All patients over 65 in a district general teaching hospital (n = 204) were screened for cognitive impairment and emotional disorder using the Clifton Assessment Procedures for the Elderly, the Mini‐Mental State and the General Health Questionnaire. Patients scoring in the disordered range were psychiatrically assessed. These procedures gave an estimated prevalence of cognitive impairment of 22 per cent in the 164 patients satisfactorily assessed. Forty‐three per cent of patients scored beyond the GHQ cut‐off, but there was a high false positive rate.Cognitively impaired patients had a significantly longer hospital stay than the unimpaired. The majority (60 per cent) of these patients could not be discharged because of lack of an appropriate place elsewhere. Amongst all elderly patients whose discharge was prevented in this way, the cognitively impaired were markedly over‐represented. The data have implications for the efficient use of hospital beds and for the welfare of elderly patients in acute hospitals.