Development and internal validation of an emergency department mortality prediction score for older adults with acute poisoning

Introduction Poisoning in older adults is an increasing clinical toxicology problem because age-related physiologic vulnerability, polypharmacy, intentional self-poisoning, and exposure to highly toxic substances can complicate early risk assessment. Existing severity scores lack age-specific validation and often require variables unavailable during initial emergency department assessment.Methods This retrospective observational study used data from six hospitals participating in the fall/slip, self-harm/suicide, and poisoning divisions of the Emergency Department-Based Injury In-depth Surveillance system maintained by the Korea Disease Control and Prevention Agency. Patients aged 65 years or older who presented with acute poisoning between January 2020 and December 2024 were included. The outcome was all-cause mortality, defined as emergency department death or in-hospital death. Candidate predictors included age; sex; Alert, Voice, Pain, Unresponsive scale; vital signs; poisoning substance category; self-harm/suicidal intent; and alcohol-related presentation. A multivariable logistic regression model was internally validated using five-fold cross-validation and bootstrap optimism correction.Results Among 1,774 patients, overall mortality was 7.2% (127/1,774). The full model had an apparent area under the receiver operating characteristic curve of 0.895, five-fold cross-validated value of 0.860, optimism-corrected value of 0.877, area under the precision-recall curve of 0.472, and Brier score of 0.049. Predictors associated with mortality included older age, impaired mental status, lower systolic blood pressure, higher pulse rate, lower body temperature, high-risk poisoning substances, and self-harm/suicidal intent. The five-item bedside warning score had area under the receiver operating characteristic curve of 0.834; among patients with a score of ≥3, mortality was 27.5%.Discussion The full model provides individualized probability estimation, whereas the bedside score provides immediate warning-level risk stratification.Conclusion An initial-assessment model and a five-item bedside score stratified mortality risk in older adults with acute poisoning. External validation is required before broader implementation.