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

Area of Science:

• Artificial Intelligence in Healthcare
• Machine Learning for Medical Triage
• Clinical Decision Support Systems

Background:

• Developing tools to alleviate medical staff burden and enhance patient care is crucial.
• Evaluating the efficacy of AI models against traditional triage systems is essential, especially during pandemics.
• The study addresses the need for efficient triage solutions in resource-strained emergency departments.

Purpose of the Study:

• To assess if Support Vector Machines (SVMs), trained on patient-reported data, can achieve triage accuracy comparable to or better than traditional score-based systems.
• To develop an automated triage system that utilizes patient symptoms, demographics, and comorbidities for classification.
• To investigate the potential of AI in improving emergency department workflow and patient outcomes during high-demand periods.

Main Methods:

• An SVM-based model was developed to automate COVID-19 triage, classifying patients into red, yellow, or green categories.
• The system processed patient-provided information, bypassing the need for laboratory tests for initial triage.
• Physician oversight was incorporated, allowing for clinical judgment to override automated scores.

Main Results:

• The SVM model demonstrated high accuracy, achieving over 98% in red-triage classification.
• The system required minimal training data, utilizing less than 2.8% of the dataset for effective classification.
• The AI tool was designed to augment, not replace, physicians' roles, integrating clinical expertise.

Conclusions:

• The SVM-based triage model shows significant potential for enhancing emergency department processes.
• The system offers improved accuracy and patient outcomes, particularly in high-demand scenarios like pandemics.
• This AI approach supports medical staff by providing efficient and reliable triage decision-making.