Recent data indicate that a substantial number of cardiology patients with chest pain and no clinical evidence of coronary artery disease suffer from panic disorder. Discriminant function analysis reveals that a self-report anxiety measure alone, or in conjunction with minimal demographic information, can predict whether such patients have panic disorder. Predictive accuracy ranges from 69 to 76% correct classification, significantly improving upon chance predictions alone. A 'split-half' design was used in order to cross-validate predictive equations, and the total sample was also analyzed so as to provide the most stable equation. Clinical and future research implications of these findings are discussed.