Development and External Validation of Clinical Features-based Machine Learning Models for Predicting COVID-19 in the...

Area of Science:

• Medical Informatics
• Epidemiology
• Machine Learning

Background:

• Timely diagnosis of emerging infectious diseases is critical for patient treatment and disease containment.
• Machine learning (ML) models were previously developed to predict SARS-CoV-2 infection using emergency department (ED) clinical data.
• External validation of these ML models in a new population is essential.

Purpose of the Study:

• To externally validate a previously developed ML approach for predicting COVID-19.
• To assess the model's performance in a distinct ED population.
• To evaluate the potential of ML for predicting emerging infectious diseases.

Main Methods:

• Retrospective data collection from US and international EDs for training and testing cohorts.
• Inclusion of clinical features, demographics, and triage information.
• Utilized gradient boosting, random forest, and extra trees classifiers for prediction.
• Evaluated model performance using the area under the receiver operating characteristic curve (AUC).

Main Results:

• 580 patients in the training cohort and 946 in the testing cohort.
• COVID-19 prevalence was 16.9% and 19.0% in the respective cohorts.
• All ML models demonstrated acceptable discrimination (AUC).
• Random forest achieved the highest AUC (0.785), outperforming gradient boosting and extra trees.

Conclusions:

• The study validates ML for COVID-19 prediction in the ED.
• ML models built on clinical features can predict emerging infectious diseases.
• This approach offers a promising solution for future diagnostic challenges in infectious disease outbreaks.