OBJECTIVES Physical restraints are often used to prevent falls and to secure medical devices in older people in hospitals. Restraint reduction has been advocated on the grounds that physical restraints have negative psychological effects and are not effective in preventing falls. The potential effect of restraint reduction on length of hospital stay (LOS) has not been investigated. This study was undertaken to compare the average length of stay of older patients in a convalescent medical ward setting before and after a restraint reduction program. DESIGN This is a retrospective study. SETTING A convalescent hospital in Hong Kong. PARTICIPANTS This study included 2000 patient episodes. MEASUREMENTS The use of physical restraint, LOS, and clinical outcomes of randomly selected patient episodes in the year before and after the implementation of a restraint reduction program were compared. The clinical outcomes included Modified Functional Ambulatory Categories and modified Barthel index. Subgroup analysis was performed on those with confusion as defined by dementia diagnosis, low abbreviated mental test score, or abnormal mental domain of Norton Score. RESULTS A total of 958 and 988 patient episodes admitted to 10 medical wards in a convalescent hospital in 2007 and 2009 were examined. There were no significant differences in the baseline characteristics of patients in the 2 years. With the implementation of the restraint reduction scheme, the rate of physical restraint use declined significantly from 13.3% in 2007 to 4.1% in 2009 for all patients. The average LOS of patients was significantly lower in the year after the implementation of restraint reduction (19.5 ± 20.7 versus 16.8 ± 13.4 days in 2007 and 2009 respectively, P < .001). On subgroup analysis, the reduction in LOS was significant in the cognitively impaired patients (23.0 ± 26.5 to 17.8 ± 15.0 days in 2007 and 2009 respectively, P < .001), but not in the cognitively normal patients. There were no significant differences between the 2 years in the incidence of fall, mobility, and activities of daily living on discharge. CONCLUSION Physical restraint reduction was associated with significant reduction in average length of stay in convalescent medical wards, especially in the cognitively impaired patients.