Risk factors for severe COVID-19 and development of a predictive model

Area of Science:

• Clinical Medicine
• Epidemiology
• Biostatistics

Background:

• Severe COVID-19 poses a significant global health challenge.
• Accurate prediction of disease severity is crucial for timely intervention.
• Existing models may lack simplicity or broad applicability.

Purpose of the Study:

• To identify independent risk factors for severe or critical COVID-19.
• To develop and validate a predictive clinical risk model for COVID-19 severity.
• To provide a tool for dynamic risk assessment in COVID-19 patients.

Main Methods:

• A case-control study design involving 410 COVID-19 patients (132 severe/critical).
• Variable selection using LASSO regression and cross-validation.
• Multifactorial logistic regression and nomogram construction for the predictive model.
• Model performance evaluation via ROC, calibration, and decision curve analysis.

Main Results:

• Five key variables identified: respiratory rate, systolic blood pressure, plasma albumin, lactate dehydrogenase, and C-reactive protein.
• The developed clinical prediction model demonstrated good predictive accuracy (AUC = 0.86).
• The nomogram effectively integrates these indicators for risk assessment.

Conclusions:

• A simple, clinically applicable prediction model for severe COVID-19 was successfully developed.
• The model aids in identifying patients at high risk for severe or critical illness.
• This tool can support dynamic assessment and management of COVID-19 patients.