Cohort study reveals accurate prediction of toxic myocarditis in acute diquat poisoning, indicating high diagnostic utility for early risk assessment.
Key Points
To identify predictors and develop a clinical risk prediction model for clinically suspected toxic myocarditis following acute oral diquat poisoning.
Conducted a single-center investigation of 159 patients (aged 18–70 years) admitted within 24 hours of toxicologically confirmed diquat ingestion.
Partitioned data into a retrospective training cohort (n = 94, 2020–2024) and a subsequent temporal evaluation cohort (n = 65, 2024–2026).
Selected candidate variables and built the multivariable logistic model using LASSO regression, with internal validation through 1,000 bootstrap resamples.
Clinically suspected toxic myocarditis occurred in 43 of 94 patients (45.7%) in the training cohort and 29 of 65 patients (44.6%) in the temporal evaluation cohort.
The final model incorporated ingested dose, lactate, neutrophil-to-lymphocyte ratio, and creatinine, yielding an AUC of 0.896 (95% CI: 0.828–0.964) in training and 0.857 (95% CI: 0.760–0.954) upon temporal evaluation.
Calibration in the temporal cohort showed an intercept of −0.642 (95% CI: −1.953 to 0.228), indicating imprecise calibration that warrants further external validation.