Objective To measure the diagnostic accuracy of the 4A's test in screening for delirium in geriatric inpatients from culturally diverse backgrounds. Methods A prospective study was conducted with patients admitted to the geriatric and orthogeriatric services of a tertiary teaching hospital. Consenting participants aged 65 years and over were screened for delirium with the 4AT by nursing staff within 72 h of admission. The diagnosis of delirium was made separately by expert assessors, responsible for the participant's clinical care, blinded to the 4AT score, within 30 min of the 4AT assessment using the DSM 5 criteria and the Confusion Assessment Method. Interpreters were used for non‐English speaking patients. The Informant Questionnaire for Cognitive Decline in the Elderly was completed by a carer/relative to assess for probable dementia. Results A total of 257 participants (mean age 85) were recruited over five months. Delirium was diagnosed in 159 (62%) by the expert assessors and 158 (62%) by the 4AT assessment. A total of 205 participants (80% of total population) had probable dementia. The sensitivity and specificity of the 4AT were 87% and 80%, respectively, in detecting delirium overall, 86% and 71% in people with probable dementia and 91% and 71% for non‐English speaking participants. The area under the receiver operating characteristic curve for delirium in the whole population was 0.92, 0.89 in the probable dementia subgroup and 0.90 in non‐English speaking participants. Conclusions The 4AT is a sensitive and specific screening tool for delirium in geriatric inpatients, including those with probable dementia or who are non‐English speaking. Copyright © 2016 John Wiley & Sons, Ltd.