Background: Paracetamol overdose may cause hepatic encephalopathy (HE). This condition demands specialized care and, in some instances, liver transplantation evaluation. No model is available for predicting HE. We aimed to set up and validate a model for predicting the occurrence of HE in paracetamol overdose.
Methods: Prospectively, 161 patients with single-dose paracetamol overdose and no HE (defined as hepatic coma grade II or more) on admission were studied during a 26-month period. Patients admitted during the first 13-month period constituted a learning set to construct a model to predict the occurrence of HE. Patients admitted in the second 13-month period constituted the validation set. Serial biochemical variables (measured twice daily), the time line after the overdose, and demographic data were used for univariate testing, and significant factors were assessed in various multiple logistic regression analyses.
Results: Thirty-two patients (20%), 15 in the first period and 17 in the second, developed HE grade II. The best model (the highest chi-square) for HE included: log10 (hours from overdose to antidote treatment), log10 (plasma coagulation factors on admission), and platelet count hours from overdose (chi-square = 41.2, P < 0.00001). In the validation set 88% (confidence interval (CI), 64%-99%) of the patients who developed HE were correctly predicted by the constructed model, whereas 90% (CI, 79%-96%) of the patients in the non-HE group were correctly predicted.
Conclusions: The constructed model for predicting HE in paracetamol overdose proved sensitive and accurate in the validation set and should be valuable for transferring high-risk patients to a liver intensive care unit/transplantation facility.