Résumé : BACKGROUND. Delirium is a common neuropsychiatric complication in patients with advanced cancer. The Memorial Delirium Assessment Scale (MDAS) is a recently developed 10-item severity rating instrument. The purpose of the current prospective study was to further assess the clinical utility, factor structure, and validity of the MDAS in a relatively homogeneous population of patients with advanced cancer. METHODS. Study entry of 104 patients occurred on their consecutive admission to a tertiary-level, acute palliative care unit in a university-affiliated teaching hospital. Patients underwent regular cognitive screening using the Mini-Mental State Examination, and serial monitoring of delirium using standardized semistructured interviews and MDAS ratings, up to the study endpoints of either patient discharge or death. RESULTS. Seventy-one patients met Diagnostic and Statistical Manual (of Mental Disorders)-IV criteria for a first episode of delirium. In 15 of 71 (21%) patients with a first episode of delirium, the first MDAS ratings were prorated because of dyspnea, fatigue, or profound delirium. In the remaining 56 patients (79%), the first MDAS ratings were rated fully and therefore evaluable. Correlations among the scale items ranged from moderate to low (correlation coefficient [r] = 0.68- 0.02). Analysis of the pattern of factor loadings identified two primary correlated factors: global cognitive (Factor II) and neurobehavioral (Factor II) (r = 0.33). Cronbach α coefficients for Factors I and II were 0.8 and 0.66, respectively, indicating a relatively high level of correlation for items within each. The Cronbach α coefficient for all 10 items was 0.78, suggesting a general underlying factor. In a larger sample of complete MDAS ratings (n = 330) a cutoff total MDAS score of 7 of 30 yielded the highest sensitivity (98%) and specificity (96%) for delirium diagnosis. The MDAS was correlated moderately with the Mini-Mental State Examination (r = 0.55). CONCLUSIONS. The authors concluded that the MDAS structure is representative of the many features of delirium, broadly grouped as global cognitive and neurobehavioral dimensions. Prorating item scores is necessary in approximately 20% of advanced cancer patients with delirium. This poses potential limitations on the applicability of the MDAS in research. Conversely, the ability to prorate item scores confers a clinical advantage to the instrument when assessing delirium in a patient population with advanced cancer. (C) 2000 American Cancer Society.