 |
|
 |
 |
        |
|
IS APACHE II A VALID MODEL IN THE ELDERLY?
Sood A, Verhulst S, and Hopkins-Price P. (2006). Is APACHE II a Valid Model in the Elderly? Proceedings of the American Thoracic Society, 3 (Abstracts Issue), A380
RATIONALE: APACHE II, the most common prognostic model used in the Intensive Care Units (ICU) across North America, has not been validated for the elderly ICU population, particularly the ‘oldest old’. This segment of the population has shown a tremendous increase in recent years. The objective of this secondary data analysis is to externally validate and compare the predictive accuracy of the APACHE II prognostic model in the various age groups admitted to the ICU. METHODS: This was a multi-center, multi-national, historical cohort study involving 40,898 patients admitted to more than 120 intensive care units across United States and Canada and enrolled in Cerner Project IMPACT. Subjects were stratified into the following age groups: the reference group (age 18-64 years), the ‘young old’ group (age 65 to 74 years), the ‘old old’ group (75 to 84 years) and the ‘oldest old’ group (85 years and older). External validation of APACHE II prognostic model was performed on this sample population using calibration tests which assessed how well the model predictions compared with the observed outcomes and discrimination tests which assessed how well the model can distinguish between observations with a positive or a negative outcome. Model calibration was assessed statistically using Hosmer-Lemeshow C- statistic and model discrimination was assessed statistically using area under the ROC curve. RESULTS: Those over the age of 65 years accounted for 51.35% of all patients admitted to the ICU, with the ‘oldest old’ group accounting for 7.45% of these admissions. The elderly populations admitted to the ICU had a smaller proportion of men, non-whites and Hispanics, and less frequently returned to baseline status after hospitalization (p < 0.01 for each comparison). The APACHE II provides for good discrimination in all age groups (area under the ROC curve between 0.75-0.85), although the discriminative power is least robust in the ‘oldest old’. The tests of calibration reveal that there is no significant difference between model prediction and observed outcomes in the ‘oldest old’ age group. (In fact, the degree of correspondence between predicted and observed survival over the entire range of risk in this population segment is significantly better than that of the 18-64 years and 65-74 years age groups). The relatively poor calibration seen in the remaining age groups is likely the result of ‘case mix’ factor. CONCLUSION: APACHE II prognostic model is a valid scoring system in the ICU for the ‘oldest old’ category of elderly in this sample population.
Please note that specific medical questions can not be answered via this website.
Copyright © 2006, Board of Trustees, Southern Illinois University - Privacy Policy
Page Last Updated June 26, 2008