Scholarlink

Background <p>Delirium frequently occurs as a severe complication among patients with acute pancreatitis (AP), contributing to extended hospital stays, higher mortality rates, and lasting cognitive deficits. The pathogenesis of delirium in this setting is strongly influenced by metabolic abnormalities, including disturbances in electrolyte balance and widespread inflammation. Although the albumin-corrected anion gap (ACAG) is a recognized indicator of metabolic dysfunction, its relevance to delirium in AP patients has not been adequately investigated.</p> Methods <p>This study utilized patient records from the MIMIC-IV database to investigate how ACAG relates to the onset of delirium in individuals with acute pancreatitis. Analytical approaches included the use of summary statistics, Kaplan-Meier survival analyses, receiver operating characteristic (ROC) curve evaluation, and both univariable and multivariable Cox proportional hazards models. To capture potential nonlinear effects, restricted cubic spline (RCS) modeling was implemented. Subgroup analyses were conducted to examine possible demographic and clinical effect modifiers. Additionally, several machine learning algorithms—such as the Random Forest—were employed to further evaluate the predictive power of ACAG.</p> Results <p>Elevated levels of ACAG were independently linked to an increased likelihood of developing delirium during both the 28-day hospitalization period and throughout the ICU stay. Results from the multivariable Cox proportional hazards analysis indicated that each incremental rise in ACAG was associated with a greater risk of delirium (hazard ratio: 1.06, 95% confidence interval: 1.02–1.10, <i>p</i> &lt; 0.001). The application of restricted cubic spline modeling verified the linear nature of this association. Among the machine learning models, the Random Forest achieved superior predictive accuracy (AUC = 0.81), and SHAP analysis highlighted ACAG as a primary determinant in model prediction.</p> Conclusions <p>The ACAG emerged as an independent predictor of delirium among individuals with acute pancreatitis, displaying a linear association with the risk of delirium onset. When compared to other commonly used biomarkers, ACAG exhibited enhanced predictive capacity for identifying patients at risk. These findings suggest that ACAG could serve as a practical clinical marker for the early detection and prompt management of delirium in this patient population.</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 between albumin-corrected anion gap and delirium in acute pancreatitis: insights from the MIMIC-IV database

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月5日

查看全文
  • Yuanshuo Ge,
  • Youran Ma,
  • Peng Lv,
  • Junhao Ren,
  • Zhe Wang,
  • Cheng Zhang

摘要

Background

Delirium frequently occurs as a severe complication among patients with acute pancreatitis (AP), contributing to extended hospital stays, higher mortality rates, and lasting cognitive deficits. The pathogenesis of delirium in this setting is strongly influenced by metabolic abnormalities, including disturbances in electrolyte balance and widespread inflammation. Although the albumin-corrected anion gap (ACAG) is a recognized indicator of metabolic dysfunction, its relevance to delirium in AP patients has not been adequately investigated.

Methods

This study utilized patient records from the MIMIC-IV database to investigate how ACAG relates to the onset of delirium in individuals with acute pancreatitis. Analytical approaches included the use of summary statistics, Kaplan-Meier survival analyses, receiver operating characteristic (ROC) curve evaluation, and both univariable and multivariable Cox proportional hazards models. To capture potential nonlinear effects, restricted cubic spline (RCS) modeling was implemented. Subgroup analyses were conducted to examine possible demographic and clinical effect modifiers. Additionally, several machine learning algorithms—such as the Random Forest—were employed to further evaluate the predictive power of ACAG.

Results

Elevated levels of ACAG were independently linked to an increased likelihood of developing delirium during both the 28-day hospitalization period and throughout the ICU stay. Results from the multivariable Cox proportional hazards analysis indicated that each incremental rise in ACAG was associated with a greater risk of delirium (hazard ratio: 1.06, 95% confidence interval: 1.02–1.10, p < 0.001). The application of restricted cubic spline modeling verified the linear nature of this association. Among the machine learning models, the Random Forest achieved superior predictive accuracy (AUC = 0.81), and SHAP analysis highlighted ACAG as a primary determinant in model prediction.

Conclusions

The ACAG emerged as an independent predictor of delirium among individuals with acute pancreatitis, displaying a linear association with the risk of delirium onset. When compared to other commonly used biomarkers, ACAG exhibited enhanced predictive capacity for identifying patients at risk. These findings suggest that ACAG could serve as a practical clinical marker for the early detection and prompt management of delirium in this patient population.