Automated triage in pandemics: Support vector machine for efficient patient classification

We aimed to develop tools to ease the burden on medical staff while improving patient care. The central question is whether Support Vector Machines (SVMs), trained solely on patients' self-reported data (symptoms, demographics, comorbidities), can achieve triage accuracy comparable to or surpassing traditional score-based systems in emergency departments under pandemic conditions.

We propose an SVM-based approach to replicate COVID-19 triage using patient-provided information rather than laboratory tests. The system automates triage scoring, classifying patients into red, yellow, or green categories. A computer program records patient data and calculates scores, while physicians have the ability to override results based on their clinical judgment.

The system integrates clinical expertise, enhancing rather than replacing physicians' roles. It achieved over 98% accuracy in red-triage classification using less than 2.8% of training data.

This SVM-based triage model shows promise for improving emergency department processes, accuracy, and patient outcomes in high-demand situations.