Predicting outcome among intensive care unit patients using computerised trend analysis of daily Apache II scores corrected for organ system failure

Abstract

Daily Apache II scores were determined prospectively on 310 consecutive adult Intensive Care Unit (ICU) patients to reflect the dynamic pathophysiological processes affecting ICU patients. Organ failure scores were derived from the Apache II scores by applying a coefficient which corresponded to the number and duration of organ failures to account for the increased mortality associated with established major organ-system failures. Computerised trend analysis of data from the first 100 patients was used to develop criteria for predicting hospital outcome of the ICU patients. The analysis took into account the absolute value of the daily scores and the rate of change relative to that of the previous day. Allowance was made for changes in scores as a result of surgery or major latrogenic complications arising after admission to the ICU. The criteria were then tested prospectively on the next 210 consecutive ICU patients. Predictions by Apache II assessments were NOT used to influence clinical decisions during the study period. It was possible to predict with a specificity of 100% 37 out of the 72 deaths in the test group. The predictive power of daily organ failure scores was superior to those obtained from a single Apache II assessment or from daily Apache II scores by a factor of 5.3 and 1.4 respectively. The estimated long-term risk of a false prediction is 1.4% at the 95% confidence level.