The field identification of the patients who should be taken to trauma centers is a major problem of trauma care. This study appears to be the first to use multivariate analysis to systematically design a field triage instrument and to evaluate its performance in terms of a meaningful and plausible criterion for which patients ought to be treated at trauma centers. Four new triage instruments were created and their performance compared to that of 2 existing measures, the Trauma Score and the CRAMS scale. The new measures appear to perform better than the best of existing instruments tested, although independent validation is necessary. The one purely physiologic instrument tested appeared to be inferior to instruments that included anatomic and historic as well as physiologic indices. Simple checklists performed approximately as well as weighted scales. No instrument was found to be high in both sensitivity and positive accuracy.