Development and validation of a predictive model for short-term symptom relief after organophosphate poisoning

BACKGROUND: Acute organophosphate (OP) poisoning outcomes vary significantly despite standard atropine/pralidoxime (PAM) therapy. Early identification of patients likely to improve rapidly could optimize resource allocation and individualized care. METHODS: This retrospective study analyzed 138 adult acute OP poisoning cases (2018-2023). Patients were randomly split into training (n = 96) and validation (n = 42) cohorts. LASSO regression analyzed 14 potential predictors (demographics, clinical presentation, labs) to select features with non-zero coefficients at the 1-SE λ threshold. These were used in multivariable logistic regression to develop a nomogram predicting clinical improvement (any reduction in the poisoning symptom score (PSS) within 48 h post-admission). Model validation included discrimination (ROC/AUC), calibration [calibration plots and mean absolute error (MAE)], and clinical utility (decision curve analysis, DCA). RESULTS: Four independent predictors were identified: baseline PSS, admission serum cholinesterase, admission Glasgow Coma Scale (GCS) score, and major comorbidities. The nomogram showed strong discrimination (optimism-corrected AUC 0.930 in the training cohort; AUC 0.905 in the validation cohort) and good calibration (MAE 0.034 and 0.066). Decision-curve analysis showed positive net benefit over treat-all and treat-none across clinically relevant thresholds. CONCLUSION: We developed and validated a practical nomogram predicting 48-h symptom relief in acute OP poisoning. This tool can enhance early risk stratification, guide clinical decisions, and support efficient ICU resource utilization.