Objective: To identify predictors and outcomes associated with frequent emergency department (ED) users. Methods: Cross-sectional intake surveys, medical chart reviews, and telephone follow-up interviews of patients presenting with selected chief complaints were performed at five urban EDs during a one-month study period in 1995. Frequent use was defined by four or more self-reported, prior ED visits. Multivariate logistic regression identified predictors of frequent ED visitors from five domains (demographics, health status, health access, health care preference, and severity of acute illness). Associations between high use and selected outcomes were assessed with logistic regression models. Results: All study components were completed by 2,333 of 3,455 eligible patients (67.5%). Demographics predicting frequent use included being a single parent, single or divorced marital status, high school education or less, and income of less than $10,000 (1995). Health status predictors included hospitalization in the preceding three months, high ratings of psychological distress, and asthma. Health access predictors included identifying an ED or a hospital clinic as the primary care site, having a primary care physician (PCP), and visiting a PCP in the past month. Choosing the ED for free care was the only health preference predictive of heavy use. Illness severity measures were higher in frequent visitors, although these were not independently predictive in the multivariate model. Outcomes correlated with heavy use include increased hospital admissions, higher rates of ED return visits, and lower patient satisfaction, but not willingness to return to the ED or follow-up with a doctor. Conclusions: Frequent ED visits are associated with socioeconomic distress, chronic illness, and high use of other health resources. Efforts to reduce ED visits require addressing the unique needs of these patients in the emergency and primary care settings.