Development and validation of nomogram models for severe and fatal COVID-19

Area of Science:

• Infectious Diseases
• Clinical Medicine
• Biostatistics

Background:

• COVID-19 presents escalating contagion and immune resistance, leading to increased severe cases and mortality.
• Effective risk stratification is crucial for timely clinical intervention and improved patient outcomes.
• Existing predictive tools require refinement to accurately identify high-risk COVID-19 patients.

Purpose of the Study:

• To develop and validate nomogram predictive models for severe and fatal outcomes in hospitalized COVID-19 patients.
• To enhance clinical management strategies by identifying patients at heightened risk.
• To reduce COVID-19 related morbidity and mortality through early risk assessment.

Main Methods:

• Retrospective analysis of 1600 COVID-19 patients, categorized into mild, severe, and fatal groups.
• Development of predictive models using univariate and multiple stepwise regression analyses on clinical and laboratory data.
• Validation of nomogram models using Receiver Operating Characteristic (ROC) curves, Hosmer-Lemeshow tests, and decision curve analysis.

Main Results:

• Nomogram incorporating age, neutrophil (NEU), lactate dehydrogenase (LDH), lymphocyte (LYM), and albumin (ALB) predicted severe COVID-19 (AUC=0.771).
• A separate nomogram for fatal outcomes identified history of cerebral infarction/cancer, LDH, and ALB as key factors (AUC=0.748).
• Both models demonstrated good discrimination and calibration for predicting severe and fatal COVID-19 cases.

Conclusions:

• Elevated age, NEU, LDH, and decreased LYM, ALB are risk factors for severe COVID-19.
• History of cerebral infarction/cancer, elevated LDH, and decreased ALB predict fatal outcomes in critically ill patients.
• Nomogram models offer valuable tools for early risk prediction, aiding timely interventions to reduce COVID-19 severity and mortality.