Related Experiment Video
Updated: Apr 11, 2026
Rapid and Specific Detection of Acinetobacter baumannii Infections Using a Recombinase Polymerase Amplification/Cas12a-based System
Published on: April 25, 2025
Development and validation of an interpretable machine learning-based model for predicting carbapenem-resistant
Yan Gao1, Guangxin Gu2,3, Ruiwen Wang2,3
1Department of Disease Prevention and Control, General Hospital of Northern Theater Command, Shenyang, China.
Background:
Carbapenem-resistant Acinetobacter baumannii (CRAB) is a major cause of healthcare-associated infections and is associated with poor outcomes in intensive care units, particularly among postoperative patients. However, predictive tools for early identification of high-risk postoperative intensive care units (ICU) patients remain scarce.
Methods:
We conducted a retrospective cohort study including 2,195 postoperative ICU patients. Clinically available demographic, treatment-related, and laboratory variables were used to develop eight machine learning models. Feature selection was performed using Boruta, and model interpretability was enhanced using Shapley Additive Explanations (SHAP) analysis. Model performance was evaluated in an independent test set using the area under the receiver operating characteristic curve (AUC), with sensitivity analyses performed using reduced feature sets.
Results:
Among 2,195 postoperative ICU patients, 694 (31.6%) developed CRAB infection. Patients with CRAB infection had significantly longer ICU stays, greater exposure to invasive procedures, higher antimicrobial use, and worse laboratory profiles than non-infected patients. Using 19 features selected by the Boruta algorithm, all eight machine learning models achieved good discrimination in the test set (AUC > 0.83). Gradient Boosting demonstrated the best overall performance, with an AUC of 0.867 (95% CI: 0.836-0.892), good calibration, and the highest net clinical benefit. SHAP analysis identified duration of mechanical ventilation, central venous catheterization, ICU length of stay (LOS), and carbapenem exposure as the most influential predictors. Sensitivity analyses showed that models using only the top 10 or top 5 SHAP-ranked features achieved performance comparable to the full model, supporting the feasibility of feature reduction for clinical application.
Conclusions:
This study provides an interpretable and clinically applicable framework for early risk assessment of CRAB infection in postoperative ICU patients, supporting targeted prevention strategies and more rational antimicrobial stewardship.
Insights
Machine learning models can now predict carbapenem-resistant Acinetobacter baumannii (CRAB) infections in postoperative ICU patients. Key predictors include ventilation duration and central line use, enabling targeted prevention and antimicrobial stewardship.
Area of Science:
- Infectious Diseases
- Critical Care Medicine
- Machine Learning in Healthcare
Background:
- Carbapenem-resistant Acinetobacter baumannii (CRAB) poses a significant threat in intensive care units (ICUs), especially for postoperative patients.
- High rates of CRAB infections are linked to prolonged ICU stays, invasive procedures, and increased antimicrobial use.
- Currently, limited tools exist for early identification of high-risk postoperative ICU patients susceptible to CRAB.
Purpose of the Study:
- To develop and validate machine learning models for early prediction of CRAB infection risk in postoperative ICU patients.
- To identify key clinical and laboratory predictors of CRAB infection in this vulnerable population.
- To create an interpretable and clinically applicable risk assessment framework.
Main Methods:
- A retrospective cohort study of 2,195 postoperative ICU patients was conducted.
- Eight machine learning models were developed using demographic, treatment, and laboratory data.
- Boruta algorithm was used for feature selection, and SHAP analysis for model interpretability.
Main Results:
- CRAB infection developed in 31.6% of patients, associated with worse outcomes.
- All eight models demonstrated good predictive performance (AUC > 0.83) in an independent test set.
- Gradient Boosting achieved the highest AUC (0.867), with duration of mechanical ventilation, central venous catheterization, ICU length of stay, and carbapenem exposure identified as key predictors by SHAP analysis.
Conclusions:
- An interpretable machine learning framework for CRAB infection risk assessment in postoperative ICU patients was developed.
- The findings support targeted prevention strategies and optimized antimicrobial stewardship.
- Feature reduction techniques demonstrated the feasibility of clinical application.