0.005). After adjusting for multiple confounders, restricted cubic spline visualization and piecewise regression suggested a potential inflection at SHR 0.94 (OR = 1.88, 95% CI: 1.69–2.06; P < 0.001). In exploratory machine-learning analyses, SHR was consistently retained as an important feature, and the association was robust across subgroup analyses.

Conclusion

Admission SHR was independently and non-linearly associated with delirium in critically ill patients. The observed inflection point (SHR 0.94) requires validation in an external independent cohort. SHR may complement existing delirium risk assessment methods, pending external validation.