Prediction of survival of critically ill patients by admission comorbidity

J Clin Epidemiol. 1996 Jul;49(7):743-7. doi: 10.1016/0895-4356(96)00021-2.

Abstract

The objective of this study was to determine how well the Charlson index of comorbidity would predict mortality of critically ill patients; and how the predictive ability of the index would compare with that of the comorbidity component (Chronic Health Points) of the APACHE II system. This prospective cohort study included in its setting an intensive care unit (ICU) and intermediate ICU (IICU) in a teaching hospital. Patients included a previously assembled inception cohort of 201 patients consecutively admitted to either unit, followed until death or discharge from the hospital, excluding patients admitted after coronary artery bypass grafting, for planned dialysis, or transferred to the IICU from another intensive care unit. Main outcome measures were recorded as death in hospital versus survival at discharge. For each patient we had prospectively obtained all data necessary to predict the probability of in-hospital death using the APACHE II system, and to classify comorbidity using the Charlson index. The Charlson index had significant ability to discriminate between patients who would live and who would die (ROC curve area = 0.67, SE = 0.05). The Chronic Health Points component of APACHE II had no significant discriminating ability (ROC area = 0.57, SE = 0.05), although the full APACHE II system was an excellent predictor (area = 0.87, SE = 0.04). Logistic regression analyses suggested that the Charlson index could contribute significant (p = 0.03) prognostic information to that obtained from the components of APACHE II other than Chronic Health, i.e., acute physiological derangement, age, and reason for admission, but the Chronic Health Points component of APACHE II could not so contribute to the rest of APACHE II (p = 0.19). Our conclusion is that use of the detailed information about comorbidity captured by the Charlson index could improve prognostic predictions even for critically ill patients.

Publication types

  • Research Support, Non-U.S. Gov't
  • Research Support, U.S. Gov't, P.H.S.

MeSH terms

  • APACHE
  • Cohort Studies
  • Comorbidity*
  • Critical Illness / mortality*
  • Diagnostic Tests, Routine
  • Humans
  • Models, Statistical
  • Probability
  • Prospective Studies