This study looked at associations of tranquilliser use and falls risk in a hospital population of confused and nonconfused patients. In a prospective observational study in a rehabilitation hospital for elderly patients, we followed 1025 consecutive patients. The number of fallers, recurrent fallers and total falls was recorded. Confused patients (p < 0.0001) and patients on tranquillisers (p = 0.001) were significantly more likely to fall than nonconfused patients and patients off tranquillisers. Confused patients on tranquillisers were significantly more likely to have recurrent falls (p = 0.026) when compared with confused patients off tranquillisers. The risk was apparent from admission, persisting throughout the first 30 days of stay. This was not noted for nonconfused patients. We identified a stratification of risk for falls with nonsignificant trends for confused and nonconfused patients on tranquillisers to be fallers and to have more falls compared with patients off tranquillisers. These data are associational and do not necessarily imply causality. There is however no evidence to recommend the routine withdrawal of tranquillisers from all patients. Any future research needs to include confused patients.