Abstract Background and Objectives: Early post-stroke delirium (PSD) is difficult to diagnose and complicates recovery. We aimed to evaluate two validated delirium screening tools, focusing on patients with stroke-related higher-order cognitive deficits (HOCDs). Methods: We prospectively analyzed the delirium features and compared the diagnostic accuracy of the Confusion Assessment Method for ICU (CAM-ICU) and the Intensive Care Delirium Screening Checklist (ICDSC) with Diagnostic and Statistical Manual -5 (DSM-5) criteria in 164 patients with acute ischemic stroke. Early PSD was defined as delirium occurring within 72 h of admission. Results: Early PSD occurred in 32.3% patients, two-thirds of whom had HOCDs such as aphasia or neglect. Individual delirium features were frequent in nondelirious patients with HOCDs, inattention (Area under curve AUC .98, odds ratio OR 28.88), psychomotor changes (AUC.95, OR 10.60), and impaired arousal (AUC.95, OR 8.57) had maximum discriminating value on logistic regression analysis. When compared with DSM-5, CAM-ICU (sensitivity 81%, specificity 84%, AUC.97) had better accuracy than ICDSC (sensitivity 70%, specificity 86%, AUC.96); however, both were noninferior. In patients with HOCDs, both had markedly reduced specificities. Adding psychomotor changes to patients who were CAM-ICU positive (CAM-ICU Plus) improved its specificity. CAM-ICU Plus (sensitivity 79%, specificity 92%, AUC.98) had the best accuracy among all the tools. Conclusions: Early PSD is common after ischemic stroke. Available screening tools have lower specificity in patients with post-stroke cognitive deficits. CAM-ICU performed better than ICDSC. CAM ICU Plus improves the specificity of CAM ICU, which needs to be confirmed by further multicenter studies. Stroke-specific delirium screening tools are required for improving outcomes.
Somarajan et al. (Tue,) studied this question.