Identifying COVID-19 cases in outpatient settings

Area of Science:

• Infectious Diseases
• Epidemiology
• Medical Diagnostics

Background:

• Accurate case identification for COVID-19 is challenging, especially in outpatient settings.
• Prioritizing patients for testing requires effective tools to distinguish COVID-19 from other acute respiratory illnesses.
• Symptom presentation and its evolution over time are key factors in diagnosing infectious diseases.

Purpose of the Study:

• To develop and validate a predictive tool for classifying COVID-19 patients based on symptom profiles and illness day.
• To provide a framework that complements laboratory testing in outpatient COVID-19 diagnosis.
• To assess the accuracy and performance of a symptom-based classification model.

Main Methods:

• Utilized generalized multivariate logistic regression on data from 236 SARS-CoV-2 positive cases and 564 controls.
• Analyzed symptom profiles and their significance in relation to the day of illness.
• Employed leave-one-out cross-validation and receiver operating characteristic (ROC) curve analysis for model evaluation.

Main Results:

• Significant symptoms included abdominal pain, cough, diarrhea, fever, headache, muscle ache, runny nose, and sore throat, varying in importance by illness day.
• The baseline model achieved a sensitivity of 0.67 at a specificity threshold of 0.95.
• Leave-one-out cross-validation yielded a high area under the ROC curve of 0.92, indicating strong predictive performance.

Conclusions:

• A tool utilizing symptom profiles and illness day demonstrates good predictive performance for differentiating COVID-19 patients from controls.
• This symptom-based approach can serve as a valuable framework to support clinical decision-making and laboratory testing in outpatient care.
• The model's accuracy was confirmed through external validation, suggesting its potential utility in real-world scenarios.