Key points are not available for this paper at this time.
Plasma diquat concentration is prognostic in acute poisoning but often unavailable in resource-limited settings. We aimed to develop a bedside model using routine clinical variables for early risk stratification after hospital admission. This retrospective cohort study included 134 patients with acute diquat poisoning (2016–2025). Predictors were identified using Least Absolute Shrinkage and Selection Operator (LASSO) regression, with independent prognostic effects verified by multivariable logistic regression. A nomogram was constructed and developed using a training cohort (n = 81), then validated on a temporally and randomly split validation cohort (n = 53). Model performance was assessed via discrimination area under the receiver operating characteristic curve (AUC-ROC), calibration Hosmer-Lemeshow test, and clinical utility decision curve analysis (DCA) and clinical impact curves. Bootstrap optimism correction was performed. The final model comprised five predictors: white blood cell count (WBC), plasma lactate (Lac), renal insufficiency, respiratory failure, and myocardial injury. It demonstrated good discrimination in the training set AUC-ROC 0.839, 95% confidence interval (CI) 0.754–0.925 and the validation set AUC-ROC 0.874, 95% CI 0.783–0.966, with satisfactory calibration (P = 0.107 and P = 0.824, respectively). The optimism-corrected C-index was 0.832. DCA suggested potential clinical net benefit within threshold probabilities of 5–75% (training) and 4–97% (validation) in this retrospective simulation. Clinical impact curves showed the model effectively stratified high-risk patients and accurately captured actual mortality events within these ranges. This study presents a clinically accessible nomogram for mortality risk stratification in acute diquat poisoning. Its promising preliminary performance warrants prospective, multi-center validation to confirm clinical utility. Chinese Clinical Trial Registry, ChiCTR2500098079 (retrospectively registered on 3 March 2025).
Zhang et al. (Mon,) studied this question.