Limitations of risk score models in patients with acute chest pain

Objectives: Cardiac multidetector computed tomography (CMCT) has potential to be used as a screening test for patients with acute chest pain, but several tools are already used to risk-stratify this population. Risk models exist that stratify need for intensive care (Goldman), short-term prognosis (Thrombolysis in Myocardial Infarction, TIMI), and 1-year events (Sanchis). We applied these cardiovascular risk models to candidates for CMCT and assessed sensitivity for prediction of in-hospital acute coronary syndrome (ACS). We hypothesized that none of the models would achieve a sensitivity of 90% or greater, thereby justifying use of CMCT in patients with acute chest pain. Methods: We analyzed TIMI, Goldman, and Sanchis in 148 consecutive patients with chest pain, nondiagnostic electrocardiogram, and negative initial cardiac biomarkers who previously met inclusion and exclusion criteria for the Rule-Out Myocardial Infarction Using Coronary Artery Tomography Study. ACS was adjudicated, and risk scores were categorized based on established criteria. Risk score agreement was assessed with weighted κ statistics. Results: Overall, 17 (11%) of 148 patients had ACS. For all risk models, sensitivity was poor (range, 35%-53%), and 95% confidence intervals did not cross above 77%. Agreement to risk-classify patients was poor to moderate (weighted κ range, 0.18-0.43). Patients categorized as "low risk" had nonzero rates of ACS using all 3 scoring models (range, 8%-9%). Conclusions: Available risk scores had poor sensitivity to detect ACS in patients with acute chest pain. Because of the small number of patients in this data set, these findings require confirmation in larger studies.