OBJECTIVES: The aim of this study is to evaluate the accuracy of non-invasive diagnosis methods for ischemic heart disease in patients with chest pain. METHODS: A retrospective cohort study was performed on new patients who have not received any diagnosis of ischemic heart disease or treatment before among those over 30 years of age who visited the cardiology outpatient clinic between 2006 to 2008 in a single medical institution. Among non-invasive diagnostic methods, stress ECG(sECG), myocardial SPECT, and CT coronary angiography(CTCA) were included in our study. Since coronary angiography(CAG) was not performed in all patients, the comparison of test accuracy was calculated by correcting referral bias to CAG depending on non-invasive test results based on the Bayes Theorem, and by reflecting the 1-year follow-up results. RESULTS: Among 4743 patients selected, 2485 patients received more than one of the following non-invasive tests: sECG, SPECT, and CTCA for differential diagnosis. sECG was performed in 853 patients(34.3%), SPECT in 997 patients(40.1%), and CTCA in 635 patients(25.6%). A total of 592 patients(23.8%) received CAG among the 2485 patients. Test indices adjusted for referral bias using the 2 methods were a sensitivity of 49% or 80% in sECG, 46% or 75% in SPECT and 57% or 88% in CTCA, a specificity of 65% or 62%, 81% or 78% in SPECT and 96% or 81% in CTCA.  A likelihood ratio for a positive test were 1.39 or 2.08 in sECG, 2.47 or 3.38 in SPECT and 14.14 or 4.65 in CTCA, and a likelihood ratio for a negative test were 0.79 or 0.33 in sECG, 0.66 or 0.32 in SPECT and 0.45 or 0.14 in CTCA. CONCLUSIONS: In conclusion, the accuracy of diagnosing ischemic heart disease was the highest in CTCA, followed by myocardial SPECT and sECG by correcting referral bias to CAG.