Establishment and comparison of machine learning models for predicting occult infection after intramedullary nail internal fixation of intertrochanteric fractures
Acta Orthopaedica et Traumatologica Turcica · 18.06.2026 · 60 (3)
DOI: 10.5152/j.aott.2026.25569 ↗Kurz gefasst
Fachartikel in der Zeitschrift Acta Orthopaedica et Traumatologica Turcica, online erschienen am 18.06.2026. Verfasst von Zhongbing Liu und 4 weiteren Autor:innen.
Abstract (Originalsprache)
Objective
The study aimed to investigate the incidence and risk factors of occult infection following intramedullary nail fixation in patients with intertrochanteric fractures and to establish a reliable predictive model for clinical use.
Methods
This retrospective study included 1000 patients with intertrochanteric fractures who underwent internal fixation with either proximal femoral nail antirotation (PFNA) or InterTan nails between January 2020 and December 2025. Occult postoperative infection was diagnosed according to the criteria for fracture-related infection. Potential risk factors were first screened through univariate analysis, followed by feature selection using least absolute shrinkage and selection operator (LASSO) regression. Three predictive models logistic regression, random forest, and XGBoost—were constructed and compared.
Results
The incidence of occult infection was 5.0%. Least absolute shrinkage and selection operator regression identified 9 independent risk factors: diabetes mellitus (P = .001), osteoporosis (P < .001), type A3 fracture (P < .001), lateral wall damage (P < .001), poor reduction quality (P = .001), intraoperative glove perforation (P < .001), prolonged operation time (P < .001), elevated body mass index (BMI) (P = .008), and increased tip-apex distance (P < .001). Among the 3 predictive models, the XGBoost model demonstrated the best performance, with an area under the curve of 0.79, accuracy of 91%, sensitivity of 62%, and specificity of 95%. The infection rate increased progressively with the predicted probability, confirming the model’s strong discriminative ability.
Conclusion
Diabetes mellitus, osteoporosis, type A3 fracture, lateral wall damage, poor reduction quality, intraoperative glove perforation, prolonged operation time, elevated BMI, and increased tip-apex distance are significant predictors of postoperative occult infection in intertrochanteric fracture patients. The XGBoost model provides effective risk prediction and stratification, supporting early identification and prevention of occult infection. Cite this article as: Liu Z, Zhu R, Ge W, Han J, Zou G. Establishment and comparison of machine learning models for predicting occult infection after intramedullary nail internal fixation of intertrochanteric fractures. Acta Orthop Traumatol Turc., 2026; 60(3), 0569, doi: 10.5152/j.aott.2026.25569.
Zitieren
APA
Liu, Z., Zhu, R., Ge, W., Han, J., & Zou, G. (2026). Establishment and comparison of machine learning models for predicting occult infection after intramedullary nail internal fixation of intertrochanteric fractures. Acta Orthopaedica et Traumatologica Turcica, 60(3). https://doi.org/10.5152/j.aott.2026.25569
Vancouver
Liu Z, Zhu R, Ge W, Han J, Zou G. Establishment and comparison of machine learning models for predicting occult infection after intramedullary nail internal fixation of intertrochanteric fractures. Acta Orthopaedica et Traumatologica Turcica. 2026;60(3). doi:10.5152/j.aott.2026.25569
BibTeX
@article{Liu2026,
author = {Liu, Zhongbing and Zhu, Ruli and Ge, Wei and Han, Jie and Zou, Guoyou},
title = {Establishment and comparison of machine learning models for predicting occult infection after intramedullary nail internal fixation of intertrochanteric fractures},
journal = {Acta Orthopaedica et Traumatologica Turcica},
year = {2026},
volume = {60},
number = {3},
doi = {10.5152/j.aott.2026.25569}
}
EUVIMED