[Hospital mortality in acute myocardial infarction: is it possible to predict using admission data?].

Purpose: To identify clinical variables on admission that are related to hospital mortality in acute myocardial infarction (AMI) and to generate a mathematic model to predict accurately this mortality.
Methods: Prospective study with 347 consecutive patients with AMI in which clinical variables related to mortality were identified by univariate and multivariate analysis. The mathematic model generated by multivariate logistic regression analysis was applied in each patient to determine his/her probability (P) of hospital death. Model's accuracy was validated by reliability and discrimination tests.
Results: Admission variables directly and independently related to hospital mortality: female gender, age, absence of history of hypertension, history of previous infarction, non-inferior AMI and Killip class. These six variables, when present cumulatively, showed increasing mortality rates. Mean P value for non-survivors was significantly greater than for survivors (43.2 +/- 31.4% vs 9.1 +/- 12.5%, p < 0.00001). Reliability of the model to predict death, assessed by stratifying patients in three risk groups (low, medium and high) or continuously (by linear regression analysis) showed excellent predictive performance. Discrimination between survivors and non-survivors, assessed by C-index (concordance probability), disclosed 85% rate of success.
Conclusion: Risk variables can be used in a mathematic model that is capable of predicting accurately in-hospital mortality of each patient with AMI. Mortality prediction can allow physicians to be more efficient in assessing risk-benefit ratios in these patients when faced with therapeutic decisions.