• P Sikka, W M Jaafar, E Bozkanat, and A A El-Solh.
• Department of Medicine, State University of New York at Buffalo, School of Medicine and Biomedical Sciences, NY 14215, USA.
• Intensive Care Med. 2000 Dec 1;26(12):1803-10.

ObjectiveTo evaluate the predictive ability of three severity of illness scoring systems in elderly patients with severe pneumonia requiring mechanical ventilation compared to a younger age group.DesignProspective cohort study.SettingTwo university-affiliated tertiary care hospitals.Patients And ParticipantsOne hundred four patients 75 years of age and older and 253 patients younger than 75 years of age enrolled from medical intensive care units.Measurements And ResultsProbabilities of hospital death for patients were estimated by the Acute Physiology and Chronic Health Evaluation (APACHE) II, the Mortality Probability Model (MPM) II and the Simplified Acute Physiology Score (SAPS) II. Predicted risks of hospital death were compared with observed outcomes using three methods of assessing the overall goodness of fit. The actual mortality of the elderly group was 54.87 % (95 % confidence interval [CI]: 45.2-64.4 %) compared to 28.9 % (95 % CI, 23.3-34.4 %) in the younger age group. There was a significant difference in the predictive accuracy of the scoring systems as assessed by the c-index, which is equivalent to the area under the receiver operator characteristics (ROC) curve, between the two groups, but not within individual groups. Calibration was insufficient for APACHE II and SAPS II in the elderly cohort as in-hospital mortality was lower than the predicted mortality for both models.ConclusionsAlthough the three severity of illness scoring systems (APACHE II, MPM II and SAPS II) demonstrated average discrimination when applied to estimate hospital mortality in the elderly patients with severe pneumonia, MPM II had the closest fit to our database. Alternative modeling approaches might be needed to customize the model coefficients to the elderly population for more accurate probabilities or to develop specialized models targeted to the designed population.

37 keywords...

hide…