Scholarlink

Purpose <p>In recent years, cervical spine surgeries performed in the elderly population have led to some postoperative complications after cervical spine procedures, particularly infections. While current research focuses predominantly on surgical site infections (SSIs), other types, including respiratory tract infections (RTIs) and urinary tract infections (UTIs), remain understudied. Additionally, comprehensive multiclass machine learning models for predicting diverse postoperative infections are lacking. This study aims to develop an optimal ML-based predictive model for identifying multiple infection risks in elderly patients undergoing cervical spine procedures.</p> Methods <p>This retrospective study included patients aged 60 years and older who underwent open cervical spine surgery at our institution between March 2011 and January 2024. We developed six ML algorithms [logistic regression, Gaussian naive Bayes, decision tree, random forest, K-nearest neighbors, and Light Gradient Boosting Machine (LightGBM)]. These algorithms were comprehensively evaluated in terms of accuracy, precision, recall, F1-score, and AUC.</p> Results <p>This study included a total of 1,151 patients, among whom 40 were diagnosed with postoperative SSIs, 30 with RTIs, and 35 with UTIs. In developing the ML models, we incorporated 54 variables, and among the various algorithms evaluated, LightGBM demonstrated the best performance in terms of accuracy, precision, recall, F1-score, and AUC for predicting different classes of outcomes. SHAP analysis highlighted the significant factors that were utilized to predict various types of infections, including actual days of hospitalization, absolute monocyte count, total cholesterol level, glucose level, APTT ratio, albumin concentration, AST, and so forth.</p> Conclusions <p>ML algorithms demonstrated exceptional performance in predicting different infections that may arise following cervical spine surgery in elderly patients. These algorithms successfully identified the key predictors of postoperative infections, enabling clinicians to effectively identify patients who are at heightened risk of developing infections after surgery in clinical practice.</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

Multiclass classification of infections after cervical spine surgery in the elderly: a machine learning approach based on preoperative and perioperative data

  • 期刊论文

  • 发表日期：

    2026年1月31日

查看全文
  • Yuezhang Ying,
  • Ziwei Fan,
  • Lequn Fu,
  • Chenyu He,
  • Zhangheng Huang,
  • Honglin Teng

摘要

Purpose

In recent years, cervical spine surgeries performed in the elderly population have led to some postoperative complications after cervical spine procedures, particularly infections. While current research focuses predominantly on surgical site infections (SSIs), other types, including respiratory tract infections (RTIs) and urinary tract infections (UTIs), remain understudied. Additionally, comprehensive multiclass machine learning models for predicting diverse postoperative infections are lacking. This study aims to develop an optimal ML-based predictive model for identifying multiple infection risks in elderly patients undergoing cervical spine procedures.

Methods

This retrospective study included patients aged 60 years and older who underwent open cervical spine surgery at our institution between March 2011 and January 2024. We developed six ML algorithms [logistic regression, Gaussian naive Bayes, decision tree, random forest, K-nearest neighbors, and Light Gradient Boosting Machine (LightGBM)]. These algorithms were comprehensively evaluated in terms of accuracy, precision, recall, F1-score, and AUC.

Results

This study included a total of 1,151 patients, among whom 40 were diagnosed with postoperative SSIs, 30 with RTIs, and 35 with UTIs. In developing the ML models, we incorporated 54 variables, and among the various algorithms evaluated, LightGBM demonstrated the best performance in terms of accuracy, precision, recall, F1-score, and AUC for predicting different classes of outcomes. SHAP analysis highlighted the significant factors that were utilized to predict various types of infections, including actual days of hospitalization, absolute monocyte count, total cholesterol level, glucose level, APTT ratio, albumin concentration, AST, and so forth.

Conclusions

ML algorithms demonstrated exceptional performance in predicting different infections that may arise following cervical spine surgery in elderly patients. These algorithms successfully identified the key predictors of postoperative infections, enabling clinicians to effectively identify patients who are at heightened risk of developing infections after surgery in clinical practice.