BACKGROUND AND AIMS: Increased gait variability is associated with a high risk of falling in older community-dwellers, but no information exists about the relationship between increased gait variability and falls occurring in older hospitalized patients. We therefore sought to determine, in an acute geriatric setting, whether gait variability in single- (i.e., usual walking) or dual-task conditions can predict inpatient falls. METHODS: Stride time variability was calculated in both single-task (i.e., usual walking) and dual-task conditions with a GAITRite-System in 13 male and 44 female patients (mean age=85.0, SD=6.6 yrs) consecutively admitted to the acute care geriatric department of Geneva University Hospitals, Switzerland. All participants were able to walk without assistive devices at day 3 post-admission. Falls during hospital stay were identified through the hospital accident reporting system. RESULTS: Ten fallers and 47 non-fallers were identified. The first fall events were significantly associated with the coefficient of variation of stride time in both walking conditions during hospital stay (OR 13.3, (95% CI 1.6-113.6), p=0.018 for usual walking; OR 8.6, (95% CI 1.9-39.6), p=0.006 for dual- task walking). Furthermore, the time elapsing between the first day of hospitalization and the first fall was significantly shorter when the cut-off value of stride time variability was calculated for dual-tasking compared with usual walking. The Cox regression model revealed that only the coefficient of variation of stride time during dual-task walking was significantly associated with the occurrence of the first fall event (p=0.006). CONCLUSION: Our results suggest that the degree of stride time variability in dual-task walking conditions distinguished fallers from non-fallers in a group of independently walking, older inpatients.