This article reports on adopting Gustafson's multiattribute utility method to develop a specific momentary severity index (vis-à-vis severity over time) for nontraumatic chest pain patients. A panel of physicians identified an appropriate set of indicators for the index. Three mathematic models were generated on the basis of these indicators with each model predicting no less than 81% of the variance in overall scores of 96 simulated patients with chest pain. High interrater reliability (0.94) was demonstrated. Congruent validity coefficients between the three model scores and the overall scores vary between 0.71 and 0.92 depending on how the panel physicians are grouped. Results indicate the feasibility of simulating physician judgments regarding patient severity in chest-pain-related cases.