Scholarlink

Background <p>Based on machine learning prediction models, we explored the anemia treatment attainment of patients on maintenance hemodialysis (MHD) and identified important factors for personalized treatment of patients.</p> Methods <p>We collected clinical data from 222 patients on MHD at West China Fourth Hospital, Sichuan University. Multiple machine learning models were applied to analyze the predictors of renal anemia treatment outcomes. A comprehensive evaluation was made in terms of discrimination, calibration and clinical utility. The top five important predictors were weighted according to the importance ranking of the variables for each model.</p> Results <p>Among the seven prediction models constructed in this study, the support vector classification model showed relatively high specificity (0.914). Logistic regression achieved the highest AUC (0.713), while stacking performed better in terms of precision (0.871) and recall (0.710) with a low Brier score (0.087). The composite results of the seven models suggested that albumin, total cholesterol, transferrin saturation, high-density lipoprotein cholesterol, and C-reactive protein were the top five significant predictors of treatment outcomes in renal anemia.</p> Conclusions <p>The important predictive factors suggested by this study may provide useful reference information for guiding treatment of renal anemia, thus improving patient outcomes.</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

Predicting anemia treatment outcomes in maintenance hemodialysis patients using multiple machine learning models

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月13日

查看全文
  • Miaoshuang Chen,
  • Menglin Chen,
  • Tao Zhang,
  • Linshen Xie

摘要

Background

Based on machine learning prediction models, we explored the anemia treatment attainment of patients on maintenance hemodialysis (MHD) and identified important factors for personalized treatment of patients.

Methods

We collected clinical data from 222 patients on MHD at West China Fourth Hospital, Sichuan University. Multiple machine learning models were applied to analyze the predictors of renal anemia treatment outcomes. A comprehensive evaluation was made in terms of discrimination, calibration and clinical utility. The top five important predictors were weighted according to the importance ranking of the variables for each model.

Results

Among the seven prediction models constructed in this study, the support vector classification model showed relatively high specificity (0.914). Logistic regression achieved the highest AUC (0.713), while stacking performed better in terms of precision (0.871) and recall (0.710) with a low Brier score (0.087). The composite results of the seven models suggested that albumin, total cholesterol, transferrin saturation, high-density lipoprotein cholesterol, and C-reactive protein were the top five significant predictors of treatment outcomes in renal anemia.

Conclusions

The important predictive factors suggested by this study may provide useful reference information for guiding treatment of renal anemia, thus improving patient outcomes.