Accuracy of the Traditional COVID-19 Phone Triaging System and Phone Triage-Driven Deep Learning Model

Area of Science:

• Medical Informatics
• Epidemiology
• Artificial Intelligence in Healthcare

Background:

• Effective phone triage is crucial for managing COVID-19 patients and hospital resources.
• Assessing the accuracy of traditional phone triage and AI models is vital for pandemic response.

Purpose of the Study:

• To evaluate the accuracy of a traditional phone-triage system.
• To assess a phone triage-driven deep learning model for predicting COVID-19 positivity.

Main Methods:

• Retrospective study of 943 suspected COVID-19 patients.
• Assessed traditional phone triage accuracy.
• Developed and evaluated a deep learning model for automated classification.

Main Results:

• Myalgia, fever, and respiratory symptom contact showed high sensitivity for COVID-19.
• Immunodeficiency, smoking, and loss of smell/taste had high specificity.
• Phone triage PPV was 48.4%; deep learning model achieved 66% accuracy and 70.5% PPV.

Conclusions:

• Phone triage and deep learning models are feasible for COVID-19 screening.
• Deep learning models can improve early medical care for high-risk patients.
• AI-driven screening aids efficient allocation of healthcare resources.