[Prognostic estimation in critical patients. Validation of a new and very simple system of prognostic estimation of survival in an intensive care unit]

Abstract

Objective: To make the validation of a new system of prognostic estimation of survival in critical patients (EPEC) seen in a multidisciplinar Intensive care unit (ICU).

Design and scope: Prospective analysis of a patient cohort seen in the ICU of a multidisciplinar Intensive Medicine Service of a reference teaching hospital with 19 beds.

Patients and method: Four hundred eighty four patients admitted consecutively over 6 months in 2003. Data collection of a basic minimum data set that includes patient identification data (gender, age), reason for admission and their origin, prognostic estimation of survival by EPEC, MPM II 0 and SAPS II (the latter two considered as gold standard). Mortality was evaluated on hospital discharge. EPEC validation was done with analysis of its discriminating capacity (ROC curve), calibration of its prognostic capacity (Hosmer Lemeshow C test), resolution of the 2 x 2 Contingency tables around different probability values (20, 50, 70 and mean value of prognostic estimation). The standardized mortality rate (SMR) for each one of the methods was calculated. Linear regression of the EPEC regarding the MPM II 0 and SAPS II was established and concordance analyses were done (Bland-Altman test) of the prediction of mortality by the three systems.

Results: In spite of an apparently good linear correlation, similar accuracy of prediction and discrimination capacity, EPEC is not well-calibrated (no likelihood of death greater than 50%) and the concordance analyses show that more than 10% of the pairs were outside the 95% confidence interval.

Conclusion: In spite of its ease of application and calculation and of incorporating delay of admission in ICU as a variable, EPEC does not offer any predictive advantage on MPM II 0 or SAPS II, and its predictions adapt to reality worse.