Dynamic evaluation of prognosis from time-dependent variables in acute myocardial infarction.

For predicting the 3 endpoints death, cardiac arrest, and cardiogenic shock within 44 days after admission for acute myocardial infarction (MI), a competing risk variant of the Cox multivariate model was developed. The population consisted of 1,140 patients with definite MI admitted within 24 hours of onset of symptoms. Prognostic variables from the entire hospitalization period were assessed. The time-dependent variables (occurrence of complications) were evaluated with occurrence up through the day before the actual prediction period started. The important prognostic variables for the endpoint death were ventricular fibrillation, age, congestive heart failure, and asystole. Variables for the endpoint cardiac arrest were congestive heart failure, ventricular premature beats, supraventricular tachycardia, extension of MI, and age. Variables for the endpoint cardiogenic shock were cardiac arrest, age, congestive heart failure, previous MI, and nodal rhythm. By using a hazard function for each endpoint and the coefficients for the variables entered, it is possible to estimate a total risk of death, cardiac arrest, or cardiogenic shock for the individual patient. This prediction can be updated during the course of hospitalization according to the occurrence of the new complications. The model can be directly utilized to assess risk.