MULTIVARIATE-ANALYSIS OF PREDICTORS OF PR OGNOSIS AFTER ACUTE TRANSMURAL MYOCARDIAL-INFARCTION

The clinical data of 722 patients admitted for acute myocardial infarc tion to the coronary care unit of the Hannover Medical School were ret rospectively analyzed. Six hundred patients survived through the fifth day of their hospital stay. We evaluated 142 variables from each pati ent, i. e., previous cardiac manifestations, drug-history, acute compl ications, laboratory data, intensive care treatment and the 1-year out come. One-hundred-sixty-nine patients underwent cardiac catheterizatio n before being discharged from the hospital. Thirty-two variables show ed to be predictive of 1-year survival in the univariate analysis, alt hough performance of logistic regression analysis revealed only seven parameters to be independent predictors: age (p < 0.0001), glycoside i ntake before infarction (p = 0.0317), acute heart failure (p = 0.0005) , late (occurring after 48 h) ventricular tachycardia or fibrillation (p = 0.0003), maximum of serum creatine phosphokinase (p = 0.0129), ne w onset of atrial fibrillation (p = 0.0116), and use of dobutamine dur ing intensive care stay (p = 0.0014). With this combination of clinica l variables alone, using a survival probability partition value of 50% , the model had a sensitivity of 39% and a specificity of 96%, respect ively, 84% overall correct classification. Predictive accuracy for dea th was 71%, compared to a predictive accuracy for survival of 85%. Dia gnostic procedures performed after infarction were highly predictive i n the individual case, but they could not improve accuracy of the stat istical model. These data emphasize the importance of multivariate met hods to find suitable predictors for outcome after acute myocardial in farction.