Development and External Validation of a Prognostic Prediction Model for Hospitalization in SARS-CoV-2-Infected...

Area of Science:

• Epidemiology
• Clinical Prediction Modeling
• Public Health

Background:

• Accurate prediction of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) hospitalization risk is crucial for outpatient triage and healthcare resource allocation.
• The study addresses the need for effective risk stratification tools during the COVID-19 pandemic, particularly during periods of variant predominance.

Purpose of the Study:

• To develop and externally validate a clinical prediction model for 30-day COVID-19 hospitalization in symptomatic outpatients.
• To identify key predictors for COVID-19 hospitalization risk.

Main Methods:

• Random forest and multivariable logistic regression models were developed using clinical variables, social determinants of health, seasonality, and air quality indices.
• External validation was performed using data from a separate healthcare network.
• Model performance was assessed using the area under the receiver operator characteristic curve (AUC) and decision curve analysis.

Main Results:

• A random forest model with clinical predictors showed an AUC of 0.83, comparable to a model including additional social and environmental factors.
• A parsimonious logistic regression model with three clinical predictors (respiratory rate, age, pulse oximetry) achieved an AUC of 0.79 internally and 0.88 externally.
• The 3-predictor model demonstrated robust calibration and clinical utility.

Conclusions:

• A simple, 3-predictor clinical model can effectively stratify hospitalization risk for SARS-CoV-2 positive outpatients.
• This model offers a practical tool for clinical decision-making and resource optimization during COVID-19 surges.
• The findings support the use of accessible clinical data for predicting severe COVID-19 outcomes.