Since 2016, diquat has replaced paraquat in China, resulting in increased diquat poisoning cases. However, understanding of diquat poisoning is still limited. This study aimed to investigate the relationship between initial diquat plasma concentration, severity index, and in-hospital mortality in acute diquat poisoning cases. This retrospective cohort study, conducted from January 2016 to July 2023 in a tertiary care hospital, used univariate logistic regression to examine the link between the initial diquat plasma concentration, severity index, and in-hospital mortality in acute diquat poisoned patients. A receiver operating characteristic curve assessed the predictive value of these parameters for prognosis. Among the 87 participants, the median age was 32 years, 35 (40.2%) were female. The overall mortality rate was 37.9%. Logistic regression analysis revealed that the initial diquat plasma concentration and severity index were associated with increased in-hospital mortality. These factors also effectively predicted the prognosis of acute diquat poisoning, with an area under the receiver operating characteristic curve of 0.851 and an optimal diquat concentration threshold of 2.25 mg/L (sensitivity 90.9%, specificity 74.1%, P < 0.05) and an area under the receiver operating characteristic curve of 0.845 with an optimal cut-off value for the sevity index of 9.1 mg/L*min (sensitivity 97%, specificity 74.1%, P < 0.05). Our results are limited by the retrospective design of this study. However, if validated, these results could impact management strategies, especially in East Asia. Further research is needed due to potential confounding factors. The findings suggest that a higher initial plasma concentration and severity index in patients with acute diquat poisoning were correlated with higher in-hospital mortality. Prospective validation will confirm the predicative value of these findings.