Objective The objective of this study was to examine the psychometric quality of the 6‐Item Cognitive Impairment Test (6CIT) as a bedside screening for the detection of dementia in general hospital patients. Method Participants (N = 1,440) were inpatients aged ≥65 of 33 randomly selected general hospitals in Southern Germany. The 6CIT was conducted at bedside, and dementia was diagnosed according to DSM‐IV. Nursing staff was asked to rate the patients' cognitive status, and previous diagnoses of dementia were extracted from medical records. Completion rates and validity statistics were calculated. Results Two‐hundred seventy patients had dementia. Cases with delirium but no dementia were excluded. Feasibility was 97.9% and 83.3% for patients without and with dementia, respectively, and decreased from moderate (93.8%) to severe dementia (53%). The area under the curve of the 6CIT was 0.98. Sensitivity, specificity, accuracy, positive predictive value, and negative predictive value were calculated for the cutoffs 7/8 (0.96, 0.82, 0.85, 0.52, 0.99) and 10/11 (0.88, 0.95, 0.94, 0.76, 0.98). The nurse ratings and medical records information had lower validity statistics. Logistic regression analyses revealed that the 6CIT statistically significantly provided information above nurse ratings and medical records. Twenty‐five and 37 additional patients were correctly classified by the 7/8 and 10/11 cutoffs, respectively. Conclusion The 6CIT is a feasible and valid screening tool for the detection of dementia in older general hospital patients. The 6CIT outperformed the nurse ratings of cognitive status and dementia diagnoses from medical records, suggesting that standardized screening may have benefits with regard to case finding. Copyright © 2016 John Wiley & Sons, Ltd.