STUDY OBJECTIVE:To assess the prognostic performance of general severity systems (APACHE II [acute physiology and chronic health evaluation], simplified acute physiology score [SAPS II], and mortality probability models [MPM II]) in coronary patients and to derive new customized indexes for coronary patients using a reduced number of variables.
DESIGN:Inception cohort.
SETTING:Adult medical and surgical ICUs in 17 hospitals in Catalonia and the Balearic Islands.
PATIENTS:Four hundred fifty-six patients with acute myocardial infarction.
MEASUREMENTS AND RESULTS:The APACHE II, SAPS II, and MPM II variables and survival status at hospital discharge have been collected. Performance of the severity systems was assessed by evaluating calibration and discrimination. Logistic regression was used to customize the MPM II(24) and SAPS II indexes. Discrimination was high enough for all of the models. However, calibration of the MPM II(24) was not as satisfactory as for the other models. The MPM II(24) and SAPS II were both reduced to five variables (MPM II(24 cor:) age, PaO2, continuous vasoactive drugs, urinary output, and mechanical ventilation; SAPS II(cor:) age, PaO2/FI(O2) ratio, systolic BP, Glasgow coma score, and urinary output). Both models showed better calibration and discrimination than the original ones.
CONCLUSIONS:Prognostic indexes developed for multidisciplinary patients show good performance when applied to patients with acute myocardial infarction, but customization can reduce the number of variables necessary to compute them without a loss of, and a possible improvement in, prognostic accuracy.
DESIGN:Inception cohort.
SETTING:Adult medical and surgical ICUs in 17 hospitals in Catalonia and the Balearic Islands.
PATIENTS:Four hundred fifty-six patients with acute myocardial infarction.
MEASUREMENTS AND RESULTS:The APACHE II, SAPS II, and MPM II variables and survival status at hospital discharge have been collected. Performance of the severity systems was assessed by evaluating calibration and discrimination. Logistic regression was used to customize the MPM II(24) and SAPS II indexes. Discrimination was high enough for all of the models. However, calibration of the MPM II(24) was not as satisfactory as for the other models. The MPM II(24) and SAPS II were both reduced to five variables (MPM II(24 cor:) age, PaO2, continuous vasoactive drugs, urinary output, and mechanical ventilation; SAPS II(cor:) age, PaO2/FI(O2) ratio, systolic BP, Glasgow coma score, and urinary output). Both models showed better calibration and discrimination than the original ones.
CONCLUSIONS:Prognostic indexes developed for multidisciplinary patients show good performance when applied to patients with acute myocardial infarction, but customization can reduce the number of variables necessary to compute them without a loss of, and a possible improvement in, prognostic accuracy.