Scholarlink

Background <p>Diabetic foot ulcers (DFUs) represent a serious chronic complication of diabetes, with substantial morbidity and healthcare costs. Despite advances in treatment modalities such as silver ion hydrogel dressings combined with negative pressure wound therapy (NPWT), clinical outcomes remain highly heterogeneous. Identifying prognostic factors and developing accurate risk prediction tools are essential for personalized management and improved patient outcomes.</p> Objective <p>The objective of this study was to identify key prognostic factors in DFU patients undergoing combined silver ion hydrogel and NPWT therapy and to develop a validated risk prediction model.</p> Methods <p>This retrospective study analyzed 200 DFU patients admitted to our hospital between March 2023 and December 2024. Data were extracted from the electronic medical record system, and patients were stratified into good and poor prognosis groups according to clinical outcomes. After collection and comparison of clinical variables, two machine learning approaches—LASSO regression and random forest (RF)—were employed to identify risk factors using an “overlapping selection” strategy. Significant predictors were further assessed by multivariate logistic analysis. A predictive model was subsequently developed and validated using R software.</p> Results <p>A total of 200 DFU patients were included in the analysis. Based on electronic medical records, 49 patients (24.50%) were classified into the poor prognosis group, while 151 patients (75.50%) comprised the good prognosis group. Significant differences (<i>p</i> &lt; 0.05) were observed between the two groups in several clinical parameters: number of ulcers, ulcer area, interleukin-17 (IL-17) levels, Th1 cell count, Th2 cell count, procalcitonin (PCT), and C-reactive protein (CRP) levels. Using two machine learning methods—LASSO regression and the random forest (RF) algorithm—different sets of risk factors were screened. By applying an “overlap coverage” approach, five common risk factors for DFU prognosis were identified: IL-17, Th1 cells, Th2 cells, PCT, and CRP. These were included in the logistic regression model. The results indicated that IL-17, Th1 cells, PCT, and CRP were all risk factors for poor prognosis in DFU patients (OR = 1.275, 1.411, 1.006, and 1.101, respectively, <i>p</i> &lt; 0.05), while Th2 cells were a protective factor for poor prognosis in DFU patients (OR = 0.001, <i>p</i> &lt; 0.05). Based on the results of the logistic regression analysis, a risk prediction model for the prognosis of DFU patients was constructed using a nomogram. The ROC curve showed an AUC value of 0.937, with a 95% CI of 0.887 to 0.986. The calibration curve indicated that the model’s predicted results were well aligned with the actual prognosis of DFU patients. The Cox-Snell R<sup>2</sup> was 0.675, Nagelkerke R<sup>2</sup> 0.454, Brier Score 0.067, model fit <i>p</i>-value 0, and statistic 33.491. The clinical decision curve was generally above the two extreme curves, indicating that the factors included in the nomogram have a high net benefit for predicting the prognosis of DFU patients.</p> Conclusion <p>There are numerous factors influencing the prognosis of DFU patients, primarily including IL-17, Th1 cells, Th2 cells, PCT, and CRP. The risk prediction nomogram model constructed based on these factors demonstrates good predictive performance for the risk of adverse outcomes in DFU patients and shows potential for clinical application in identifying high-risk individuals for targeted management strategies. Clinicians should identify high-risk populations early and optimize treatment strategies accordingly to improve 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

Factors associated with the prognosis of diabetic foot ulcers treated with silver ion hydrogel dressings combined with negative pressure wound therapy and the construction of a risk prediction model: a study of two machine learning methods, LASSO and random forest algorithms

  • 期刊论文

  • 发表日期：

    2026年3月13日

查看全文
  • Sheng Ding,
  • Yin Gong,
  • Mingming Song,
  • Zaiwen Guo,
  • Ran Sun,
  • Yunfeng Liu,
  • Zili Ge

摘要

Background

Diabetic foot ulcers (DFUs) represent a serious chronic complication of diabetes, with substantial morbidity and healthcare costs. Despite advances in treatment modalities such as silver ion hydrogel dressings combined with negative pressure wound therapy (NPWT), clinical outcomes remain highly heterogeneous. Identifying prognostic factors and developing accurate risk prediction tools are essential for personalized management and improved patient outcomes.

Objective

The objective of this study was to identify key prognostic factors in DFU patients undergoing combined silver ion hydrogel and NPWT therapy and to develop a validated risk prediction model.

Methods

This retrospective study analyzed 200 DFU patients admitted to our hospital between March 2023 and December 2024. Data were extracted from the electronic medical record system, and patients were stratified into good and poor prognosis groups according to clinical outcomes. After collection and comparison of clinical variables, two machine learning approaches—LASSO regression and random forest (RF)—were employed to identify risk factors using an “overlapping selection” strategy. Significant predictors were further assessed by multivariate logistic analysis. A predictive model was subsequently developed and validated using R software.

Results

A total of 200 DFU patients were included in the analysis. Based on electronic medical records, 49 patients (24.50%) were classified into the poor prognosis group, while 151 patients (75.50%) comprised the good prognosis group. Significant differences (p < 0.05) were observed between the two groups in several clinical parameters: number of ulcers, ulcer area, interleukin-17 (IL-17) levels, Th1 cell count, Th2 cell count, procalcitonin (PCT), and C-reactive protein (CRP) levels. Using two machine learning methods—LASSO regression and the random forest (RF) algorithm—different sets of risk factors were screened. By applying an “overlap coverage” approach, five common risk factors for DFU prognosis were identified: IL-17, Th1 cells, Th2 cells, PCT, and CRP. These were included in the logistic regression model. The results indicated that IL-17, Th1 cells, PCT, and CRP were all risk factors for poor prognosis in DFU patients (OR = 1.275, 1.411, 1.006, and 1.101, respectively, p < 0.05), while Th2 cells were a protective factor for poor prognosis in DFU patients (OR = 0.001, p < 0.05). Based on the results of the logistic regression analysis, a risk prediction model for the prognosis of DFU patients was constructed using a nomogram. The ROC curve showed an AUC value of 0.937, with a 95% CI of 0.887 to 0.986. The calibration curve indicated that the model’s predicted results were well aligned with the actual prognosis of DFU patients. The Cox-Snell R2 was 0.675, Nagelkerke R2 0.454, Brier Score 0.067, model fit p-value 0, and statistic 33.491. The clinical decision curve was generally above the two extreme curves, indicating that the factors included in the nomogram have a high net benefit for predicting the prognosis of DFU patients.

Conclusion

There are numerous factors influencing the prognosis of DFU patients, primarily including IL-17, Th1 cells, Th2 cells, PCT, and CRP. The risk prediction nomogram model constructed based on these factors demonstrates good predictive performance for the risk of adverse outcomes in DFU patients and shows potential for clinical application in identifying high-risk individuals for targeted management strategies. Clinicians should identify high-risk populations early and optimize treatment strategies accordingly to improve outcomes.