Detection of delirium in the intensive care unit: comparison of confusion assessment method for the intensive care unit with confusion assessment method ratings

OBJECTIVES: To compare the Confusion Assessment Method (CAM) and CAM for the Intensive Care Unit (CAM-ICU) methods for detecting delirium in alert, nonintubated older ICU patients. DESIGN: Comparison study. SETTING: Fourteen-bed medical ICU of an 800-bed university teaching hospital. PARTICIPANTS: Twenty-two patients aged 65 and older admitted to the ICU. MEASUREMENTS: Two blinded, trained clinician-researchers who had undergone interrater reliability testing interviewed patients separately, usually within 10 minutes of each other (up to 120 minutes). Each researcher examined patients for the four key CAM criteria: acuteness, inattention, disorganized thinking, and altered level of consciousness. One researcher used the CAM method with the Mini-Mental State Examination and Digit Span; the other researcher used the CAM-ICU method with nonverbal cognitive and attention tasks. RESULTS: Rates of delirium were 68% according to CAM and 50% according CAM-ICU. Comparing the two methods, agreement was 82%, with a kappa of 0.64. Using the CAM as the reference standard, the CAM-ICU had a sensitivity of 73% (95% confidence interval (CI)=60-86) and specificity of 100% (95% CI=56-100). There were four false-negative ratings using the CAM-ICU. Reasons for disparate results were that the CAM used more-detailed cognitive testing that detected more deficits (3 patients) and the time elapsed (90 minutes) between ratings in one patient with markedly fluctuating symptoms. CONCLUSION: CAM and CAM-ICU agreement was moderately high. Although the CAM-ICU is recommended for ICU patients because of its brevity and ease of use, the standard CAM method may detect more subtle cases of delirium in nonintubated, verbal ICU patients.