| Sumario: | Background: Administrative health data are widely used for suicide attempt surveillance yet concerns remain about accuracy. The Victorian Emergency Minimum Dataset (VEMD) Human Intent Descriptor is an administrative coding system for classifying self-harm and suicidality in emergency department (ED) presentations. This study evaluates its accuracy in detecting suicide attempts by comparing it to clinician-applied Columbia Classification Algorithm of Suicide Assessment (C-CASA) ratings from medical records. Method: This cross-sectional validation study examined 607 ED presentations referred to psychiatric triage across three hospitals in August 2020. C-CASA classifications were compared with corresponding VEMD Human Intent Descriptor data. Sensitivity, specificity, predictive values, likelihood ratios, and Cohen's kappa were calculated. Receiver operating characteristic (ROC) curves assessed overall discrimination. Results: The VEMD descriptor demonstrated high specificity (99.0%) but low sensitivity (25.0%–27.3%), indicating many false negatives. The ROC analysis showed poor discriminatory ability (area under the curve = 0.62–0.63). Forty percent of missed cases were captured in ED diagnoses, highlighting gaps in coding accuracy. Conclusion: While the VEMD descriptor reliably confirms suicide attempts, its poor sensitivity limits its utility for surveillance. Findings underscore the need for improved coding protocols and alternative detection strategies to enhance suicide attempt surveillance in ED settings.
|
|---|