The objective was to examine differences in symptom severity assessment by emergency department (ED) patients and by emergency physicians (EPs) and to relate these assessments with case management and disposition. The design was prospective convenience sample of ED patients. The setting was a U.S. university hospital ED with an annual ED patient census 28,000. The participants were all ED patients registered when first author was in ED; excluded were patients treated by the major trauma response team and those with a psychiatric chief complaint. All patients were interviewed by the first author and asked to classify their symptoms as emergent, urgent, or nonurgent; the EP attending classed patients' symptoms at presentation and after work-up was complete. Three hundred-one cases were entered in the study from May to August 1996. Although 28% of ED patients self-rated their symptoms as nonurgent, 5% of this group required hospital admission. Of this group 35% were assessed by the EP attending as having required emergent or urgent ED care. Of this group 5% also rated by the EP initially as nonurgent had their case severity upgraded after work-up. Reliance on either patient symptom self-assessment or physician screening assessment by telephone to determine appropriateness of an ED visit is not reliably safe for at least 5% of presenting patients. Even prospective ED visit severity assessment does not reliably identify "unnecessary" ED visits.