In the intensive care setting, delirium is a common occurrence that comes with subsequent adversities. Therefore, several instruments have been developed to screen for and detect delirium. Their validity and psychometric properties, however, remain controversial.

In this prospective cohort study, the Confusion Assessment Method for the Intensive Care Unit (CAM–ICU) and the Intensive Care Delirium Screening Checklist (ICDSC) were evaluated versus the DSM–IV–TR in the diagnosis of delirium with respect to their validity and psychometric properties.

Out of some 289 patients, 210 with matching CAM–ICU, ICDSC, and DSM–IV–TR diagnoses were included. Between the scales, the prevalence of delirium ranged from 23.3% with the CAM–ICU, to 30.5% with the ICDSC, to 43.8% with the DSM–IV–TR criteria. The CAM–ICU showed only moderate concurrent validity (Cohen's κ = 0.44) and sensitivity (50%), but high specificity (95%). The ICDSC also reached moderate agreement (Cohen's κ = 0.60) and sensitivity (63%) while being very specific (95%). Between the CAM–ICU and the ICDSC, the concurrent validity was again only moderate (Cohen's κ = 0.56); however, the ICDSC yielded higher sensitivity and specificity (78 and 83%, respectively).

In the daily clinical routine, neither the CAM–ICU nor the ICDSC, common tools used in screening and detecting delirium in the intensive care setting, reached sufficient concurrent validity; nor did they outperform the DSM–IV–TR diagnostic criteria with respect to sensitivity or positive prediction, but they were very specific. Thus, the non-prediction by the CAM–ICU or ICDSC did not refute the presence of delirium. Between the CAM–ICU and ICDSC, the ICDSC proved to be the more accurate instrument.