Prognostic value of the Thrombolysis in Myocardial Infarction risk score in a unselected population with chest pain. Construction of a new predictive model

Abstract

Introduction The Thrombolysis in Myocardial Infarction (TIMI) risk score (TRS) has proven to be a useful and simple tool for risk stratification of patients with chest pain in intermediate- and high-risk populations. There is little information on its applicability in daily clinical routine with unselected populations. Aims The aims of the study were to prospectively analyze the predictive value of the TRS in a heterogeneous population admitted for chest pain and to construct where possible a new modified model with a greater prognostic capacity. Population and Methods Seven hundred eleven consecutive patients were admitted over a 1-year period to the cardiology unit for chest pain without ST-segment elevation. Thrombolysis in Myocardial Infarction risk score variables, relevant medical history variables, in-hospital examination results, and therapy information were collected. Cardiac events at 1 and 6 months were recorded. Results Seventy-one (9.8%) patients had a compound event (myocardial infarction/revascularization/cardiac death) at 6 months. On multivariate analysis, the variables associated with cardiac events were left ventricular ejection fraction (EF) of <35% (hazard ratio [HR] = 2.9, P = .002), diabetes (HR = 1.8, P = .02), and TRS (HR = 1.3, P = .007). Events at 6 months were 2.3% for a TRS of 0/1, 4.2% for 2, 10.2% for 3, 11.0% for 4, and 18.7% for a score of more than 5. A new modified scale was constructed to include EF and diabetes as independent variables, and this yielded an increase of 44% in the combined event at 6 months per score unit increase (HR = 1.44, P = .001). The modified scale showed a greater predictive capacity than the original model. Conclusions The TRS is an important short- and long-term prognostic predictor when applied to an unselected population consulting for chest pain. The inclusion of diabetes and EF as variables in the model increases predictive capacity at no expense to simplicity.