Letter to the editor: Derivation and validation of a prediction rule for sedative-associated delirium during acute respiratory failure requiring mechanical ventilation.

Prendergast and colleagues developed a prediction rule for sedative-associated delirium in mechanically ventilated patients, with good discriminatory ability (AUROC 0.83 in derivation, 0.70 in validation). However, we suggest three improvements: First, the model should specify the decision timepoint, as sedation is frequently adjusted after intubation. Second, using cumulative doses alone may not fully reflect sedation strategies, and incorporating sedation depth (e.g., Richmond Agitation-Sedation Scale) could enhance its relevance. Third, the tool's transportability depends on the delirium assessment method. Recalibration methods, such as adjusting intercepts and slopes, should be considered to ensure consistent performance across settings. Incorporating sedation regimen choices, such as dexmedetomidine versus lorazepam, would also improve clinical applicability. These refinements could strengthen the model's role as a decision-support tool for individualized sedation management.