Aim A substantial number of elderly patients take benzodiazepines (BZD) regularly despite concerns about toxicity and possible dependence, and there are relatively few data to guide clinicians' decisions regarding discontinuing benzodiazepines in the elderly. Methods We carried out a retrospective chart review of 75 elderly patients admitted to a chronic medical/psychiatric unit who were taking a standing dose of benzodiazepines on admission, comparing 40 patients who discontinued benzodiazepines versus 35 who did not discontinue. Purpose We examined the association of BZD discontinuation versus continuation with clinical outcomes on discharge, and further examined clinical characteristics associated with BZD discontinuation. Results Discontinuers had shorter length of stay without evidence of worse cognitive and functional outcome except a trend toward increased incidence of agitation. Logistic regression models suggested anxiety, higher age and higher dose of antidepressants at the beginning were significantly related to successful discontinuation during admission after regression. Conclusion These data imply that BZD withdrawal during admission can be safe and feasible in many elderly frail patients, and that withdrawal might be associated with shorter duration of chronic hospitalization. Geriatr Gerontol Int 2014; 14: 388–394.