Personalized risk score for post-COVID-19 condition: Bayesian directed acyclic graphic approach

Area of Science:

• Epidemiology
• Computational Biology
• Public Health

Background:

• Post-COVID-19 condition (PCC) is a growing concern in the post-pandemic landscape.
• Accurate risk assessment is crucial for managing PCC and its impact on individuals and healthcare systems.

Purpose of the Study:

• To develop a personalized composite risk score (CRS) for Post-COVID-19 condition (PCC).
• To estimate PCC probability across different SARS-CoV-2 variants using a Bayesian directed acyclic graphic (DAG) model.

Main Methods:

• Utilized a Bayesian DAG model on meta-analysis data from 41 studies (over 860,000 COVID-19 cases).
• Incorporated 215 combinations of demographic and health-related risk factors.
• Developed a CRS ranging from 0 to 500, categorized into risk quartiles.

Main Results:

• The CRS accurately predicted PCC risk across various SARS-CoV-2 variants (Wild/D614G/Alpha, Delta, Omicron BA.1/BA.2).
• External validation confirmed predictive accuracy, with minor deviations in the BA.5 Omicron subset.
• The model demonstrated adaptability for emerging SARS-CoV-2 subvariants.

Conclusions:

• The Bayesian DAG model provides an adaptable framework for PCC risk prediction.
• The CRS facilitates individualized medical care and prioritization for early PCC diagnosis.
• This approach supports informed healthcare resource allocation for high-risk populations.