OBJECTIVE: To compare three methods for rating legitimate use of psychiatric emergency services (PES) in order to develop criteria that can differentiate appropriate from inappropriate PES service requests. METHOD: Ratings of PES visits by treating physicians and ratings of the same visits made during review of medical records. STUDY DESIGN: Two previously used methods of identifying justified PES service use were compared with the treating physician's rating of the same: (1) hospitalization as visit outcome and (2) retrospective chart ratings of visit characteristics using traditional medico-surgical criteria for "emergent" illness episodes. DATA EXTRACTION METHODS: Data were extracted through use of a physician questionnaire, and medical and administrative record review. PRINCIPAL FINDINGS: Agreement between the methods ranged from 47.1 percent to 74.1 percent. A total of 21.7 percent of visits were rated as true health "emergencies" by the traditional definition, while 70.4 percent of visits were rated as "necessary" by treating physicians, and 21.0 percent resulted in hospitalization. Acuteness of behavioral dyscontrol and imminent dangerousness at the time of the visit were common characteristics of appropriate use by most combinations of the three methods of rating visits. CONCLUSIONS: The rating systems employed in similar recent studies produce widely varying percentages of visits so classified. However, it does appear likely that a minimum of 25-30 percent of visits are nonemergent and could be triaged to other, less costly treatment providers. Proposed criteria by which to identify "legitimate" psychiatric emergency room treatment requests includes only patient presentations with (a) acute behavioral dyscontrol or (b) imminent dangerousness to self or others.