Accurate prediction of the outcome of pediatric intensive care. A new quantitative method

We surveyed nine pediatric intensive care units (ICUs) to compare patient populations and to test prospectively the hypothesis that differences in mortality rates were due to differences in severity of illness. Age, clinical service, the reason for admission (emergency or scheduled), and the seriousness of the underlying chronic disease were recorded on admission. The severity of illness was assessed on the day of admission with a physiology-based measure, the Physiologic Stability Index. The resulting score was used to group patients according to mortality risk. The observed numbers of ICU survivors and nonsurvivors in each mortality-risk group from eight of the pediatric ICUs were compared with the predicted numbers of survivors and nonsurviors calculated from a mathematical function (logistic model) derived earlier from data on 822 patients at one of the institutions. Patient populations in the ICUs differed significantly with respect to age (range of medians, 15 to 36 months; P less than 0.0001), medical admissions (range, 39 to 81 percent; P less than 0.0001), emergency admissions (range, 53 to 91 percent; P less than 0.0001), and the percentage of patients with serious underlying chronic disease (range, 18 to 48 percent; P less than 0.0001). Mortality rates also differed significantly (range, 3.0 to 17.6 percent; P less than 0.0001), as did the Physiologic Stability Index scores (P less than 0.0001). The mathematical function based on the Physiologic Stability Index score and on age reliably predicted the outcomes in all ICUs. We conclude that differences in mortality rates among pediatric ICUs can be explained by differences in the severity of illness.