Predicting deaths among intensive care unit patients

Abstract

Daily APACHE II assessments to reflect the dynamic pathophysiologic processes affecting ICU patients were carried out prospectively on 212 adult ICU admissions for the purpose of predicting hospital outcome. Trend analysis of daily APACHE II scores was used to develop the first set of criteria for predicting deaths among the ICU patients. The analysis took into account the absolute value of APACHE II score of each day and the rate of change relative to that of the previous day. One hundred consecutive adult ICU admissions were used to determine the criteria that were tested on the following 112 consecutive adult ICU admissions. Daily APACHE II scores identified correctly 16 of the 34 patients who died. A second set of criteria was next derived by raising the values of the original limits in order to build-in greater margins of safety without too great a loss of sensitivity. The predictive power of daily APACHE II scores using the second set of criteria was superior to that of a single APACHE II score by a factor of 4. Predictions were not used to influence clinical decision-making during this study.