CLABpredICU---AI-driven risk prediction for CLABSI in intensive care units based on clinical and biochemical parameters - 11/07/25
, Jutang Babat Ain Tiewsoh, MD b, Amarjeet Kumar, MD bRésumé |
Background |
Central line--associated bloodstream infections (CLABSI) are major causes of morbidity and mortality in intensive care units. This study aimed to develop an artificial intelligence-driven predictive model for CLABSI within 2 calendar days of central line insertion using routine biochemical parameters for early detection.
Methods |
A retrospective analysis of adult intensive care unit patients with central lines was conducted. Demographic and biochemical parameters were collected. Feature selection using Recursive Feature Elimination identified key predictors. Four models---Extreme Gradient Boosting (XGBoost), logistic regression, support vector machine, and random forest---were trained and validated.
Results |
Among 234 patients, 39 were CLABSI-positive. Support vector machine demonstrated the highest predictive power (area under receiver operating characteristic curve = 0.91) and diagnostic odds ratio = 45.34. Seven key predictors were identified: prothrombin time days 1 and 2, international normalized ratio day 2, sodium day 1, potassium day 2, neutrophil-to-lymphocyte ratio day 1, and urea/creatinine ratio day 2. Decision curve analysis showed an estimated risk stratification at a 23% cutoff “clabpredicu.netlify.app/”.
Conclusions |
The developed artificial intelligence model shows strong potential for early CLABSI prediction using routine blood parameters. Future studies should focus on external validation and broader clinical application to enhance early infection prevention, particularly in resource-limited settings.
Le texte complet de cet article est disponible en PDF.Highlights |
• | The Support Vector Machine model achieved the highest AUROC (0.91) and a diagnostic odds ratio of 45.34 for early prediction. |
• | Seven biochemical parameters, including prothrombin time (days 1 and 2), INR (day 2), sodium (day 1), potassium (day 2), NLR (day 1), and urea/creatinine ratio (day 2), were found significant. |
• | Decision curve analysis showed a positive net benefit (0.235), and a web-based risk calculator “clabpredicu.netlify.app/” enables real-time risk estimation. |
• | The model’s integration into electronic health records and bedside decision-support tools can enhance early intervention. |
Key Words : Central line--associated bloodstream infections, Machine learning model, Artificial intelligence, Early prediction, Intensive care unit
Plan
| Ethics approval: This study was approved by the Institutional Ethics Committee of All India Institute of Medical Sciences, Patna (Ref. No. AIIMS/Pat/IEC/UG-STS/MBBS 2021/Dec24/15). The requirement for informed consent was waived by the committee. All data were anonymized to maintain patient confidentiality, and the study was conducted in accordance with the ethical standards of the Declaration of Helsinki. |
|
| Conflicts of interest: None to report. |
Vol 53 - N° 8
P. 875-880 - août 2025 Retour au numéro
Article précédent
- Implementation of a secure electronic form for bloodborne pathogen exposure reporting associated with increased reports among healthcare workers: A quasi-experimental study
- Erin Bammann, Himgauri Nikrad, Deborah Aragon, Heather Young
- Ex vivo modeling of extraluminal migration of pathogenic microbes in the presence of a hemostatic agent
- Katelyn Gough, Tim R. Dargaville, Amanda Ullman, Nicole Marsh, Elise S. Pelzer
Bienvenue sur EM-consulte, la référence des professionnels de santé.
L’accès au texte intégral de cet article nécessite un abonnement.
Déjà abonné à cette revue ?