STUDY OBJECTIVETo enable emergency physicians to make well-informed triage and treatment decisions, accurate tools to evaluate the severity of diquat poisoning are needed. This study establishes severity indices for diquat poisoning (SIDPs) in assessing the risk of death for patients with acute diquat poisoning for triage purposes and 28-day mortality.METHODSThis multicenter cohort study involved 204 patients. Predictors identified by the Burota algorithm and stepwise Cox regression were incorporated into Cox proportional hazards models to develop SIDPs, one for triage and one for prognosis (SIDP-T and SIDP-P, respectively). SIDP-T predictors were based on self-reported information at emergency department (ED) presentation, and SIDP-P predictors included additional biomarkers obtained in the ED. Models were developed using data from one hospital (n=106), followed by internal validation using bootstrapping and external validation using a data set (n=98) from 35 different hospitals.RESULTSSIDP-T found age, estimated diquat amount, heart rate, and Glasgow Coma Scale score to be the key predictors, achieving a C-index of 0.79 (0.70, 0.88), positive predictive value of 0.86 (0.49, 0.99) and negative predictive value of 0.76 (0.66, 0.83) in external validation. SIDP-P included age, initial plasma diquat concentration, white blood cell count, and aspartate aminotransferase, with C-index of 0.82 (0.74, 0.90), positive predictive value of 1 (0.51, 1) and negative predictive value of 0.74 (0.65, 0.82) on the external validation set.CONCLUSIONOur derived severity indices can provide rapid mortality prediction. SIDP-T uses self-reported information and basic vital signs at ED admission, and SIDP-P adds biomarkers and accurately predicts 28-day outcome.