Scholarlink

Introduction <p>Vancomycin is a widely used antibiotic for the treatment of serious Gram-positive bacterial infections. However, its clinical utility is often limited by the risk of nephrotoxicity, typically reflected by abnormalities in serum creatinine levels, which may indicate the occurrence of acute kidney injury (AKI). Timely identification of patients at increased risk is essential for early intervention and improved clinical outcomes.</p> Aim <p>This study aimed to identify clinical risk factors associated with vancomycin-induced abnormalities in serum creatinine levels and to develop predictive models capable of identifying high-risk hospitalized patients during vancomycin therapy.</p> Method <p>We conducted a retrospective cohort study including 1,008 hospitalized patients who received vancomycin treatment between January 2018 and June 2022 at the First Affiliated Hospital of Shandong First Medical University. Patients were grouped based on the presence or absence of serum creatinine abnormalities, defined as an increase of ≥ 26.5&#xa0;μmol/L or ≥ 50% from baseline. Multivariate logistic regression was applied to identify independent risk factors. Five machine learning algorithms—logistic regression, random forest, support vector machine, extreme gradient boosting, and gradient boosting machine (GBM)—were trained and compared. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity.</p> Results <p>The incidence of serum creatinine abnormalities was 9.22%. Chronic kidney disease, respiratory failure, pancreatitis, pneumonia, and mechanical ventilation were identified as significant risk factors (all <i>p</i> &lt; 0.05). Among the models tested, the GBM algorithm showed the highest predictive performance with an AUC of 0.783, along with good balance between sensitivity and specificity. The final model was deployed as a freely accessible web-based prediction tool using the R Shiny framework.</p> Conclusion <p>Abnormalities in serum creatinine levels during vancomycin therapy remain a clinically significant concern, especially in patients with comorbidities or critical illness. The machine learning-based predictive model developed in this study offers a practical tool for individualized risk assessment, enabling early risk stratification and proactive management. Incorporating such tools into clinical workflows may enhance patient safety and optimize antibiotic use.</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

Clinical risk assessment of serum creatinine abnormalities during vancomycin therapy: a retrospective study using machine learning models

  • 期刊论文

  • 发表日期：

    2025年8月19日

查看全文
  • Yilei Yang,
  • Haiying Yan,
  • Xiangyue Wang,
  • Jiahui Lao,
  • Ruiqiu Zhang,
  • Zhaoyang Chen,
  • Shiyu Ma,
  • Yan Li,
  • Xiao Li

摘要

Introduction

Vancomycin is a widely used antibiotic for the treatment of serious Gram-positive bacterial infections. However, its clinical utility is often limited by the risk of nephrotoxicity, typically reflected by abnormalities in serum creatinine levels, which may indicate the occurrence of acute kidney injury (AKI). Timely identification of patients at increased risk is essential for early intervention and improved clinical outcomes.

Aim

This study aimed to identify clinical risk factors associated with vancomycin-induced abnormalities in serum creatinine levels and to develop predictive models capable of identifying high-risk hospitalized patients during vancomycin therapy.

Method

We conducted a retrospective cohort study including 1,008 hospitalized patients who received vancomycin treatment between January 2018 and June 2022 at the First Affiliated Hospital of Shandong First Medical University. Patients were grouped based on the presence or absence of serum creatinine abnormalities, defined as an increase of ≥ 26.5 μmol/L or ≥ 50% from baseline. Multivariate logistic regression was applied to identify independent risk factors. Five machine learning algorithms—logistic regression, random forest, support vector machine, extreme gradient boosting, and gradient boosting machine (GBM)—were trained and compared. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity.

Results

The incidence of serum creatinine abnormalities was 9.22%. Chronic kidney disease, respiratory failure, pancreatitis, pneumonia, and mechanical ventilation were identified as significant risk factors (all p < 0.05). Among the models tested, the GBM algorithm showed the highest predictive performance with an AUC of 0.783, along with good balance between sensitivity and specificity. The final model was deployed as a freely accessible web-based prediction tool using the R Shiny framework.

Conclusion

Abnormalities in serum creatinine levels during vancomycin therapy remain a clinically significant concern, especially in patients with comorbidities or critical illness. The machine learning-based predictive model developed in this study offers a practical tool for individualized risk assessment, enabling early risk stratification and proactive management. Incorporating such tools into clinical workflows may enhance patient safety and optimize antibiotic use.