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

Post-COVID-19 condition (PCC) has gained traction currently in the post-pandemic era. To address this, we utilized a Bayesian directed acyclic graphic (DAG) model to develop a personalized composite risk score (CRS) for PCC, based on the tabular data derived from a comprehensive meta-analysis. Our risk assessment model incorporates 215 combinations of risk factors, including personal demographic and health-related profiles, across 41 studies involving over 860,000 COVID-19 cases. The CRS ranges from 0 to 500, categorizing patients into risk quartiles and estimating PCC probability across SARS-CoV-2 variants of concerns, including Wild/D614G/Alpha, Delta, and Omicron BA.1/BA.2. External validation demonstrated accurate predictions, though higher risk scores showed slight deviations, particularly in BA.5 Omicron subset. The risk assessment model is not only adaptable for incorporating new evidence as SARS-CoV-2 subvariants emerge but also very valuable in facilitating the optimal individualized medical care for PCC patients and prioritizing a spectrum of risk groups for early PCC diagnosis. Notably, the adaptability of Bayesian DAG model enhances PCC risk prediction, enabling data integration for evolving SARS-CoV-2 contexts and informing healthcare resource allocation for high-risk groups.