Related Experiment Videos
Development and Prospective Validation of a Machine Learning Tool for Predicting Multidrug-resistant Organism
Nadheem M Shajeef1,2, Teresa M Sobi1,2, Chithra Jayaprakash3
1Department of Pharmacy Practice, Nirmala College of Pharmacy, Muvattupuzha, Kerala, India.
Summary
Machine learning effectively predicts multidrug-resistant organism (MDRO) infection risk in critical care units (CCUs). This tool aids early intervention and antimicrobial stewardship, improving patient outcomes and resource management.
Area of Science:
- Critical care medicine
- Infectious diseases
- Health informatics
Background:
- Multidrug-resistant organisms (MDROs) pose a significant threat in critical care settings, leading to increased morbidity, mortality, and healthcare costs.
- Irrational antibiotic use, invasive procedures, and diagnostic delays contribute to MDRO prevalence.
- Early identification of patients at risk for MDRO infections is crucial for effective management.
Purpose of the Study:
- To develop and pilot validate a machine learning (ML)-based tool for predicting MDRO infection risk in Critical Care Unit (CCU) patients.
- To utilize baseline clinical data for early risk stratification.
- To assess the tool's real-world predictive performance.
Main Methods:
- Retrospective analysis of 323 CCU patients' data, including demographics, comorbidities, and medical device usage.
- Development and evaluation of five ML models using Python.
- Pilot prospective validation on 49 new CCU admissions within 48 hours of admission.
Main Results:
- 76 out of 323 patients (23.5%) developed MDRO infections.
- Random forest model showed 96.0% training accuracy and 75.3% test accuracy.
- Prospective validation achieved 81.8% sensitivity, 100% specificity, and 95.9% overall accuracy.
Conclusions:
- The ML tool effectively stratifies MDRO risk in CCU patients.
- Despite potential overfitting, high prospective sensitivity and specificity support its use as a bedside aid for antimicrobial stewardship (AMS).
- Early risk stratification facilitates timely infection control and optimized resource allocation.
