Sign in Subscribe
Machine learning

Machine Learning Algorithm for Surgical Site Infection Prediction Showcased by Peking University Researchers

Strategic Insights

— 2 min read

Researchers from Peking University recently conducted a network meta-analysis to compare the performance of various machine learning algorithms in predicting surgical site infections (SSI). The study, published in the journal Computers in Biology and Medicine, aimed to identify the best-performing algorithm for building SSI predictive models. The researchers found that models based solely on surgical type outperformed those without discrimination of surgical type, and that mixed-use of structured and textual data-based models outperformed models solely based on structured data. Notably, the study concluded that Boosted Classifiers may be the best algorithm in SSI prediction.

Key Takeaways:

  • The study compared the performance of 10 machine learning algorithms in predicting SSI, including Logistic Regression, Decision Trees, and Random Forest.
  • Models based solely on surgical type outperformed those without discrimination of surgical type, with a Relative Diagnostic Odds Ratio (RDOR) of 2.71 (95% CI: 1.25-5.90, P = 0.01).
  • Mixed-use of structured and textual data-based models outperformed models solely based on structured data, with an RDOR of 8.70 (95% CI: 3.65-20.75, P = 0.0001).
  • The study found that Boosted Classifiers may be the best algorithm in SSI prediction, with a Superiority Index (SI) of 1.42 (95% CI: 0.68-2.97).
  • The researchers analyzed 493 articles and identified 40 articles that met the inclusion criteria, with 10 algorithms and 84 SSI prediction models included in the review.
  • The study was conducted by researchers from Peking University, led by Xiaoyuan Bao, with additional authors Jiao Shan, Bin Wang, Yanbin Wang, Yan Wang, Meng Lv, Wei Huai, Yicheng Jin, Yixi Jin, Zexin Zhang, and Yulong Cao.

Statistics:

  • 493 articles were identified in the systematic search
  • 40 articles met the inclusion criteria, with 10 algorithms and 84 SSI prediction models included in the review
  • The study found a Relative Diagnostic Odds Ratio (RDOR) of 2.71 (95% CI: 1.25-5.90, P = 0.01) for models based solely on surgical type
  • The study found an RDOR of 8.70 (95% CI: 3.65-20.75, P = 0.0001) for mixed-use of structured and textual data-based models
  • The Superiority Index (SI) for Boosted Classifiers was 1.42 (95% CI: 0.68-2.97)

Sources:

  • The best machine learning algorithm for building surgical site infection predictive models: A systematic review and network meta-analysis. Computers in Biology and Medicine, 2025;192:110286.
  • Pergamon-elsevier Science Ltd, The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, England. (Elsevier - www.elsevier.com; Computers in Biology and Medicine - www.journals.elsevier.com/computers-in-biology-and-medicine/)
  • Xiaoyuan Bao, Medical Informatics Center, Institute of Advanced Clinical Medicine, Peking University, Beijing, People's Republic of China.

Read more