We conducted two studies with medically hospitalized cancer and acquired immunodeficiency syndrome (AIDS) patients to assess the reliability and validity of a new measure of delirium severity, the Memorial Delirium Assessment Scale (MDAS). The first study used multiple raters who jointly administered the MDAS to 33 patients, 17 of whom met DSM III-R/DSM IV criteria for delirium, 8 met diagnostic criteria for another cognitive impairment disorder (for example, dementia), and 8 had non-cognitive psychiatric disorders (for example, adjustment disorder). Results indicate high levels of inter-rater reliability for the MDAS (0.92) and the individual MDAS items (ranging from 0.64 to 0.99), as well as high levels of internal consistency (coefficient alpha = 0.91). Mean MDAS ratings differed significantly between delirious patients and the comparison sample of patients with other cognitive impairment disorders or no cognitive impairment (P < 0.0002). The second study compared MDAS ratings of 51 medically hospitalized delirious patients with cancer and AIDS made by one clinician to ratings on several other measures of delirium (Delirium Rating Scale, clinician's ratings of delirium severely) and cognitive functioning (Mini-Mental State Examination) made by a second clinician. Results demonstrated a high correlation between MDAS scores and ratings on the Delirium Rating Scale (r = 0.88, p < 0.0001), the Mini-Mental State Examination (r = -0.91, P < 0.0001), and clinician's global ratings of delirium severity (r = 0.89, P < 0.0001). Thus, our findings indicate that the MDAS is a brief, reliable tool for assessing delirium severity among medically ill populations that can be reliably scored by multiple raters. The MDAS is highly correlated with existing measures of delirium and cognitive impairment, yet offers several advantages over these instruments for repeated assessments which are often necessary in clinical research.