Poststroke delirium incidence and outcomes: validation of the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU).

OBJECTIVE: To describe the epidemiology and time spectrum of delirium using Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria and to validate a tool for delirium assessment in patients in the acute poststroke period.
DESIGN: A prospective observational cohort study.
SETTING: The stroke unit of a university hospital. PATIENTS: A consecutive series of 129 patients with stroke (with infarction or intracerebral hemorrhage, 57 women and 72 men; mean age, 72.5 yrs; age range, 35-93 yrs) admitted to the stroke unit of a university hospital were evaluated for delirium incidence.
INTERVENTIONS: None.
MEASUREMENTS AND MAIN RESULTS: Criterion validity and overall accuracy of the Czech version of the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) were determined using serial daily delirium assessments with CAM-ICU by a junior physician compared with delirium diagnosis by delirium experts using the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria that began the first day after stroke onset and continued for at least 7 days. Cox regression models using time-dependent covariate analysis adjusting for age, gender, prestroke dementia, National Institutes of Stroke Health Care at admission, first-day Sequential Organ Failure Assessment, and asphasia were used to understand the relationships between delirium and clinical outcomes. An episode of delirium based on reference Diagnostic and Statistical Manual assessment was detected in 55 patients with stroke (42.6%). In 37 of these (67.3%), delirium began within the first day and in all of them within 5 days of stroke onset. A total of 1003 paired CAM-ICU/Diagnostic and Statistical Manual of Mental Disorders daily assessments were completed. Compared with the reference standard for diagnosing delirium, the CAM-ICU demonstrated a sensitivity of 76% (95% confidence interval [CI] 55% to 91%), a specificity of 98% (95% CI 93% to 100%), an overall accuracy of 94% (95% CI 88% to 97%), and high interrater reliability (κ = 0.94; 95% CI 0.83-1.0). The likelihood ratio of the CAM-ICU in the diagnosis of delirium was 47 (95% CI 27-83). Delirium was an independent predictor of increased length of hospital stay (hazard ratio 1.63; 95% CI 1.11-2.38; p = .013).
CONCLUSIONS: Poststroke delirium may frequently be detected provided that the testing algorithm is appropriate to the time profile of poststroke delirium. Early (first day after stroke onset) and serial screening for delirium is recommended. CAM-ICU is a valid instrument for the diagnosis of delirium and should be considered an aid in delirium screening and assessment in future epidemiologic and interventional studies in patients with stroke.