A Practical Clinical Score Predicting Respiratory Failure in COVID-19 Patients

Area of Science:

• Medical Informatics
• Critical Care Medicine
• Epidemiology

Background:

• The COVID-19 pandemic strained healthcare systems, challenging patient triage and flow management.
• Existing tools like the National Early Warning Score (NEWS) have limitations for triaging COVID-19 patients.
• There is a need for improved tools to identify patients at risk of respiratory failure.

Purpose of the Study:

• To develop a practical, automated triage tool for COVID-19 patients.
• The tool aims to rapidly distinguish patients at high risk of respiratory failure.
• Utilizes readily available clinical data for efficient screening.

Main Methods:

• Analysis of electronic medical records from 385 COVID-19 patients.
• Utilized a big data platform for data extraction and exploration.
• Compared a novel model incorporating NEWS and clinical data against NEWS alone for predicting respiratory failure.

Main Results:

• The novel model demonstrated superior performance (AUC 0.92) compared to NEWS (AUC 0.63) in predicting respiratory failure.
• The model accurately identified patients requiring supplemental oxygen who were prone to respiratory failure (AUC 0.86).
• High sensitivity (0.81) and specificity (0.89) were achieved by the improved model.

Conclusions:

• A new clinical data-based model effectively predicts negative outcomes in COVID-19 patients.
• This tool shows excellent potential for initial screening and triage in emergency settings.
• The model's performance suggests it can aid in managing patient flow during surges.