Background and aims: There has been no reliable severity system based on the prognosis to guide therapeutic strategies for patients with pyrrolizidine alkaloid (PA)-induced hepatic sinusoidal obstruction syndrome (HSOS). We aimed to create a novel Drum Tower Severity Scoring (DTSS) system for these patients to guide therapy. Methods: 172 Patients with PA-HSOS who received supportive care and anticoagulation therapy in Nanjing Drum Tower Hospital from January 2008 to December 2020 were enrolled and analyzed retrospectively. These patients were randomized into a training or validation set in a 3:1 ratio. Next, we established and validated the newly developed DTSS system. Results: Analysis identified a predictive formula: logit (P) = 0.004 × aspartate aminotransferase (AST, U/L) + 0.019 × total bilirubin (TB, μmol/L) − 0.571 × fibrinogen (FIB, g/L) − 0.093 × peak portal vein velocity (PVV, cm/s) + 1.122. Next, we quantified the above variables to establish the DTSS system. For the training set, the area under the ROC curve (AUC) (n = 127) was 0.787 [95% confidence interval (CI) 0.706–0.868; p < 0.001]. With a lower cut-off value of 6.5, the sensitivity and negative predictive value for predicting no response to supportive care and anticoagulation therapy were 94.7% and 88.0%, respectively. When applying a high cut-off value of 10.5, the specificity was 92.9% and the positive predictive value was 78.3%. For the validation set, the system performed stable with an AUC of 0.808. Conclusions: The DTSS system can predict the outcome of supportive care and anticoagulation in PA-HSOS patients with satisfactory accuracy by evaluating severity, and may have potential significance for guiding therapy.