BACKGROUND: Falls remain a significant cause of morbidity in PD. Risk factors are not well understood. OBJECTIVE: In this study we explore risk factors for falls in PD utilizing the cross-sectional, baseline data in the National Parkinson Foundation Quality Improvement Initiative (NPF-QII) database. Subjects are being followed prospectively, and this study will provide the basis for subsequent longitudinal analyses. METHODS: A cross-sectional analysis of data from 2,876 ambulatory patients with Parkinson disease enrolled in the NPF-QII at 18 sites. Main outcome measure was falling history in the 3 months preceding assessment. The following were considered as possible predictor variables: age, sex, height, weight, body mass index, disease duration, age at disease onset, investigator's confidence in the diagnosis, Hoehn and Yahr stage, rest tremor, ability to stand unassisted, coexistent pathologies (cardiovascular, respiratory, diabetes, cancer, neurological, osteoarthritis, and "other" comorbidities), anticholinergics, antidepressants, antipsychotics, cognitive enhancers, deep brain stimulation surgery, timed-up-and-go, semantic fluency, and 5 word recall. Variables with associations to the outcome measure in univariate analyses were analyzed in multivariable models using logistic regression. RESULTS: 37.2% of subjects experienced falls. In the multivariable regression model the following variables were found to be independently associated with falls: disease duration; Hoehn and Yahr stage; absence of rest tremor; cardiovascular, arthritis, or "other" comorbidity; antidepressants; deep brain stimulation surgery; timed-up-and-go; and, semantic fluency. CONCLUSION: Disease duration but not age is independently associated with falls in Parkinson disease. Timed-up-and-go accurately reflects falls risk. Impaired semantic fluency is independently associated with falls, while verbal memory is not. Comorbidities, antidepressants, and deep brain stimulation also contribute to falls risk.