Predictive modelling of transport decisions and resources optimisation in pre-hospital setting using machine learning...

Area of Science:

• Emergency Medicine
• Health Informatics
• Data Science in Healthcare

Background:

• Pre-hospital care systems face global challenges in resource management.
• Machine learning (ML) offers potential for analyzing complex data to improve patient care and resource optimization.
• Accurate prediction of patient transport needs is crucial for efficient emergency medical services.

Purpose of the Study:

• To predict which emergency medical service cases require patient transportation versus those that do not.
• To leverage machine learning techniques for efficient resource allocation in pre-hospital care.
• To enhance the quality and efficiency of emergency medical services through data-driven insights.

Main Methods:

• Utilized machine learning algorithms (Random Forest, SVM, XGBoost, AdaBoost) to predict patient transport decisions.
• Analyzed a dataset of 93,712 emergency calls from a Middle Eastern national pre-hospital emergency medical care provider.
• Incorporated demographic and clinical variables to improve the accuracy of predictive models using R programming language.

Main Results:

• All trained ML models, particularly XGBoost (83.1% accuracy), accurately predicted patient transportation decisions.
• Identified statistically significant patterns for targeted resource deployment.
• Achieved high specificity rates (RF: 97.96%, XGBoost: 95.39%), minimizing false positives for 'Transported' cases.

Conclusions:

• Machine learning algorithms demonstrate transformative potential for improving pre-hospital care quality in Qatar.
• High predictive accuracy offers actionable strategies for day/time-specific resource planning and patient triaging.
• Findings support data-driven quality improvement interventions and future operational strategies in pre-hospital emergency care.