Scholarlink

Purpose <p>This study was designed to evaluate the postoperative frailty status of patients with non-small cell lung cancer, identify influencing factors, establish a machine learning-based prediction model, and explore the correlation between frailty status at 3&#xa0;months and early recovery at 1 month postoperatively.</p> Methods <p>This retrospective analysis included patients with non-small cell lung cancer who underwent surgery at our hospital from 2021 to 2024. Clinical variables, including demographics, tumor characteristics, treatment, and laboratory tests, were analyzed. Feature selection and model construction were performed by using LASSO regression. Cross-validation assessed the accuracy of the models. Frailty at 3&#xa0;months and quality of recovery at 1&#xa0;month postoperatively were measured by using the Tilburg Frailty Index and Quality of Recovery (QoR-15) scales, respectively.</p> Results <p>A total of 1,013&#xa0;patients were included. The initial model achieved an AUC of 0.833, accuracy of 0.854, recall of 0.382, and F1 score of 0.502 in the training set, and an AUC of 0.786, accuracy of 0.857, recall of 0.242, and F1 score of 0.364 in the validation set. Of the patients, 190 (18.8%) developed frailty at 3&#xa0;months postoperatively. After applying Synthetic Minority oversampling Technique to balance the data, the model’s performance improved (area under the curve [AUC] 0.850, accuracy 0.791, recall 0.818, and F1 score 0.795 for the training set; AUC 0.819, accuracy 0.778, recall 0.762, and F1 score 0.781 for the test set). Additionally, we developed a nomogram to visually represent the predictive model, enabling clinicians to easily assess frailty risk in individuals based on key factors. Correlation analyses showed that frailty at 3&#xa0;months was moderately negatively correlated with early recovery at 1&#xa0;month (correlation coefficient = − 0.370).</p> Conclusions <p>This study developed a predictive model of postsurgical frailty in lung cancer, providing insights into personalized patient management and early recovery improvement. Further studies should explore the clinical application of the model.</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

Development and Validation of a Machine Learning-Based Predictive Model for Postoperative Frailty in Patients with Non-Small Cell Lung Cancer and Its Relation to Early Recovery

  • 期刊论文

  • 发表日期：

    2025年5月14日

查看全文
  • Xue-e Su,
  • Cui-liu Lin,
  • Huai-gang Wang,
  • Jing-Liu,
  • Cheng-bao Peng,
  • He-fan He,
  • Shanhu Wu,
  • Xu-feng Huang,
  • Shu Lin,
  • Bao-yuan Xie

摘要

Purpose

This study was designed to evaluate the postoperative frailty status of patients with non-small cell lung cancer, identify influencing factors, establish a machine learning-based prediction model, and explore the correlation between frailty status at 3 months and early recovery at 1 month postoperatively.

Methods

This retrospective analysis included patients with non-small cell lung cancer who underwent surgery at our hospital from 2021 to 2024. Clinical variables, including demographics, tumor characteristics, treatment, and laboratory tests, were analyzed. Feature selection and model construction were performed by using LASSO regression. Cross-validation assessed the accuracy of the models. Frailty at 3 months and quality of recovery at 1 month postoperatively were measured by using the Tilburg Frailty Index and Quality of Recovery (QoR-15) scales, respectively.

Results

A total of 1,013 patients were included. The initial model achieved an AUC of 0.833, accuracy of 0.854, recall of 0.382, and F1 score of 0.502 in the training set, and an AUC of 0.786, accuracy of 0.857, recall of 0.242, and F1 score of 0.364 in the validation set. Of the patients, 190 (18.8%) developed frailty at 3 months postoperatively. After applying Synthetic Minority oversampling Technique to balance the data, the model’s performance improved (area under the curve [AUC] 0.850, accuracy 0.791, recall 0.818, and F1 score 0.795 for the training set; AUC 0.819, accuracy 0.778, recall 0.762, and F1 score 0.781 for the test set). Additionally, we developed a nomogram to visually represent the predictive model, enabling clinicians to easily assess frailty risk in individuals based on key factors. Correlation analyses showed that frailty at 3 months was moderately negatively correlated with early recovery at 1 month (correlation coefficient = − 0.370).

Conclusions

This study developed a predictive model of postsurgical frailty in lung cancer, providing insights into personalized patient management and early recovery improvement. Further studies should explore the clinical application of the model.