Early Risk Stratification of COVID‐19 Associated Cytokine Storm Using a Four‐Cytokine Cumulative Score: A Retrospective Cohort Study

Cytokine storm (CS) is a major immunopathological driver of severe illness and death in COVID-19, yet practical tools for its early identification remain limited. This study aimed to identify early predictive biomarkers, characterize their lead times and develop a cumulative cytokine score for early risk stratification of COVID-19-associated CS. In this single-center retrospective cohort study, 295 hospitalized patients with COVID-19 were included. Admission cytokines and inflammatory markers were analyzed. Cox proportional hazards models were used to identify predictors of CS and in-hospital mortality. Receiver operating characteristic analysis was performed to determine optimal cut-offs, and the interval between biomarker elevation above the cut-off and CS onset was defined as the early warning interval. A four-cytokine cumulative score (0-4 points; IL-6, IL-8, IL-10, TNF-α) was constructed. Internal validation was performed using 2,000 bootstrap resamples. Cytokines increased approximately 4-5 days prior to CS onset, earlier than CRP and d-dimer. In the multivariable model integrating clinical variables, a cumulative cytokine score of 3-4 was independently associated with increased CS risk (HR = 4.078, 95% CI: 1.180-14.101, p = 0.026) and higher in-hospital mortality (HR = 22.026, 95% CI: 2.569-188.846, p = 0.005). The score demonstrated good discriminative performance for CS prediction (AUC = 0.795). A cumulative cytokine score integrating multiple inflammatory pathways may offer a theoretical basis for the early auxiliary diagnosis and risk stratification of COVID-19-associated CS.