Study objectives: We determine the diagnostic accuracy of a noninvasive clinical screening instrument for the detection of acute coronary syndrome (ACS) in emergency department (ED) patients. The screening instrument is composed of clinical data input into a computer to estimate the probability of ACS. Methods: We aim to enroll 1,000 patients with complete data who were observed in an ED chest pain evaluation unit at Carolinas Medical Center, Duke University, and the University of California–Davis, and prospectively collect data to determine the pretest probability (PTP) of ACS for each patient using 3 methods of PTP assessment: (1) a computer-assisted, database-derived method; (2) the acute cardiac ischemia time-insensitive predictive instrument (ACI-TIPI) method; and (3) physician estimate to the nearest 1%. This information will be used to compare which method of PTP assessment categorizes the highest percentage of patients as being very low risk for ACS (ie, PTP Results: We completed enrollment on August 1, 2004, and follow-up and adjudication of outcomes continues. We enrolled a total of 1,154 patients and expect to obtain valid follow-up in 85% of subjects. Complete pretest probability are data available for 723 patients. Only 12 (1.6%) patients with a positive outcome were identified. The mean (median) pretest probability estimates by the 3 methods were 5.2% (4.0%) for physician estimate, 6.4% (4.3%) for attribute matching, and 13.6% (9.9%) for ACI-TIPI. The percentage of all patients estimated to have a pretest probability of ACS of less than 2% was 19% (95% confidence interval [CI] 16% to 22%) for clinicians, 34% (95% CI 31% to 38%) for attribute-matching, and 9% (95% CI 7% to 12%) for ACI-TIPI. No patient with a pretest probability of less than 2% estimated by any of the 3 methods developed ACS. Conclusion: At interim analysis, the rate of positive outcomes for ACS was unexpectedly low in this sample of patients evaluated in ED chest pain units. Attribute matching categorized significantly more patients as having a very low pretest probability compared with clinician estimate or ACI-TIPI.