Dynamic biomarkers and Cox regression with time-dependent covariate for mortality prediction in severe fever with thrombocytopenia syndrome

Severe fever with thrombocytopenia syndrome (SFTS) is a fatal tick-borne infectious disease that lacks effective treatments. Dynamic analysis that reflects changes in the SFTS patient’s condition is needed. This study aimed to evaluate the time-dependent predictive performance of key biomarkers using a time-dependent Cox regression model. A retrospective multicenter cohort study was conducted on 440 SFTS patients hospitalized in South Korea between 2013 and 2024. Time-dependent Cox regression and time-dependent receiver operating characteristic (ROC) analyses were applied to assess the prognostic value of Blood Urea Nitrogen (BUN), Prothrombin Time (PT), and Activated Partial Thromboplastin Time (aPTT). Missing data were handled using multiple imputation. aPTT consistently demonstrated high predictive accuracy (AUC > 0.90) throughout the disease course, indicating its sustained role in coagulopathy. PT exhibited strong early-stage predictive power (AUC = 0.86 on day 2) but declined over time, reflecting its utility for early monitoring. BUN showed a progressive increase in predictive performance (AUC = 0.70 on day 2 to AUC = 0.78 on day 8), supporting its relevance in later stages of disease progression. Non-survivors exhibited significantly higher levels of BUN, PT, and aPTT compared to survivors. This study demonstrates the utility of time-dependent analysis for evaluating dynamic biomarker changes in SFTS patients. aPTT is a robust predictor throughout the disease course, while PT is valuable for early-stage assessment and BUN for later-stage management. These findings suggest the importance of integrating dynamic biomarker monitoring into clinical decision-making to improve prognosis in SFTS patients.