We sought to identify clinical risk factors for falls in people with advanced Alzheimer disease (AD) in a prospective longitudinal observational study set in specialized AD care units. Forty-two patients with probable or possible AD were recruited. Age, sex, Mini-Mental Status Examination, Clinical Dementia Rating Scale, Neuropsychiatric Inventory/Nursing Home, Morse Fall Scale (MFS), modified Unified Parkinson's Rating Scale (mUPDRS), and gait parameters using a GAITRite Gold Walkway System with and without dual-task performance were examined. Time to a first fall was the primary outcome measure, and independent risk factors were identified. Participating subjects were old (non-fallers age, 82.3 +/- 6.7 years; fallers: 83.1 +/- 9.6 years; p = 0.76) and predominantly women (36 female/6 male). Fallers did not differ from non-fallers on any parameter except the MFS (non-fallers: 35.6 +/- 26.1; fallers: 54.4 +/- 29.8; p = 0.04), the UPDRS (non-fallers: 4.75 +/- 3.98; fallers: 7.61 +/- 4.3, p = 0.03) and cadence (steps per minute: non-fallers: 102.3 +/- 12.3; fallers: 91.7 +/- 16, p = 0.02). Fallers and non-fallers were equally affected by dual-task performance. The hazard ratios for MFS, UPDRS, and cadence were not affected by adjusting for age, sex, MMSE, or NPI scores. In conclusion, falls in advanced AD can be predicted using simple clinical measures of motor impairment or cadence. These measures may be useful for targeting interventions.